# MIT HEALS

Data: 11-01-2025 21:39:13

## Lista de Vídeos

1. [MIT HEALS Launch: Governor Healey remarks](https://www.youtube.com/watch?v=hglvsF6nAh4)
2. [MIT HEALS Launch: President Kornbluth remarks](https://www.youtube.com/watch?v=uw5e1XbPFfU)
3. [MIT HEALS Launch: Phillip Sharp keynote](https://www.youtube.com/watch?v=7HH1DDNpVkw)
4. [MIT HEALS Launch: Revolutions plenary session](https://www.youtube.com/watch?v=JFzCn1wTZ8w)
5. [MIT HEALS Launch: Expansion plenary session](https://www.youtube.com/watch?v=UKEF6ehVuWE)
6. [MIT HEALS Launch: Systems, spoken word performance](https://www.youtube.com/watch?v=EOR114NFNpo)
7. [MIT HEALS Launch: Brains breakout session](https://www.youtube.com/watch?v=lvMWYeM7PJI)
8. [MIT HEALS Launch: Biosphere breakout session](https://www.youtube.com/watch?v=kIZuU-vSSLA)
9. [MIT HEALS Launch: Healthcare breakout session](https://www.youtube.com/watch?v=eJhszcNUeYs)
10. [MIT HEALS Launch: Immunology breakout session](https://www.youtube.com/watch?v=V-XjgwVENq0)
11. [MIT HEALS Launch: Health plenary session](https://www.youtube.com/watch?v=AxFQ9chLlVI)
12. [MIT HEALS Launch: Entrepreneurship panel session](https://www.youtube.com/watch?v=iy7Hry92Xzw)
13. [MIT HEALS Launch: AI plenary session](https://www.youtube.com/watch?v=WCkj484Lw7Y)
14. [MIT HEALS Launch: Translation plenary session](https://www.youtube.com/watch?v=kVmqwAnp7Oo)
15. [MIT HEALS Launch: Collaboration panel session](https://www.youtube.com/watch?v=1GFQHdI_qsQ)
16. [MIT HEALS Launch: Connections plenary session](https://www.youtube.com/watch?v=aJDpsCJgx8Y)

## Transcrições

### MIT HEALS Launch: Governor Healey remarks
URL: https://www.youtube.com/watch?v=hglvsF6nAh4

Idioma: en

[MUSIC PLAYING]
I think we're really at the
beginning of a whole new era
of biological
understanding, and one
where I think MIT
is really extremely
well-positioned to lead.
The planet is changing, and
new diseases are coming.
The things that we
thought we were treating
are changing in real time, which
as inventors, is an opportunity.
But it's really
a call to action.
Things that might have
taken years in the past
can be done in weeks,
if not days, using
the power of computation.
And what better place to
invest in technology than MIT.
At MIT, it's not
simply the discovery
that's up at the top
of our ambition list.
It's also applications
of those discoveries
to making the world
better, to coming up
with better products,
better ways to do things.
One of the things that
we're excited always
to collaborate with our
colleagues at MIT around
is how do we make big leaps?
How do we make daring
departures from what
is obvious and adjacent?
And that's where many of
the breakthroughs come from.
We want to find a
mechanism to bring people
from lots of different
areas together so we
can get this synergistic
effect to solve
really important problems.
There's been a
natural confluence
of the science, the
medicine, the capital,
and the entrepreneurs.
And that doesn't exist
anywhere else in the world.
And MIT has been focal to that.
Being in this environment
allows us then
to tie into that ecosystem and
to learn from them in terms
of understanding
how best the work
we do can be applied outside.
But vice versa, it allows
them to have access
to how we're thinking
about pushing
the boundaries of the future.
And so that's one
of the many reasons
that we are launching MIT HEALS.
Anyone who has spent
any time at MIT
knows this place
has a secret sauce.
And I love so much that
we can now combine that
with the strength that
the hospitals, with all
of the remarkable industry
connections in Kendall Square
to really move health and
life sciences forward in ways
that we can only imagine.
[MUSIC PLAYING]
All right.
Good morning again.
So I'm Sally Kornbluth,
President of MIT,
and I'm delighted to welcome you
to the launch of the MIT Health
and Life Sciences Collaborative,
or simply MIT HEALS.
I'll have more to say about
MIT HEALS in a moment.
But first, it's my honor to
introduce a very special guest
this morning.
Since she was elected to the
Commonwealth of Massachusetts
just two years ago,
Governor Maura Healey
has focused on making the
state a leader in clean energy,
a more affordable place to live,
and a great place for businesses
of all kinds to start,
grow, and succeed.
Here at MIT, we
especially appreciate
her personal
enthusiasm and advocacy
for the state's interlocking
innovation ecosystems,
from climate technology
and advanced manufacturing,
to AI and the life sciences.
So we cannot be more delighted
to have her with us today as we
launch MIT HEALS.
Please join me in offering a
warm welcome to Massachusetts
Governor Maura Healey.
[APPLAUSE]
Well, good morning.
And thank you so much,
President Kornbluth.
Belated happy birthday to you.
Look, I got word that
this was happening,
and I never want to miss an
opportunity to visit MIT.
And I mean that.
So to all of you who are
here, Madame President,
distinguished faculty,
scientists and leaders
from across this
great university,
thank you for all that you do.
And thank you for
all that you're
going to do because today is
about what you're going to do.
And it's so, so exciting.
Congratulations on this
launch of MIT HEALS.
I love that title and
what it encapsulates.
I was also delighted
to have a moment
with your co-chairs of
this morning's symposium,
Dr. Amy Keating and
Katharina Ribbeck.
Lovely to meet you
both, although Amy and I
go way back because we went
to that other school down,
down the way.
One of the things as governor
that I really appreciate
is the opportunity
to see so many
of our state's
accomplished scientists
and brightest minds
come together,
work together, and
forge a new commitment
to improving human life.
It's even more exciting when
you think about this convening
to think about all the
amazing cures and treatments
and discoveries that
will result from it.
And I'm proud to say, and
I really believe this.
This is something that can
only happen in Massachusetts.
There's no place that has the
ecosystem that we have here.
We must fight hard to
always protect that
and to nurture that.
But we have the
greatest scientists
in the world, the greatest
universities, the greatest
hospitals.
We're number one for
innovation, for education,
and for health care.
And we draw top talent
from all over the globe.
We believe in solving
hard problems.
We, in fact, embrace
solving hard problems.
And this university
has always been
about thinking of ways to
reduce suffering, help humanity,
and further the common good.
We also are home to-- and
I find this remarkable.
Four of the 12 Nobel
Prize winners this year.
Four are from Massachusetts
and of course, two
are from this great university.
And it's just another
indication of Massachusetts
and this ecosystem always
punching above its weight,
so to speak.
This is what gives us
and distinguishes us
among global competition, both
for talent and for innovation
and for output.
It's why I also worked hard to
go out and recruit and appoint
Yvonne Hao as a leader
in an innovation economy,
to come and lead economic
development for the state.
I appointed Kate Walsh to be our
Secretary of Health and Human
Services, a leader of one of our
world-class teaching hospitals.
It's why we worked
hard, many in this room,
to go out and get after and
win the Investor Catalyst
Hub in ARPA-H, which is
the federal government's
new medical discovery agency.
And it's why in our new economic
development law that I recently
signed, we are making a
transformative commitment
as a state to the Mass
Life Sciences Initiative.
This has made us a global
leader and our charge
is to lengthen that lead.
And it's one of the
reasons I'm here
to say thank you because
what you are doing
in furthering aligns with
our state's interests
and where we want to go.
That includes also investments
in making Massachusetts
the global hub for climate
technology and for applied AI
and all sorts of things that
you are already ahead of us on.
But we are looking forward
to sync up and support
what you are doing.
And it's not by accident.
This is who we are
in Massachusetts.
We don't just believe in
science, we lead in science.
And we also know that that
requires intentionality
and effort every single day.
And that's what you do in
your classrooms, in your labs,
and other spaces.
And when I say "we,"
make no mistake about it.
I mean, you-- the scientists,
the brains, the researchers,
the students of
the Massachusetts
Institute of Technology.
And as President
Kornbluth knows,
MIT and the state
of Massachusetts,
the Commonwealth
of Massachusetts,
have a long and very
connected history.
So thank you for
all that you do.
I think that the MIT
HEALS initiative really
reflects the strengths
of our state,
in some really profound
and important ways,
and in ways that we
know are particularly
critical at this moment as we
once again affirm collectively
our commitment and indeed
reverence for science,
for knowledge, for
innovation, for collaboration,
and for partnerships
that will really
help us transcend in time.
In this moment in our nation's
history and our world's history,
your leadership and
the work that you do
is more important than ever.
I cannot wait to see
the results of it.
I look forward with my team to
getting out to various things
you have happening related
to this wonderful, wonderful
effort.
And I just thank you for
being here in Massachusetts
and making not only our
state better but our country
and indeed our world better.
So enjoy the rest of this
wonderful, wonderful day.
And thank you,
President Kornbluth.
[APPLAUSE]

---

### MIT HEALS Launch: President Kornbluth remarks
URL: https://www.youtube.com/watch?v=uw5e1XbPFfU

Idioma: en

Well, Thanks to Governor Healey.
I so appreciated her remarks.
So let me start my own remarks
with a few essential thank yous.
First, I have to say, in
reading through today's program,
I expect that we all
had the same experience.
Holy Toledo, this
lineup is spectacular.
The caliber of the speakers,
the boldness of their thinking,
the scale of their
accomplishments,
the range of frontier
subjects being covered
is just astounding, sort of
like a fantasy football team
if biologists played football.
In any case, I want to start
by thanking the incredible team
that persuaded all of
these brilliant people
to bend their calendars
to make this day possible.
Special shoutout to our Dean
of Science, Nergis Mavalvala,
to Dean of Engineering
and Chief Innovation
and Strategy Officer,
Anantha Chandrakasan,
to the entire Faculty Program
Planning Committee co-chaired
by Professors Amy Keating
and Katharina Ribbeck,
and especially to Marsha
Warren and the events team,
who orchestrated every detail.
We're all in your debt for
creating this extraordinary day.
And of course, I want to thank
all of our speakers from inside
and outside of MIT for
offering this amazing 14-course
intellectual banquet.
The talent assembled here--
from those who will
speak, to every one of
you-- has produced some truly
towering accomplishments.
But also, and I believe
more importantly,
you represent a deep well
of creative potential
for even greater impact, which
brings me to the insights that
inspired MIT HEALS.
Since I still have
fairly fresh eyes
on the wonders of this
multi-institutional community,
let me start with a
little perspective.
So I started my career as a
cell biologist and then a cancer
biologist.
And my husband, Danny Lew,
also leads a research group
here focused on pure
curiosity-driven science.
Well, we both had
long, happy careers
at another fine university.
But I will never
forget the expression
on Danny's face the day I
mentioned that I might interview
for the president's job at MIT.
MIT?
It was like those cartoon
dogs, where the eyes spin.
He was thrilled-- we
both were-- to think
we might join a place
with such a wall
to wall appreciation
for fundamental science.
And you can imagine
when we got here--
like kids in a candy store.
But pretty quickly, you
realize it's not a candy store.
It's a candy factory--
not only MIT, but the whole bio
ecosystem, from Kendall Square
to the seaport, to the
legendary hospitals
and the countless corporations,
startups, universities,
venture capitalists,
incubators, and philanthropists
across the region.
You see it any time you meet a
clinician or a faculty member.
Even students, they're exploring
techniques or technologies
right at the frontier
of their fields.
And then they casually
tell you, oh, yeah, I also
started a company
or two or more.
As a source of new knowledge
and new tools and new cures,
and of the innovators
and innovations that
will shape the future of
biomedicine and health,
there is just no place like it.
And today, we're here
to talk about how
we can make it even better.
That is the ambition
behind MIT HEALS.
Our goal with MIT HEALS is to
inspire, accelerate, and deliver
solutions at scale to
some of society's most
urgent and intractable
health challenges.
So with that goal in mind
and with expert guidance
and leadership from
many of you here today,
including Anantha and Nergis
and many other faculty leaders,
we've identified
three priority areas
where we have a
vital opportunity
to make a real difference.
In shorthand, these
three priorities
are talent, collaborations,
and bridges.
First, talent-- the caliber
of talent in this room today
is the caliber of talent that
this ecosystem needs every day
forever.
If we want to be the biotech
capital of the world,
to spawn the next
biotech revolution too,
we can't take this concentration
of talent for granted.
But nothing attracts
talent like talent.
So a crucial part of MIT
HEALS will be finding ways
to support, mentor, connect,
and foster community
for the very best minds at
every stage of their careers.
The second priority
of MIT HEALS is
to spark new cross-cutting
collaborations
between individuals, both
within MIT and with allies
in the hospitals and industry.
To encourage faculty to pursue
high-risk, high-impact ideas,
were going to be providing
substantial MIT funding.
So why is this?
Now, certainly because
daring, unproven ideas
can often be the
most transformative,
but also because federal
dollars are only shrinking.
Federal funding seems to
get more and more cautious
all the time.
I mean, I remember
this frustration--
and I'm sure many of you
experienced this frustration
firsthand-- to get a grant.
Sometimes people prefer you've
actually finished or almost
finished the work that
you're proposing to do.
This doesn't lead to
transformative discoveries
necessarily.
I also want to emphasize that
this internal funding will
support not only translation,
but deep, fundamental,
curiosity-driven research.
For people outside
academia, it's
very easy to see the
value of translation.
They see the cures that
make it to the bedside.
Yet for some outside of
the scientific community,
it can be a mystery why we care
so much about basic research.
But it is fundamental
curiosity that
has driven the discovery
of so much of what we know
and now take for granted
in the life sciences.
And I have to say--
I always say this-- even for
those focused on translation,
you can't translate nothing.
You have to think about the
fact that today's cutting-edge
approaches-- immunotherapy,
AIDS therapy, the mRNA vaccines,
the new sickle cell cure,
thanks to things like CRISPR,
things discoveries made based
on just fundamental curiosity--
all of them sprang
from a robust pipeline
of breakthrough knowledge.
So we're investing
enthusiastically
to keep that pipeline strong.
And of course, in
terms of translation,
we're eager to empower
MIT's unique community
at the juncture of
engineering and bioscience
in an era powered by
incredible new tools.
We want to give these faculty
members the confidence
to collaborate freely
with colleagues
from across disciplines-- from
AI, policy, and economics,
to the humanities, to design.
We want them to take
on the hardest problems
and pursue their biggest ideas.
We see great potential
for producing
practical, scalable solutions
in a wide variety of fields
such as AI and life science,
low-cost diagnostics,
neuroscience and mental health,
environmental life science, food
and agriculture, the future of
public health and health care,
and women's health.
Third and finally, MIT HEALS
will prioritize building bridges
between institutions, developing
even stronger, more effective
relationships with the
region's world-class hospitals
and biotech and
pharma industries.
Several of our
faculty, many faculty,
have established very
fruitful research programs
with hospitals
across the Charles.
I think about the work of
Regina Barzilay and J-Clinic
as standout examples.
But we want to
make it much easier
to make these relationships
spring up and flourish.
We want them to be
in reach for everyone
so that each PI doesn't
have to reinvent the process
and re-navigate the same
barriers around data sharing,
computing power, and more.
And at the same time,
while MIT has always
had a distinctive openness
to working with industry,
we believe that
there is much more
we can do to connect
our researchers
with the full power
of the ecosystem,
from biotech to pharma.
I have to say, I can speak
from my own experience--
even though it's receding
into the distant past--
working on basic science of a
particular drug-resistant breast
cancer.
We published a paper in
Science Translational Medicine.
We even patented our findings.
But it felt absolutely
impossible to take it further.
We didn't have an
ecosystem to plug into.
Here we definitely
have the ecosystem.
And MIT HEALS will help to very
deliberately improve and deepen
those connections, too.
Overall, MIT HEALS
is an opportunity
to make our work as individuals
and institutions collectively
add up to more than
the sum of our parts.
And I believe this
effort could not be
more timely or more important.
A key next step will be
selecting a faculty member
to serve as a director.
And a team of faculty
from across the institute
are conducting the
search right now.
In each of the
areas I described--
talent, collaboration
between individuals,
and bridges between
institutions,
we feel that action
is imperative.
And it gives me
great pleasure to let
you know that MIT
HEALS has already
attracted major support from
several sources, for instance,
in the area of talent.
An important aspect
of MIT HEALS is
supporting our graduate students
to explore new directions
in life science, particularly
interdisciplinary research.
I'm pleased that MIT
will be able to provide
new internal support
for an initial cohort
of graduate fellows
for MIT HEALS.
And we're delighted
that the founder
and CEO of Flagship Pioneering,
Noubar Afeyan, who's
here today--
one of our long-time
collaborators,
a member of the Executive
Committee of the MIT
Corporation, and a former MIT
Course X PhD student himself--
has committed to supporting
additional graduate fellowships
as part of MIT HEALS
So thank you, Noubar.
[APPLAUSE]
In the same spirit of
supporting great talent and as
another key part
of MIT HEALS, we
will launch the MIT Health and
Life Sciences Collaborative
Fellows Program, a
world-class program
for postdoctoral scholars.
This highly competitive program
will give fellows the freedom
to explore some of the
most pressing issues
in life science
and human health.
Fellows will receive a four-year
fully funded appointment at MIT
and benefit from faculty
mentorship, as well
as participation in
community events focused
on careers in academia,
industry, and entrepreneurship.
We are delighted to announce
a gift from the Biswas Family
Foundation to establish
the inaugural cohort
of postdoctoral fellows.
The Biswas Fellows will
conduct cross-cutting research
in areas such as AI
computation and health,
low-cost diagnostics, nanoscale
therapeutics, neuroscience,
women's health, and
curiosity-driven life sciences.
Thank you so much to the
Biswas Family Foundation.
[APPLAUSE]
As an example of sparking
new collaborations,
I'm also thrilled to announce
another integral piece of MIT
HEALS--
the new Hood Pediatric
Innovation Hub.
It was established
through a catalytic gift
from the Charles
H. Hood Foundation,
a leader in supporting
groundbreaking and innovative
pediatric research.
Led by Professor
Elazer Edelman--
the Edward J. Poitras Professor
in Medical Engineering
and Science and a world-renowned
cardiologist-- the hub
will bridge the
translational gap
for innovations in pediatrics.
Currently, the major
market incentives
are medical innovations
intended for adults.
As one example,
children are often
treated with medical
devices and therapies that
don't meet their needs because
they're simply scaled down
versions of the adult models.
The Hood Pediatric Innovation
Hub aims to overcome such
persistent barriers to bringing
life-saving technologies
to market by leveraging MIT's
state-of-the-art technological
expertise and a network of
health care stakeholders across
the country.
It will also strengthen
collaborations
with many hospitals, including
with an important partner
in Boston Children's Hospital.
So huge thanks to
the Hood Foundation.
[APPLAUSE]
And finally, I'm
excited to share
that one of the flagship
components of MIT HEALS
will be a collaboration
framework between MIT and Mass
General Brigham.
It will bring together these
two world-class research
institutions to advance
technology and clinical research
for transformative
changes in patient care.
Through the MIT-MGB
Seed Program,
we are garnering
and will continue
to seek philanthropic
funding to establish
seed grants for joint research
projects led by MIT and MGB
researchers in mutual
areas of interest.
And I'm extremely
pleased to announce
that Analog Devices will
help launch the MIT-MGB Seed
program with gifts
to both MIT and MGB
to establish the Analog Devices,
Inc. Fund for Health and Life
Sciences.
This visionary gift, which has
been described by Analog CEO
Vince Roche as a
once-in-a-generation
opportunity, will help
accelerate the adoption
of critical sensing, digital,
and AI technologies in clinical
settings.
And the new framework
overall will
be a model for how we can
all work together more
effectively in the future.
So thank you to MGB, and
thank you to analog devices.
[APPLAUSE]
So this is fantastic.
I mean, you can see there's
fuel, energy, power behind this.
We have very high
hopes for MIT HEALS.
And the momentum is
already beginning
to make these aspirations real.
Now, before I close, I have
to particularly call out
the Dean of Engineering
and Chief Innovation
and Strategy Officer,
Anantha Chandrakasan, who
has had the vision and
drive behind bringing
this initiative to life.
So I really want to
thank you again, Anantha.
[APPLAUSE]
And to all of you, I'm
so grateful that you've
joined us for this
important moment,
for this outstanding program.
Look forward to
working with all of you
as we work to seek new
knowledge, bold new solutions
to real problems in human
health in fields from cancer
to climate change.
So thank you all, and I hope
everyone enjoys the day.
[APPLAUSE]

---

### MIT HEALS Launch: Phillip Sharp keynote
URL: https://www.youtube.com/watch?v=7HH1DDNpVkw

Idioma: en

Good morning, everybody.
It's really nice to
see you all here.
Thank you for joining us.
I'm Nergis Mavalvala and I'm the
Dean of the School of Science
here at MIT.
I share President
Kornbluth's excitement
and wonder at all that we have
achieved in the life sciences
at MIT over the past decade.
MIT has been a place where
fundamental knowledge
and understanding
of biology happens,
where rapid conversion
from ideas to discovery
to translation happens, where
computation and policy meet
science and engineering
comfortably,
where innovation is
a natural destination
on the journey from
discovery to societal impact.
As well as we do all of this,
as you heard from Sally,
we know we can do better.
And I couldn't be more energized
to be here with you today
to celebrate and
dream of the future
that the MIT HEALS
Collaborative will help create.
By catalyzing collaborations
in fields like immunology,
neuroscience, AI,
sustainability, and healthcare,
and strengthening MIT's
ties with Boston's
extraordinary hospitals
and industry leaders
in biotech and pharma, the
MIT HEALS Collaborative
aims to spur discoveries
with broad and lasting
real-world impact.
Now, you couldn't come
up with a better example
that aligns with those ideals
than our very own Dr. Phil
Sharp.
He has been at the vanguard, not
only of scientific discovery,
but also MIT's collaborations
with biotech, biopharma,
the healthcare industry,
his entire career.
Now, before I tell you
a bit more about Phil,
let me say a few words about how
the MIT HEALS Collaborative came
about.
It does involve Phil,
as many wonderful things
that happen in the life
sciences at MIT does.
Little over three years ago,
at the instigation of many wise
colleagues and friends of MIT--
and many of you are in the room
here today--
Dean Chandrakasan and I
charged a faculty-led committee
to study what we
might do better,
to coordinate all the fabulous
but disparate work in the life
sciences being
carried out at MIT,
and also how might we
better nurture and curate
collaborations with
the life sciences
research being carried out
in the biotech and hospitals
around us.
That study, led by professors
Tyler Jacks and Kristala
Prather, culminated
in the VITALS report.
VITALS stands for Vision to
Integrate, Translate and Advance
the Life Sciences.
A key finding of that report was
the hunger among MIT researchers
for more opportunities
to collaborate
across disciplinary boundaries.
The opportunities identified
and the recommendations
of the VITALS report were
embraced by President Kornbluth.
And with the prodigious
energy of our chief innovation
and strategy officer,
Dean Chandrakasan,
here we are today embarking
on a truly vital endeavor.
Upon hearing of the faculty's
interest in collaborating
with each other and with
our award-winning hospital
systems, the Kendall Square
biotech ecosystem, and beyond,
Phil set to work helping
us build those bridges.
A pivotal moment for me was a
gathering where a group of us
were brainstorming about what
this collaborative effort might
look like and what big
ideas would we lead with.
Lots of ideas were
being batted around.
At one point, Phil
challenged the group.
What do we do best
at MIT, he asked.
Everyone in the room must have
had their favorite answer,
but only Phil had an answer
so succinct and fundamental
and yet so universal
for us here at MIT.
We measure, he declared.
As a physicist, I can
admit I have never
directly peered into a
microscope and seen a cell.
But I do know measurement, and
that resonated tremendously.
Our obsession with
inquiry, with observation,
quantitative characterization,
experimentation, discovery,
and invention were all
wrapped into those words.
At MIT, we measure.
It's part of MIT's secret sauce
that Nancy Andrews alluded to
in the video you saw at the
beginning of our session.
It's what makes me so optimistic
about the unimagined discoveries
waiting to be made when we
bring seemingly disparate ideas
and tools together
through cross-disciplinary
collaborations.
This is the promise of the
MIT HEALS Collaborative.
And I'm very grateful to
Phil for his clarity, wisdom,
and encouragement as we
embark on this endeavor.
For those of you who
don't yet know Phil,
let me introduce him to you.
Dr. Phil Sharp is an
Institute professor
at MIT in the Koch Institute
for Integrative Cancer Research.
He is also the chair of the
advisory board for the MIT Abdul
Jameel Clinic for Machine
Learning in Health, or J Clinic.
In 1993, he shared the Nobel
Prize in physiology of medicine
for the discovery
of split genes.
These are genes that--
the genes in DNA are not
one contiguous segment
of information as
they are in bacteria.
In fact, they are split.
And that messenger RNA
has to be processed
by splicing before it can be
translated into a protein.
This fundamental discovery
would alter the course
of how we understood genetic
disease and the development
of potential treatments.
In 2004, for this
work, he was awarded
the National Medal of Science.
You'll learn more about that
discovery from Phil shortly.
At MIT, he joined the Center
for Cancer Research in 1974,
serving as the director
from 1985 to 1991,
before becoming the head of
the Department of Biology
for eight years.
He was also the
founding director
of the McGovern Institute
for Brain Research
from 2000 to 2004.
Phil co-founded Biogen,
serving on its board for nearly
30 years, as well as Alnylam
Pharmaceuticals, a leader in RNA
interference-based therapies
with Professor Bartel
and Professor Emeritus Paul
Schimmel and MIT postdocs.
Phil is an elected member of the
National Academies of Sciences,
the National
Academies of Medicine,
the American Academy of Arts
and Sciences, the American
Philosophical Society, and
the Royal Society of the UK.
His accolades, his advisory
board service, his awards,
and most importantly, his
500-plus research papers are too
numerous to recount in
the time that we have.
Even with our ample
measurement skills at MIT,
it would be truly hard to
measure his impact on biology
at MIT and beyond.
Please join me in welcoming
Professor Phil Sharp.
[APPLAUSE]
Wow.
[LAUGHTER]
What a day.
What a way to start
a beautiful day.
I couldn't be more excited.
I couldn't be more enthusiastic
about MIT, MIT HEALS,
and what it's going to mean for
patients, people, the country,
Massachusetts, and MIT.
What I want to do in
the next few moments
is to give you a brief
overview of a journey
that I think tells you why this
initiative is so important.
I know I'm going to be
looking backwards and setting
the stage for what you're going
to hear today, looking forwards.
But I do think it
illustrates why we're here
and what we hope
continues to come out
of this great collaboration,
of which I am sure it will.
So I'm going to look
back over 50 years.
50 years is when the
Center for Cancer research
was established here at MIT.
I'm going to bring that
journey to where we're at now.
And I'm trying to emphasize what
changes, what major initiatives,
had this impact.
I have conflicts of interest.
Alnylam is the only one that I
will mention some relevant data
to.
Let's look back
at the late 1950s.
MIT's leadership looked at
the Department of Biology
and decided that in the
future, biology at MIT
will be gene-oriented, molecular
biology, driven mechanistically,
incorporating chemistry and
physics into this marvelous
discovery of the structure
of DNA by Watson and Crick
in Rosalind Franklin's data.
And that led to the recruitment
of Salvador Luria to MIT.
Salva was a founder
of molecular biology,
the mentor of Jim Watson, and
came to MIT in the late 1950s
with other faculty and people
at MIT to create what is
the Department of Biology today.
In early 1970s, Salvador Luria
led a number of people at MIT
to apply for a basic cancer
research center here.
This is after the National
Cancer Institute Act
in Congress.
Leading that initiative was
Luria and Dave Baltimore
in molecular cell
biology and virology,
Herman Eisen in immunology, and
Phil Robins in cell biology.
Here in the early '70s,
the pillars of much of what
we're talking about in health
care, molecular immunology,
cell biology, was established
here as a basic cancer center.
And remarkably, the culture
of this cancer center
is a culture of MIT.
It's a culture of
collaborative interaction,
of people sharing ideas,
sharing initiatives.
What is shown here on
the slide is the group
of molecular cancer biologists
in the Cancer Center with Dave
Baltimore peeking
down at the left,
Bob Weinberg over his shoulder.
I'm up on the right side.
But this was a collaborative
group of sharing ideas--
an MIT mode of science.
And it had an enormous effect.
Dave Baltimore received a
Nobel Prize in 1975, one year
after the establishment
of the Cancer Center.
And then Susumu Tonegawa, who
was recruited to the Cancer
Center--
shortly thereafter
another Nobel Prize.
After that-- and
then Bob Horowitz
at MIT, who was in the
Department of Biology
and a member of the
cancer center, also
received a Nobel Prize in the
period of the Cancer Center
in its site in the chocolate
factory on Ames Street.
I want to illustrate why this
initiative of the Cancer Center
was so timely.
Here in the early 1970s,
there was the beginning
of the understanding that
the pillars of biology--
which was the central
dogma illustrated here--
as DNA information flowing
to RNA flowing to protein,
had just been established
by Jacoby Meno in 1961.
And what you see here
is a simple diagram--
that the RNA is being copied
from the DNA in a colinear way
and proteins translated from it.
And there is a sugar that is
metabolized by these proteins
and that sugar regulates
the expression of this gene.
This is how we understood
molecular biology in the early
1970s.
But we thought that the world
was more complex than that.
When we started talking about
multicellular organisms--
was the gene and how it
was expressed the same
in multicellular organisms?
And there were hints-- and
I won't go through those--
that there was
something different.
And that, for example, led to
the understanding and discovery
that genes in our systems
have different structures.
They are split.
And they are expressed
by an intermediate step
of the splicing of the RNA,
removing the dashed lines shown
at the top of the slide,
and joining the black boxes.
The black box is the sequences
that code for the protein.
So here, the complexity of
the genome and complexity
of how it's expressed
and how it's regulated
was much more complex
than it was in bacteria.
And it led us to
understand that we
were in a new territory of
understanding disease processes.
I also want to note, the
splicing pattern on those genes
vary from cell to cell and
condition to condition.
And therefore, even though
we have 23,000 genes,
we have many more proteins.
They vary between cells
from those proteins
by alternative splicing and they
carry out different functions.
And we're just in the
last several years
able now to look at a
single cell and say,
this is exactly what the gene
is expressed at in those cells.
But also in the
same time period,
in 1975, we had this
convergence or confluence
of remarkable discoveries that
created synthetic biology--
the ability to make organisms
that had never existed before.
And that was this
recombinant DNA technology--
DNA sequencing and
chemical synthesis.
So we could take DNA and
we could make new genes
and we can put
genes in organisms.
This technology was
recognized to be
a transformational technology.
And the first biotech
community or company
that was established
in 1976, Genentech,
was by Bob Swanson
and Herb Boyer.
Swanson, an MIT graduate--
Boyer, as a scientist
at San Francisco--
led to this
foundation of biotech
and the whole biotech industry.
And I had the pleasure of
collaborating with Wally Gilbert
at Harvard and six
European scientists
in establishing Biogen in 1978.
The idea here is, this
technology had been
developed in academic labs.
And it wasn't going
to benefit anyone
until it was translated
into the private sector,
gained capital and
investment, recruit people,
and make a difference
in people's life.
Be innovative in
this technology.
And that led to
Biogen first being
established in Geneva in 1978.
When the situation became a
little clearer in Cambridge
as to what you could do with
recombinant DNA In the early
'80s, this moved to Cambridge.
And what I show you here is what
Cambridge looked like in 1975.
What you see to
the right of that
is Main Street and
MIT across the street.
But you see a large parking lot.
A parking lot, for
historical reasons, was open.
And Biogen began in a little
building on Binney Street
on the left side of that slide.
So here, Cambridge was the site
in which this new technology
was beginning to nucleate.
Genzyme and other companies
established in the same area
shortly after or
at the same time.
But then it grew
and it consolidated
and it collaborated.
And that became the Kendall
Square area in 2012,
where that same
space is now occupied
by companies that are
making marvelous advances
in important ways.
And as Sally said,
it happened because
of the great universities,
the hospitals
across the street with
remarkable faculty
and clinicians, and the New
England finance and venture
capital acceptance.
This continued to grow.
We're the most concentrated
biotech community in the world.
Life science-- it's
a remarkably dynamic,
collaborative, networking
community that is reflecting
a very important part.
And I forgot to mention,
it wouldn't have happened
if it wasn't for MIT engineers.
MIT engineers allowed us
to take these products
and make new components of it.
So all of this happened.
And now we see what
Cambridge is like.
I should illustrate
here in 2018 that
large pharmaceutical companies,
research stations, research
institutes, have moved
into the community.
And there's probably
50 or more small to
medium-sized biotech companies.
Biogen has grown.
And Alnylam, which I'm going
to talk about in a moment,
and then Broad--
I'm going to talk
about immediately--
as additions to
this community that
empowered a lot of
remarkable advances.
So we move to 2000, the
turn of the century.
And biotech took on
another acceleration
in the country because of the
development of the human genome
initiative and the
human genome sequence.
And it's interesting to think
about the origins of that work.
Whitehead was
established in 1985.
It was a new institute
relationship to MIT.
MIT was open to bringing new
resources and new structures
to being part of
the community at MIT
at the Whitehead Institute.
Excuse me.
Dave Baltimore recruited
Eric Lander, who
started the genome initiative.
And that genome initiative then
moved with additional resources
to the Broad Institute, which
is across the street from MIT--
associated with MIT and Harvard.
Again, a collaboration.
And that led in 2003, with an
international initiative highly
influenced and led by
the Broad Institute,
to the whole human
genome sequence.
This was a accelerator because
then we had in our hand
the sequence of every gene--
the sequence of the genome.
And we could begin to
integrate complex genetics
and molecular tools into
this identification of--
this powerful technology
of recombinant DNA.
This allowed us to more
rapidly translate our science
and our technology
into patients.
And just to illustrate
how that happened
or the consequence of that,
I will speak about a subject
that appeared in 2000, and that
is with the genome sequence.
In previous science,
it was recognized
that our cells encode another
500 or more genes that we
didn't know about, that
encoded small RNAs.
These small RNAs were
called microRNAs.
The initial discovery
of the first
was in 1993 out of Gary Ruvkun
and Victor Ambros' [INAUDIBLE]
network.
It was done initially
with collaboration
in Bob [INAUDIBLE] lab.
And out of that came this
discovery of microRNAs.
And as you will note, that was
recognized by a Nobel Prize
in this year, 2024.
This microRNA pathway is
present in every cell.
It was totally unrecognized
until the 2000s by David Bartel
and others.
And it has become a major
regulatory-- involved in cancer,
involved in many processes.
And then at the
same time, in 1998,
there was a discovery by Fire
and Mello of the fact that
double-strand RNA--
most RNA is single
stranded in the cell.
But you can see double strand.
And double-strand
RNA from a gene
will silence the expression of
that gene in many organisms.
Their work was done in
the worm C. elegans,
but it was soon recognized
that that work interdigitated
with the microRNA work.
And this led us to
an understanding
that there was another
biochemical system
in the cell we didn't know.
Andy Fire was a grad
student here at MIT,
but did the work
after he left MIT.
Using that work, and with
David Bartel and Phil Zamore
and Tom Tuschl, postdocs, and
Paul Schimmel's influence,
it became recognized
that this whole process
of silencing a gene
with RNA offered
a new therapeutic modality.
We could use the natural pathway
in the cell, design a drug--
siRNA-- from the
sequence, and then
treat a patient if the
modulation of that gene
would have an impact.
So that was just another way of
possibly benefiting patients.
And that led to then the
establishment of Alnylam,
and the confrontation
that the big problem here
was delivery of the RNA.
We turned to engineers and other
chemists to solve that problem.
Delivery of RNA into
cells is difficult
because RNA is hydrophilic
and your cell has
a membrane that's hydrophobic.
And therefore you have to
transport the RNA into the cell.
But once it's in the cell, it
will enter the microRNA pathway
and carry out-- if it's
complementary to the messenger
RNA--
destruction of the messenger
RNA and then be catalytic.
So the big issue
here was, could you
introduce this RNA into cells?
And that led to the development
of lipid nanoparticles, which
look like a virus.
They have a cationic
lipid of a certain type.
Pieter Cullis taught us--
that is a Canadian scientist
who made major contributions
to this.
Impacting the RNA into
these nanoparticles
would, with the right ionic
pH and PK, give you then
a change in conformation
of those lipids
as the particle enters the cell.
In your blood system,
you have a pH of seven.
In the cell, it's nearly a
pH of 5 in this endosome.
And that creates
this transition.
That same technology--
similar technology
was used in the
vaccination in COVID that
rescued us from the pandemic.
It has then been used to
treat billions of patients.
Here again, you see
the collaborations
between molecular
discovery, to innovation,
to the translation
to benefit patients.
And then other people using
those insights to further solve
a major societal
problem in the pandemic.
So that brings me
to looking forward
at what you're going to hear
in the next several hours
from my remarkable
colleagues here at MIT.
And that's what we've
called convergence.
With the discoveries of the
structures of sequences of DNA
and the genome pathways
and the cell mechanisms
by which the cell carries out
communications between cells,
we have the makings--
not a complete understanding,
but the makings--
of how the cell works
and the components.
And if we look back over time,
that's a product of research
from the mid '50s--
as recruitment of Luria--
the establishment
of the institutes,
research advances,
discoveries of multiple types.
And then we are at a
stage in which engineering
can take those components
and add to our tools
and understanding
these processes.
Adding AI can change how we
can empower our scientists
to benefit patients
to solve problems.
And that's what we've
called convergence.
And it's what MIT HEALS is
a major component in making,
and it is the future
of life science.
And it's a future of benefiting
a large number of patients.
The Koch Institute is one
of the examples that have
been established around it.
You will hear more
today about the Koch
Institute and its technology.
But its faculty
are a combination
on every floor of engineers,
life scientists, chemists,
and physicists.
We are a very interdisciplinary,
integrated institute.
But if you look at the McGovern
Institute in Brain Research--
the Picower in brain research,
the Reagan Institute,
you see this integration
of technology
and molecular cell
and cell biology,
and the whole premise of
looking outward and impacting
on patients.
And it is a remarkable,
remarkable--
and I should include
the Reagan Institute--
a remarkable set of institutes
associated with MIT.
No other institute
that I know of
has been able to create this
body of collaborative institutes
in a community with such a
powerful biotech community
and hospitals that is looking
forward, as our governor said,
in making MIT--
Massachusetts-- keeping it an
absolute leader in this space.
So innovation continues.
Here is just the number of
companies over the last--
from 2007 to now.
That faculty in the Koch has
continued to use technology--
to make discoveries
in technology
and translate to patients.
And that is the continued
engagement with society,
changing the outcome
of our patients
in our country with
this technology.
So I am really grateful to have
this chance to speak with you.
Everything I've
talked about is a, we.
Everything has been done by
collaboration with many people--
with faculty, with our
students, our postdocs.
I show you here
reunions of my lab
over the last 30 or so years.
All of MIT and all
of my colleagues
have contributed to
making this moment.
And I am terribly excited
about what MIT HEALS means
for MIT and for the country.
Thank you.
[APPLAUSE]

---

### MIT HEALS Launch: Revolutions plenary session
URL: https://www.youtube.com/watch?v=JFzCn1wTZ8w

Idioma: en

All right, good
morning, everyone.
My name is Robin
Wolfe Scheffler.
I'm from the MIT program
on Science, Technology,
and Society.
And it's my pleasure
to introduce this panel
on revolutions.
As you've seen this
morning, people commonly
talk about the life
sciences in the modern era
as one of revolutions--
the revolution of DNA,
the revolution of informatics,
the revolution of gene editing.
As a historian of
the life sciences,
however, I'm not
quite sure if this
is the metaphor we want for
talking about the work we
do here at MIT.
A revolution, if
you'll remember,
implies a violent turning
over of the world.
Indeed, philosophers of science
who have looked at revolutions
have gone so far as to suggest
that the worlds of science
before and after
a revolution are
incommensurable with each other.
There can't be any dialogue,
only misunderstanding.
Speaking in revolutionary
terms, therefore,
risks obscuring the rich history
of work at a place like MIT
and dismissing its importance
to the present day.
And I think this gives us
a shallow understanding
of the types of
innovation and cooperation
that make our Institute unique.
A better metaphor than
revolutions, if all be forgiven,
is that of waves.
If you think about a stream
of drops landing on a pond,
the force of each
wave might dissipate.
The discovery of DNA
becomes ubiquitous as a tool
for research over time.
However, it's still there.
It still shapes every
subsequent type of work.
New drops on the
pond add and interact
with others that came
before it, creating
new and unexpected patterns.
And I think this
sounds much more
like the work that's
happened at MIT,
as we've heard from
Professor Sharp
and we will hear from the
speakers in this panel.
Ideas and institutions
build upon each other,
rather than violently replacing
or superseding one another.
New people and new ideas
find new uses for old tools,
and they combine them
with their new knowledge.
The story is not
one of displacement
but of constructive interference
and intermingling of science
and engineering to produce
unexpected and new ways
of knowing life that
could only happen here.
MIT is so special precisely
because our pond, if you will,
has seen so many
drops land upon it.
So, without further
ado, it's my pleasure
to introduce three
speakers who have both
made their share of
waves at MIT and have
seen many wash upon its shores.
First, Tyler Jacks, who
is the David H. Koch
professor of biology,
a Daniel K. Ludwig
scholar, the co-director of
the Ludwig Center at MIT.
And he's going to speak to
the long history of assembling
genetics, informatics, and
technology to study cancer.
Second, Douglas
Lauffenburger, who's
the Ford Professor
of Engineering
and a founding professor
of the modern Department
of Biological Engineering.
He's going to speak to the
entry of engineering approaches
into the study of life as a
novel way of studying and doing
science.
And finally, Ruth
Lehmann is the director
of the Whitehead Institute
and a professor of biology.
She is going to highlight the
long work of the Institute
in combining
approaches, old and new,
to understand the complicated
and rewarding question of life's
complexity itself.
So, without further
ado, I'd like
to bring Tyler
Jacks to the stage.
[APPLAUSE]
Well, thanks, Robin.
And welcome, everybody.
Thanks for joining us.
I'm personally extremely
excited that we're here today,
celebrating the
launch of MIT HEALS.
As you heard, I was
involved with Kris Prather
in leading a faculty
committee that
led to a series
of recommendations
that arguably brought
us to this point.
And I'm just so excited about
what this collaborative will
do for our community, for our
region, and really for the world
as it relates to
advances in life science
and, importantly, in health.
I've been asked to talk a little
bit about cancer research at MIT
and tell the MIT cancer
research story, which
I'm very happy to do.
It's a long story.
It goes back, as you heard
from Phil Sharp, 50 years now.
In fact, we're celebrating
the 50th anniversary of cancer
research on the MIT
campus this year.
It began modestly in this
converted candy factory
on Ames Street,
where MIT decided
to invest in molecular biology
and recombinant DNA technology
to understand the
mysteries of cancer.
And that effort, as
you heard from Phil,
was massively successful and
really laid the foundations
for modern-day
molecular oncology.
Over the years, many,
many MIT faculty
got interested in
cancer research.
It really touches every
corner of our University.
And in response to that, we've
created the Koch Institute,
which embodies this
spirit of collaboration,
as Phil called it,
convergence, bringing together
multiple disciplines to
address and hopefully overcome
the challenges of cancer.
The Koch Institute
is a physical entity.
But it also connects many
aspects of the MIT community
around these most
important problems.
And I'll touch on
some of the activities
that the Koch Institute faculty
are working on today and project
on into the future.
But I want to begin looking
backwards again to the Cancer
Center.
This is a picture similar
to one that Phil showed you
of a group of investigators who
were recruited to the Cancer
Center, including Bob
Weinberg, Nancy Hopkins, who
was showed in the previous
image, David Baltimore, Salvador
Luria, and Phil Sharp
himself, thinking about,
how do cancer cells
arise from normal cells?
What are the changes
at the molecular level
that drive that process?
And Bob Weinberg,
who's circled here,
was interested in the question
of whether normal genes are
altered through mutation in
the development of cancer cells
from normal cells.
And he set about to find
the first cancer gene.
And after some years of
effort, he and his lab group
published three really
landmark papers in 1982
describing the first
human cancer gene.
It was called HRAS, which is
mutated in bladder cancer.
That was where it was found.
But it and its
cousins are now known
to be mutated in many different
very important human cancers,
including lung cancer and
colon cancer and pancreas
cancer, and more.
It really is an important
driver of the cancer process.
It becomes an oncogene
through a very subtle change
in the DNA, a single
nucleotide change, which leads
to a single amino acid change.
And that converts the
protein to an oncogenic form.
And the hope of
that research was
that, if we could understand
cancer at this level,
we could actually address it
molecularly and maybe find
new drugs that
target this altered
protein and similar
altered proteins
and bring about more effective
treatments for the disease.
That has happened.
And I'll touch on that briefly.
It happened here too,
but it took a long time.
RAS turned out to be a
very difficult problem.
And it actually took
nearly 40 years,
investigators from
academia and then industry,
to target altered RAS.
But in 2001, first Amgen and
then Mirati, now owned by BMS,
successfully developed
KRAS-targeting drugs,
which have been approved
by the FDA in lung cancer
and more recently
in colon cancer.
And this represents
a major shift
in our ability to control
those cancers that I
mentioned that have
KRAS and other RAS gene
mutations as drivers.
So that was then, a
single cancer gene, 1982.
This is now, where, with the
advent of high-throughput
DNA-sequencing technologies, the
sequencing of the human genome,
as you heard from Phil--
major contributions by the Broad
Institute and others at MIT--
we now have collectively
sequenced probably close
to a million cancer
genomes by now
and discovered not just one
human cancer gene but probably
500 cancer genes,
genes that are altered
by mutation in human cancers.
And there's more than a hundred
molecularly targeted drugs
like these that I'm
showing you here
that have been
approved by the FDA
for the treatment of cancers
with those alterations.
So the promise from 1982
has actually borne true
in spectacular fashion.
These genes are
mutated sometimes
rarely in a given
cancer but sometimes
commonly across many cancers.
This is an analysis
done by investigators
in a large international
collaboration,
including the Broad
Institute investigators,
looking at 26 commonly mutated
cancer genes across here,
26 different cancer types.
And again, this allows
for molecular diagnostics
to identify particular cancers
with particular alterations that
allows them to be treated with
targeted, more precise cancer
medicines that are having
significant benefit.
Now, at MIT, we're
approaching the complexity
of cancer, this myriad of
mutations in cancer genes,
in lots of different ways.
We have expertise in
my lab and elsewhere
at the Institute in modeling the
molecular alterations in cancer
in genetically
engineered mouse models.
And we can do this using
a variety of techniques
to very accurately model what
happens in human cancer cells
but in a tractable experimental
model system, the mouse.
And these tools
are useful not just
to understand the molecular
development of the disease
but also what one
can do about it.
How can you use these models to
detect cancer at earlier stages
or treat it more effectively
or maybe intercept
its progression from its early
stages to its later stages?
And this is very active and
fertile ground here at MIT.
The technologies for making
genetically engineered mouse
models have improved
dramatically over the years.
I started this work myself
here at MIT in 1988.
And it was very difficult
to do back then.
Now, with tools like CRISPR, we
can do this very, very easily.
And the most recent
version of CRISPR tools
that we use in this
process is called
prime editing, a method
developed by David Liu
at the Broad Institute,
which takes the Cas9 enzyme,
modifies it so that it no longer
makes a double-stranded DNA,
but rather a single-strand
nick, and then fuse that nickase
to a piece of reverse
transcriptase,
an enzyme that will
take an RNA template
and convert it into a DNA form.
This was actually
a discovery made
by David Baltimore, which
was the thing that won him
the Nobel Prize in 1975.
And you can therefore feed this
enzyme a long extended guide
RNA that will bring it
to a place in the genome
but also have an extension on
it that has a mutation, shown
in red here.
And the reverse
transcriptase activity
will then convert that
mutation in RNA form into DNA.
And that DNA will then be
transferred into the genome
at that specific position.
This is an incredibly
powerful technology.
We've used it in
collaboration with David's lab
to make mice that
carry the prime editor.
And now, in a
simple system where
one designs a guide RNA
carrying a mutation of interest,
one can introduce that into
animals and mutate genes
at any position in vivo.
And we've done that
in one example,
again bringing us back
to the RAS oncogene,
to look at different
alleles of KRAS
and how they affect tumor
development here in the lung.
And I won't belabor
this point, but suffice
it to say that different
mutations cause
different phenotypes.
They all cause cancer, yes.
But some of them, G12R
here, causes a much more
profound cancer phenotype than
the other three for reasons
that we don't understand.
But in all likelihood,
this is teaching
us something very important
about the biology of RAS
and how, ultimately, we
will want to target it.
We can do this in a gene-by-gene
and allele-by-allele fashion
today.
But we are also able to use
massively parallel methods using
CRISPR-based libraries to
mutate many genes simultaneously
or create many alleles
of genes simultaneously.
This is the work of
Francisco Sánchez-Rivera,
assistant professor
in the Koch Institute,
who has developed methods to
introduce, via prime editing,
whole libraries of guide RNAs to
mutate a given gene of interest
at every position, to create a
massive number of alleles across
that gene, ones that are
known to be mutated in cancer
and others that perhaps are
not yet known but might be
discovered in the future.
And by this method,
by introducing
a range of mutations in
a pool of cancer cells,
one can then subject them to
a variety of functional tests
to determine the consequences
of these individual alterations
and, again, help
predict what will
happen when we find
these mutations
in the clinic in the future.
Now, I've spent most
of my time this morning
talking about biology
and molecular biology
and genetic engineering.
But as I mentioned,
the Koch Institute
is also about the convergence
of engineering and science.
And so I want to give a few
examples of technologies
from the engineering domain
that are actively being pursued
and are impacting how
cancer patients are treated
and also how we discover
cancer in the future.
Phil mentioned that one of
the focuses of our research
is to find new ways to deliver
medicines to cells, including
cancer cells.
And we have great
expertise in nanotechnology
in Dan Anderson's lab, Paul
Hammond's lab, Bob Langer's lab.
They're leaders in this space.
But we're also
interested in using
new engineering-based
methods to deliver
more conventional
drugs, including
conventional chemotherapy.
You might be surprised to learn
that chemotherapy is used today,
and the dose of
chemotherapy is used today,
based on formulas
that were developed
more than a hundred years ago
that simply asked the body
size of the individual
who's being treated based
on their height
and their weight.
And by this very simple
formula, regardless
of the particular
individual, whether they
be very tall, short,
and heavy, they
might get the exact
same dose of drug.
And the distribution of the drug
in those two different people
is likely to be
very, very different.
And their genetics is
likely to be very different.
And their metabolism is
likely to be very different.
But the formula doesn't pay
attention to any of that.
And so Louis DeRidder,
a graduate student
in Bob Langer's lab
and Gio Traverso's lab,
working with collaborators
at the Dana-Farber,
decided to try to
bring technology
to solve this problem in
a very simple fashion,
to develop a closed-loop
system that would actively
measure the concentration of a
drug, like a chemotherapeutic,
while the patient
is being treated
to test the concentration
of that drug in real time
and then adjust the dose as
a function of the measurement
of the drug in the bloodstream.
They call this system CLAUDIA.
And it's not just an idea.
They've actually built it.
Again, the concept
with CLAUDIA dosing
is that one can actively measure
and therefore monitor and adjust
dosing so that the dose of the
drug in that individual patient
will be at the desired amount.
Whereas with the body surface
area dosing shown on the lower
left, one would
overshoot in some cases,
undershoot in others, depending
on the various factors
that I mentioned previously.
They've built this device.
They've tested it in
an animal model so far.
And what you can
see from these data
is that, compared to the
standard approach, which
overshoots in this
particular example,
leading potentially
to overdosing,
with the CLAUDIA system, they
can get very, very consistent
dosing across long
periods of time.
So it's my prediction
that, before very long,
this technology will
be used in the clinic
to improve efficacy and avoid
toxicity in cancer patients.
So for today, we need to be
able to treat cancer patients.
We need to develop new
precision medicines.
But as we look to
the future, I think
we can imagine a better time
when we can avoid cancer
altogether, when we can
predict more accurately who's
going to develop
cancer and intercept it
at a very early stage.
And here I want to
close with an example
from Regina Barzilay,
which does exactly
that, using the power of
artificial intelligence
to predict the future of cancer.
And this is a personal
story for Regina
because she herself developed
breast cancer in 2014.
And she, during that
journey, asked her doctors,
why couldn't you have
known this was coming?
You take an image of
my breast every year.
Surely there was something
there at an earlier stage
that would have told you that
I was going to get cancer.
And they put their hands up and
said, we do the best we can.
But Regina wasn't
satisfied with that answer.
And so she wanted to
know, perhaps there
are latent signals in
mammographic images
that would tell you that
there's a cancer coming
at some future time.
And so she then collaborated
with colleagues from the Mass
General Hospital and
elsewhere to use AI technology
and deep learning algorithms
to ask the question,
if you train computers
with thousands of images
from women who didn't get breast
cancer in the future, ones who
did in three years
or in five years,
can you actually
develop an algorithm
that can accurately, or at
least more accurately, assess
the risk of an individual?
And I'm happy to say
that this has worked.
In a paper published
a few years ago,
they demonstrated
that they can develop
a neural network which is quite
predictive in assessing risk.
It dramatically
outperforms the standard
used in the field
today, which are
based on things like family
history of disease or age
but nothing about the
particulars of the individual.
The AI-driven
model, called MIRAI,
greatly outperforms
that technology.
And because it was
developed initially
with a data set from a
limited number of hospitals,
they've now moved beyond and
tested these systems with data
from seven different
hospitals across three
different continents.
And the algorithm
works equally well,
if not a little bit better, with
those diverse data sets, which
is very, very encouraging.
And this has now been
implemented across the world.
This slide says there's
1.7 million mammograms that
have been screened by MIRAI.
I checked this morning.
We're up to 1.9 million.
So this is being
dramatically rolled out
to improve the doctor's ability
to accurately predict and then
monitor patients who are at
the highest risk of developing
breast cancer in the future.
So I'll close there
and remind you
that we're celebrating 50 years
of cancer research at MIT.
It's been a long and
remarkably successful journey.
But I think the best
is still yet to come.
We at MIT have
tremendous strengths
across our great University
to bring technologies
from far and wide to apply
to these difficult problems.
And I think with the
MIT HEALS initiative,
we have potential to
do that much more.
So I'll stop there and thank
you for your attention.
And I'll now introduce my
colleague, Doug Lauffenburger,
who will take it from here.
[APPLAUSE]
Morning.
I'm very appreciative to be able
to participate here and provide
a complementary partnership
perspective from engineering,
celebrating alongside the
remarkable past of Life Sciences
at MIT and anticipating
the surely even
more remarkable future ahead.
So I need to start from this
engineering perspective of what
engineering contributes.
Engineers create
technologies to be
useful to solve problems and
address challenges in society.
They do this by
bringing together
manipulation of components
and making new components
to construct these technologies.
But they do it in
a manner that's
going to be predictive rather
than trial and error, which
involves quantitative
measurement and modeling, so
synthesis and analysis.
And as was pointed out
by our Dean of Science,
you can't translate
anything from nothing.
And so the engineering
technologies
are grounded in particular
branches of science.
And that's really what defines
the discipline of engineering,
is what is its
foundational science
that it's going to
measure and model
and then it's going to
manipulate and make from?
So examples include mechanical
engineering and electrical
engineering, firmly grounded
in the realms of physics,
and chemical engineering,
materials engineering,
firmly grounded in the
science of chemistry.
Those are very well
established examples.
And, of course,
what happens then
is these engineering
disciplines are
aimed at a spectrum
of application realms.
Could be manufacturing, energy,
transportation, environment,
including medicine and health.
And so engineers
will bring together
their analysis and synthesis
of their branches of science
to address these problems.
And in fact, in the engineering
world, ever since the 1950s,
these classical engineering
disciplines have been applied
very effectively to
medicine and health,
with outcomes such as
prosthetics, kidney dialysis,
heart pumps, imaging,
drug delivery.
And these all came
from the physics-
and chemistry-based
engineering disciplines that
existed at that point in time.
And so if you look
back, you could
call this-- this really was mid-
to late-20th-century convergence
of engineering disciplines
with medicine and health,
just that minimal
biology was involved.
And so when people talked
about biomedical engineering,
it was a big M for medical and
a very, very tiny b for bio
because it really was not
involved in a substantial way.
Nonetheless, it was
the right convergence
for that time pre
the revolutions
that Phil talked about.
So since that time, since
the '50s, '60s, '70s types
of convergence, there have been
indeed these two major life
science revolutions or waves.
I like the notion of waves.
Molecular biology,
being the first,
permits biological science
to be fused with engineering.
Engineers can use molecular
and cellular components
to make new biological
technologies.
So that enables the
biological engineering.
The second one being the
genomic biology revolution.
And that, in fact,
requires bioscience to be
fused with engineering on the
highly quantitative measurement
scale modeling to
understand and model
in a predictive way
these very complex
molecular cellular systems
so that the new biological
technologies synthesized from
molecular biology revolution
can now be analyzed in
a predictive way based
on the genomic
biology revolution.
So you see, those
two revolutions
laid the foundation for a
brand new kind of engineering
that was not
possible before them.
Not surprisingly,
MIT is the place
at which this new
biological engineering
discipline was created.
In 1998, it was established,
became a full-fledged department
in 2005, and became course 20.
And what you see is
that it now stands
alongside the physics-based and
the chemistry-based engineering
disciplines.
But now its science foundation
is molecular biology
and genomic biology.
And from that foundation,
it can make technologies
in a predictively
designed way to apply
to manufacturing, energy,
environment, defense,
and medicine and health.
So it's alongside those
in being able to address
the challenges across the
entire spectrum of society.
Now, I want to make sure I
say something about education
because you really can't
make long-standing waves
or have a sustained
revolution without educating
the next generation of people
who now think in the new ways
and have capabilities
in the new ways.
So a very profoundly
important thing
about biological
engineering here
is it became course
20, the first new MIT
undergraduate
major in 39 years--
entirely new curriculum.
Students were rooted in classes
from the biology department
and genetics and biochemistry
and molecular biology.
But all the engineering,
analysis, synthesis,
measurement, and modeling
is brought to bear purely
and solely on the molecular
and genomic biology
in an entirely new way that has
never existed at any institution
worldwide.
And now these
students are going on
to become the next
generation who
are addressing these
problems in ways
that the physics-based
engineering
and chemistry-based
engineering just don't
have the capabilities to do.
And so now we can add to the
armament that the world needs.
I will note, at MIT,
education is a huge priority.
And in fact, I need to spend
happily, unfortunately,
most of my day in
my office having
office hours for my
undergraduate students
who have an exam
tomorrow morning,
and they want to discuss the
questions and the problems
and the topics.
And to me, that's
a huge priority,
as it is for faculty
across the Institute.
So what are the
products of this,
along with new types of
ideas and technologies?
Of course, not surprisingly,
companies have been spun out.
And I loved Phil's
list and could
put biological
engineering-spawned companies
alongside.
I'm just going to
note a few that
are widely recognized
as preeminent
in their particular fields.
So Adimab out of Dane Wittrup's
lab in biologics discovery
and engineering.
Synbio out of Linda
Griffith's lab in Oregon
on chip technologies.
Pivot Bio from Chris Voigt's lab
in agricultural biotechnologies.
Applied BioMath, now part
of Certara in systems
pharmacology, from myself
and a number of colleagues.
Also note Strand
Therapeutics at the bottom
now bringing synthetic biology
into the realm of engineering
immune cells to
be more effective,
again, in a
predictively defined way
by Ron Weiss and Darrell Irvine.
To end in the last few minutes
in a forward-looking way--
but, of course, to look forward,
you say, where are we going?
And, of course,
where you're going
depends on where
you're already rooted.
And so one of the
major things that's
coming out of this
discipline is addressing
what I think is one of the
long-standing challenges that
has not been addressed
effectively yet.
President Kornbluth talked
about the hard, hard, hardest
problems.
That's what this is.
You all are very well aware
of the attrition in the drug
discovery and
development pipeline,
that a very, very small
fraction of therapeutics
that enter the clinic succeed.
And the biggest
problem of that is
that the major costs
of the pipeline
are in the clinical trials.
So you're wasting the
vast majority of the money
that you're investing.
So this is a critical problem
that's been around for decades.
And again, people say, well,
we're doing the best we can.
We need to do better.
Society needs it.
So what is at the
root of this problem?
There's multiple
issues, of course.
But from our point of view, it's
the very poor translatability
from animal models
to human patients,
that one can do very, very,
very effective things in animals
that just don't work in humans.
That's what's got
to be improved.
And why is that?
Well, traditional
animal-to-human translation
focuses on using
genomics largely
and other quantitative
measurements,
identifying pathways and
processes involved in a disease
or lack of response
to a therapeutic that
are observed from very well
controlled animal experiments
and presume that,
if those are what's
most important in the
animals, then if we can find
some hint of them in the
humans, they're also going
to be important in the humans.
But this is incorrect.
In the vast majority
of cases, the things
that are most important
in the animals
are not what's most
important in the humans.
Therefore, things that
work very well in animals
are not effective in humans.
And this is the
long-standing lament.
Again, you're all familiar
with of curing disease
after disease in mice with no
consequent benefit in humans.
I call this the
"tip of the iceberg"
problem because what you're
looking at is what's most
important in the animal
with very little guarantee,
and, in fact, very minor chance,
that it's actually what's also
most important for
the human patients.
How do you address that?
Well, this is where
the combination
of biology and engineering
inherent in this new discipline
really holds a lot of promise,
that if you understand,
on the one hand, what
these biological pathways
and processes are, what
the data represent,
and, on the other
hand, you're capable
of the sophisticated
computational analysis
and modeling-- one
can bring together,
for instance in the
example I'm showing here,
using a semi-supervised AI
machine-learning approach
in which the secret sauce
is not the sophistication
of the computation, but it's
the understanding of how
the data map and what the data
mean biologically-- then you
can find pathways and
processes that are, in fact,
most important for
humans even when
they seem to be of negligible
importance in the animals.
So I call this, instead of
being focused on the tip,
you can actually find the
iceberg underneath that's
most important for the human
by this combination of omics
data, understanding biological
pathways and processes,
and the computational
machine learning.
These are projects we've
been pursuing not only
with a number of the biopharma
companies in this area
and elsewhere, but the
Department of Defense, the Gates
Foundation for vaccines, and,
not surprisingly, the FDA,
which wants to improve this
and the methodology for this
strongly.
Some examples have to do
with Humira resistance
in Crohn's and colitis with
a very well known biopharma
company, with the Army--
protection against
neurocognitive trauma--
and I can name others.
And most importantly,
this approach
is now being adopted into
industry therapeutics pipelines.
And so I believe that we will
see, over the coming decade
or so, significantly
improved translatability
from the preclinical
studies to clinical.
So that's just an
example of rooter
where we are now in
biological engineering,
looking at the kind
of challenge that
has been too hard to do before
but is no longer too hard.
With that, I thank you
for your attention.
I thank you for being here
at this momentous event.
And it's a great pleasure to
introduce our next speaker, Ruth
Lehmann, who's a professor
in biology and director
of the Whitehead Institute.
[APPLAUSE]
Hello, good morning.
It's a real pleasure
for me to be here today
and to celebrate HEALS.
Like Sally, I am also
a recent recruit.
And I also have to say that
it is just invigorating
and empowering to
be in an environment
where the University is without
limits and the environment,
the ecosystem of Kendall
Square could not be better,
and the hospitals
across the river
give us really an
opportunity to think big.
What I want to tell you
about is a little bit
about the Whitehead Institute.
And I'm calling it a
radical experiment.
This is actually
a title of a book
about the history of
the Whitehead Institute
that Gerry Fink just completed.
And at the beginning, when the
Whitehead Institute was founded,
it was clearly a
radical experiment,
and it wasn't always taken with
total positivity, let's say,
because it was a bold approach
having an Institute which
was independent in its
research, government,
and finance, solely associated
with an institution,
with MIT, where all the
faculty are teaching at MIT
and are in departments at MIT.
The founding vision
was very bold, as
bold as MIT is, to assemble
the best scientific talent,
to catalyze their
work with support,
and to encourage them
to go their own ways.
There were many
early breakthroughs.
And actually, you heard
about quite a few.
You heard several times
about the breakthrough work
of Bob Weinberg.
And that was at the
Whitehead Institute
in the field of cancer,
work in STEM cell biology,
transcription, and other fields.
In particular-- and again,
you heard this already--
one third of the
human genome was
sequenced at the Whitehead
Institute in collaboration
with the NIH and MIT.
And that led then,
by Eric Lander,
starting the Broad
Institute as a spin-out.
What's also really important is
shaping a new generation, not
just as teachers, but also
the Whitehead Institute
established the Fellows
Program for newly minted PhDs.
And the first notable Fellows
are David Page and Eric Lander,
and others are George Daley,
Angelika Amon, and Peter Kim,
and also David Bartel,
who was mentioned earlier.
So what are we about today?
Today, we are 18 members--
14 of them are tenured.
And we have six Fellows.
And we have a large number
of graduate students
and undergraduates and postdocs
working in our laboratories.
And probably, one would say
from just the recognition,
the experiment was a success.
But it's still an
experiment as I see it.
So what are we looking for?
And when I speak here
about the Whitehead,
I think I speak about biology.
I speak about basic sciences.
We really, today, want to
understand how organisms
perform their tasks in life.
And that is to encapture
the complexity.
We believe that that
has to be achieved
by a mechanistic
understanding-- as we heard
Phil Sharp saying, measuring.
We want to understand
it mechanistically.
But then we do want to
deconstruct and reconstruct
to really understand
complexity of life.
And that can be done only
by collaborative work.
And that is to draw from
biology, physics, chemistry,
engineering, and
increasingly computer
science, AI, and machine
learning to develop new tools
and designs.
In general, we don't want to be
afraid of asking any questions.
We don't want to be afraid
to change directions
if that is important.
And I think HEALS
will empower that
through the collaborations
that will spark new ideas
and new points of inflection.
So two points, when we
think about history,
we can jump at today.
We can analyze cells now
at a level of resolution,
and we can ask and model
cells in a way, that
was unforeseen before, leading
to a project spearheaded
by Jonathan Weissman, and also
with the work of Ian Cheeseman,
to describe actually
a virtual cell,
to know every component,
every aspect of what
is happening in a cell.
But going one step
further, how can we
understand a whole organism?
How do we understand the parts?
How do we bring neurobiology
and immunology together
to understand how an
organism actually functions,
how it feels an infection?
So these projects are
supported by some initiatives.
And these initiatives, they're
trying to be very bold.
And they're trying to really
address some of the pressing
issues of today's challenges.
So the Whitehead Initiative of
Biology, Health, and Climate
Change addresses
questions directly
of agricultural
questions in terms
of drought-resistant
seeds, but also,
how does temperature
affect the cell?
How does a cell work at
different temperatures?
And you will hear later
in the next session
from Sinisa Hrvatin
about hibernation.
The Women's Health Initiative
brings fundamental science
together with
really understanding
why is there a health bias.
And I'll talk to you a
little bit later about this.
And all of this could
not be possible,
grasping this
complexity, without using
AI approaches to lead the way
but also to be led by the data
that we are accumulating.
We are not alone.
I always see the Whitehead
a little bit as a hub.
We have lots and lots of spokes.
Many of those spokes go
across the street to MIT.
But also, we have lots of
interactions with hospitals
and other universities.
We seek out collaborations,
and collaborators seek us out.
So today what I
want to talk to you
is actually give you five quick
vignettes about some science.
And they're pretty much
all going on right now.
These vignettes are also
highlighting how we're perhaps
addressing, somehow, Doug's
criticism to basic science
that we're doing this all in
model organisms because, as you
can see, much of what
we're learning here
is directly either applied
to humans, in humans, or not
in humans but in plants.
And so I will be talking about
these five challenges in biology
and actually for human
health and the planet.
And so let's jump into it.
First, I'm going to talk to
you about the Lourido lab.
The Lourido lab
works on Toxoplasma.
And many of you know how
dangerous Toxoplasma can be,
as it infects a large
percentage of the population
and can have really
serious consequences
for pregnant women, infants,
and immunocompromised people.
So what Sebastian and his group
have done is they took all
the 8,000 genes of Toxoplasma
and made gRNAs against them
so that he can interfere with
those RNAs and then infect
the organism where normally
the infection occurs.
And then they're able
to use different tissues
and different-- and look
at what of these cells
are now surviving.
Those cells which are
surviving are obviously
not the ones whose genes are
required for the function.
And so he can directly
select with this whole genome
mutagenesis for genes which
affect fitness and has
identified a number
of genes which affect
the fitness of the infection.
And this is clear
how it's directly
applicable in the future.
The next, we go to another
challenging problem.
And that is, how do we make
better drought-resistant plants?
How can we deal with the
issue of food supply?
And so what most of you know
is that many seeds are actually
produced by crossing
different parent plants.
And then the seed itself is much
better than the parents were.
The problem is that
those seeds can not
be propagated continuously.
And so if they're
just grown, then there
will be a large variety.
So what Mary Gehring
is addressing
is to ask the
question of, could we
actually make seeds which
can be asexually propagated?
And that actually happens
sometimes naturally.
And so she is using these
kinds of seeds which sometimes
make that mistake of not needing
fertilization to find out
what factors are required.
And then can we up the
opportunity for seeds
to develop asexually and
then produce populations
of seeds that can actually
be used without having
the hybrid-vigor problem?
The next question is going
directly to human biology.
And that is addressed
by David Page.
His lab really is puzzled by
this striking and still very
unexplained sex bias in the
prevalence, the severity
of diseases between
males and females,
also, certainly,
drug sensitivity.
And so what they found is really
surprising and very interesting.
So when we just look at
all the cells in our soma,
in our body, not the cells
in our reproductive organs,
there are two types of
chromosomes, sex chromosomes,
the X and the Y.
And when we look
at the expression of the
active X chromosome in males
or females, so XX
versus XY people,
the active X is
expressed the same way.
So all the genes are--
there's no sex difference.
But the sex differences
come from the genes which
are expressed on the so-called
inactive X, which is actually
not inactive but expresses 38%
of the genes on the active X.
And also, there are genes
expressed on the Y chromosome.
And these could really influence
the expression of these genes
are making a cell different,
any cell in the body different,
whether it is in an XX
organism or an XY organism.
And so the idea is that this
can have an influence on disease
prevalence and sex bias.
The next issue is
another big question.
And that's the big question
of neurodegeneration.
And there is so little
that we know at this point
where we can really do something
about neurodegeneration.
A particular type
of neurodegeneration
is neurodegeneration
caused by prion disease.
Prion disease is usually a
very swift neurodegeneration.
And the death is due
to the accumulation
of misshapen proteins.
And so Jonathan
Weissman designed
CHARM, which is where a
fusion protein recruits
a methyltransferase specifically
through a DNA-binding domain
to a specific gene.
And by that, it methylates that
gene and silences that gene.
And so this can be
delivered into the brain.
And in mice, this reduces 80%
of the prion-gene expression.
And mice, which are
affected by prion disease,
live longer after
this injection.
And so my final
vignette is perhaps
the most biggest
challenge in human health,
and that is chronic disease.
There are many types
of chronic diseases.
And the problem with
chronic diseases
is they're not due
to a single gene.
So we cannot start
with a single gene.
But they have
multifactorial causes.
And we don't even exactly
know what the causes are.
So what Rick Young
and his colleagues
found is that often through
elevated oxidative environment,
through the pathogenic factors,
proteins are less mobile.
They are less mobile because
their cysteines are altered
so that they are now
interacting with other proteins
more through
cysteine-cysteine bridges.
And he phrases this
as proteolethargy.
And this proteolethargy
may, first of all,
lead to reduced
function and output.
And they showed this very
clearly with insulin receptor
mobility being reduced.
And it is also a target
which can actually be
more globally and widely used.
So I hope that with these
vignettes that I just presented,
you got an idea of where
basic science meets
translational science.
And that couldn't be
more better illustrated
than the many companies
which have sprung out
of these initiatives that
have started at the Whitehead
Institute.
I'm really looking
forward for HEALS
to further enlarge the
interactions that we have.
And I think the
possibilities for science,
both at the mechanistic
level and understanding
the complexity for
health and the planet,
is really, really great.
And I look forward to that.
Thank you very much.
[APPLAUSE]

---

### MIT HEALS Launch: Expansion plenary session
URL: https://www.youtube.com/watch?v=UKEF6ehVuWE

Idioma: en

I'm Katharina Ribbeck.
And I'm Amy Keating.
And it was our pleasure to chair
the program planning committee
for today's symposium.
It's incredibly exciting
to see it coming to life.
The hardest part of
this job, by far,
was putting together
just a single-day program
from the wealth of
ideas and innovation
that surround us at MIT.
It is, yeah, really this wealth
of discoveries and innovation
that makes MIT so exceptional.
Here we are not just
exploring new frontiers,
we are actively creating them.
Now we have the
privilege of hearing
from four remarkable
early-career-stage faculty who
exactly embody this spirit.
These rising stars
from chemistry,
biological engineering
and neuroscience,
are not only expanding
our understanding.
They're taking discoveries
in bold new directions
and really influencing the way
we think, the way we innovate,
and the way we collaborate.
So professors Oleta Johnson,
Bryan Bryson, Sara Prescott,
and Sinisa Hrvatin
will share with you
their visions for the
future in four short talks.
Please join us in
welcoming them.
[APPLAUSE]
Good morning.
I am excited to be here today.
I'm an assistant professor.
This is my second
year here at MIT,
and my lab is broadly
interested in using chemistry
to direct protein
traffic in cells.
Before I tell you
what that means,
I want to start by
asking you a question.
What do these five
people have in common?
I'll give you three
seconds to think about it.
All right, if you thought
that these people are
leaders, innovators, or icons,
I'd totally agree with you.
As it pertains to my research,
they share, unfortunately,
another commonality.
All five of these
people were diagnosed
with a devastating
class of disorders
known as neurodegenerative
diseases, that ultimately robbed
them of the physical and
cognitive faculties that
enabled them to have such
a huge impact on the world.
With that in mind, I invite
you to imagine a world
where a diagnosis with
one of these diseases
is not a timer, counting down
to some inevitable decline
in the way that you engage
with those around you.
My lab is working right now to
turn that world into a reality,
and I'll give you an overview
of how we plan to do this,
but first, I want to
be specific about what
these five different diseases
actually have in common.
If you look at the photos
at the bottom of the slide,
these are all photos of tissues
from patients that are diagnosed
with these diseases.
And you'll notice
these dark spots
that are on the
slide in these cells.
Those are proteins
that are aggregated,
a common feature of
all five diseases.
What that means at
the molecular level
is that proteins
that are normally
pretty flexible in a cell
to do their normal function,
they are somehow--
they somehow get
clumped together
in what we call these
protein aggregates.
We don't always know how they
happen or why they happen,
but I like to think of some
cellular protein highway
where one path gets cut off,
leading to this cellular protein
traffic jam.
Similar to if you were to say
shut down the Sumner Tunnel.
And now there's
a giant aggregate
of cars on the way
to the airport.
With that in mind, it's
important to consider
ways to prevent these
cellular protein traffic
jams from happening
in the first place.
Luckily, there's a protein
that I care about called DNAJB6
that is already very good
at directing protein traffic
and preventing these aggregates.
But as you've seen, it doesn't
always work in these diseases.
Based on what we
know about DNAJB6,
my lab has three
big-picture goals.
We want to develop
new chemical tools
that we can use to detect
these aggregates earlier
than we can right now.
We want to learn more
about the mechanisms
that DNAJB uses to direct
protein traffic to see
if there's any clues there on
what we could do as chemists.
And we want to discover
chemicals, small molecules that
can work with DNAJB6 to either
activate or mimic its activity
and prevent aggregates
from happening.
And I was really lucky to
recruit three fabulous graduate
students in my first year
to work on these problems.
Cordiana is working on a
fluorescence-based platform
for early detection
of protein aggregates.
Aaron is taking really
complicated DNAJB6 structures
and using chemistry to
make them more simple,
to make it easier
for them to study,
for us to study how they work.
And Isha is looking for chemical
detectors that activate DNAJB6
by acting as glues, recruiting
it to a protein of interest
in a disease of interest.
And that's just the beginning,
because last month I
recruited three more
fantastic graduate students
in my group who
are going to work
on three new projects
towards our goals,
and together we hope
that that will result
in better diagnostics and better
therapeutics that, had they
existed 20 or 30 years ago,
would give us more time
to learn from the
wisdom of experience
of people like Rosa Parks,
Muhammad Ali, or Stephen
Hawking.
But most importantly, we hope
that the tools that we develop
in my lab could be used to
improve the quality of life
for people who are diagnosed
with these diseases.
Ordinary everyday
people like you and me.
Thank you.
[APPLAUSE]
Good morning, everyone.
I'd like us to imagine a
world without tuberculosis.
This is the Granville mummy,
estimated to be found in 600 BC.
At the time when they
unearthed this mummy,
they thought that the cause
of death for this mummy
was ovarian cancer.
However, in the early 2000s,
researchers found that this
mummy had had tuberculosis.
Fast forward to
the 19th century,
where it's estimated that one in
seven individuals was dying from
tuberculosis.
This is an ad from the Christmas
Seals asking you to donate money
to stamp out tuberculosis.
But unfortunately, I'm here to
report that the Christmas Seals
were unsuccessful in this
campaign because today alone,
4,000 people will die from TB.
OK, so what would
be transformative?
What would be transformative
for TB is a protective vaccine.
And this is something
that we do not yet have.
We have one vaccine for TB.
It is over 100 years old, and it
doesn't meet the product profile
needed to end this pandemic.
However, if we did, it
would change history.
So looking at this graph, this
is TB incidence per million
on the y-axis and
time on the x-axis.
If we do nothing and we
just keep standard of care,
that's the red line.
If we had a 60% protective
vaccine and a mass vaccination
campaign, look at
what could happen.
All right, so the question
becomes, how do we get there?
How do we make a
protective TB vaccine?
All right, so
that's the question
that my lab decided to tackle.
So why aren't we there yet?
We heard about measurement.
I would like to submit to you
that this is a measurement
problem.
So the question
becomes, how do you
design a vaccine
in the first place?
We might all be experts
about designing vaccines
for viral pathogens,
but tuberculosis
is caused by a bacterium.
So when I think about this
problem as an engineer,
I think about the
following question, what
proteins from the bacterium
does the immune system see?
All right, this is an
especially hard problem
in the context of
infection with bacteria
because viruses have
tens of proteins.
TB has 4,000.
So it's not that simple just
to say which protein do I pick.
It's actually a
really hard problem.
And to make that problem even
just a smidge more difficult,
we have to do this
in a biosafety level
three environment, which
means that everything just
goes much slower, much more
complicated, and much more
arduous, because this isn't some
place where you can just say,
I want to try something
new and do it tomorrow.
This takes a lot of
effort and energy.
So despite those roadblocks, I'm
really excited to share with you
that we've made significant
inroads to this problem.
OK, so how do we do this?
So work led by an
incredibly talented graduate
student in collaboration
with Forrest White.
We set out to address
the measurement problem,
asking specifically
in human cells
across diverse human
genotypes, if we
infect with TB, what
proteins from the bacterium
are seen by the immune system?
So I'm going to call
this problem solved.
We can do it.
We can do it within days now.
OK, So now the next
question becomes,
if you can make
these measurements
and ask what the immune
system sees with bacteria,
can we ask what does
the immune system see
when you give a vaccine,
including those proteins
from the bacterium?
And that's what we've
been able to do now.
So now we can design
vaccines as mRNAs
and within weeks get an answer
about a particular design
that we've made.
And to give you an example
of what we've done just
in the last few years, is we've
made all these different vaccine
prototypes, and I don't have
the time to go through all
the details, but what we can
do now with these measurement
technologies and say, what if
we change just one parameter
about this mRNA vaccine?
How does it change
immune output?
So that's what you're looking at
here on the right graph, where
I'm just showing two vaccine
designs where we just
made a small 100 amino acid
change to the mRNA vaccine,
and we can change immune
output by orders of magnitude.
So while this has seemed like a
millennia-old problem for a very
long time, I'm very optimistic.
I'm incredibly optimistic
because I think we can do this.
So I'll leave you with
a provocative question,
which is, will the
next TB vaccine come
from Kendall Square?
And I think it will.
I'm very encouraged by this
because in the last six years
I think we've done so
much to answer really
a millennia-old problem.
But I think we can
do so much more.
And I think the
momentum is on our side.
Thank you.
[APPLAUSE]
Imagine a world where we
understood how our daily stress
was damaging our health.
We could track it
with simple tests
and intervene to prevent or
reverse its harmful effects.
In this world, stress
would no longer
silently fuel chronic
disease, wreak havoc
on our mental health, and
reinforce social inequities.
We all face stress
in daily life,
but globally, stress levels
are now at an all-time high.
This Gallup poll asked
people around the world
to self-report
their daily stress
levels as part of a
Negative Experience Index,
and for over a
decade, stress levels
have been consistently
rising, driven
by factors like financial
insecurity, social isolation,
and workplace burnout.
These same studies show that
stress levels are unequally
distributed across different
countries and demographics.
Within the US, stress
levels are highest
in college-age adolescents,
in women, and low-income
and minority groups.
Looking at these
statistics, it's
hard not to feel that we are
living in a pandemic of stress
with marginalized
communities and our youth,
the very future of our nation,
bearing the heaviest burden.
The consequences
of this on society
are massive, damaging not
just our mental health
but our physical health as well.
Children, for example,
who experience
severe early life stress as
measured by high ACE scores,
lose over a decade
of life expectancy
and are at higher
risk for conditions
like cardiovascular disease,
infertility, and cancer.
This association between stress
and worse health outcomes
is also seen in animal models.
For example, early life stress
reduces survival in wild baboons
and cause mice that are
raised in lab environments
to develop predictable
pathologies,
like making their
very hair turn gray.
These observations tell us
that these associations between
stress and worse health outcomes
cannot be explained entirely
by diet, smoking, and access
to health care services.
And it begs the question of what
biological processes link stress
to increased disease
and mortality risk.
I'm a neuroscientist,
and my lab studies
how the brain controls essential
functions like breathing
and heart rate.
And we see stress not
just as a feeling,
but as a biological program with
specific purpose and patterns.
When we face a threat, the
brain primes the body for action
by triggering the
sympathetic nervous system
to release norepinephrine,
causing rapid breathing,
a racing heart, and sweating.
These circuits also
signal endocrine centers
like the adrenal
glands to produce
cortisol and other hormones,
keeping the body on high alert.
Cells throughout the body
receive these signals
and are forever changed.
With new tools, we can
now monitor and target
these pathways with
unprecedented precision
to ask which precise
circuits are disrupted
by chronic and severe stress.
How do stress signals
affect different cell types
throughout the body, and how
are these changes remembered
at the molecular scale?
Through this lens,
we can view stress
as a predictable program that
can be tracked and treated.
To understand this, we're
collaborating with other Boston
area labs, using animal models
to control stress exposures,
and then using advanced
imaging and molecular tools
to track within
individual animals
how this rewires brain
activity and disrupts
physiological systems.
These comprehensive data sets
capture a more holistic view
of stress and are helping
us predict causal brain
changes that alter
bodily states and lead
to stress-related disease.
Looking ahead, addressing
this health crisis
requires new perspectives
and solutions.
As for cancer research, we
must both learn core principles
while also embracing
stress' complexity,
and endotypes if we want to
advance precision treatments.
This effort will rest on
pillars of collaboration
across disciplines to better
understand the molecular scars
of stress and to build
evidence-based diagnostics
and interventions
within our community.
And I for one am looking
forward to this bright future.
Thank you all for
your attention.
[APPLAUSE]
Good morning.
My name is Sinisa Hrvatin.
I'm honored to be here.
I want you this
morning to imagine
a world in which suspended
animation were possible.
A world in which a 10-year-old
girl suffering an incurable
disease, like
maybe glioblastoma,
instead of imminently dying,
has the choice to enter a state
of suspended animation for
a few years, maybe a decade,
and then wake up at a time
in which there is a treatment
or perhaps even a
cure to that disease.
I know this is currently
science fiction, pretty much
complete science
fiction, but what
I want to share with you today
is that there are animals
in nature that have the
ability to enter similar states
and that have the ability
to enter something
like suspended animation that I
think we can learn so much from.
And then learn how
to eventually develop
suspended animation for humans.
So I'll share with you examples
of a few of these animals.
This is the Arctic
ground squirrel.
This animal lives
above the Arctic Circle
at very low temperatures,
and during most of the year
it has a normal
body temperature,
just like you and me,
around 37 degrees Celsius.
But during the winter, it can
survive by entering hibernation,
and it can lower its body
temperature all the way down
to around 0 degrees Celsius,
entering a sleep-like state
that we call torpor.
My laboratory studies
how animals enter torpor
and also how do
their cells survive
at these very low temperatures.
And I want to share with
you a couple of things
that we've recently found.
We identified a
population of neurons
in the brain that
controls torpor.
And remarkably, we can actually
initiate a torpor-like state
even in animals that
don't naturally hibernate,
just by stimulating
these neurons.
So here you can see an
example of a laboratory mouse.
This is a thermal camera image.
It has a body temperature
around 37 degrees Celsius.
But when you stimulate
these neurons,
you can lower the
body temperature
of this animal all the way down
to around 25 degrees Celsius.
Now, this is not as low
as a natural hibernator,
but it's actually
really cold for a mouse.
And we can do this not
only for minutes or hours,
but actually for days to weeks,
showing for the first time
that we can induce a
long-term torpor-like state
in an animal that doesn't
naturally hibernate.
We've also, in a
parallel study, started
looking at how do
hibernator cells survive
at these low
temperatures and seeing
whether we can learn from their
mechanisms to improve human cell
cold survival.
And here you can see an example
of a handful of human cell lines
that don't fare very well
at 4 degrees Celsius.
But then when we take
small molecules that
mimic some of the adaptations
that hibernators have
during hibernation and
treat those cells with them,
we can dramatically improve
their cold tolerance.
And you can
immediately imagine how
this could be useful for organ
storage or organ transplantation
and also how
eventually we could use
this to learn how to lower the
temperature of the human body
to lower and lower temperatures.
And finally, perhaps
the greatest example
of suspended animation in nature
are these tardigrades, cute
little organisms that actually
live all over our globe,
from the Arctic
to the Antarctic,
and they have an
incredible ability
to survive being
frozen all the way down
to -150 degrees Celsius.
And they can stay in that
state for months to years
and then come out of it
essentially unharmed.
We're trying to understand
how tardigrades survive this
and whether we can transfer
some of those adaptations
to human cells.
Now life on Earth has evolved
incredible adaptations
to survive in
extreme environments,
and I believe that by
studying these animals,
we'll learn to harness
some of the mechanisms
behind these adaptations
to advance medicine,
perhaps even one day open
the door to fantastic
possibilities of space travel.
Now, this work needs to start
with fundamental biological
discoveries.
But in order to develop
this as a technology,
we also need to collaborate with
engineers, chemists, physicists,
physicians.
And this is precisely why
there is no place like MIT
to pursue this dream.
Thank you.
[APPLAUSE]

---

### MIT HEALS Launch: Systems, spoken word performance
URL: https://www.youtube.com/watch?v=EOR114NFNpo

Idioma: en

Victory Yinka-Banjo was born
and raised in Lagos, Nigeria
and moved to the United States
on a student visa in 2021
to attend MIT.
She's majoring in Course 6-7, or
computer science and molecular
biology.
And she's done
undergraduate research
with Professor Jonathan Weissman
at the Whitehead Institute
and Professor Caroline Uhler
at the Broad Institute.
Currently, Victory is
preparing for graduate school,
where she'll prepare--
pursue her interest
in molecular medicine.
And for those of
you here who are not
involved in this
process, I'll mention
that the applications
to most graduate schools
were due Sunday.
So between that and
this performance,
I'm not sure how restful
Victory's Thanksgiving break
was.
Victory excels in science,
in computational thinking
and also in the arts.
She's involved with the
MIT Black Theater Guild
and has composed and
performed spoken word poetry
for MIT student groups and
for the celebration of Sally
Kornbluth's inauguration.
For today's celebration
of the life sciences,
Victory will present
her piece "Systems,"
which will weave together ideas
about systems of molecules,
disciplines, people
and countries.
Please join me in
welcoming Victory.
[APPLAUSE]
M. I. T. As my words
meet your minds and minds
meet your hands and your
hearts, I need you to zoom out.
Adjust the magnification
of the lens
that you walked into this
room viewing the world through
and give me a chance to
communicate what I argue
is an extended breakthrough.
Because if there was one
thing that my pending biology
degree has shown
me, it is that there
is power in thinking beyond
what exists in our proximity.
Now, travel with me to
a world observable only
by electron microscopy
and you will find
meaning behind this philosophy.
Now, under a microscope adjusted
to its highest magnification,
you might see a nucleus, a
mitochondrion, a ribosome,
individual organelles packed
with biological power.
But if you dare to zoom
out, what you will find
is the subcellular
components operating
in ways far more powerful
because together they are parts
of a whole, subparts of a cell.
In synergy, these organelles--
they form a system of whole cell
complexity.
Now, zoom out again and you'll
find that these cells come
together as tissue, and
tissue together as organs,
and organs with systems
of blood vessels,
and fluids and
skeletal structure.
They create something
systematically
beautiful and dependent as
every human in this room.
To see that you don't need an
external lab tool in accounting
for this larger
picture, our eyes
do more than a
microscope's view see.
Gone are the days of
reductionist biology
of taking pieces apart,
considering them individually.
We are at MIT, so we
build with mind and hand,
put pieces together.
This collaborative is not just
about one part, one subject,
just about biology.
It is interdisciplinary and
to claim otherwise is to fail
to zoom out and look back on
our rich history of revolutions,
be ignorant of our promising
future of expansions,
a history of fundamental
discovery decorated
by the sequencing
of the human genome,
unleashing a scientific
era broader than the Broad
Institute, a history of sharp
Nobel laureates transforming RNA
splicing and medical imaging,
a history that birthed HST,
solidifying relations from
MGH and Harvard Med to MIT,
and a future of expansions
with breakthrough synergy that
advances cell and gene
therapy, uses AI to accelerate
diagnostics and fight antibiotic
resistance of mechanical
engineering building 3D-printed
hearts that pump life into new
treatments, a future of
memories governed by McGovern's,
collaborations with the Picower
and understanding of brains
fueled by molecular power.
This collaborative is
not just about biology.
It is a celebration of
an institute and industry
identical to a table where
everyone can find a seat.
A celebration of a
system built by parts
where everyone
finds their niche,
whether that's in
science and engineering,
where you can
interface technology
into clinical medicine
to develop a therapy
and design its delivery.
But for that therapy to
become a patient's reality,
we need people who
innovate beyond science
to overcome policy, understand
the nuances of clinical trials
and testing.
We have to think about
the markets governed
by a for-profit motive.
And so we need economists
for investor connecting
and for strategized
research funding.
But we also need entrepreneurs
for commercialization advancing
and for biotechnology
proliferating.
And, of course,
we need physicians
to understand the anatomy
of the very people impacted
by this pipeline
of R&D to ensure
the final life-changing delivery
that brings all of this back
to the ordinary people going
about everyday life activities,
see all these [INAUDIBLE]
must synergize
to go from bench to bedside.
But to claim that this
collaborative is only
about medicine is to ignore
the achievable global scales
of science where living
systems can power
the sustainability of life
on Earth as we know it.
Enzymes for plastic waste
degradation, bacteria
for decarbonized food
production, [INAUDIBLE] scales
that allow us to zoom out
of continental borders,
considering systems
of land bridged
by oceans, networks
of nations and synergy
to tackle global
health issues, networks
of nations that come together
to deliver breakthroughs
like a record time vaccine for
a pandemic that was unforeseen,
networks of nations
extending from the Americas
to the ends of the
Earth, including Asia,
including Europe, including
Africa, where I come from.
Gone are the days where science
just stays within our proximity.
To be at MIT is to take
our innovations and science
to a world that needs it
but also bring the world
to a science that needs
it, bringing people like me
to a campus where
they're welcomed.
Would you have imagined that a
Nigerian-born visa holder would
discover at MIT her superpower,
that at the crossroads
of computation and
biology she would tower,
would break glass ceilings
and surmount racially
discriminatory systems,
a woman who 100 years ago
would have been a
victim but today
is contributing
to the advancement
of multiple disciplines.
To be at MIT is to strive to
increase the representation
of minority groups,
where even when
we have declining
diversity in admissions,
we can hold steadfast to our
exemplary all-women president,
provost, and
chancellor positions
and hope the next
time we do better.
To be at MIT is to contribute
to an era of infinite impact.
It is to look beyond the
microscope, zooming out
to embrace the grander scope.
To be at MIT is
to latch onto hope
so that in spite of a global
pandemic, we fight and we cope.
We fight with science and
policy across clinics, academia
and industry for the betterment
of our planet, for our rights,
for our health.
We fight with might.
To be at MIT is to
pursue translation
for more than just Cambridge,
for more than Massachusetts,
for more than one nation,
for a network of nations,
for the globe.
It is right here,
right now that we
set this foundation and
this continuous pacesetting.
Oh, it goes far beyond
differential equations.
It will take collective actions
today, our minds and our hands
to ensure a future that
outlasts us tomorrow.
MIT.
To be here is a privilege,
one that we can pay forward.
So will you exist as yourself,
as an individual cell
unnoticed without a
microscope or will you
become visible by
staying connected
to the far-reaching life
science and health system?
Will you zoom out.
[APPLAUSE]

---

### MIT HEALS Launch: Brains breakout session
URL: https://www.youtube.com/watch?v=lvMWYeM7PJI

Idioma: en

I think we'll go ahead and
get started since we're
starting a little bit late.
I'm Emery Brown, I'm from BCS,
Brain and Cognitive Sciences
Picower Institute.
I'm going to be the moderator
for this session on brains.
So why think about the brain?
It's this fantastic organ.
It only weighs 3 pounds,
but it consumes almost 20%
of the oxygen in
the cardiac output.
And studying the brain is a
challenge because almost--
but what's cool about it
is that everything we learn
can help us with new diseases
relevant to the brain.
And the critical factors,
as was mentioned earlier
in today's first session, is
that, just like the other areas
of investigation
that were discussed,
brain research and brain related
improvements in health care,
there are three things
that converge here
at MIT, advances in science,
engineering innovations,
translational applications.
And you'll hear that
today from our speakers.
So these are our speakers.
We're very happy to have
Ed Boyden from the Picower
Institute, Mark Baier, also
from the Picower Institute, Josh
McDermott from the
McGovern Institute,
and also Evelina Fedorenko
from the McGovern Institute.
And we just made a
little chart to show you
that indeed, these
three areas will
be touched throughout with the
talks that you're going to hear.
So with that, I'll turn
it over to my colleagues.
Ed.
[APPLAUSE]
All right, yeah,
thank you so much.
This is really a
fun day and it's
great to see so many
convergent fields
in this amazing, important area.
I direct a group that
tries to confront
through technology the spatial
and temporal complexity
of the brain.
And this is a great
arena for innovation.
I mean, the brain is
a gigantic object.
And brain cells are
gigantic objects.
But the wiring of the
brain is nanoscale, right?
The wiring is extremely tiny.
And the connections
between brain cells
are also very, very small.
Also, we have molecules, right?
The human genome has, what?
30,000 or so genes encoding
for countless biomolecules.
And we need to know what they
do and how they work together
and how they go wrong
in disease states.
But it's even more
complicated because there's
a time dimension, too.
If you think about Alzheimer's
disease or learning or aging
or almost any process that's
important to us as people,
there's an extended component of
time related to these functions.
Yet, the elementary
neural codes of the brain
are very, very brief,
millisecond timescale electrical
pulses.
So we build tools to try
to help neuroscientists
conquer these differences
in space and time.
Now, why is it important?
Well, lots of
technologies exist.
On the left, MRI scans,
on the right, microscopes.
But it's very hard to
image a large object
like a brain down to those
nanoscale dimensions.
So in our group, we practice a
lot of creativity skills, which
we think can be
learned and practiced,
and one idea is to do
the opposite of what
people are doing.
And for literally
300 years, the way
that biologists
magnify images of cells
is with some kind
of lensing effect.
We started wondering,
what if you
could magnify cells directly?
And so the basic idea, we
call expansion microscopy,
we chemically install a
dense, spiderweb like mesh
of, basically, baby diaper
polymer inside the brain.
Not a living brain,
but a preserved one.
And if you do it just right,
you can weave that chemical mesh
inside cells and outside
cells, around biomolecules,
in between biomolecules,
and you can
magnify the brain physically.
So in panel B here is a
piece of the mouse brain
several millimeters on the side.
Panel C, the same piece of
brain tissue about a day, day
and a half later, but we've
blown it up by 100 times
in volume.
And the polymer starts
out super dense,
like in the upper
left in that cartoon,
and ends up like
in the lower left.
Here's a little movie of a
piece of brain being expanded.
So we formed the baby
diaper polymer earlier,
and it's half an hour sped
up to about half a minute.
But we add water right there.
And I hope you can see that
this piece of brain tissue,
which is permeated by
the baby diaper polymer,
is physically growing right
before your very eyes.
That's great.
So now, we can use
imaging tools and see
tiny things that
are too small to be
seen by regular technology.
And that's important because,
again, molecules are tiny
and the wiring of the
brain is also really tiny.
And yet, you want to understand
it across an extended 3D spatial
extent.
So here's an actual
piece of brain tissue.
This is from the
mouse hippocampus,
which is involved
with many functions,
including related to memory, but
lots of other things as well.
And the brain cells are
expressing fluorescent proteins.
So these are proteins borrowed
from jellyfish, coral,
and other creatures.
And it gives you a color
code or barcode, if you will,
so you can tell the cells apart.
We deliver the genes for
these proteins using viruses.
We then preserve the brain
and expand it using the method
that I told you about.
And now, we can see the final
wiring much better than before.
So we're going to fly through
this piece of brain right now.
And I hope you can see that
these individual processes,
which have distinct colors,
are much, much easier
to see than one of my guests.
And then, of course,
we can Zoom in and out.
It's intrinsically
a multi-scale way
of visualizing the
brain circuitry.
So we're excited about
this idea and how we can
use it to map entire brains.
But of course, this doesn't
work on a living thing.
And so we have to
really switch gears here
to tackle the second
question, which is time.
How do we control these fast
electrical pulses of the brain?
If we could control
them, then you
could drive certain brain cells
and figure out what they do.
Do they trigger a
behavior or a pathology
or perhaps the remedy
of a pathology?
And so the idea that
we brought forth
is this concept that we
now call optogenetics,
because opto is about
light, and genetics,
because we use genes to
mediate light sensitivity.
The basic concept is
basically like this.
Suppose you had
solar panels that
convert light to electricity.
If you could somehow
install them on brain cells,
shine light on the solar
panels, convert to electricity,
you could control the brain.
And then, you could bring
light into the brain.
Electrodes are brought into
the brain all the time.
Because it doesn't feel pain
we can bring optical fibers
in as well.
So how do we do it?
Well, we have to make
the neurons sense
light, which means we have
to find those solar panels.
And it turns out, all
over the tree of life,
you can find molecules
that do that, right?
So here we are looking at a
single celled green algae.
And it has an eyespot in the
back there, which helps it
sense light to control
those flagella,
those tails that help it swim.
We just zoomed in
to the eyespot,
and it's chock full of these
proteins that convert light
into electrical
signals, ion fluxes.
Kind of like those solar panels
that we were just talking about.
So what we can do is take
advantage of the fact
that this is a protein.
It's genetically encoded.
It has the right temporal
and other properties.
It's a fast response to light
because you just open up a pore
and instantly get
ion flux across.
And so the idea that
we brought forth
was, let's borrow this
gene and see if we
could put it into the brain.
And so this is where we had
to count on serendipity.
We're very excited now, by
the way, to try to use AI
and other methods to do on
purpose what we did accidentally
here.
Can we go out and look for
treasure in the natural world
and not just wait around
for it to show up?
Anyway, we can
take the gene that
encodes for this light
activated protein,
we can install it using
a gene delivery method,
like a gene therapy vector
into the brain cell.
The brain cell will manufacture
the light activated protein.
And then, again,
serendipity kicked in again.
It was safe.
It worked.
It didn't have to work.
The algae could have been a
happy home for the protein
and the neuron could have
died or malfunctioned.
Anyway, it was really
surprising in a pleasant way
that when we shine light
from a laser or an LED
on these brain cells, they
fired electrical pulses,
not unlike the ones
happening in your own heads
right now as I say these words.
So there's lots of
things you can do.
You can take different
parts of the brain,
put these light activated
molecules in them,
try to see what happens.
So I'll just pick one example
because it has a nice movie.
But dopamine is a very important
neuromodulator in the brain.
In the popular press, it's
often called the pleasure center
of the brain.
Could you find out
whether reinforcing
the activity of these neurons
also reinforces behavior?
And so it's a very
simple experiment here.
Mice go and poke their
nose in a little portal.
And if they go to
the right one, they
get a pulse of light that aims
at these dopamine neurons, which
are expressing the
light activated protein.
If they go to the left
portal, nothing happens.
And so you can see the
result. Lifts its nose,
gets the pulse of light,
does it again, does it again.
And so this mouse is,
basically, working for light.
So this is very widespread.
We share all of our
tools freely, by the way.
So expansion microscopy
that I talked about before
is in use by thousands
of research groups,
almost 900 experimental
results coming out already.
Optogenetics is in use by
thousands of research groups
as well.
And interestingly,
these molecules
are starting to be
used in people as well.
So many people are blind.
They've lost their light
sensors in their eyes.
A team in Europe got
one of our molecules,
discovered here at MIT, and was
able to deliver the gene for it
into the eye of a person who
lost his normal light sensors.
And they converted the
spared cells of the eye
into, well, kind of a camera.
And so he was able to
recognize household objects.
He could see doors on a hallway.
Not perfect vision, couldn't
recognize faces very well,
struggled to read text,
so more work to be done.
But still, a very
interesting advance
in an otherwise extremely
difficult to confront condition.
Going forward, we want
to integrate these tools.
We've always shared
the tools freely,
but we want to
start combining them
towards this goal
of making computer
simulations of the brain.
If we could simulate
the brain, of course,
that'd be very interesting from
an AI perspective, potentially.
But also, we could maybe
look under the hood
and poke around in silico--
in software and
try to figure out
where in a network to intervene
to help with the disease state.
And of course, if we could
simulate brain computations
of different kinds,
it could become a tool
that we could use to probe
different scientific questions.
So in the coming years, our hope
is to start with small brains,
like fish and worms,
and start going
to larger and larger brains.
And this is sort of
a complicated slide,
but it outlines one
of our ways of trying
to put together all the
different approaches that we
take.
So the upper left, we
want to look at the brain
while it's in action,
the blue part.
In the lower left
with yellow, we're
going to use the
expansion method
to make maps of the brain.
And then, the three
blocks to the right
are ways of making different
models, machine learning
models, biophysical
models, and so forth.
So to conclude, we're very
excited about these tools
we're building to see, map,
control, and understand
the brain.
We share them all very freely.
But in the time to
come, we're very
excited to try to integrate them
into a workflow that hopefully
can help us build computational
models as well that
can be of great use for others.
Thank you.
[APPLAUSE]
Just let me grab my jacket here.
All right, sounded like an
unruly classroom earlier,
but I'm glad
everybody's calmed down.
A sad fact of life is that
diseases, deprivation,
and injuries can
impair brain function,
and it's extremely difficult
to restore brain function,
particularly in adulthood.
And a very classic
example of this
is a condition called
amblyopia, which
is a condition that is a
visual impairment that develops
in children that don't have high
quality vision during infancy.
And that's what's shown
here on the left side.
This is a child that has
a cataract in one eye.
The cataract is easily
detected and can
be surgically
removed and replaced
with an artificial lens.
But as a consequence of that
altered visual experience
during infancy, there's a
permanent visual disability.
So there's a very severe visual
disability in the amblyopic eye,
the eye that had been deprived,
and a loss of stereoscopic depth
perception that has a major
impact on the child's life.
Now, there is a treatment
for this type of disease,
it's called patch therapy.
It was introduced well
over 100 years ago
and is still the
medical standard
of care, which is something of
an embarrassment, I would say.
But patch therapy can work.
So the patching occurs
in the other eye,
which we call the fellow
eye, and it forces vision
through the eye that
had been deprived.
The limitation of patch
therapy is that it only
works very early in life.
In this case, on
the left in humans,
if the patching is not
done before the age
of four months of age, then
there is no recovery of vision.
So if the cataract was not
detected early enough in life
or-- and corrected, this is
really a very poor prognosis
for that child.
We've been modeling this for
over 50 years in animal studies,
and that's what's shown on the
right is just a similar story.
This is a cat model of
deprivation amblyopia as well.
So I didn't mention, but
the cause of amblyopia
is an impaired
synaptic development
in a part of the brain
called visual cortex.
So what we'd like
to do is to harness
the principles of experience
dependent synaptic plasticity
to promote recovery
of brain function,
particularly in adults.
So over, again, the last
half century or so, we've
learned a tremendous
amount about the mechanisms
of synaptic
plasticity, plasticity
at excitatory
synapses in the brain.
We know that synapses are
bidirectionally modifiable.
And one of the critical
triggers for this plasticity
is a neurotransmitter--
whoops, can I go back one?
A neurotransmitter receptor
called an NMDA receptor.
And what we've learned is that
activation of NMDA receptors
can trigger both
synaptic depression
and synaptic potentiation.
And the critical
variable is the level
of postsynaptic response at the
same time those receptors are
activated.
And so what you see on
the right is a function
where we can map the
consequences of NMDA receptor
activation depending on the
level of postsynaptic response.
So there's a critical site here.
Again, go back please.
My laser pointer doesn't work.
One forward.
Thank you.
So you can see strong NMDA
receptor activation can yield
a synaptic potentiation and weak
NMDA receptor activation yields
synaptic depression.
So what is controlling
this postsynaptic response?
Well, obviously, it's
how many synapses
are active at the
same time and what
the strength of those
synapses is, and also,
the concurrent level of
inhibition in the brain.
So we'd like to
understand how we
can promote the
conditions that will
lead to synaptic strengthening.
So a lot of our
work was influenced
by a theory of
synaptic plasticity
that was introduced
by Leon Cooper.
Leon was a mentor, friend,
and colleague of mine.
Sadly, he died just
a few weeks ago.
He's a very famous physicist.
He won the Nobel Prize for a
theory of superconductivity.
But after that, he
went on to become
interested in synaptic
plasticity in the brain.
And he developed a theory that
is now called the BCM theory,
or Bienenstock, Cooper,
and Monroe theory.
It's a very simple idea.
And what he proposed is
that the synaptic strength
varies as a function of the
product of the input activity
and a concurrent level
of postsynaptic response,
some function of the concurrent
level of postsynaptic response.
And that function is
shown here in the middle,
and it should bear
a close resemblance
to what I just showed you.
In fact, Leon's theory
inspired the experiments
that revealed the properties
of bidirectional synaptic
plasticity in the brain.
A critical variable
in Leon's theory
was that crossover point
between synaptic depression
and potentiation doesn't
have a fixed value.
But rather, it's free to
slide back and forth depending
on the average
activity of the cell.
So in other words, the
rules of synaptic plasticity
are themselves plastic,
and that's a property
we now call metaplasticity.
So the question is, can
we slide that threshold
to the left to promote
synaptic potentiation
and allow recovery of function?
Next slide.
I guess I can do that.
OK, so the way we
addressed this was
to take advantage
of a neurotoxin that
is derived from Puffer fish.
If you've ever had fugu,
it's a delicacy in Japan.
It's a type of sushi that is
prepared from Puffer fish,
but it can be quite dangerous.
Because in the liver and ovaries
of pufferfish is a lethal toxin
called tetrodotoxin.
The reason it's lethal is
because it blocks all action
potentials.
It blocks all nerve impulses.
It can be used in experiments
to block electrical activity
in the brain.
So the experiment
that we conceived
was to inject tetrodotoxin
into the eyes of an animal that
had been rendered
amblyopic, shut off
the activity of
the visual system,
allow that threshold
to slide to the left,
and then restore vision and ask,
do the synapses regain strength?
And I wouldn't be telling
you this if it didn't work,
and the results were
really quite dramatic.
So this is experiments
that were done
in a cat model of amblyopia.
We've done the same
thing in mice as well.
And the graph on the left
here shows the acuity
of either the
amblyopic eye in blue
or the fellow eye,
the good eye in green.
And so this animal had been
deprived of vision in one eye,
in the amblyopic eye
for a period of time.
That eye was opened.
There's no spontaneous
recovery of vision,
just as the case in the humans.
But at the time indicated with
the ttx and the downward going
arrows, this drug was
injected into the eyes.
And then, of course, that
shut off all the activity
in the fellow eye.
But now, we wait and
watch what happens.
So when the ttx wears off,
the fellow eye response
returns to normal.
But amazingly, the amblyopic
eye response comes bounding back
and goes right up
to a value that's
comparable to the fellow eye.
So a complete
reversal of amblyopia.
And this experiment was done at
an age when traditional therapy
is no longer effective.
So that's what's shown
on the right side
here, that downward going
arrow, that this is occurring
at about 14 weeks
of age, which is
well beyond that sensitive
period when patch therapy works.
So we were very excited
by these results.
It seemed it was a
theoretically inspired study.
But it really was
based on an assumption.
And the assumption was is
that ttx actually shuts off
activity in the visual system.
However, more
recently, we discovered
that if we can shut off
activity in the retinas,
that the relay nucleus
that brings information
from retinas to cortex
is not quieted at all.
In fact, the activity in
that relay nucleus, which
is called the lateral
geniculate nucleus,
that activity starts to fire off
in bursts, spontaneous bursts,
high frequency
bursting activity.
And that bursting activity
occurs both in the neurons
that are postsynaptic
to the inactivated eye
and to the other eye as well.
So it's a qualitative change
in the type of activity
that's going to the cortex.
So the question then
became was, geez,
are these bursts really
required for this recovery?
So I have a wonderful graduate
student, Matty [INAUDIBLE],
who tested this by
knocking out the gene that
was responsible for
this bursting activity
in the lateral
geniculate nucleus.
You can eliminate bursting.
And when she did so, she
eliminated the recovery
that was caused by
the ttx injection.
So we have to
really rethink what
the mechanism for
this recovery is.
And I don't have a
good answer for you.
But what we do know is that
the bursting activity sets up
a pattern of very highly
synchronous activity
in the brain and also impairs
inhibition in the cortex.
So we think we've achieved
what we hoped to achieve,
not by sliding the
threshold necessarily,
but by evoking a stronger
postsynaptic response.
So the mechanism is
still, obviously,
under intensive
investigation in the lab.
But we can already
start to think
about how we might apply this
in a therapeutic setting.
So the next step for us was
to go to a primate model
of deprivation amblyopia.
So we've repeated
these experiments now
in monkeys that were rendered
amblyopic by early life
monocular deprivation.
And the results are,
again, very dramatic.
So we can get a very
substantial improvement
in the electrophysiological
measure of visual function
in monkeys.
But this also-- there's been--
nature has provided this
experiment in humans as well.
There are amblyopic humans
that, unfortunately,
have suffered damage to their
fellow eye in adulthood.
So they have relied on
their fellow eye for vision
for their whole life.
But then, because of
injury or disease,
the fellow eye has to be
removed or is otherwise
incapable of supporting vision.
And what's observed is
that, in many cases,
there is an unexpected recovery
of vision in the amblyopic eye.
So this means that plasticity
is there in our brains.
We just need to find a
way to tap into that.
Now, losing the fellow eye
is not a therapeutic strategy
to promote recovery
from amblyopia.
But what is is temporarily
inactivating that fellow eye
and, essentially, rebooting
the visual system.
So we think this is an
exciting future ahead.
And with that,
thank you very much.
[APPLAUSE]
Our next speaker
is Josh McDermott.
[APPLAUSE]
Greetings, everyone.
All right, so we
rely on our senses
to perceive the world around us.
But as Mark just told you,
our senses are fragile.
So in fact, over
39 million people
worldwide are blind and
over 70 million people
worldwide are deaf.
And we would love to be
able to restore the senses
in these individuals.
Now, certainly, the treatments
for these conditions
have improved over
time, but we're
far from being able to fully
restore sight and hearing.
All right, so the
first thing you all
need to all have
receptors in your ears
called hair cells that are
responsible for turning sound
into electrical signals,
and they're shown here.
So they sit on top of
a membrane in your ear
that vibrates in
response to sound.
And the hairs on the cells
get bent when they move around
and that causes changes in
their membrane potential.
And those signals are
sent to your brain
via the auditory nerve
that's shown there.
Now, deafness is typically
caused by hair cell dysfunction.
All right, the second
thing you need to know
is that, in the last
five years, there
have been huge
advances in our ability
to build computer models
of sensory systems.
And this has been
really largely driven
by progress in artificial
neural networks.
So when we do this in
our lab, we typically
start with a model
of the ear that's
rooted in everything that
we know from biology.
And then, we put on top of that
an artificial neural network
that gets optimized
to do different kinds
of auditory tasks, like
recognizing words or localizing
sounds.
And what this long stream
of fantastic students
has shown over and
over again is that when
you build models in this way,
they reproduce human behavior
to a pretty remarkable extent
across many different domains.
And this has really
transformed our ability
to understand and
study perception.
So our goal now is to use these
models to understand and improve
bionic ears and eyes.
So our lab study is
the sense of hearing,
so we're focused on bionic ears.
But we think many
of the insights
will be equally
relevant to vision.
So the cochlear implant
is a bionic ear.
It aims to cause hearing
via electrical stimulation.
And it is the primary treatment
for deafness at present.
There are over a million of
these devices in use worldwide.
So the way this works
is there's a microphone
outside the head and that--
the sound signal
that gets recorded
gets turned into
electrical signals,
and those are sent down that
wire that goes into the cochlea
and wraps around in that
spiral, and the wire
has got a bunch of
electrodes on it.
And so the key concept is
that if you have hair cell
dysfunction that
causes deafness,
but if the auditory nerve
that would normally convey
signals from the hair cells to
your brain, if that's intact,
then that can be
electrically stimulated
and that should cause
you to hear something.
So you bypass the hair cells to
create the sensation of sound.
So ideally, the
electrical stimulation
would replicate the nerve
patterns of activity
that would normally happen
from hearing electrically.
But that's really hard to do.
And so although these devices
are in some sense miraculous,
they're still not close to
restoring normal hearing.
And so one example of
that is shown here.
So I'll show you a number of
graphs that will look like this.
So this graph plots how well
people can recognize speech
in noise.
So the y-axis is word
recognition accuracy,
and the x-axis is the
signal to noise ratio.
So as you move to
the right, the speech
gets louder relative
to the noise
and it gets easier to recognize.
So the black curve shows how
well people with normal hearing
can recognize words.
And when the speech is loud
enough, they're at ceiling.
The red curve shows
what happens in humans
that have cochlear implants.
OK, so on the one hand, the fact
that they can score about 60%
correct is amazing because
these are individuals who would
otherwise be completely deaf.
But you can also
see that there's
a large gap between the red
curve and the black curve,
and that's what we'd
really like to close.
All right, so what limits
outcomes with these devices?
Well, there's at least
three potential factors
that we know of.
One is that the device
strategies are almost surely
suboptimal.
Another is that if you
undergo a period of deafness,
the neurons in your
auditory system die.
So there's neurodegeneration,
to varying extents,
in the auditory nerve or
beyond and what we call
the central auditory system.
And there's also the potential
that your brain is suboptimally
decoding the signals that
it gets from these devices.
All right, so what
limits our research
into understanding
these factors?
Well, one is that we have
very limited access to what's
actually happening in the brain
of a human implant listener.
So it's not like you can
just look at a person
and tell whether they
have neurodegeneration
or tell whether their
brain is suboptimally
decoding those signals.
But there are two
other major obstacles.
One is that the best
outcomes typically
occur after somebody has lived
with an implant for a very
extended period of
time, months to years.
And the other issue
is that the outcomes
are highly variable
across individuals,
as I'll show you in a moment.
So it's not like you can
get a new idea for a device
and give it to one person
and then really extrapolate
much about how it
will work in general.
So this makes it very
challenging to experiment
with alternative
device strategies.
And so most of the major
implant manufacturers
have all pretty much
been doing the same thing
for the last 10, 20 years.
All right, now, what
we can do nowadays
is build very good simulations
of the nerve responses that
would result from
a cochlear implant.
And Anisha Banerjee is
a fantastic grad student
in our lab who has built
one of these simulations.
The details here
don't really matter.
It's a computer program that
predicts the nerve responses
from electrical stimulation.
And so one example
is shown here.
So the top picture shows
an example, nerve response
from a cochlear implant.
The different rows represent
different nerve fibers
and the x-axis is time, OK?
What's shown on the bottom
is an analogous simulation
from a normal ear to the
same speech utterance.
So the point is that the
cochlear implant is intended
to produce nerve responses that
look like what you would get
from the normal ear, but
you can see that they
look really different.
And that's probably
part of the issue.
All right, so how can we
understand these outcomes
with the models that
we have of the brain?
So one other
important fact to note
is that human implant users
improve pretty substantially
over time.
So this is a really cool graph
that shows word recognition
scores as a function of
the number of months post
activation.
So how long someone has lived
with their device, and each line
is a different individual.
The different
colors don't matter.
So you can see that
all the lines go up
and they go up over
a period of time.
So people are learning
to do something
with their brains
that makes them
better at decoding
the information they
get from the device.
But the other thing
to note here is
that the asymptotic performance
is really highly variable.
So some people end up doing
great and others much less so.
And so what we don't know
is whether that variation is
due to differences in
how well people can learn
or whether, for instance,
it might be differences
in neural degeneration
or some other factor
that we don't know about.
So these models that we build
using machine learning we think
can help reveal the potential
of brain plasticity.
So we can take these models
of the auditory system
and build what we call a
static model of hearing
through a cochlear
implant where you take
a model of the normal
auditory system
and you just give it input
from a cochlear implant.
But we can also
build what we would
call a plastic model, where
you completely re-optimize
the neural network to deal with
input from a cochlear implant,
simulating what would happen
if your brain could completely
readapt itself to this
altered kind of input.
We can also potentially look
at the impact of degeneration
by looking at what
happens if you just
delete a bunch of nerve fibers
in the input to the model
or delete a bunch of neurons.
All right, so what happens?
So here's one
example of a result.
Again, this is the
same kind of graph
that shows word recognition
accuracy versus the signal
to noise ratio.
The black line shows a
model of normal hearing
that is very similar
in its performance
to humans with normal hearing.
And the green
curve shows a model
that gets input from
a cochlear implant,
but that is fully plastic.
So you can see that the
green curve is actually
pretty close to the
Black curve, right?
It's a little bit worse,
but really, pretty close.
Now, the red curve
shows the results
of humans that have
cochlear implants that we
tested in our lab on the
exact same experiment.
And you can see that
they're not doing nearly
as well as the model of the
green curve that is, right?
So that suggests that,
in principle, you
could do a lot better with
the input from the implant,
if you could fully
reoptimize your brain
to process those signals.
Can we simulate the possibility
of incomplete plasticity?
Well, one thing
we can do is just
reoptimize some of the
parts of that neural network
to deal with input from
the cochlear implant.
And when you do that,
you get the blue curve,
and that's now within
striking distance
of what you see from humans.
Now, by contrast, I'm
going to show you now
what happens if you delete 50%
of the neurons in these models.
And those are the dashed lines.
And the effect of that is
actually remarkably modest.
So we see huge effects of
the extent to which you
can reoptimize your
brain for this altered
input and really modest
effects of neurodegeneration.
So it's consistent
with the idea that
plasticity is a dominant
factor underlying outcomes.
And it raises the possibility
that increasing plasticity
in the human auditory
system, doing things
like what Mark was
just talking about,
could potentially
yield big benefits.
All right, so the next steps
that we're excited about
are to use these models
to derive new device
strategies that could
potentially further improve
outcomes.
And so we think the models can
be very useful here in revealing
the best case outcomes
for candidate devices,
potentially enabling
large scale screening.
So you can try out lots
of different ideas.
And then, the most
promising options
are things that you could
potentially explore in humans.
So I think this
is a nice example
of basic science
with impact where
the models that we
build of the brain
and refining those models
will lead to better therapies
that we hope will help people.
So thank you.
[APPLAUSE]
And with that, it
is a great pleasure
to introduce the world's expert
on the neural basis of language,
my colleague, Professor
Evelina Fedorenko.
Thank you.
All right, so let's
see if this works.
So language is--
I study language,
and language is this
incredible superpower we have.
Using language, we can
build deep relationships
with each other and restore
international political order.
We can send submarines to
the bottom of the ocean,
rockets into space,
and cure diseases.
We can bring about
political and social change
and record the history of
our planet and our species.
So how do our brains achieve
this remarkable feat?
So using studies of
patients with lesions,
non-invasive imaging like fMRI
and intracranial recordings,
and patients undergoing surgery
for brain tumors or epilepsy,
we've learned that there's
a whole bunch of areas
in your brain that
help you do this.
And I assure you, all of these
areas are working in your brains
as you're listening
to me right now.
These areas look something like
this in any individual brain.
And interestingly, they
support both comprehension
and production.
This makes us think that
these areas, basically,
store our knowledge of
language, so what words mean
and how to put words together.
And of course, we need this
knowledge to both take a thought
and encode it into
a word sequence,
or process somebody
else's word sequence
and infer the meaning
that they intended.
What I just said should make
it clear to some of you who
may have heard of things like
Broca's area and Wernicke's area
that these language
areas I mentioned
are actually different from
these areas classically
discussed in the
neurobiology of language.
It turns out that these Broca's
area and Wernicke's area
are lower level areas.
So Broca's area
is an area that's
important for planning
speech motor movements,
and Wernicke's area is important
for perceiving speech sounds.
So, for example, if I now switch
to talking in a language that's
unfamiliar to you,
your Wernicke's area
will be working
just as hard as it's
working right now
because it's still
getting its preferred input,
which is speech sounds.
But your purple areas,
your language areas,
the activity in
them will plummet
because you would no
longer be able to extract
any meaningful
information from them.
These areas, the
language areas, are also
distinct from areas that
support thinking and reasoning,
as my group and a few
others have shown.
Of course, our language
system interacts closely
with our reasoning abilities,
but nevertheless, the language
areas are not the areas that are
doing thinking in your brain.
And you can lose
your language system,
as in cases of severe
aphasia, and still
be able to think and
reason in complex ways.
A couple other things.
The system seems
quite robustly similar
across diverse languages.
There is over 7,000 languages
spoken and signed across
the world.
We've tested this in
about 50 languages.
Looks similar.
For those of us who have
multiple language systems living
in our brains, like bilingual
and multilingual speakers,
they all live in that system,
and developmentally, the system
is there and shows adult like
topography by about age three.
Now, our focus
today is on health.
So losing language is an
incredibly devastating
condition.
Not having full access to
language and development
leads to all sorts of cognitive
delays and difficulties
in establishing relationships
with your caretakers
and losing language
in adulthood.
In cases of
degeneration or stroke
aphasia is life changing in
all sorts of horrible ways.
So in trying to understand
language development,
we're trying to
ask questions like,
why is the language
system sometimes
doesn't develop properly as,
for example, in cases of autism?
Can we predict language
development trajectories?
And can we intervene
in high risk cases
to ensure that the language
system give it the best
chance to develop typically?
For acquired
disorders, we're asking
questions like, in what ways
can a language system break?
Can the language system
fix itself, to some degree,
in cases when it
starts malfunctioning?
And what can we do externally
to intervene on the brain
to ensure that the language
system recovers as much as
possible?
So we work on both developmental
questions and questions
of language loss in
adulthood, and I'll
give you a glimpse of
the latter line of work.
So this is a collaboration that
we've had going for a few years
now with Swathi Kiran, who is at
BU leading the Center for Brain
Recovery, and the lead junior
scientist is [INAUDIBLE], who
is now a postdoc at Harvard.
So understanding
recovery from aphasia
has been going on for many
years, centuries really, now.
So a lot of people
have focused on trying
to understand the trajectory,
the timeline of development.
And very consistently,
a lot of groups
have shown that most
recovery happens
within the first few months.
But in many cases,
recovery does continue even
after the first year,
although at a slower pace.
Many other groups have
studied questions like,
what makes some people
recover better and worse?
And one big finding is
that the younger you are
and the better your
brain health is
at the time the stroke
happens, the better off you're
going to be.
But also, it matters
where the damage happens
and how extensive it is.
In our work, we've been
trying to understand
the recovery mechanisms that
underlie recovery from aphasia.
So to the extent that
a patient recovers,
what's happening in the
brains of those patients?
What are the changes that lead
to the restoration of some
of the linguistic function?
And there have been two big
ideas for how this might happen.
One idea is what's
referred to as homeostasis.
So it's basically the idea is
that you have some remaining
parts of your language
system and they're
trying to recover function
by somehow reorganizing
the function among
the remaining bits.
So you're trying to go back
as much as possible to the way
things were before
the stroke event.
A very different idea
is reorganization.
So this is a more drastic
takeover of function
by areas that weren't
doing language before,
but are now going to
be doing language.
And a good candidate for
this kind of takeover
is the system that many people
call the multiple demand
system or the executive
function system.
It's a very cool
system you all have.
It supports so-called
executive abilities
like working memory,
attention, inhibitory control,
and it's active across
many, many different kinds
of behaviors.
One way to think
of this system is
it's a system that evolution
gave us to be able to adapt,
right?
We have all sorts of specialized
mechanisms, like for hearing
or for vision or for
language, but we also
have resources to be
able to deal with change.
And this system, the idea was
that this system is flexible
enough so that if some
specialized mechanisms get
damaged, the system
can take over.
Why does it matter?
Well, if we know that it's
a language system that
helps you recover, the remaining
parts of the language system,
or this very
different system, we
can intervene on these systems.
We can either stimulate them
or do behavioral therapy that
targets this particular system.
So it really matters for
trying to boost the recovery
mechanisms that are in place.
So we tried to
evaluate these two
ideas in a group of patients
with chronic aphasia
and a bunch of controls.
So what we did was record
their brain activity
as they listened to
language to understand
which parts of their brain
are processing language.
And we also measured
their linguistic ability
in all sorts of ways using
standard assessments.
And so then, we're, basically
asking, OK, to the extent
that these patients vary in
how good they are in language.
What's different about how
their brains process language
so that some are doing
better than others?
And it turns out that
it's really homeostasis
that's doing this work.
So it's the same system that
was doing language in your brain
before stroke that--
whose ability, whose
activity is best predictive
of how well you're
doing after a stroke.
So now, individuals who show
most typical like patterns
in that system are
the ones that do best.
So what we're
doing now is trying
to evaluate a
range of therapies,
some based on behavior.
So for a while, people
got really excited
about this idea of
takeover of function.
And a lot of
therapies got shifted
to trying to increase general
ability, executive function
ability.
Let's make people
better at working memory
or at keeping
attention in the hope
that it helps restore
these loss functions.
It turns out that, no, you need
to really focus on language
and keep training the
language system up.
We're also doing work with
electrical stimulation
of the language
system remaining parts
to see if it boosts recovery.
We're also doing this very
hard thing of longitudinal
scanning trying to--
very little is
understood, actually,
about the acute stage of what
happens in the first few months
because most people don't want
to do research when they just
suffered a stroke.
But it's really,
really important
to understand how things change
during those first few months
after injury, so
we're trying that.
And much like what
Josh was talking about,
we're also building
computational models
based on language models
like ChatGPT, including
creating kind of aphasic
variants of these models
to get a more mechanistic
level insight into how
things can break, and
hopefully, how we can fix them.
So thank you very much.
[APPLAUSE]
Nice job.
Thank you.
Empty chair.
It is a big chair.
So we have time
for some questions.
Are there any that
get it started off?
Tony.
Use a mic back there.
In animal studies,
does language have
to exist for thinking to exist?
This is something you presented.
And I've read some studies that
say that animals actually think,
even if they cannot speak.
Is there a biological
reality to that?
I mean, animals certainly do
all sorts of complex cognition.
Whether their thinking
mechanisms are exactly the same
as humans, well, probably not.
But whether they're just
kind of-- whether humans
have simply bigger and better
versions of those mechanisms
is potentially likely.
And a lot of people
are working on this.
But Yeah, you can definitely
think without language,
including in animals.
Yeah.
Any other questions?
You mentioned something
about language development
with the very young.
And you said around three
years of age things change.
But previously, I've
heard a lot about much
later than that learning
a foreign language up
to age 10 or 11,
somewhere like that.
How do these two relate?
Sorry, I didn't quite
hear everything you said.
So how will we
learn later in life?
You mentioned in the
talk that around, I
believe it was the
age of three years,
that the way the language
acquisition learning was taking
place dropped off at that point.
No.
I was just saying that the
system I was talking about
is there by age three.
Of course, learning can
happen much later than three.
Yeah, it still works.
Because I'd heard like
10, 11 years old that--
Into the teenage years,
you can learn language
in a native like way.
Yeah.
I have a question.
In terms of the scope
of your research,
do you also look into
cognitive, I guess,
impact caused by other diseases
such as cancer entering
the brain, for example?
Thank you.
Good question.
I guess some of the
tools that we developed,
we do work with brain
cancer neurosurgeons.
This is not about cognition,
but our microscopy methods
if you can find brain tumor
cells more accurately.
We had a paper earlier
this year where
we showed that you could get
much more accurate counts
of aggressive brain tumor cells
with our new imaging methods
than with older ones.
So not exactly
conditioned, but the tools
can be definitely applied.
Yeah, I was more referring
to metastasis of other cancer
into the brain.
Yeah, I don't think any
of us work on that per se.
I mean, I'm very
generally interested
in what happens as people age.
We do a lot of work on the
consequences of hearing loss.
And that is also
accompanied with other kinds
of cognitive changes.
So that's one thing that's
a little bit related
to what you said.
And of course, there's things
like Alzheimer's and Parkinson's
that play into that as well.
Thank you.
Thank you.
So Mark, going back to your
discussion about amblyopia,
what's the best you
can do right now
for someone who has amblyopia?
Well, as I mentioned, patch
therapy is the standard of care.
The severity of amblyopia
depends on the cause.
So there are more benign causes.
More benign.
Less severe causes that can
be treated as late as age 8, 7
or 8.
So again, it's patch therapy.
But in recent years,
there's been introduction
of what's called dichoptic
therapy, which is using
binocular visual stimulation.
And they're getting
very good results.
But again, the
effects so far are
restricted to this early
developmental period.
Go ahead.
What causes a loss
of language ability
with age, hunting
for words, and so on?
And are there other ways
to recover from that?
That's a great question.
So I'm also very interested
in aging, like Josh is.
And we've just finished
a study looking
at how the language system in
the brain changes with aging.
And we can't find any
ways in which it changes.
So neurally, it
looks just as good
as a system in young brains.
It's as strongly
active, as strongly
lateralized to the
left hemisphere,
responds similarly
selective to language.
And so there is no
obvious changes there.
So I suspect a lot of the
word finding difficulties
that people report
with aging have
to do with more
general memory systems.
And of course, those
are very susceptible
to disorders like
Alzheimer's, and in general,
decay with healthy aging.
And those changes you can see
very well, even with methods
like non-invasive imaging.
I don't know if there
is more to that.
If you don't mind, I
have two questions.
One is, with the
cochlear implants,
do you find that there's more
neuroplasticity in people
who lost their hearing
at a later age as opposed
to people who never
were able to hear?
Do they adapt to a
cochlear implant better?
Yeah, it's a little
bit complicated.
So I mean, in general,
the best outcomes
happen with people who are
implanted at a very early age.
And one plausible
explanation for that
is that, in general,
the brain tends
to be more plastic at
younger ages, right?
But there are also--
there is also evidence
that there are benefits to
having some hearing experience.
And so that probably
factors in as well.
Yeah, that's a short answer.
My second question
is about language.
You talked about these
centers of the brain
are for all multi-language,
bilingual, multilingual.
When people lose language
and become aphasic,
do they always lose
all of their languages,
or can they lose one
of their languages
if they're multilingual?
And how does that play
into your studies?
That's a great question.
So in most cases, yes, you
lose multiple languages,
but there are interesting
effects where sometimes people
report that the language
that's not a native language,
you tend to preserve
a little bit better.
And we think we
just recently got
a glimpse of a possible
mechanism for that.
I always found this
really confusing.
Not making sense.
But one thing that we see
is that native languages
tend to be processed
a little more focally.
So they don't always
engage this whole system.
There is a core part
in the posterior
in the left temporal
cortex that seems
to really, really be important.
But for all of your later--
so for me, English is
a non-native language,
and English engages this whole
network very consistently.
And so if you have vocal damage
in the left temporal cortex,
it is more likely to wipe
out your native language,
and the remaining parts may
work better in the later learned
languages because it's just
more consistently recruited
in this way.
Fantastic.
Hi.
Thanks for sharing
your research.
They are fascinating.
I'd like to know if there
is any studies showing
the impacts of the use--
the early and frequent use
of technology, social media,
and generative AI by
children and teenagers
in core cognitive functions?
Because attention,
learning, memory,
because we've been hearing
a lot about the impacts
of social media and screens in
the mental health, increased
levels of anxiety and
eating disorders in girls,
but what about the
other functions?
Using technology can
harm in the long-term
our attention, language
acquisition, memory, learning.
Do we have any robust
evidence on that?
Thank you.
Yeah, to my knowledge,
there isn't, actually,
robust evidence one
way or the other.
I mean, there is a
literature that--
I mean, Ev may remember
this better than I do.
But there there were
a lot of studies
for a while showing that you
could see some advantages
from people that played
a lot of video games,
like on benefits on visual
attention, for instance.
And I'm not up to date on what
the current status of that is.
Did that stuff hold up?
I can't remember.
Yeah, and that research
program on video games
started because people thought
that video games are destroying
our children's brains.
And they started looking
at all sorts of functions
like attention, memory.
And it turns out everything
they looked at is actually
a little bit better in kids.
So that's where that went.
And yeah, I have not
heard of any research
other than the mental
health effects of exposure
to social media that it
affects attention or memory
in any negative way or language.
I still keep my kids off it.
[LAUGHTER]
Hi.
Thank you so much.
The brain is so amazing
because there's still
so much we don't know.
I've heard about how
people are trying to--
I mean, I guess all of
you guys in your own way--
but are trying to
map each circuit.
And to me, it sounds like
the kind of impossible dream
that we once thought mapping
the whole genome would be.
So do you think that's
something that we'll
see in any of our lifetimes?
I mean, just understanding how
all the different pieces fit
together.
Or is that way out there?
Thank you.
I think everybody should
take a shot at that one.
Go ahead.
Yeah, well, I think
one of the great things
about being here at MIT is
the technology collaborations.
So great engineering school,
great science school,
great cross-talk.
And the influx of everything
from nanotechnology
for neural interfacing to
new noninvasive strategies
for delivering
energy and recording
information from
the brain, I think
that's presenting a huge amount
of almost engineered serendipity
between different fields
that is opening up a of--
a bit of a Moore's law in
terms of improving our ability
to look at and control
brain circuits.
So being an engineer, maybe
it's how I think about things.
But I think being at
MIT, you're surrounded
by people who are always asking
for what the next big problem is
that they could work on.
And they may have
revolutionized one field,
and they're looking for
the next big challenge.
And for many people,
that's the brain.
Mark, any thoughts?
No thoughts.
[LAUGHTER]
No.
I mean, I think there's
value in simplification.
I'm a lumper, not a splitter.
And looking for
general principles
seems to me a lost art that
needs to be reinvigorated
and I hope to see that continue.
Yes, I think, I mean, one
thing I would say is--
I mean, I don't
know how realistic
it is to think that
we will actually
understand the human brain
fully at the circuit level.
And I'm not sure that
we will necessarily
need to for a lot of the
things that we want to do.
So for instance, if we end up
really understanding learning,
and if learning is the
key to a lot of the things
that we want to do, then
that may be sufficient.
And having really
detailed explanations
of what happens at the
circuit level, I mean,
may or may not be critical for
what we actually want to do.
So I think there will be
examples where we have really
circuit level
explanations for things,
but I'm not sure
that will generally
be the case for the human
brain just because it's
so big and so complicated.
Hello.
I'm from bioengineering
department,
and we do run the clinical trial
with the long COVID participants
and chronic Lyme participants.
And I've been always very
fascinated about the patients
complain about their
brain fog symptoms
as number one major
symptom complaints.
And that's why I'm here.
And I got very fascinated
with the language aphasia
presentation.
Thank you so much.
And I would love to-- because I
don't have much basic knowledge
in the brain science.
But I want to ask this
question for my patients
actually here that this is
all new symptoms in terms
of being lost in the
middle of the speech
when they're trying
to engage interaction
with the summoning conversation.
And the way they describe
brain fog and cognitive deficit
is oftentimes related to their
language of difficulties.
And I would love to
know your insight
in terms of the brain function
associated with the infection
triggers or source.
Not so much a stroke
issue, but perhaps,
what are the insights with
the infection triggers?
And can this kind of a brain
fog complaints description
of language difficulty
be something
to look at in the case of
Wernicke's or Broca's area,
or what are the insights
that we can look further
to expand our research?
I would love to
hear your feedback.
Thank you.
Yeah, I don't know if I have
too much to say about this.
I mean, from my understanding,
the brain fog like symptoms
in long COVID patients are not
quite specific to language.
The language may
be affected, but it
seems like it's a pretty
wide spread systemic effects
on multiple systems.
So it's possible that insights
from specific disorders,
like language or audition,
related things may be helpful,
but it seems like there is some
more global things to understand
there.
And maybe the
deficits are actually
in this very vulnerable system
to aging, which we mentioned
the system that affects
memory and is very
affected in Alzheimer's,
so the system that
has big hippocampal components.
But yeah, sorry, I don't
think I have anything
more specific than that.
Thank you so much.
So a question for
you before you leave.
So what do you measure to
document the patient's brain
fog?
What kind of tests
do you have them do?
So currently, in our study,
we do three different measures
that we add.
So typically, a day before
person comes to MIT visit
for study visit, we
assign the cognitive test
called brain check.
It's equivalent
version of a battery
that can be done remotely.
So it takes strobe and trail
making and memory function,
immediate recognition
type of testing.
And once they come to
MIT for study visit,
we do two different
neurocognitive testing.
One is called right eye, which
we add a visual sensitivity
testing.
So the person will--
we're going to look
at the eye coordination
and pursuit and saccades
and the brain reaction time
and physical reaction time
by using different images
and how quickly they react.
And the other one we also use
equivalent version of EEG called
[? Wave. ?] It's actually
a commercial testing,
but we look at the
brain activity,
look at the alpha peak wave
during the P300 testing,
we also do the flanker
testing as well
just to see the difference
in terms of reaction time.
Great, thanks.
We can learn from
the audience as well.
[LAUGHTER]
So Hi, I'm [INAUDIBLE], and I'm
a first year master's student
at Harvard.
So my question may be
a little bit naive.
So in my undergrad
study, I did projects
related to understanding
molecular mechanism of autism.
And I also have some experience
dealing with the neuroimage.
So my question
would be, well, it
is possible to have a model,
like machine learning model,
to actually predict both
the genomics and the--
I mean, the outcome of
diseases or the hearing
aid devices for both
genomics level and also
the circuit level.
So if that is possible,
so what is the challenge
that you may see in this?
So we have less
than two minutes.
You want to try that.
Josh that's you.
So, I mean, we're
definitely not there, right?
I mean, I think the means by
which genes lead to circuits,
I mean, that's a pretty big and
hard and complicated problem.
I don't know.
I mean, machine
learning could certainly
be useful probably
in understanding
that if you had enough data.
And my guess is there probably
will be lots of headway on that.
I mean, there's
definitely lots of people
who are very excited
about building
models of the entire brain
at the circuit level.
I mean, this is I think
what Ed maybe aspires to.
And where that's going to lead,
I think, we still don't know.
But it's a really
exciting direction
And, it could
become within reach.
OK, thank you.
So one last question.
Thanks.
Thanks for your talk.
I'm curious.
I have actually two
questions, if I might.
With regard to language,
are the same centers
for language processing
identical regardless
of the input via hearing,
sight, signing, et cetera?
Yes, Sorry, I didn't
explicitly say this.
They're AI model,
so it doesn't matter
if you're reading or
listening or processing
sign language if your sign.
OK, my second
question is, how broad
is the definition of language?
Does it encompass
situational awareness
when you're in traffic,
traffic lights, where
cars are, danger on the street?
Is that part of the
language function?
Well, so I certainly
wouldn't put--
I mean, language is a very
difficult construct in the sense
that it interacts with all
sorts of other systems.
But it so happens that we have
a set of these areas that are
quite specialized for language.
And then, a lot
of ongoing work is
trying to understand how
these language circuits work
with low level
audition circuits,
how they work with
memory circuits,
how they work with
attention circuits.
Because, of course, we use
language in its real life
complexity where
there's a lot of systems
that come together and
have to talk to each other.
But it does seem that there
is some sense in which there
is a system that's specialized
for just this encoding decoding
aspect between thoughts
and word sequences.
I guess a follow on to that is,
if I lost my language centers,
would I lose my ability
to understand situational?
No, you would not.
In fact, there is a large body
of work with severe aphasics
and their cognition--
so it relates
to the first question as well--
their cognition is totally fine.
They can navigate the world,
they can socially reason,
they can keep their attention,
they can play chess,
they just can't express
their thoughts and language
or understand others.
Thank you.
So to be in sync with the
remaining part of the program,
let's thank our speakers.
[APPLAUSE]
Thank you.
And I guess they're going
to reconvene over in the--
across the way
there very shortly.
OK, great.

---

### MIT HEALS Launch: Biosphere breakout session
URL: https://www.youtube.com/watch?v=kIZuU-vSSLA

Idioma: en

OK, folks, we are
going to get started
with this breakout session,
as we continue this fantastic
day of science and engineering
here at MIT, for the HEALS
Initiative.
I am delighted to get
to introduce and share
the biosphere breakout
session today.
My name is Matt Shoulders.
I'm a professor of
chemistry, and really
happy to see all of you here
for this important topic.
You made your way
all the way over.
We really, really appreciate it.
And it's a beautiful
new space here,
so I hope you're enjoying it.
Today, I'll start with just a
few brief comments of my own
and then hand it
over to a fantastic
cast of four speakers
whose names you
can see on the prior slide.
But we'll introduce
them along the way,
to tell you about the topics.
To tell you about
really amazing research
and the future of this field.
Understanding how living
systems power life on Earth.
And I'm really excited
to hear their talks.
When the talks are done--
and this is important--
when the talks are
done, we're planning
to have about 15 minutes,
10 to 15 minutes for a panel
discussion.
That's why we have chairs here.
So think about the
questions that you
want to ask these
fantastic faculty,
because if you don't, I'll
have to ask all the questions
and your questions will
be far more interesting.
So we'll have time
for that afterwards.
OK, so why do we have
a biosphere breakout
at the MIT Health and Life
Sciences Collaborative?
I hope and think that the
answer should actually
be obvious, right?
But the interface between
life sciences and health
is critically important.
But the health of people, the
health of the environment,
the health of the globe, the
health of the planet, life
sciences are central to
solving the major problems
that health in all those
spheres depends on.
Life sciences are absolutely
central to solving
the key challenges that we face.
So these range, of
course, from rapid climate
change that we have
to adapt to, and we
have to figure out how
to also halt or slow.
Life sciences are
these challenges
that face the life sciences,
including enormous increases
in demand for food, for
resilient agriculture,
and for healthy food.
We also face the damage done
to deteriorating environments,
and the effects on well-being
caused by all of that.
And really, the
life sciences are
key to solving these challenges.
Whether you talk about
synthetic biology,
plant science, microbiology,
environmental science, ecology,
the viable solutions, the
globally deployable solutions
to these types of challenges,
to the health of our world
and the health of
humans are going
to come in a meaningful
way from the life sciences.
And we also, along the way, as
we come up with these solutions,
we have to think about
ethical and equitable means
of deployment right
from the start
as we think about
inventing these solutions.
So MIT is a leader
in this space.
You've been hearing about lots
of amazing biomedical research
and tremendous
discoveries being made.
But at this interface
between the life sciences
and the biosphere, the life
sciences and the climate,
life sciences in
agriculture, there's
also huge opportunities
for impact on health.
In fact, even sort of small
incremental improvements
in these spaces
can have the sort
of massive effect on human
health and planetary health that
would elude most types
of drugs that you
could dream of inventing.
So we think this is super
important, super relevant
to the HEALS Initiative.
And today, you're going
to hear from these four
faculty about both
challenges and solutions
to problems at this
interface between life
sciences and the environment
and sustainability.
And you're going to hear them.
You're going to hear them
talk about challenges
that are facing our globe.
And all the way down to
individual communities.
And the talks today
are going to start
at the molecular
scale, as we first
start by talking
about photosynthesis
and the basics of
photosynthesis and how
it powers the entire planet.
How we can maybe
understand it better
and improve on it to
find great solutions.
We'll go from the molecular
scale to the scale of crops,
and actually providing
food for the world.
We'll talk also today
about how the changing
climate and the interface
between the climate
and green spaces plays
huge impacts, has
huge impacts on human health,
and human health equity as well.
And then finally,
at the end, we'll
go back to the global
scale and we'll
think about how tiny
microorganisms play
a huge role on Earth.
And it's a very fascinating
area there as well.
You're in for some
great treat here
with these four
wonderful faculty,
and I should get off the stage
so you can hear from them.
Our first speaker is Professor
Gabriela Schlau-Cohen.
She's the Robert T.
Haslam and Bradley
Dewey, Professor of Chemistry.
[APPLAUSE]
All right.
Well, thank you, Matt.
And thanks to all of you who
made this journey over to this--
enjoyed this view
and these talks.
So Matt introduced
us to photosynthesis,
and photosynthesis occurs in
a wide variety of organisms
and a wide variety
of environments
to collectively generate
energy on an enormous scale.
Now, behind this feat is a
series of remarkable features.
For example, if we take this
corn and this crop field
here, one question that you
may have wondered yourself,
and is actually a central
question in research
in my group, is why is it
that plants like this one
don't get sunburned?
So plants don't have to put on
SPF 50 when they go outside.
They stay in the sun all day.
But yet, except in a
few limiting cases,
they don't get sunburn.
Now, the reason for this is
they have their own built
in molecular machinery.
It serves the
function of sunscreen.
In my group, we're interested
in understanding the mechanisms
behind this molecular machinery,
both for fundamental knowledge
and because of this plot here.
Shown are yields
for four major cash
crops as a function of time.
And what we can see
over the next decades
with predicted demand
in dotted lines
and predicted supply
in dashed ones,
is that there is a
growing mismatch.
And not having enough food is
a huge issue for human health.
It turns out we need to eat.
And so that means that
one of the biggest health
challenges facing us is figuring
out how to bridge this gap,
so the supply meets the demand.
One way to do that is
to increase the supply.
Fundamentally, the food supply
comes from photosynthesis.
But photosynthesis operates
at a measly 1% to 3%
efficiency for biomass
or food generation,
with the rest of
the energy getting
lost to things like metabolic
processes, inefficiencies
or environmental responses.
For example, under certain
conditions, up to 70%
of the absorbed light gets
converted to heat and lost.
This is that sunscreen
process that I was telling you
about before.
And so in many
conditions, it's actually
that conversion to heat
that limits crop yields.
So if we want to
understand this energy
allocation and these limits, we
can go back to our crop field.
And we can zoom in.
And what we find are some of the
most sophisticated nanomachines
on Earth.
These are the photosynthetic
super complexes
that collectively
capture light and then
convert that light
into chemical fuels.
These chemical fuels are then
used for biomass generation.
Now, even in the first
steps of this process,
this energy allocation
is already occurring.
Specifically, this is where we
see the allocation of energy
into these environmental
responses converted to heat
and lost.
But the size of these
slices of pie, they
actually change
with many factors.
And there's been decades of work
in plant biology, spectroscopy,
plant science, that have
all led us to understand
that, for example, with
something like weather,
this amount can
decrease dramatically.
But if what we want
to do is understand
what it is that's
controlling this,
so we can begin to think
about manipulation,
we really want to understand the
specific proteins responsible.
So, for example, in
green algae and moss,
these are good model systems for
understanding energy allocation
in green plants in general.
The specific protein responsible
is one that looks like this.
Here's the structural model.
It's called light harvesting
complex stress related,
or LHCSR.
This is a protein that
actually takes that sunlight
and converts it into heat.
And the level of that conversion
gets controlled by that protein
by structural or
conformational changes.
So if what we need to
understand is energy allocation
in photosynthesis,
what we want to know
is what is the conformational
switch in proteins
like this one that activate the
dissipation of energy as heat.
And so if we want to understand
this dissipation of energy
as heat, a really important
reporter is fluorescence.
So fluorescence is
something that's
used to study photosynthesis
on all sorts of length scales.
This is a wide
interdisciplinary community.
So what happens is proteins
like this one absorb light.
The amount of light
that comes in then
can come back out
as fluorescence.
So the more light that
comes out, the less energy
is getting converted
into heat and vice versa.
Now based on this
inverse relationship,
this is how fluorescence is
used to study photosynthesis.
And like I said, on many
different length scales,
from the kind of individual
proteins experiments
that we do in my group, all
the way up to the global scale
where fluorescence is measured
in satellites from space.
And that really allows us to
connect these different scales
using these different approaches
to understand energy allocation.
But say we want to understand
these individual proteins
and this conformational
switch, that question.
I posed before.
Well, one challenge in a
typical biochemical experiment
is we may take proteins like
this one and measure around 10
to the fourth individual
proteins at a time.
Well, the challenge with
that kind of experiment
is that each one
of those proteins
may not be in the
same conformation
or may actually be undergoing
a conformational change.
And so experiments
had traditionally
averaged over this, making
it impossible to identify
these conformations and the
different energy allocation.
Now, it's because
of that limitation
that over the past
decades, single molecule,
or in this case, single
protein spectroscopy,
has emerged as a powerful tool.
This is one of the major tools
that we use in my research
group, where we can zoom in and
measure the fluorescence of one
protein, and then
another and then another,
to collectively understand all
of their fluorescent properties.
So here's an example of
some data from my lab
where we look at the
fluorescence distribution.
And so what we see are
two peaks, one of which,
whereas more of light harvesting
confirmation, and the other
is our dissipative confirmation,
where the energy is
being converted to heat.
So this one has less
fluorescence coming out.
And so this data was taken under
conditions that mimic low light.
But we can change how
we're doing our experiment,
understand how energy allocation
in photosynthesis changes.
And so if we do the same
thing, mimicking high light,
what we see is a change
in our distribution
and an increase in
the level of energy
going to heat by around 50%.
So more energy gets lost.
So collectively, this allowed
us to identify these two peaks
and see how this
switch gets turned.
So we now have
this understanding
of these individual protein
conformations, how they change,
and how they're
controlling energy
allocation in photosynthesis.
And so with this fundamental
understanding of the protein
level properties,
we can now begin
to think about that gives us a
blueprint for what's happening
in organisms like this one.
So we can think about
reallocating this,
redistributing the energy
potentially to increase biomass,
and maybe even one day
increase our crop yields.
And with that, I'm
going to turn it over
to our next speaker
who will talk
about continuing to think about
how to change these organisms.
Dave Des Marais from Civil
and Environmental Engineering.
[APPLAUSE]
Thanks, Gabriella.
So I'm a plant biologist, and
so Gabriella has just given you
a fantastic introduction to one
of the most amazing and central
things that plants do for
the entire biosphere, which
is to take solar energy that's
abundant but transient source
of energy, and to capture that
energy through the process
of photosynthesis,
interacting with CO2,
with water to make sugars.
And as probably even most
elementary school students know,
the other thing, of course,
that comes from this
is also very important
for the biosphere, which
is oxygen, biomedical
oxygen, which we can breathe
and which allows
respiration to be
such an efficient process for
harvesting and using energy.
So what I like to teach
my introductory students
in biology, that we're taking
this transient source of energy
and we're trapping that as the
bonds in the sugar molecules,
right?
So now we have this physical
thing that a plant or an animal
can use to transport,
to move around,
to use as energy when
they would like to.
Not necessarily when
the sun is shining.
But as you probably know, sugars
play a second equally important
role, which is those
initially fixed
carbon molecules form the
backbone of everything else.
And I literally mean
everything else.
Everything that the
plant makes, everything
that an animal might eat
from that plant, everything
that a fungus might use as
it decomposes that plant, all
of those organic biomolecules
begin as carbon photosynthate.
But there's also very
important other molecules
that are part of that
fixation process.
So you know we need
to eat protein,
because protein contains
things like nitrogen and iron.
Our lipids, of course,
are often very complex,
large macromolecules.
And not only the composition
of those molecules
depends upon other nutrients,
but also the processes
to generate those molecules.
The very action of
doing photosynthesis
requires all sorts of
interesting elements
that are largely
gained from the soil.
So what is the ultimate source
of those important components?
Well, there's many, many
elements that organisms need.
Plants and us.
Nitrogen is often
was thought about,
but there's also countless
mineral molecules.
Where do those come from?
Well, ultimately, those
are weathered off of rocks.
There are abundant
sources of many
of these minerals
in the lithosphere
and as rocks weather,
those enter the soils,
those can be captured by
the biosphere, by capture,
by plants and microbes
to build their biomass.
Nitrogen is also
incredibly abundant
but is abundant
in our atmosphere
as nitrogen gas, N2 gas.
And you may know that nitrogen
gas is incredibly hard
to break apart those
two nitrogen molecules.
But breaking apart those
two nitrogen molecules
is critical to
turn that nitrogen
into usable forms of nutrients
like ammonia nitrate, which
plants can uptake,
turn into proteins,
thus providing us
with a food source.
Now in natural
communities, which
I have grown up studying as
an evolutionary biologist
and ecologist, in
natural communities where
plants and animals and microbes
are found and where they thrive
is really a function of
what's in those environments.
If there's rocks nearby
that can weather out
these important minerals, well,
then the kinds of plants that
can make use of those minerals
or that do well when they're
in excess or in deficit, and
similarly, for nitrogen, where
there's enough
nitrogen in the soil,
there's plants that will
thrive in those places.
And if they don't
thrive in those places,
in those natural
communities, well,
then they can move someplace
else, or they might go extinct.
So natural communities
have evolved
over hundreds of thousands
and millions of years
to have compositions
of plants and animals
and fungi that
form the biosphere.
Some plants die, they rot.
They create nutrients
for microbes
which produce and sort of
continue the cycle of life.
But in agricultural
systems, we don't
have that same nice
balance, unfortunately.
It's not simply a matter of
having the plants grow someplace
else, because we want to
have those plants grow
at very high yield,
in the places where
we need them to be growing.
And so that creates a real
problem when any of these things
become out of balance.
And this has led to what
has been called the nitrogen
problem.
So we know that nitrogen is a
key limiting nutrient for most
of our agricultural production.
Most of our calories
in the world
come from cereals, either that
we eat directly, corn and wheat,
or that we feed to our animals,
which we then eat the animals.
And to maintain these very
high yields from cereals,
we need to have a lot
of nitrogen fertilizer.
As I told you, those
things can become limiting.
If you imagine a common
agricultural year,
you're pulling up all that
biomass, you're shipping it off
and the nitrogen
is going with it.
So we're constantly adding more
and more nitrogen fertilizer
back to the soil.
The predominant source of the
nitrogen in the developed world
is through industrial
fixation of nitrogen
through the Haber-Bosch process.
Here's a few figures to
make the point to you
that this is an incredibly
resource intensive process
to capture that
atmospheric nitrogen
and turn it into ammonia that
we can feed to our plants.
So just huge proportions of
the total global carbon budget
come from the process of
making industrial nitrogen.
But even more
depressing is that most
of the nitrogen we
put onto our soil
doesn't end up in our plants and
doesn't end up in us as food.
It's running off, or
it's being lost back
to the atmosphere as
climate changing gases.
So this is the nitrogen problem.
Now, you're probably
aware that there
are plants out there that have,
in a sense, figured this out.
So in places where nitrogen
in the soil is very limiting,
microbes can fix that
nitrogen. Atmospheric nitrogen.
And many plants
have figured out how
to coexist with those microbes
and make use of that nitrogen.
So there's this
really nice exchange
of carbon from the plant and
nitrogen from the microbes.
And of course, legumes like
soybeans and many other beans
are the most famous
and widespread example
of this in our food system.
But cereals don't
have this trick.
They don't have-- and they
haven't figured out how to have
that symbiosis, how to get these
relationships with microbes
to provide nitrogen. So this has
really been sort of a moonshot
genetic engineering project
since at least the 1970s.
And a number of proposals
have been brought forth
about how we might try to
introduce this nitrogen capture
process into our cereal crops.
So at base, what
this issue is, is
how do we get the nitrogen
out of the atmosphere,
into a microbe, and into the
plants we want those microbes
to be associated with?
And in turn, what
is the plant then
giving back to those
microbes in exchange
for this resource, this
valuable, energetically
intensive resource capture?
So this is the central question.
And this is a
question, a challenge
that we've been tackling as
part of the MIT Climate Grand
Challenge Program, led by Chris
Voigt, a number of faculty
from across the Institute
where we're trying to address
this nutrient use in agriculture
problem in a very holistic way.
And what I want to
share with you today
is a small piece
of that project,
which is trying to optimize this
particular little interaction.
How do we get those microbes
to capture the nitrogen,
give it to the plants that
we want them to give it to,
under the conditions that we
need them to do so, and then
have the microbes in a sense
cultivated by the plants.
So Professor Chris Voight
here in MIT's bioengineering
department, is really focusing
on this first question.
How do we genetically
engineer bacteria
to fix the nitrogen when and
where we need them to do it?
My group is really interested
in the plant side of this.
My group is comprised of
plant physiologists, plant
geneticists, and we
think very deeply
about how plant performance
is driven and affected
by environmental conditions
and environmental change.
And one thing we've
been focusing on
in the last few years
is trying to understand
under what conditions
do plants provide
these cheap carbohydrates,
organic acids and sugars
to the microbes in the soil.
Is there genetic
variation in this process?
How does the
environment drive this?
And critically,
could we understand
at a mechanistic level how
to manipulate that process?
These are some really key
questions we're interested in.
What molecules are
there being produced?
Does the environment
drive this process?
And critically,
can we manipulate
the production of what
we call these accidents?
Now to do this requires
a tremendous amount
of technological know how.
So we're looking at roots,
so they're below ground.
We can't just walk
up to the plant
and measure them without
disrupting itself.
So we're collaborating
with a Professor Sixian You
here in MIT's research lab
for electronics and EECS,
who's developed non-label,
non-invasive imaging
technologies for living tissue,
where we can look at the roots
in situ, as they're growing,
and look at their metabolism,
look at their chemistry,
look, see how they're
growing at the cellular level.
And my PhD student,
Chloe Heitmeyer,
has developed a
very simple system
to grow these plants under
controlled conditions
where we can actually
then image them
in situ, and critically
sample those exudates
from living plants.
So as I introduced
the beginning,
I'm a plant biologist, right?
And when I talk to my
colleagues at the universities,
they say, well,
what on Earth are
you doing in this environmental
engineering department at MIT?
And this is what I'm doing here.
The opportunity to collaborate
with electrical engineers
and to recruit the best
PhD students in the world
to come and work in the
lab is why we're here.
It's this interdisciplinary
fearlessness
that I think so many of us
here at MIT bring to the table.
And I hope that we can use
this integrative approach
to tackle this very
simple, very complex,
mechanistic problem in a
way that we can hopefully
advance food security and
ultimately human health.
So with that, I'd like to
introduce Professor Janelle
Knox-Hayes who is a professor
of economic geography
and the planning head for MIT
architecture and planning.
All right.
All right.
Good afternoon, everyone.
And thank you so much for
joining this panel today.
I'm going to present
some research that's
looking at the interface,
taking a scale out
at the city level, the
interface of climate
impacts like extreme heat
and health dimensions.
And so this is based
out of work that
was completed a
couple of summers
ago in the height of
the COVID pandemic,
with Cynthia Rosenzweig at
Columbia University and Juan
Camilo Osorio at
Pratt University.
And we were really
interested in understanding
what was happening
with the pandemic
and how city officials could
respond to interventions
such as extreme heat in the
summer of 2020, and 2021,
when the pandemic cases
were at their highest.
But people were not able
to be brought together
into cooling centers.
A lot of public infrastructure
wasn't operational.
And so we wanted to
look at those dynamics
and see how extreme
heat is really
factor of a range of different,
not just natural conditions,
but built environmental,
socio-environmental conditions.
So our questions
for the study were,
what are the places of highest
risk in heat for COVID?
And we focused on
New York City in part
because Cynthia and Juan Camilo
are based in New York City.
We had really good access
through their network works
to a range of data, including
COVID health data and then
also this satellite imagery.
But New York City was
also the ground zero
of the pandemic for
the United States.
And what we quickly discovered
is that the pandemic
unfolded in waves.
And we could track those waves.
They were very distinct
within New York City.
So we basically mapped
three waves of the pandemic,
and that allowed us to
look at the relationship
between the environmental, the
built social infrastructure
and the sociodemographic
characteristics.
And then we also
wanted to understand
the neighborhoods that were
most impacted by COVID-19.
We were able to
show that there's
a direct correspondence with
compound risks of extreme heat.
And while it's not yet
published in this work,
we also looked at factors like
flood risk and air quality risk
and show that those
risks, they're compound.
If you're vulnerable
to one of these risks,
you tend to be vulnerable
to all of them.
And so in sum, we're looking at
a range of economic inequality,
health, built
environmental factors,
and natural environmental
factors, including
heat, extreme--
or sorry, extreme heat,
air quality and flooding,
and creating a compound
risk index for the city
that they can use to make
different planning decisions.
So there's a lot of data here.
I know I don't
have a lot of time
to try to walk
through this, but I'm
going to try to walk through
some of the key findings
from each of the
waves of the pandemic
and then look at
our index and how
that can shape different
public understandings of how
to respond to things
like the COVID pandemic.
So the unique challenge,
again, the cooling strategies
such as cooling centers, did
not meet the physical distancing
requirements of COVID.
Vulnerable communities had
less access to green space.
The other scientists
on the panel
have been talking about how
significant different factors
of plants are for generating
food, for remediating heat,
for generating energy.
And we absolutely saw that
in terms of vulnerabilities
to COVID-19.
We know a lot about
the impact of crowding
in the spread of respiratory
illnesses like COVID-19,
and that there's a
general relationship
between density, urban
density and spread.
What was really
interesting about our data,
though, is we were able
to parse urban density
and show that the most
dense areas of New York City
actually weren't
the most vulnerable.
There are different
kinds of densities.
And so I'll walk through
what those results look like.
And then we looked at the
social determinants of health,
including racial dynamics.
We created a social
vulnerability factor index
that included things like
language, isolation, access
to resources, education
level, income level,
and compared that as
well with COVID cases.
So just to take a quick
look at New York City,
we're mapping the five Boroughs.
And I don't know if I
have a pointer on here.
Maybe I do.
Yeah.
Where we can see Bronx, Queens,
Brooklyn, Staten Island.
And then we combine for to
understand vulnerability
the first two waves,
because there's
different factors that occurred
across all three waves.
And the first wave,
initially people
were being told to stay
indoors, not to mask.
Later that was reversed, and we
realized actually being outside
is much better.
If the weather is good and you
have access to green space,
get out into fresh air, don't
be crowded into indoor spaces.
And masking became essential.
In the second wave
of the pandemic
there were lockdowns
across the cities.
And you can see here
the timing where we
were able to map each of these.
Sorry, my pointer isn't quite--
first wave between January and
July of 2020 was a summer wave.
The second wave, between
August and July of 2021,
included the peak
in a winter wave.
And then between the second
and third waves, the biggest
intervention.
This is when the vaccinations
became available.
So you had a much
more extreme variant
of the virus in the third wave.
But the we'll look at how
vulnerability played out
through vaccinations.
All right.
So this study
design, really this
was taking inspiration from
an epidemiological study,
Madrigano, who created a
heat index for New York City
and creating a spatial
variation of that.
So our data are all associated.
We had more than 30 variables.
We associated them
spatially rather
than trying to think about cause
and effect of the pandemic.
We were looking at
spatial correspondence.
We created this
compound risk index
and then we evaluated
the interventions.
And the way we did this was
by combining the variables
into different factor groups.
So we looked at
vegetation density
to understand this
relationship of green space.
We looked at impervious
surface, because again, it's
not a matter of total density.
And you really in New
York City see that
with the impact of Manhattan.
Manhattan is incredibly
dense, but it
has, with Central Park in the
middle, really good access
to green space, and then other
sociodemographic conditions
that made it less
vulnerable, at least
in the first and second
waves of the pandemic that
helped us create this idea of
a mixed density variable, which
we differentiated from
rental overcrowding.
So it's not just a
matter of density.
The mixed density in some
of the places in Manhattan,
it allows for this concept
of a 10 minute city.
People who are living
in those conditions
had access to childcare,
to food resources,
to quality green space.
They might be within 10
minutes of their workplace.
And that had a severe
impact or high impact
in reducing the
severity of COVID cases.
And then also the distance,
this was kind of critical,
especially for those that didn't
have access to air conditioning,
to public cooling.
And then finally, the
social vulnerability,
we combined different factors
to look at social vulnerability.
So some of the map
on the left here
shows the top 15 most
vulnerable neighborhoods.
And a lot of these
were widely reported
in the news, Kensington
Heights on the north.
And then I'm not sure why
this isn't quite working.
Corona was one of the most
impacted neighborhoods
in Queens, and then
Sunset Park in Brooklyn
were some of the areas that
were getting a lot of attention.
And one of the first
things that you notice
is the distance of
these neighborhoods
from quality green space.
So we used a number
of different factors
to map and measure
green space and look
at where our communities
and what kind of access
do they have.
And then this is,
again, quite different
from impervious surface.
It's one of the core
vulnerabilities.
You can see how physically
dense Manhattan is,
the relative impervious surface,
but that doesn't necessarily
correspond with
quality green space
because you have good
access to green space there.
I'm going to skip
through the regressions.
I think there's interesting
results in the paper,
if you're curious
to dig through that.
And just quickly highlight
some of the key findings.
In the first wave,
temperature and COVID-19 cases
concentrated in the
same neighborhoods.
So green space is not
only for mediating heat,
it's also for mitigating
the impact of the pandemic.
It was highly significant,
especially in this summer wave
where people needed access
to quality outdoor space.
Social vulnerability was highly
significant across all three
of the waves.
And again, for those
that are curious,
you really can see
this impact here.
There were really
interesting differences
across the waves for access
to cooling infrastructure,
as you can imagine with
significant lack of air
conditioning was significant
in predicting vulnerability.
And then of course,
distance to open
space and social vulnerability.
In the switch between the
first and second waves,
we identified a really
fascinating transition.
It was known that Black
and African-American
populations were particularly
vulnerable in the first wave.
Hispanic populations were
also particularly vulnerable,
but for whatever reason didn't
get the same level of media
coverage.
And we were able to
demonstrate with our data
a shift that occurred
between waves one and two
where that vulnerability
was largely addressed
for Black and African
populations in the second wave,
and it was not addressed
for whatever reason
for Hispanic populations.
We looked into the
employment records
and there's actually, for both
African-American and Hispanic
populations, a number
a higher majority
that are working in
essential services.
For the Black population,
that was primarily in transit.
For the Hispanic population,
that was primarily
in cleaning services.
So there may be
some factor there
that that's not fully
explained in our data that
should be looked into as
to why one population was
more vulnerable.
And then the rental
overcrowding.
We saw a huge impact
there with the difference
between rental overcrowding
and mixed density.
And so this map, I just
want to quickly illustrate
the differences here.
This is the first
wave mapped over
Black and African populations.
You can see how strong
the direct correspondence
between COVID cases
and those percentages.
And then this map
is the blue hashes
are showing the second wave
of the pandemic and Hispanic
populations.
And again, you see, especially
in Brooklyn and Queens, a shift
away from predominant Black
and African populations,
but the direct impact on
the Hispanic populations.
All right.
And I'm almost out of time.
I know there's so much
data in here to cover.
Rental overcrowding,
again, a huge
overlap in prediction with COVID
vulnerability, which was, again,
very different for mixed density
in the first and second waves.
What's interesting
is in the third wave,
we really see the impact
of the vaccination rate.
This had a profound shift
in moving the vulnerability.
So you see the
areas in Queens that
were deeply impacted in
the first and second wave,
are less impacted
in the third wave.
Manhattan finally had a bigger
impact in the third wave.
And this may be as a
consequence of people's
choices around vaccination
and where that was directed.
And so just to conclude,
I know I'm out of time,
but the way that we utilize this
is we combine all of the data
and created this risk
index for the city
so that they could see where the
areas that are most vulnerable.
This part, the oceanfront
areas around Long Island,
are very well understood
for flood risk, especially.
But this Back Bay
of New York was not
well understood, how vulnerable
these populations are here.
And then we went
through and looked
at the existing
interventions for heat,
open spaces, pools,
open and cool streets,
and mapped the correspondence
of areas that are high risk
and have low access to services.
The areas that are high
risk but have good access
to services, and
then the areas that
are low risk and low
distance to cooling,
and presented these
findings to the city
so that they could look at
mapping their interventions
and understanding where were
existing vulnerabilities
and where should they
better be targeted.
All right.
So I'll leave it at that.
Thank you very much.
[INAUDIBLE] And happy to pass
it now to our next speaker
from Biology Institute,
Professor Penny Chisholm.
Thank you.
Thank you.
I have a couple of props here,
which I'll explain in a minute.
Well, it's sort of
obvious, but so I
am going to focus on the
big picture with this talk.
And we're going to go
back to photosynthesis.
What I'd like to do
is, first of all,
show you my favorite
slide on life on Earth.
And I'm going to describe
my passion for phytoplankton
over the last 40 years.
And I'm going to use that as an
example of how curiosity driven
research can yield really
interesting understanding
of life on Earth.
And this is just going to be
one organism I'm talking about,
and there are millions of others
that we should be studying.
So that's my pitch.
And then finally, I'll make a
big pitch for that, for MIT.
So first of all, life on Earth,
this is my abridged version.
We can't forget that
life all depends
on this process
of photosynthesis.
Well, there are some microbes
that can make biomass
from deep sea vents,
but if you track that,
they need oxygen, which
comes from the surface.
So all of life depends
on this photosynthesis,
taking solar energy, CO2 and
turning it into food and fiber,
and then all of the
organisms on Earth
that can't do that, take
that, we eat and we respire,
and use that chemical
energy to live.
And I feel so
passionate about this.
I actually collaborated
with a friend of mine.
We wrote a children's book
called Living Sunlight,
and it starts out-- it's
narrated by the sun--
and it starts out with
a child with its hand.
It said, the sun says, "Put your
hand on your heart and feel.
Feel your heart pumping.
Feel how warm you are.
That is my light inside of you."
And I showed this to a friend
of mine when we were drafting it
and she said, that's
a beautiful metaphor.
And I said, "It's
not a metaphor."
I said, "That is sunlight
energy in there."
And I just think it's so
important to think about that.
Because half of the MIT--
well, I digress.
But half of the MIT plant
scientists are in this room,
and we need more of them.
That's one of my messages.
OK.
So now on today's Earth,
these two reactions
are fairly balanced, roughly.
But if you integrate over
millions and millions of years,
there's more productivity
than there is respiration.
And we have this
fossil fuel, which
is basically ancient plants.
That's buried sunlight.
Another book over there.
And what we are doing
in this 150 years
is taking that and burning it,
same process there, respiration,
and that's the CO2 that's
increasing dramatically
in the atmosphere.
Hence climate change,
hence MIT's huge initiative
on climate.
So all of these initiatives
are in this one slide.
But today I'm going to
focus on that neglected
little green thing
up there, which
is the microscopic plants, the
phytoplankton, that do actually
roughly half of the
photosynthesis on Earth, if you
round up.
Maybe a little bit.
And you can't see
this very well,
but it's from another children's
book on the phytoplankton.
But there they are, these
little microscopic green plants.
It's in the upper 200
meters of the ocean.
This is roughly to scale.
And they feed all of the
life in the ocean ecosystem,
even the life in
the deeper waters.
And that is through the
settling of the biomass that
is not eaten in the surface.
We call this marine snow.
And so it's dead stuff
and mucus and everything
that's colonized by bacteria.
And as it settles, of course,
there's no photosynthesis down
there, so there's no
consumption of carbon dioxide,
but there's respiration.
So the bacteria are
respiring carbon dioxide.
And what that does is it sets
up this concentration gradient.
It's pumping CO2
into the deep ocean.
This is called the
biological pump.
So phytoplankton
serve this function
in the oceans of constantly
keeping the CO2 concentrated
in the deep ocean.
And just as a thought
experiment, if they didn't exist
and you mixed all
the ocean, so all
that CO2 equilibrated
with the atmosphere,
there would be 2 times as
much CO2 in the atmosphere.
We'd be really cooked.
OK, so when I came to MIT, I was
already studying phytoplankton.
And the most of the different
species, here are some pictures,
are roughly were between 10
and 100 microns in diameter.
In 1979, a colleague of
John Waterbury in Woods Hole
discovered these tiny little
1 micron phytoplankton
is called synechococcus.
They're shown here fluorescing
orange because of the pigments
they have, and they
were discovered
because a fluorescence
microscope was suddenly became
available and you could see
these tiny little things.
And I was lucky enough
to have a team of people.
To work with a team of people.
We took a flow cytometer, which
is a laser based instrument
on a ship, and Rob Olson,
brilliant young scientist,
was able to see even smaller--
whoops-- even smaller
than synechococcus.
You can't really see it.
There's two little
red dots there.
That's also chlorophyll
fluorescing.
That circle.
This little organism
called prochlorococcus.
It's the smallest and most
abundant, photosynthetic cell
on the planet, it turns out.
It's less than a
micron in diameter.
It's the size of the wavelengths
of light that it absorbs.
And I've been studying
it ever since.
It's been my muse.
And we now know that
there are three billion,
billion, billion prochlorococcus
in the global oceans.
They're everywhere in the
warm mid-latitude oceans.
Their annual photosynthesis
is equal to that
of all crops on land.
And over the years, we've
been able to isolate them
into culture.
And after the human genome
was finished being sequenced,
we were able to start.
They were one of the first
microbial genomes sequenced,
because they're so small, you
could do it in half a year.
Now, you can do
it in a half day.
It's very exciting.
So we're knee deep
in genomes now.
So genomes, transcriptomes.
We use all the
technologies that are
developed for
biomedical research,
we apply to prochlorococcus.
And now we have a
database, we can go.
These are transects in
the world oceans where
we have a database of
prochlorococcus genomes
from all over the global oceans.
And so over the
years, I just let
prochlorococcus point the way in
terms of what we should study.
And this is a beautiful
model organism
to use for what I call
cross-scale biology, which
is really ecology.
But that word is loaded.
So we call it cross-scale
system of biology.
And so my goal was just to
use this organism to study it,
not only at the molecular
level, but all the way
to the global biosphere
level, the oceans.
Which you can do with
this simple little cell.
And some of the things-- these
are just some of the stories
that we've learned
across, along the way.
The first thing we learned is
that although this cell has
about 2,000 genes, every
time he sequenced a new one,
you'd find 100 new genes
that you didn't see before.
And the projected pangenome
of the collective of all
prochlorococcus in the
oceans is about 80,000 genes.
That's four times the
size of the human genome.
So this tiny,
simple little cell,
it's really a superorganism.
And we can debate whether
it's a single species,
but by many measures
it's a single species.
So it's incredibly diverse.
And that's how it
maintains these populations
throughout the oceans.
Also one of my postdocs recently
discovered a novel gene transfer
agent, which he named
[INAUDIBLE] after a goddess
of good fortune or something.
Anyway, these serve
to move genes around
within the superorganism
to continuously create
the diversity in the system.
They also make these
novel metabolites,
secondary metabolites.
They make hundreds
of these molecules
that in most model microbes that
are studied make one or two,
and we don't know what
their function is at all.
That's-- somebody else has
got to figure that out.
But it's really exciting
because the cells are
trying to tell us something.
There are thousands of these
molecules in the ocean,
so we see them in the cultures.
You go in the ocean,
look at the sequences
and you can see that there
are thousands of variants.
Also we've studied vesicles.
They pop out little
lipid bound vesicles
that contain DNA and RNA.
So then you go to the
oceans and you say,
are there vesicles
in the oceans?
Yes, there are.
There are thousands
of different species
are making these
vesicles in the oceans
and putting them out there.
It's like dissolved information
going around in the oceans.
And also studying interactions
between prochlorococcus
and other organisms, a
new evolutionary theory
was developed.
Not in my lab, but
in a colleague's lab,
called the Black
Queen Hypothesis,
which is really interesting.
So some people say, well,
that's all very interesting.
It's basic research,
but how can we use them?
And I get this
question all the time.
And I say, well, they're
out there doing their job,
keeping the biological
pump and all of that.
But I look to the Sand County
Almanac for the answer from Aldo
Leopold and hopefully
not insulting anybody
in the audience who
wondered how to use them.
But I'm sure that in
time, they will be used.
They are the smallest
amount of information
that can do this amazing
photosynthesis reaction.
OK, so what do we
need more of at MIT?
I think we need more of
this curiosity driven
research on life on our planet.
And I'm all for human health.
And that's what this
initiative is about.
But also, I think
we need to broaden
our scope of life sciences
to embrace, most importantly,
plants.
I think we need
more of that at MIT.
So life is a complex
adaptive system.
And this-- here's my laser--
this is where we've
focused a lot of energy
here in biology at
MIT, for good reason,
and with amazing results.
I think that focus is
incredibly important.
But what I like to remind us is
that this is how things work.
But this, at this level, the
ecosystem level, the population
level, is why they work
that way, because that's
where evolution is operating.
And there's feedback
mechanisms from these.
This level of
organization feeds back
on this level of
organization and shapes it
through natural selection.
So we need to look at both
how and the why of the system.
And this movement is underway,
I'm very excited to say.
I say let's not miss out.
Molecular biologists, let's--
these are papers that just came
out recently.
"Let's Reconnect With Nature,
Integration Across Scales."
This one is actually
from a colleague at MIT.
Mick Follows.
"A Metabolism Shared
Across Scale."
So there is a movement
to go in this direction.
So with that, I
think prochlorococcus
and all the people that
contributed to this,
but most importantly
two foundations
made all of this work
possible, the Moore Foundation
and the Simons Foundation.
I never would have been able to
do this without that support.
It's extraordinarily important.
And finally, you probably
can't read this, but this.
I was playing around
with the ChatGTP
and said, what would
MIT and Kendall Square
look like in 25 years if
my fantasy world existed?
And that's what we
hope will happen.
And it came up with
plants all by itself.
So it understands.
I'm supposed to sit down.
[APPLAUSE]
OK.
Fantastic.
Let's see.
So I hope-- I really
enjoyed these talks.
I hope you did too.
I think it really captures how
life science, broadly defined,
is going to play a huge role in
many aspects of human health.
And these are leaders
in that space.
So this is your chance
to ask them questions.
And I heard there's a prize.
First question gets a free book.
But you do have to come up to
the mics to ask your question.
So if you can come up to the
mics and ask your questions,
that way the interpreter
can hear and interpret.
Thank you.
OK I have actually
three questions
for three different people.
And they're quick.
One is, given that the
nitrogen fixation requires
a combination of a
plant and a microbe,
is anybody talking
about simultaneously
engineering both of them--
both kinds at the same
time to communicate?
Second question.
The density that you
cite in New York,
do you consider per
area or per volume?
I mean, New York
is really built up,
so that might make a difference.
And the final question
is, in deep sea vents,
is it really
dependent on the sun?
I guess we can
take them in order.
Yeah, sure.
We can take them in order.
Dave and then Janelle.
Yes, so the way we're
envisioning this
is sort of a
co-optimized system.
We're not using any sort of
directed evolution process where
we're--
the process is not
so much co-optimized
as the goal to have a
co-optimized system where
we understand what the plants
want, what the microbes want,
and how they're
talking to each other.
So yes, the idea is to
have them synergistic
and sort of designed in
tandem to be together.
Yeah, and on the
density question,
it's a terrific density--
or a terrific question.
We were looking at density in
terms of the total population
density.
And then for mixed
density, it was including
building type and building use.
So mixed density included both
residential and commercial.
And when you map out what
that variable where it is,
again you see the
areas like Manhattan
have a lot of mixed density.
Rental overcrowding was also
taking total population density
and combining it
with housing type.
So type of family, is it a mixed
family, a multiple family unit
complex, and then number of
residents in that in each unit.
So, so rental
overcrowding showed
areas where you have
rental use, but then high--
more density in
each rental unit.
So kind of I think
that is both combining
the area and the volume
for that question.
The deep sea vents,
the bacteria down there
are oxidizing hydrogen
sulfide, and so they
need oxygen to do that.
To fix CO2.
Right, Dave?
Yeah.
So it's coming
from the atmosphere
up above, which is
coming from the plants,
and finding its way down
to the deep ocean and that.
So in effect, they need
oxygen from plants.
Thank you.
Hi, my name is Amy Nurnberger.
I'm from data and specialized
services and research data
management.
Thank you, each of you, for
such delightful presentations
showcasing the
variety of work that
is done around the
biosphere and plants.
Given that necessary
interdisciplinarity,
the focus on collaboration
in this initiative.
Can you each speak
briefly about what
you would like to
see from yourselves,
from your colleagues,
from others on the stage
to increase the effective
sharing, and reuse
of the data and other
associated products
that your research creates, so
that you can more fruitfully
grow from each other?
Thank you.
Well, I think one of the
themes of this session
is across scales.
So I think the key to
that understanding,
and Penny did a great job
of wrapping it up with that,
is understanding how to
get the output of one scale
to merge the input of
the other and vice versa.
And I think sometimes
fields are very
specific in the type of
data that they produce,
and how they look at it.
But I think finding
metrics, like in some ways
actually, I think
photosynthesis is
a great example for this
because of fluorescence,
like we both talked about.
That actually allows
a kind of cross scale.
Because you can
look at the process
itself in a way that you
can't with other things.
You need an exogenous reporter.
And so I think while we have
that in what we measure,
we maybe don't save it in
the same way or report it.
And I think having
catalyst type funding that
allows you to build across
fields allows that kind of thing
to happen.
Yes.
I mean, your question kind
of keeps me up at night,
to be honest.
So we've heard a lot
about omics this morning.
Penny mentioned different omics.
I do a lot of
genomics and we've had
20 some years now to think about
what to do with genomic data.
In a sense it's kind of
easy and that there's only
four letters that
make up-- five if you
want to count Uracil for RNA.
But it's something that's pretty
easily stored and manipulated
and we more or less
understand what it is.
And we've now gotten
to the point where
we can collect
all the other omic
data much more easily than we
can analyze it or even handle
it.
So I am very interested
and very concerned
with how we would standardize
metabolomic, proteomic
phenotypic data,
that kind of stuff
that Gabrielle's lab generates.
These are optical things.
Like how do you quantify and
store that and compare it?
So thinking about ways
to integrate big data
and be clever about how
we generate the big data,
I'm very bullish on.
And think we need to spend
a lot more time on doing it.
And I'm optimistic the
people at MIT can this out,
but it's something that
we're very focused on.
Yeah, so another area of my work
I wanted to talk about today
but didn't really have time is
Indigenous community planning.
And I've been building
a group in urban studies
and planning that's
really focused
on trying to weave traditional
ecological knowledge
into conventional sciences.
So for example,
the nitrogen cycle.
That's well known in
Indigenous practice
of growing three different
species together,
the corn, beans, and rice,
to help fix the nitrogen.
There's a lot of knowledge
about plant urban
and natural systems that comes
from Indigenous communities.
So the green space we
showed in the COVID study,
how significant that is.
It has a cooling factor,
but it isn't just
the physical cooling factor.
There's a lot of well
demonstrated health literature
about how being around
high quality green space
or high quality density of
trees helps human health.
And one of the things--
one of the students in my lab
has started studying is seeds.
The way in which they have
actually acoustic resonance.
So she played sounds
to seeds, and then has
demonstrated that the
seeds actually respond.
There is sound that comes back.
And she's done some really
amazing architecture
and art displays,
mapping and showing
what that sound looks like.
But we're partnering now with
Professor Nicholas Makris, who's
in aeronautic and
acoustic engineering,
to look at that sound
resonance across a broader
spectrum of trees.
And it could
revolutionize the way
we think about urban planning
in relation to plants.
So there's a whole, not
just within the seeds,
but mature plants, a whole sound
biology that exists around them.
And we're in planning, often
thinking just about sight.
Visually, what are we seeing?
Or even the light impact
is somehow a site impact.
But acoustically what is
happening both in our natural
and our urban environments
is critically important.
And so this kind of combination
of Indigenous knowledge,
what is known about the
species, what they're
speaking, what they're saying.
The urban planning and
then the life sciences
of understanding the physics
and the engineering of that
could be really critical
in unpacking and creating
new areas and ways in which
we think about planning
and development and health.
And so I think this initiative
really captures that spirit.
And that's critically
what we need more,
is to put people from
different disciplines
together, from different
ways of knowing together
and combine that knowledge
to better effect.
Be efficient.
I'll just second what Dave said.
Because we have just this
extraordinary database
of prochlorococcus
sequences from all
over the global oceans just
screaming with studies,
with things to tell us.
And it's just a huge challenge
getting it all organized
so that others can use it.
So we have time for one
more question today,
which is fantastic.
You guys didn't have to
listen to any of my questions.
So well done as an audience.
Go for it.
So my question is about
the COVID-19 research.
I'm curious.
I think it's Lydia
Bourguiba's lab
that showed a real
correspondences between humidity
in the immediate environment
and COVID transmission
and that in tropical
locations, it's
when the humidity goes
above a certain point that
COVID transmission increases.
And in climates
like ours, it's when
we start to spend a
lot of time in dry heat
in the winter that COVID
transmission increases.
I was really struck by your
mapping of flood zones and heat
zones with COVID.
And I'm curious if that is--
if there's any
collaboration there?
If that's another factor,
and if that sort of humidity
might be another
interventional lever that we
could use in public health?
Absolutely that's
a terrific point.
And thank you for highlighting
that research within MIT
as well.
It's something we haven't
published the flood risk
paper yet, the flood
risk version of this,
but something that was really
significant in type of building
was basement use.
Where are people
living in basements?
It's a huge vulnerability.
I mean, you can
understand for flood risk.
Obviously, we've seen
that in New York City
before with some of
the major storms.
But then also with COVID risk.
And so humidity might be one of
the factors where again, where
are people living
in crowded spaces,
rental spaces with
suboptimal urban conditions.
And then how is that
increasing the compound risk
or the compound vulnerability.
A great point.
Thank you.
Thank you.
Thank you guys very much.
Thank you for being here.
You have 20 minutes to
make your way back over.
So thank you, guys.
[APPLAUSE]

---

### MIT HEALS Launch: Healthcare breakout session
URL: https://www.youtube.com/watch?v=eJhszcNUeYs

Idioma: en

Hello, everybody.
We're going to start the
health care session now.
Thanks so much for being here.
We have a really exciting
session for you today.
My name is Jon Gruber.
I'm the chairman of the
Economics Department
here at MIT.
And my work is in health care
economics and health care
policy.
We also have speaking
Kate Kellogg, who's
the David J McGrath junior
professor of Management
and Innovation and a professor
of Business Administration
at the MIT Sloan School.
Kate's work is helping knowledge
workers and organizations
develop and implement
narrow AI and generative
AI solutions, on-the-ground
and everyday work
to improve decision-making,
collaboration, and learning.
We have Joe Doyle,
who's the Erwin H Schell
Professor of Management
and applied econometrics
at MIT Sloan School as
well as the co-chair
of the health sector
at the JPAL Action Lab
and co-principal investigator
of the NBER Roybal
Center for Behavioral
Change and Health.
Through all these
many initiatives,
Joe partners with large health
care providers and payers
to conduct randomized controlled
trials aimed at improving
health care delivery.
And then finally, we
have Michael Yaffe,
who's the director of the MIT
Center for Precision Cancer
Medicine, the David H
Coke professor in science,
professor of the MIT departments
of Biological Engineering
and Biology, intramural
faculty at the Koch Institute,
and also, a real doctor.
So if any of you are
having pains or whatever,
you can ask him
during the session.
His laboratory studies how
signaling cancer cells focusing
on cell stress, DNA
damage, inflammation
control the response of tumors
to conventional and novel types
of cancer treatment.
And he's particularly
interested in building
the bridge between
basic biological science
and clinical treatments
for human disease.
So great panel here today--
we're just going to talk
for about 10 minutes
and then hopefully have a
little time for Q&A at the end,
depending on how much
we abuse our privileges.
So I will go first.
And I'm going to talk--
if you can put my
slides up, I'm going
to talk today about risk
sharing in the life sciences.
So we're just waiting,
I think, for my slides.
Anyway, so there we go.
Insurance, paying for health
insurance in the life sciences,
great.
So let me see.
I think if I go like this,
it'll go to the-- no,
that's the pointer.
If I go like that--
how do I get to the next one?
How about that?
There you go.
So today, I'm going to
focus on risk sharing.
This is a term
economists use glibly.
But let me be clear
on what I mean,
which is that medical risks
are the greatest risk facing
most Americans.
They're difficult to
prepare for in that they're
large and unexpected.
So what do we do as society,
as developed societies
all around the world?
We spread that risk
through insurance.
How does insurance work?
We prepay a fixed
amount in return
for coverage of our medical
costs when we need care.
OK, so that's how we share the
risk of getting hit by a car
or having a heart attack
or developing cancer.
We prepay now.
And then later, those
costs are covered.
Now, in that world,
we have to recognize
that life science innovation
is incredibly expensive.
New innovations in life sciences
are expensive, becoming more so.
The estimates suggest that the
cost of developing a new drug
is $2 billion.
That results in high prices to
the end users of those drugs.
Those high prices are what
we need insurance for.
And we need it because we are
making miraculous steps in life
sciences innovation every day.
We can talk about a moderately
expensive miracle, which
is GLP-1s, these
new drugs, which
can be used to control weight
and help with diabetes,
have been shown in large
trials to be massively
successful at curbing weight
and increasingly shown impacts
on curbing other behaviors.
And that's moderately expensive.
That's $15,000 a year.
Then, we have incredibly
expensive miracles,
which is cell and
gene therapies, which
are coming online, which are
literally curing the incurable.
These are diseases which through
loss of a genetic lottery
was killing people.
We can now literally
cure them and let
people who would have died
at age two live to old age.
They're miraculous.
They're unbelievably expensive.
A typical cell and gene therapy
costs millions of dollars.
So you talk about something
you need insurance for.
You need insurance against
losing a genetic lottery
and having to face a
$2 million expense.
But the US insurance
system has key holes.
The first hole is
discrimination.
So before the Affordable
Care Act was passed in 2010,
it was completely
legal in the US
to discriminate against those
who were sick in your insurance
decisions.
You could deny
insurance coverage
to someone who was sick.
You could charge a sick
person or a woman more
for health insurance
than you'd charge
a healthy person or a man.
Or you could exclude those
with pre-existing conditions
from coverage.
That ended with the
Affordable Care Act.
We banned discrimination
in insurance.
Now, insurers have to
charge men and women
and the sick and healthy the
same price for health insurance.
This is critical in a
world where we increasingly
predict risk.
Increasingly, I can tell
who's going to be sick
before they walk in my door.
What that means is any
profit-maximizing insurer
is going to want
to avoid the sick.
So in a world where we know how
sick people are going to be,
it's critical that we don't
allow insurers to discriminate.
Now, the ACA solved
this problem.
It seems safe for now.
But there are still
opponents of this protection
now in power in the government.
And that's something
we need to monitor.
I think the ACA is pretty safe.
But this would be a critical
loss for risk protection
if this went away.
There's also the
uninsured, however.
About half of 25 million
Americans lack insurance.
About 25 million Americans,
sorry, lack insurance.
This was cut in about
half by the ACA.
But it could rise significantly.
There are policy decisions
that matter here.
The number of uninsured
covered by the ACA
went down by about four
million under the first Trump
administration, rose
by about 5 million
under the Biden
administration, and is
anticipated to fall again.
Who are these remaining insured?
Many of them are actually
eligible for free insurance
today if we could just
automatically enroll them.
Some are young
invincibles, people
who feel that they could just--
they don't need insurance
because they're healthy.
And some are
undocumented immigrants,
which is by itself a
political minefield.
So it's going to be challenging
to get these remaining
uninsured.
But it's an important goal if
we want to once again protect
against medical risk.
There's also the fact that
we have a multi-payer system.
Individuals move across
different payers.
So if I'm an insurer, and I
have a decision about investing
in preventive care for you,
but by the time you're sick,
you'll be with a
different insurer,
I don't want to make
that investment.
Why should I pay that $2
million for a cell and gene
therapy that's
mostly going to save
money for some other
insurer down the line.
So insurers don't want
to provide investments
because people might leave
and go to other insurers.
On top of that,
insurers are worried
about these really
expensive risks,
these things like
cell and gene therapy
that are incredibly expensive.
So what do they do?
Insurers get insured.
They buy reinsurance.
There's a contract
that says, look,
you're insuring everyone at MIT,
but if someone has costs above
$50,000, we'll pick
up the difference.
They buy it from big companies
that offer that reinsurance.
The problem I
realized recently is
the ACA did not fix
the reinsurance market.
It's completely legal
to discriminate.
So MIT could go to
a reinsurer and say,
look, we're going to cover
the cell and gene therapy.
And the reinsurer can say, nope,
I'm not going to reinsure that.
You're on your own for that.
So there's still
holes in the system
even for those insured
we need to address.
So let's do an
important case study.
We're going to talk about
innovation a lot today.
Let's talk about one of
the most innovative areas
in medical care, which is
cell and gene therapies.
The adoption has
actually been slow.
There are some new cell and
gene therapies being introduced.
And the adoption has been
slow despite the fact
that they're literally
miraculous and life-saving.
Why is that?
Well, insurers are very
wary of paying for them.
First of all, as
I said, they have
to make a major investment
in paying for them,
but they might not get
the lifetime benefits
of people staying on the plan.
Second of all, these
drugs are miraculous,
but the trials are so
far very short-term.
We don't know if
five years from now,
they're going to stop working.
And insurers are like, I
could pay all this money,
and five years from now, it
could wear out and stop working.
And so as a result, I'm wary
of making that investment.
And finally, reinsurers won't
cover these drugs in many cases.
So insurers have
been slow to adopt.
When they do adopt,
they try very
hard to limit who can get
these new innovations.
The problem is people can't
afford them without insurance,
and life-saving opportunities
are being missed.
And this can in turn feed back
to investment in these new areas
because people are reluctant
to invest in cell and gene
therapies because they're
afraid they won't be covered.
So what do we do?
Well, single-payer
health care is not
walking through
the door in the US.
It's too politically
difficult. But we do actually
have single-payer
health care in the US.
People don't realize it.
We actually have
single-payer health care
in the US for a particular
disease, renal failure.
For those with
renal failure, they
are almost called the end-stage
renal disease program.
And it's a universal
government-provided single-payer
insurance.
Why couldn't we do this
for other diseases,
say genetic abnormalities?
Why couldn't we say for those
born with genetic abnormalities,
they will be on a
government-funded single-payer
insurance system?
That would solve
all these problems.
And basically, as
a society, we'd
all pay a little
higher taxes because we
won the genetic lottery
to support those
who lost the genetic lottery.
It's a perfect case for
government intervention
where we as a society
share the risk.
Alternatively, if
that's not feasible,
we can think about developing
private sector models
to spread the cost with things
like value-based reimbursement
that allow us to have
the manufacturers share
the risk of drugs not working
and subscription models that
put the risk on manufacturers.
These are things that
I'm working on, trying
to see if we can push
these in the private sector
in the absence of a government
solution to this problem.
Now, one of the last
thing I'll talk about
is what about pricing.
These treatments
are very expensive.
You might say, well,
why do they have to be?
Why don't we just tell
these manufacturers
they can't charge much?
And the answer is because
innovation won't happen.
Innovation does respond
to financial incentives.
So the traditional
approach to dealing
with this around the world is
we measure something called
comparative effectiveness.
All around the world they have
organizations which say, look,
if a drug only delivers
this much of health value,
we'll only pay this much for it.
The US is the only country
that doesn't do that.
As a result, we pay dramatically
more for our medical treatments
than anywhere else in the world.
The typical medical
treatment costs about twice
in the US for innovative
treatments as it does in Europe.
So that's something we
could do in a step that
was started with the
Inflation Reduction Act.
The problem is for these
incredibly expensive new drugs,
they're actually worth the
price they're charging.
So they're charging $2 million.
And you know?
They're delivering $2
million of benefit.
So dealing with pricing
in this new sector
is going to be much
more challenging.
And we're going to need to
think about other mechanisms
to deal with that.
So finally, I guess the
last thing I want to say
is, how do we finance
this innovation?
This is another area
economists are thinking about,
which is basically we
need to not only cover
these things for insurance
but actually promote
more R&D funding of
these valuable things.
This is discussed in my
book, Jump-Starting America,
where we talk about the role
of public funding of research
and development, which
actually at its peak
amounted to 2% of US GDP
and is now only 0.5% today.
This funding complements private
R&D. It promotes private R&D
and has huge economic returns.
Yet, as I said, investments
have been falling.
So the bottom line is we need
to think about creative solution
to get more innovative
funding in this area
and to help bear or share
the risks that people face
with expensive new treatments.
I'll stop there.
Thank you very much.
[APPLAUSE]
I am delighted to be here today
to talk to you about my research
on narrow and generative AI.
And both of these forms
of AI have great potential
to transform health care.
Here's an example of narrow AI.
Some of my coauthors at the Duke
Institute for Health Innovation
worked with hospital doctors to
develop a highly accurate tool
using machine
learning which detects
unexpected patterns in patients
in the ER to predict sepsis.
And sepsis is a life-threatening
illness infection
where you need to
predict it quickly
in order to treat it quickly.
Here's a generative
AI example developed
by some of my coauthors at
NYU Langone Health where
they use GPT-4 to read
clinical notes of patients who
are inpatients and create
patient-friendly discharge
summaries for patients.
And discharge
summaries have been
shown to be really important
to transitions of care
from inpatient to outpatient.
And yet, 88% of current
discharge summaries
are unreadable to patients.
So clearly, these
AI solutions can
result in improved quality,
reduced costs, and increased
revenues.
And that's often what
developers are focused on.
However, what they
often do not focus on
is the workers who need to
implement the solutions.
And based on my work
with coauthors at Duke,
we found that narrow
AI tools often
offer few benefits to workers,
require laborious development,
and also, are a threat
to user autonomy.
And with colleagues
at Langone, we
found that generative
AI solutions often
have unclear applications,
require laborious iteration,
and result in harmful outputs.
And workers don't
like these solutions
that make their lives worse.
So how can developers
design AI solutions
for successful implementation?
I've done research at Duke
looking at their narrow AI
solutions and NYU Langone
looking at their generative AI
solutions.
And what we found is that
each of these solutions
have particular
characteristics that
raise challenges for workers.
So the challenges
with narrow AI are
that it's predictive,
laborious and prescriptive,
and with generative AI are that
it has unclear applications,
is laborious, and
results in new risks.
So developers need to design AI
tools with worker implementation
in mind.
So with narrow AI, let's look at
this predictive characteristic
first.
At Duke, what they found
is that specialists
who treat patients for
particular diseases
often asked the developers to
develop tools to flag patients
at risk of developing
these diseases.
So, as Jonathan just
told you, AI tools are--
in medicine, we're going to be
able to predict a lot of things.
And so this is great.
We can now intervene
earlier with patients
with all of these disease states
in order to have higher impact.
The problem is it means
that upstream clinicians are
expected to use these tools.
And now, they're bombarded
with a host of narrow AI tools
that they didn't ask
for in the first place.
So what they do at Duke is they
identify the true end users,
in this case, the
primary care physicians.
And they develop solutions
to address their pain points.
So in the case of a
narrow AI tool which
can detect chronic kidney
disease early, what they found
is that the
challenge for PCPs is
that they already have limited
time to prepare for patients.
So they don't have time
to use another tool,
and they also don't have time
to ensure appropriate follow-up
for patients who get
flagged by the tool.
So the Duke development
team worked for solutions
where it was the
nephrologist who
asked for the tool who
actually are the ones who
use it to detect this problem.
They send their recommendations
to an EHR system,
which gives doctors the
recommendations directly
before the patient visit.
And they provide
dedicated care managers
so that any patients who
are flagged to be at risk
get appropriate follow-up.
Another issue with narrow AI is
that it's laborious to develop.
It requires a lot
of back and forth
between clinicians
and developers
before you can make
these tools work.
And so specialists came to
Duke developers and said,
help us detect low-risk
pulmonary embolism in the ER
because a lot of patients
with pulmonary embolism
are getting admitted
to the hospital who
don't need to be admitted.
And it's the emergency
room physicians
who need to use this
tool who were initially
asked to help develop it.
And they said, we've got
a lot of other things
we're already working on.
So Duke then went to
the downstream vascular
and cardiology
specialists and got
them involved in this
laborious development
to do things like validate
outcome definition,
make sure that model
inputs were fit for use,
and do a silent trial to test
the performance of the tool.
Finally, narrow AI
is prescriptive.
It tells doctors what to do,
and that can threaten autonomy.
In the case of the sepsis
tool I told you about,
it was rapid
response team nurses
who wanted ER doctors to use
this tool to detect sepsis.
And initially, they were
the ones using the tool.
And ER doctors
didn't like having
these people far away from
the ER all of a sudden
being the ones to detect sepsis
on the ER doctor's patients.
So the rapid
response team nurses
started, whenever an AI tool
flagged someone with sepsis,
they would do a chart review.
But they would call
the doctors in the ER
to make sure the doctors were
the ones who did the diagnosis
and placed all the orders
for patients with sepsis.
So Duke developers have
learned to increase benefits
for true end users, reduce
labor for end users,
and protect the autonomy
of true end users.
Generative AI similarly
raises particular challenges
for workers.
At NYU Langone, one challenge
is that generative AI
is general purpose,
so it has a million
different potential
applications.
And what they do is they
involve the workers themselves
in widespread decentralized
experimentation
so that it's the workers,
not the developers, who
are deciding what are
the applications that
are going to be most valuable
for the health care workers.
And one way they do
this is prompt-a-thons,
where they provide workers with
basic education in generative
AI and hands-on ability to use
the generative AI with health
care data sets in order
to get a feel for things
and begin to develop ideas
for good applications.
One developer said, "GPT has
these fascinating moments where
it just does stuff."
And that's motivating for people
working with generative AI.
Another challenge is
that these solutions
require laborious iteration.
So for that tool I
told you about with
the patient-friendly
discharge summaries,
that required clinicians to go
through dozens of patients who
had been in inpatients
and their clinical notes
to figure out how to
draft an accurate prompt
that resulted in a good output
for the patient-friendly
discharge notes.
What the developers
do is they need
to sustain worker motivation
for doing this iteration on top
of their regular job.
And so they provide on-demand
assistance in office hours,
provide help with what they
call "prompt tricking,"
which in this case involves
things like with GPT-4 using all
capital letters when
giving instructions,
and even telling GPT-4 that
you'll give it a tip if it gives
you good outputs.
Finally, generative
AI presents new risks
because of hallucinations, bias,
and lack of interpretability.
So developers use a
Gen AI risk screen
where they screen all
potential applications up front
to see which ones
are worth developing.
So in one department, clinicians
came to developers and said,
we really want to use generative
AI to translate consent forms
into other languages.
But what the
developers discovered
is that GPT-4 tends to
miss important nuances.
So it's not a good technical
fit with generative AI,
and it also raises
regulatory concerns related
to this bad interpretation.
So they stopped the
development of that tool.
So at Langone, they
catalyzed this decentralized
experimentation
with prompt-a-thons,
used this on-demand
technical support,
and also, use a risk
score to prioritize which
projects are going to be best.
So in sum, what I want you
to take away from today
is we often focus on how these
tools are great for quality,
costs, and revenues.
But we forget that
they often present
problems for the workers
who need to implement them.
Developers need to
address these challenges
to design AI solutions for
successful implementation.
Thank you.
[APPLAUSE]
All right, hi, everyone.
My name is Joe Doyle.
I'm an economist at MIT
Sloan, and a lot of my day job
is partnering with
payers and providers
to try to get patients
to be healthier
and providers to be
more guideline adherent
and do a better job.
So I help run our Health
Systems Initiative at Sloan.
And we have three
pillars of research
backed by about 30 faculty
members and a bunch of PhD
students working on health
analytics, healthcare
operations, and health
care incentives.
On the analytics side, we
build new frontier tools,
but also, we have
people like Kate
who help think about how do
you get those tools adopted
in a meaningful way.
Operations are to get more
efficiency in our health care
system.
And then, as an
economist, I think
a lot about these incentives.
How do you get people
to behave healthier?
How do you set up the right
procedures and incentives
to get that to happen?
So let me talk about one
part of my research, which
is randomized trials in
health care delivery.
Most drugs or medical devices
will have some randomized trials
behind them to show you
that they're effective.
But when we go to change how
we're going to deliver care,
there's much less that's
done in that rigorous way
that we can learn whether
things work or don't work.
And so I'm here to evangelize
that we need more of this.
So let me talk a bit about
the economics underlying it.
Here's a map of Boston.
And if you think
about the Bus Route 1,
if you get on in Roxbury, the
life expectancy is 59 years old.
If you go a couple of
miles north to Back Bay,
life expectancy is 92 years old.
So that's 3 miles buys you
30 years of life expectancy.
So you often hear people
say there's genetic code.
There's zip code.
This is the zip code part of it.
Through poverty and
other social risk
factors are major
drivers of whether people
are healthy or not.
So health care in
the clinic needs
to think a bit more
broadly and think, well,
what can we do upstream?
How can we address the issues
that are really driving
a 30-year drop in
life expectancy?
So one way we
think about it is--
and John talked about
the need for insurance
if you get unlucky.
Well this graphic here shows
you that 1% of the population
drives about 30% of health
care costs in any given year.
And we're spending $4.5 trillion
in health care in the US.
So 1% of people are driving,
have over $1 trillion
of spending on them.
If we could figure out how
to treat those patients
in a healthier way, in a
more effective way, more
efficient way, we could
save a lot of money
and improve their lives.
So in the health care
delivery literature,
these people are called super
utilizers or health care
hotspots.
And there was this really
interesting article
in The New Yorker
a few years ago
about hotspotting could make
people's lives better and save
money at the same time.
So it was a profile
of Camden, New Jersey.
It's a high-poverty city
outside of Philadelphia.
And Dr. Jeff Brenner, who's
sort of a hero of mine--
I'd say he's a very
inspiring physician there
who organized the collection
of data from all the health
care providers into one
health information exchange.
And then, they would
try to figure out,
who are these hotspots, and
how can we treat them better?
Their idea was to create
a team of about 10 people,
including nurses, psychologists,
social workers who
would make home visits, I
think about 5 to 10 home visits
to the super utilizers,
send nurses with them
to their appointments, and
try to navigate that system
in a more effective way.
If you had all the
health care problems
the people in the 1% have,
1% of health care spending,
then you would really appreciate
having this team of people there
to help you.
So here are the number
of admissions leading up
to entry into that program
and then afterwards.
And you can see in the quarters
before, you see a ramp-up.
People got unlucky.
They lost the genetic lottery
or some kind of lottery,
and they are going
to the hospital more.
That's how you get
into the program.
And then, after they're in
the program, their costs fall.
And that's exactly what The New
Yorker article and hotspotting
programs that were popping
up all across the country
were pointing to this
graph and saying,
we could save a lot
of money and improve
the lives of these
people at the same time.
Now, economists and
health service researchers
wonder, well, what
would have happened
if they didn't get that program,
what we call the counterfactual?
Against the fact that
they got the program, what
would have happened?
Well, the nice thing
about a randomized
controlled trial,
you flip a coin
and figure out who's in
the treatment group, who's
in the control group.
The control group gives
you the counterfactual,
what would have happened
if we flipped the coin
and got heads instead of tails.
And so in the
orange bars, that's
exactly what happened
to the control group.
It exactly mirrors what happened
in that treatment group that
got lavished care, all these
home visits, phone calls,
appointment collaborations.
And so what we learned
from this experiment
is that the drop in health care
spending was a bit of a mirage.
It wasn't the program
that did that.
When people get
healthy and unlucky,
they tend to get luckier
than they were before.
And they sort of
revert to the mean.
Now, I wanted to go
into this program
and say, well, let's show
that this thing works
so that it will spread even
faster than it's already
spreading.
This is a very inspiring
program, helping people
both in their clinical
needs and their social needs
to improve their health.
And what I learned
was that it's not
going to be sustained by
lower health care costs.
We might want to pay for
it for other reasons,
but we can't make
the claim that we're
saving money at the same time.
And it might have felt good
to pat ourselves on the back
and say, we're saving
money at the same time.
But if you're not
actually doing it,
then you're just going
to chase your tail.
We need to know
what works in order
to have that spread so we
can make real progress.
So the paper came out
just before COVID.
There was a lot of attention
devoted to it because this
was a very popular program,
like I said, springing
up all across the country.
And what folks said
across the country was,
well, I used to assume
that this works.
This seems like it should work.
In our own data,
our own dashboards,
people get healthier after
they go on the program.
And what they had to do was
go back to the drawing board
and figure out, does it
work in my area or not?
So going from a
presumption that something
works to actually let's
figure out if it works,
that's what we do at MIT.
We actually try to figure out
what works in a rigorous way.
So then, we can march down the
path of progress as opposed
to window dressing
that feels good when
you say that you're doing it.
So I partnered with Geisinger
on another study that
targeted patients
who had diabetes,
and they were uncontrolled
according to the clinicians
there.
They had a blood sugar
that's HBA1C level over 8.
So to put it in perspective
would be diabetic.
And over 8, they call
it an uncontrolled.
On average, people had
a score of over 10.
And if your doctor tells
you have an A1C over 10,
that's kind of a scary number.
So this program
was also inspiring.
It gives lots of healthy food
to these low-income patients
with diabetes.
It gives about 10 meals per
week, not only for the patient,
but for their entire family.
They go to a clinic every week
to pick up their groceries.
And at the clinic, there's
a dietitian, a nurse,
and a community
health worker that
will try to address their
needs through foot exams,
have dietitian consultations,
have cooking classes,
diabetes self-management
training.
And these are brand new
clinics that are well-lit,
staffed with
super-friendly people.
People love going
to these clinics.
But what I found was, again,
a mean reversion story,
that when they were targeting
the people with very high levels
of A1C, the treatment
group got healthier,
but the control group
also got healthier.
And this is back to this
mean reversion story.
And so what am I taking away
from this sort of now growing
evidence that when you target
people with a pretty sick,
some measure that
they're pretty sick--
so it was a high blood
sugar level or high spending
in the last year--
you really have to be worried
that they might mean revert.
So one idea is you can't
just trust those dashboards
that you're seeing that
people are getting healthier.
If mean reversion is a story
that you have to worry about,
you can't just rely on
a pre-post comparison.
You need a credible
control group.
The other idea is to target
durably eligible people.
So the research
I'm doing right now
is I'm finding that if you could
predict who would get healthier
regardless of the
program, then you
could target the scarce
resources of this program
toward the people who would
actually need the program.
And when I did a reanalysis
of the people who
predicted to not get
healthier on their own
based on what we could
observe when they enrolled
in the study, those
people, their HBA1C
would fall by more than
a point if you gave them
the program, which is
better than a lot of drugs,
so getting people off of drugs
toward healthier food, which
was the initial
goal of the program,
but learning how to target it
in a way that actually makes
that difference as opposed
to just hoping that you're
making the difference.
And also, just back
to that bus route map,
social risk factors are
obviously important.
So there's just
more work that we
need to do to learn
how to address them.
We don't just give
up on those 1%.
We think about, well, what
we were doing before wasn't
working to save money.
Let's see if we could--
are there other programs
you could do, like
better remote monitoring,
other technological solutions
that you'll be seeing here
at MIT throughout the day.
Just really quickly,
just to point out that
I work with Geisinger on these
large-scale messaging campaigns.
So you might get these
text messages reminding you
of your appointment.
Well, what should
be in those messages
to get you to do healthy
behaviors like get your cancer
screenings on time, get
vaccinated, and so on?
And with machine
learning, we can
try to target those
messages, figuring out
exactly what types
of messages get
you to behave in the
healthiest ways, so at least
giving you that
advice so that you can
choose to do so for yourself.
At Sloan, we're working
with Quest Diagnostics,
which takes blood from
almost every American
every three years.
Everybody will go into
there at some point.
And we're doing randomized
controlled trials
to see what we could do to get
people to be healthier when
they get those results back.
And then I just wanted to end
on you don't need to always do
a randomized controlled trial.
Sometimes with data
analytics, we can do more.
So just my last example here is
the x-axis here is birth weight.
And at 1,500 grams,
3 pounds 5 ounces,
newborns below that are
labeled very low birth weight.
And what you can see here
is that as you get lighter,
we spend more on you.
And you see a discrete jump
if you go across 1,500 grams
because especially at
lower-level hospitals,
they increase the amount
of treatment they give you.
If you get categorized
in that way.
And then, what we see
is as you get lighter,
the mortality rate is rising.
But it falls against
the trend if you
cross that magical threshold
that got you the extra care.
And so here, if we're rooting
around for waste and value
in health care, here's a
place where people born just
above and below that threshold
is like a genetic lottery
or a birth weight lottery.
Some people win the lottery,
and they're actually too light.
They're just below
the threshold.
They get more health
care, and actually, they
get better outcomes.
And then, this was replicated
in other countries like Chile.
Here's the mortality
result for Chile.
But here's eighth grade test
scores or first to eighth grade
test scores.
If you were born too late, you
do better in school in Chile.
And you do better
in school in Norway.
But in high school,
you're better off
if you were born a few
grams less at the time
that you were born.
So there's a lot to do.
And I really appreciate you
guys spending the day with us
here at MIT to figure out
how we can do it together.
Thanks very much.
[APPLAUSE]
I'm Mike Yaffee, and it's a
great pleasure to be here today.
As you heard, I'm a
professor in biology
and biological engineering.
And I direct the MIT Center
for Precision Cancer Medicine.
But I'm also a
practicing clinician.
I'm an intensive care physician,
and I'm a practicing surgeon.
And what I'm hoping to do in
the next nine minutes or so
is to connect this session on
health care delivery with some
of the basic science
and engineering concepts
that we heard about during
the first half of this,
during this morning's session.
And so to start off, I'm
going to focus on cancer.
And just for fun, I took every
paper that had been published
in Cell or Nature 43 years ago
in 1981 that contained the word
"cancer."
And I used deep learning
to build a word cloud.
And the words that
jump out at us
are things like gene, protein,
DNA, virus, transforming,
principle.
And in fact, that fit
with 1981's view of cancer
as a genetic disease in which
oncogenes were turned off
or tumor suppressor
genes were silenced.
And MIT was at the very
center of this work.
In fact, many of those papers
in science, in Cell or Nature,
were authored by the luminaries
here at MIT, Bob Weinberg
and Phil Sharp, David Houseman,
Rudy Jaenisch, Sheldon Penman,
David Baltimore, all
the people that you
heard alluded to this morning.
The problem is, let's
look 42 years later.
There are many more papers
about cancer in Cell or Nature.
And so I limited myself
to the first 160 papers
that I could identify.
And now, all of a sudden, the
main text in the abstracts
are things like patients, human
treatment, therapeutic response.
This should be very
worrisome for us
here at MIT because we
don't have a hospital.
We don't own a medical school.
Fortunately, however, if
you look a little deeper,
there are many ways
that MIT can contribute
because underlying this is a
much more systems-based approach
where instead of
just oncogenes, now
we see the immune
system and the tumor
microenvironment, metabolic
defects, and signaling defects.
And so using that, I want to
talk about how in the future
we're going to bridge the divide
between the clinical world
and the basic science world.
And I'm going to use
my own lab's research
just as an example of
how to illustrate this.
So I want to leave you with
three take-home messages.
The first is that the future
of precision health care
has to be an active and
ongoing dynamic collaboration
between clinicians
and basic scientists.
In 1981, the basic
science was revolutionary.
And it was directing
clinical care
or beginning to
illuminate clinical care.
Now, we have to go both
ways in order to do this.
And an example that
I'll use for this
is our lab's research
related to tissue injury.
We've been very interested
in tissue injury
in the setting of both trauma
and cancer, and in particular,
how tissue injury is sensed
by the innate immune system
and how that interacts with
the blood clotting pathways
to further stimulate the immune
system for or against tissue
injury.
And this was work that is
the typical kind of thing
that a biologist
does in his lab,
doesn't think too much
more other than the system
in which we're looking at until
the COVID pandemic came about.
And I say that because during
the COVID pandemic, as you know,
many additional
intensive care units
were opened across the city.
All of us that are intensive
care physicians got
recruited to take care
of those diseases.
And it was clear
to us that this was
a very different type of
respiratory failure caused
by this virus.
It was not the same
respiratory failure
that we saw in patients
with pneumonia or patients
with sepsis that
Kate alluded to.
This was a disease where
the mechanics of the lung
were completely normal.
But for some reason, the lungs
were unable to transfer oxygen
into the blood.
And that turned out to be due
to a blood clotting problem that
was induced by the virus.
In fact, what the
virus was doing
was activating the
immune system and causing
microvascular
thrombosis, something
that's been referred
to as immunothrombosis.
Well, this turned out to
be exactly the process
that we had been studying in
a completely separate context
related to cancer
and tissue trauma.
And this ultimately then led
us to use clot-busting drugs
like alteplase, a drug that
we use for strokes and heart
attacks, to treat
these patients who
had COVID-induced
respiratory failure.
That was the STARS trial that
grew out of the basic research
here at MIT.
Let me spend a
minute and tell you
how a failed
operation completely
changed my view of inflammation,
tissue injury, and cancer.
In 1987, when I was
a medical student,
I was invited to
participate, which
really means hold retractors,
in a case where a patient had
pancreatic cancer.
And in those days in 1987,
we didn't have the imaging
that we have now.
And every patient who
had pancreatic cancer
went to the operating room, and
we would make a big incision
and take a look and make
a decision based on what
we saw about what to do.
And this operation,
unfortunately,
was what we called
a "peek and shriek."
We made a big incision.
We took a look.
There was cancer everywhere.
It was obvious there was no
benefit to operating further
on this patient.
And so we simply
closed the incision
and told the patient
to go home and enjoy
the last few weeks of her life.
Two years later,
when I was an intern,
this patient reappeared in the
emergency room with a bowel
obstruction, two years later.
Now, no one wanted
to operate on her.
But eventually, we had to.
And when we operated on
her, her bowel obstruction
was due to a single
adhesive band.
There was not a spot of cancer
left anywhere inside her body.
And this idea that injury and
trauma combined with live tumor
cells because we hadn't
resected any of the tumor
could somehow activate the
immune system and result
in a complete cure has motivated
a fair amount of our research
ever since.
And here's a paper that we
published just two years ago
that shows how the injury
response in live tumor cells
can promote anti-tumor immunity.
And I think that this ability to
go back and forth between what
we see in the clinic and
what we see in the laboratory
is going to lead
to new approaches
where we can combine surgery,
chemotherapy, and immunotherapy
more effective.
Now, of course,
we're always taught
as physicians above
all else, do no harm.
And so this has
also made me wonder,
maybe the tissue damage and
inflammation and wound healing
response from surgery is making
some of my patients do worse.
Perhaps, I'm adding
fuel to their cancers.
What determines if surgery or
injury or inflammation results
in cancer cure or
cancer progression?
We don't know.
But those are the
types of things
that we should be working on.
The second take-home message
I want to leave you with
is we need to think about
diseases and treatments
not in terms of single genes
but in terms of cell networks.
Now, some of you may
have a cardiologist,
and maybe you have
a pulmonologist,
and maybe you also have
a gastroenterologist.
And you know how
frustrating it is
when you go to see
those physicians
because your
cardiologist only wants
to treat your heart problem.
And he doesn't understand or
doesn't think about the fact
that when you treat
the heart, you also
affect the kidneys in the lungs
and the brain and everything
else.
And as an intensive
care physician,
I don't have the luxury to
treat just one organ system.
I have to think
about my patients
in terms of how all those
organ systems work together.
The same is true moving forward
with how we think about cancer.
We can't think of cancer as
just a mutation in an oncogene
because the cancer
cells are living
surrounded in their
microenvironment by other cells,
including normal--
and this is colon
cancer shown here--
including normal cells and
additional stromal cells.
And the same is true
with genes in cancer.
We can't think of a particular
gene as being mutated.
We have to think about
how that gene is working
in the setting of a network.
And this leads to a
new concept, which
I'm calling network medicine.
And that's because the
connection between genes
and proteins is different
in a disease state
than it is in a
non-disease state.
And this gives us
a new way to be
able to use
disease-specific drug
combinations because if
certain nodes in this network
are connected together
in a disease state,
I could use two drugs that
target two different nodes.
And those drugs would be
particularly effective in cells
in which this network
has been rewired,
whereas it would be much less
effective in the normal state
where only one of
those two nodes
is functionally wired together.
In our own laboratory,
we've discovered
a way in which this works.
We found a
vulnerability that seems
to be cancer specific
that relates to how
cancer cells undergo division.
And this has now led to
two clinical trials and one
proposed.
One clinical trial uses
two different drugs
to treat castrate-resistant
prostate cancer.
Another uses two different
drugs to treat metastatic colon
cancers that contain
a KRAS mutation.
And a proposed
trial is now going
to look at using two
particular drugs in a way
that we think will
dramatically benefit patients
with ovarian cancer.
And the very third point
that I want to leave you with
is the fact that many of the
drugs and treatments we now use
work differently than we
previously have thought.
Now, oftentimes, when
we prescribe a drug,
we're fully aware that in
addition to the way this drug we
think works, there are a variety
of side effects of that drug.
But sometimes, the side effects
could actually be the mechanism.
And as an example of this, we
were working on a drug called
5-fluorouracil.
This is a drug that's
been around for 40 years.
It is a standard of care that is
used to treat advanced-stage GI
cancers like pancreatic
cancer and especially
colorectal cancer.
And in fact, the
textbooks will all
tell you that this drug works
very effectively because it
blocks DNA synthesis.
It targets this enzyme
called thymidylate synthase.
And that's how this drug works.
There's just one
problem with this.
And that's if you look
objectively at patients
who respond to the drug or don't
respond to the drug, responders
in red, non-responders
in blue, and you
look at the level at which
this target is expressed,
there's absolutely
no difference.
How can the drug
work in some people
and not work in other people
if the target of the drug
is equally expressed in both?
And I'll simply cut
to the chase and say,
a few months ago, we were
able to publish a paper
and show that, in fact, that's
because the drug does not
work this way.
The drug works effectively
in cancer patients
because of a side effect.
The drug gets incorporated into
RNA, what Phil Sharp talked
about this morning.
And by being
incorporated into RNA,
it blocks the production
of functional ribosomes,
the machinery that
makes proteins.
And these functional
ribosomes turn out
to be essential in
certain GI cancers.
And that's the reason that this
drug works particularly well.
Now, my friends in
computer science
like to say it's not
a bug, it's a feature.
And the same is
true with medicines.
Sometimes, the side
effects aren't bugs.
Sometimes, they're
actually the mechanisms
of how the drugs work.
So I hope I've left you with
these three take-home messages
for how I think MIT as part
of HEAL can move forward.
We have to have
better collaborations,
active collaborations
with clinicians.
We have to start
thinking about diseases
in terms of network medicine.
And we have to recognize
and explore the fact
that many drugs we use probably
work differently than we think.
Thank you very much.
[APPLAUSE]
All right, so we are
going to sit on the stage
and do Q&A. I've got a
couple of questions teed up,
but I'd rather hear from you.
So I'm going to start with
one to get us started.
And we're all going to answer.
But I really would hope we
can hear from all of you.
So if you have a question,
you can come on up
and stand at the mics.
And I'll call on you if there
people that are interested.
But let me start
with one question.
And Mike, you sort
of hit on this,
but I'll let you flesh
it out a little bit more,
which is what makes MIT such a
unique place to do what you do?
I think the thing that's
so special about MIT
is how incredibly smart
all of our colleagues
are, the students, the postdocs,
the staff physicians, the staff
scientists and engineers,
and particularly
all of the other faculty.
Really, a great piece of
advice I got from Phil Sharp
was if you really
want to be successful,
just surround yourself
with smart people.
And you can't find that
anywhere at the density
that you can find
them here at MIT.
Kate?
For me, as someone
who studies AI,
MIT is just an
embarrassment of riches.
Literally, in every department
across the institute,
we have experts studying AI
development and implementation.
And that's great
because it creates
this ecosystem of researchers
not only within MIT.
But because of conferences
and speaker series,
it means we get lots of
industry practitioners
and other researchers
from around the globe
to work with on the problem.
Yeah, I'll just echo this.
When it comes to
innovation, it's
that levels of expertise at
such world-class level across so
many different domains,
they work together
to make those innovations
first get developed
and then get implemented.
That's pretty unique.
I think, simply, MIT is
just a no-bullshit place.
I mean, basically, you can't
do good work of the type we
talked about unless
you're incredibly
rigorous and self-critical.
And that happens more at
MIT than any place I know.
That's why our students
sometimes are unhappy.
And basically, we are just a
place that believes enormously
in scientific rigor.
You saw it in
Joe's presentation,
the willingness to take
on strongly held beliefs
and challenge them.
And you saw in the
incredible science
that's being done up here that
MIT is just the perfect place
to be working in areas where
lives depend on it because we
don't get so full of our
beliefs that we're not
willing to challenge them
and take them seriously.
So I would add that as well.
It seems like we might
have some questions.
Please, go ahead.
Sure, [INAUDIBLE] Walter
Bender [AUDIO OUT]
Scream it.
I'll scream it, and
then you can echo it.
Yeah, no, we hear you now.
Oh, OK.
My question has to do--
let me make an
assertion first, which
is scientists
write to scientists
or write for scientists.
When I open up a
nature journal article,
in order to understand
that article,
I've got to have
some background.
And I noticed that a lot of
you and a lot of the speakers
today have behind their
name, MD, comma, PhD.
So when you talk
about communication
between the scientific community
here at MIT and the clinicians,
the practitioners across
the river, the ones that
don't have that PhD
as well as the MD,
how are we going to bridge
that communication gap?
Well, Michael, you're a natural
person to start on this.
I think that's a
brilliant question,
and I don't have an
easy answer for it.
But something that I've been
arguing for for a long time
is we need to embed
PhD scientists
for at least internships or
short visits into the clinic
itself.
The only way that
people are going
to learn to speak
the same language
is if they live in
the same community.
And so I think some interchange
in which basic scientists spend
some time with
clinicians and clinicians
have the opportunity to spend
some time in our laboratories
here at MIT--
in fact, one of my colleagues,
one of my surgical colleagues,
is sitting here in the third
row because he's decided
that he wants to understand.
He wants to get a
deeper understanding
of health economics.
And I think that ability
to go back and forth
and to embed in each
other's cultures
is the only way we're
going to make progress.
Kate, Joe?
Yeah, I mean, I guess
I would just add,
so I'm an ethnographer, which
means I'm an anthropologist.
So I train my students to go do
exactly this where they embed.
But I think one other
thing we try and teach
them is to really
understand what
are the problems of the people
in the setting, not problems
that you come in with, what
problems that they're really
dealing with because
that's where the most
interesting work happens.
I'll just say really briefly,
at the Poverty Action Lab
here at MIT, we put out a lot
of policy insights and things
that are geared toward policy
makers as well as practitioners,
and something that you guys
can look up after today.
Yeah, I just want to add.
The question was talking about
clinicians and researchers,
but let's extend that
further, which is let's
talk about Kate's work and
about how AI needs to speak
to the people using it.
And we need to be able to--
and this is something I
think MIT needs to work on.
I think that's why I'm glad the
School of Humanities and Social
Sciences is involved with
this health and life sciences
collaborative, which we need
to get better communication.
We need to just not
create the new science.
We need to explain to
people why it's important
and what we do is important,
how it can change lives.
And I think that
communication needs
to go not just PhD to MD but
PhD and MD all the way down
to people who don't have a
college degree so that everyone
can really understand why we
do and why it's important.
Yeah, please?
Yeah, thank you all so much
for the insightful comments
and for the wonderful
work you do.
I was curious about this
conflict between AI and autonomy
because I think there have
been some studies published
that show that these AI models
are able to, for example,
diagnose diseases better
than a radiologist.
But also, there is this human--
as radiologists, you are
trained to do the same job.
So I was curious
about, should AI
be taking over the
role of radiologists
if they are able
to predict better?
And sort of, where do
we draw the balance?
I guess I can start with that.
So what we know from other
technological revolutions
is what we see is that some
job categories get completely
displaced and that humans learn
to do higher-order thinking,
higher-order tasks.
And so what's going to
happen with AI, I think,
depends on how fast the
technology progresses
and what frictions there
are for deployment.
But one thing, for sure,
is that in the near term,
many AI solutions are going
to require human and AI
working together.
And so what we're going to see
is role reconfiguration where
humans used to do everything.
Now, they need to figure out
what is the AI best at versus
what are they best at.
Let me just add, the Economics
Department and the Sloan School
were blessed with the
recent Nobel Prize
jointly in economics.
And the two people won the Nobel
Prize, Daron Acemoglu and Simon
Johnson, wrote a book called
Power and Progress, which
is very much about this history
and exactly this set of issues.
So I urge people to
take a look at that.
Yeah?
Obviously, from
your presentations,
health care is a
very complex area.
There's so many
different stakeholders,
and there's
competing incentives.
It seems like health
care, we're very reactive.
We wait for all these
trends and these diseases.
Someone has cancer,
obesity, diabetes.
There's cost disparities.
There's an aging population.
Has there been any efforts
to look at health care
at time of birth?
I know it's the long term, and
we're very reactive society.
I want instant
gratification, and we're
dealing with the crisis
when the thing is burning.
But is there any
analysis or any appetite
for dealing with health
care at an early stage
so we're not dealing with such
a snowballing crisis effect?
Michael, do you want to--
I would say the
extent to which we've
been able to address
health care at birth
is really through
early disease testing.
So there's some
immune deficiencies
that are now standard
of care in all newborns
so that we can
diagnose those early
rather than wait
for those patients
to then develop some
immune-related disease, at which
point, it becomes
much more expensive.
Now, as genomics becomes
more personalized,
I think we will be in a better
position to at least risk assess
and decide who should get
a test and who shouldn't.
As you're probably
aware, we're all
in this dilemma about
Prostate-Specific Antigen, PSA.
Should we use it?
Should we not use it?
And it only works if the
prevalence of the disease
is high.
So I do think the
place this will
be helpful is once personal
genomics really catches on,
we should be able to risk
stratify and better decide what
tests people should get when.
But I can't tell you about
the economics of that.
Well, when we think about the
incentives of the repair shop
gets paid, there's going
to be a lot of focus
on repairing as opposed to if
the prevention shop gets paid,
there'll be more
focus on prevention.
And so the move
in the US has been
toward trying to put
more risk on providers
so that if they get
people healthier,
they would actually benefit
from that as opposed
to welcoming them
back through the door.
And that's a slow-moving process
because of all the stakeholders.
But it's, I think, slowly but
surely going that direction.
What do you think?
One of my colleagues,
Nathan Hendron,
is running something called
the Policy Opportunities Lab.
And what this lab--
Policy Impacts Lab, I'm sorry.
What this lab does is actually
quantify the social value
of different interventions.
And the ones that
are most valuable
are the ones on kids, things
that intervene at birth
and early age are incredibly
socially valuable.
And I think we do
need to be directing
our focus in that area.
Joe talked about some
research on low birth weight,
which does that.
I talked about cell
and gene therapies,
which are really for these
birth defect diseases.
I think we need to be shifting
our focus that way, for sure.
Yeah?
Building up on a
comment, building up
on the comment on the
stakeholders misalignment
and the reimbursement
model, we have
been talking about value-based
care for a long time now
and have not been able
really to accomplish,
at least in oncology,
very significant progress.
Where do you see is the best
way to try to accelerate that?
I'm sorry, I didn't
quite understand.
The value-based care model.
Oh, value-based payment.
Yeah, so basically, this is a
really interesting and important
issue.
Like I said, we've
got these drugs that
look incredibly good
in early trials,
but we're not sure
how well they'll do.
And I think that, actually,
the government has recently
put together an initiative
on this for sickle cell.
There's new genetic
treatments for sickle cell.
And the government
just today announced
that they've got
a new initiative
to work with states to set
up essentially a rebate
system where if the
drug doesn't work,
the drug companies will rebate
the money that the states paid
in.
So I think we're making
good progress there.
But I also think
it's been somewhat
of a failure of-- the
economist said, like I said,
just change the incentives,
and value-based payment
will take care of itself.
And it turns out
it's incredibly hard.
You have to put a lot of work
into it, and that takes time.
So it wasn't going
to happen overnight.
I don't think-- people
who knew this area knew
it wouldn't happen overnight.
But I did think it was going
to happen faster than it has
because I thought the providers
would know where the waste is
a bit more than the payers did.
And it turns out, we all
need to figure it out.
But changing the incentives is
the first step, not the last.
Last question.
Hey, sorry, this is kind
of a technical question,
and we've got a
broad panel today.
But still, I'd like to ask it.
The sobering statistic
that I've come across
in a couple of different places,
including with Michael's lab,
is that maybe only about
a third of patients
benefit from putting
all of our best data
together for
personalized medicine,
so in the context of the best
care we can bring to patients.
And as Michael
knows, I'm sort of
betting on the computational
biology AI racehorse
to really help us there.
But that's only a piece
of the picture, I think.
So from a collaboration
standpoint,
I just wonder if the rest
of you, and together,
Michael, just have some
initial ideas about how
we may make progress through
collaboration on this really,
again, very sobering fact
that all of our best efforts
with all this great
technology, it's
only helping a small portion
of our patients right now?
All the omics, proteomics,
genomics, all the rest of it
combined, it's not getting
us nearly as far as we
need to as fast as we want to.
So I'd like your
thoughts on that.
So I think part of the reason--
so what Scott Ritter Bush is
talking about is the fact that
at least using genomics,
right-- genomics hasn't really--
only 10% or so of
cancer patients
really benefit from genomic
analysis of their tumor.
Only 10% or less actually
have a change in therapy
that results in improvement.
And part of that, though,
is because we've been
so heavily focused on genomics.
So I do think just integrating
in in a patient-specific manner
other types of omic analysis,
looking at metabolism,
looking at immune infiltration,
looking at proteomics-- now,
that's harder to do
because we have not
evolved the technology
that allows us to do this
in a rapid, high-throughput
way on patient samples
like we can do with genomics.
That's exactly the kind
of thing we can do at MIT.
But the other part of it is
we can't do it in isolation.
We have to do this
together with looking
at patient historical records,
risk assessment, family history.
And that's even
harder to incorporate
into that type of thing.
Yeah, I would just add-- it's a
great question, great comment.
I would just add,
as a society, we
play a role in directing
technological advance.
And I think we need to be
thinking about directing
technological advance to
things which are helping
a narrow slice of people
very well versus a broader
set of people maybe through
birth improvements less well.
And that's the kind of
trade-off that economists
love to think about and
that we'll be working on.
So I think we just stop there.
Thank you very much for
your comments and questions.
[APPLAUSE]

---

### MIT HEALS Launch: Immunology breakout session
URL: https://www.youtube.com/watch?v=V-XjgwVENq0

Idioma: en

All right, with that,
I'm going to get started.
Welcome to the immunology
breakout session.
My name is Stephanie
Springer, and it's
my pleasure to chair
this session here
on immunology at MIT.
What is immunology and
why do we care about it
in health and disease?
So with that, I would like to
give a very brief introduction
on what the immune system
is and why we care about it
when we think about physiology
as well as disease states.
The immune system, and
this is a slide from class,
can be very simply
depicted as a seesaw.
And this is a very
simple oversimplification
where the immune
system is trying
to balance fighting
pathogens, but also avoiding
to attack the body itself.
However, as I just said,
this is a very drastic
oversimplification of what our
immune system is actually doing.
So this seesaw is
probably more accurate
because we are not
living in isolation
and our immune system is
not acting in isolation.
Lifestyle, diet, stress,
all of these factors
affect how our immune
system is working,
and diseases push our immune
system every day to react,
and our immune system
still has to keep
at bay not attacking ourselves.
And an imbalance within
the immune system
can cause diseases,
but likewise,
allow the resolution of diseases
like pathogenic infections.
And it can also affect our
health and our well-being,
and likewise, our
well-being, being happy,
will affect our immune system.
Now, this is a human
body depicted as dots.
One dot is 10 to the
power of 9 cells,
and one large square
is 1,000 dots.
Now, how big is
our immune system
if this is our entire body?
You see here in red
and in dark red,
the two major subtypes of
cells that make up our body.
The light red cells
are erythrocytes.
These are bone
marrow-derived cells
that carry around our
oxygen. They also have
some level of immune function.
The dark colored cells
are all bone marrow
derived immune cells.
So these are the cells that
fight pathogens, fight cancer,
but also are responsible for
inducing autoimmune diseases.
And as you can see here, in
contrast to the other organs,
our immune system
is, by and large,
the largest organ in our body.
It is likewise not a
single solid organ.
It is disseminated
throughout our body.
It has a network of nodes,
hubs, meeting points
called lymph nodes that
are spread throughout,
and this is where our
immune cells actually
communicate with one another.
And our immune system can
be very broadly subdivided
into different elements.
We have the innate
immune system,
and this is the part of the
immune system that senses
if there's danger somewhere.
And then, we have the adaptive
immune system in, specifically,
two arms, the cellular
arm of the immune system
and the humoral arm of the
immune system mediated by T
cells and B cells respectively.
And these are the cells
that instill memory
in our immune system.
Meaning, if you get a
prophylactic vaccine
against flu, these
are the cells that
memorize that they've
seen flu before
and are better at eliminating
the virus once you see it.
These two arms of the
adaptive immune system
act slightly differently.
The T cells recognize
specific patterns on cells
and eliminate those cells
versus the humoral arm
is mediated mostly
by antibodies.
And throughout this
session, we will
hear from experts across
all of these areas
of our immune system.
Specifically, Facundo
will talk to us
about the humoral immune system.
Aroup, later on, will talk
about B cell responses.
Jessica will talk about
the innate immune system
and how patterns are hardwired.
Hari will put many
of these contexts
into, one, how cells
interact with one another.
And Michael, at the
very end of the session,
will talk to us how we can
use those basic principles
and use it to engineer
our immune system.
This is the summary of
these five professors
titles, which I won't
repeat for time constraints.
And I'm really excited
about this session.
And with that, I
would like to welcome
Facundo Batista as our
first speaker to the stage.
[APPLAUSE]
Yes, good morning, everyone.
And as Stephanie said, my
name is Facundo and yeah,
there we go.
And that is a typical
Argentinian name.
And that is why I am going to be
speaking with his strange accent
all the way.
So when people is
asked about what
has been the best biotechnology
discovery in history,
many will argue that it
will be wine or beer.
I, and many others, will argue
that it is going to be vaccines.
And essentially, vaccines
has been the source
of protection over the
last just five decades
and saving lives for about
150 million human beings.
And although we know
that vaccines work,
we don't know how they work.
But we know that these
molecules, the antibodies that
are able to bind pathogens
and trigger function,
they are the main
responsibilities
of active protection or
a correlation with them.
And antibodies are
not responsible
of active protection,
but they are also
critical in terms
of treating disease.
And I am using this example in--
just to point out that
in giving an antibody
to a person, an antibody that
recognizes a malaria parasite,
it will protect
that person for one
year just by injecting
the antibody into them.
And these antibodies
have been the basis
of many therapeutic cures to
cancer and infectious disease,
and as such, it makes them a
key target for bioindustry.
And the industry of
antibodies is, at the moment,
in the order of
$300 million a year
and is estimated
to be in one dollar
billion in about 10 years.
That is why we are particularly
interested in trying
to understand how these
antibodies are generated.
And in these regards,
antibodies are
produced by a particular type
of cells that are plasma cells.
And these cells are
really antibody factories
that they produce
millions of molecules
per hour every single day.
But plasma cells are the
result of a particular pattern
of differentiation of B cells,
and antibodies initially
are produced on the
surface of naive B cells.
And those cells are
able to recognize
either viruses or either
antigens that they
are given through vaccination.
And once they do that, they
will do something rather unique.
They will get together
and start to divide
and they will form these
that are in structures that
are known as germinal centers.
And here, they will
do something magic.
What they are going
to be doing is
they are going to mutate
the original gene that
carries that protein in order
to make that antibody much more
potent over time.
And this antibody will
become much more specific
for a particular
antigen or a tumor.
And the way that this
happens is that, initially,
when we get vaccinated
or infection,
we produce very low affinity
or not high quality antibodies.
But then, because of the
germinal center formation
and this specific
maturation that I mentioned,
the antibodies gain high
affinity and specificity
over the differentiation
of the cells that
then relies into the long-term
survival of this antibody
production.
And just to give you
an idea, we are all
being vaccinated
for measles here.
And those antibodies that
are-- those plasma cells that
are generated in the very
early stage of our life,
they will live with us for about
80 years until we pass away.
So my lab, what we do is we
ask fundamental questions
about how B cells, biology,
and antibody responses happen.
And indeed, I will
show some examples
on one particular
discovery that we
did that really kind of
enabled us to do really
absolutely cutting edge.
What we have been
doing over the years
is to manipulate or try to
create a humanized mouse.
In this mouse, what
we do is we change
the genetic code in
different chromosomes
to now express B-cell
receptors, or antibody
on the surface of naive cells,
that are not any longer mouse,
but they are humanized.
And as such, we can study the
immune resposne of these mice
as these B cells will
be practically human.
So we use this technology
to develop and unravel
several key questions in
terms of B-cell maturation
and biology.
And one of them was,
for example, this
is how a plasma cell looks.
And as I told you,
this plasma cell
lived with a very long
time in our models.
And these plasma cells
are able to produce
about 100 million molecules
of antibody per cell per hour.
And they need to produce
that over the course
of many, many years.
And in order to do this, what
we show, in particular, a PhD
student in the lab,
Stephanie, what is--
Sophie, what she has done
is she has particularly
looked into the code of
translation of that protein.
For those of you that
are not familiar,
proteins are
expressed as an RNA.
They are translated
by this adapter that
constitutes the amino acid, or
bring the amino acids to this
that they are the
shape of the protein.
And what Sophie found
was that these tRNAs
are critical in
programming the plasma
cell to produce this large
level of antibodies there.
We are not only look
into plasma cells,
we also look very much into what
antibodies do into our response.
And I know Arup is going to
talk a little more about this.
But essentially, what
we show is that we try
to represent this as a castle.
Germinal centers, that
I mentioned to you
before, they exist in the body.
And antibodies that
we produce, they
are able to block the
entrance of particular cells
or to facilitate the
entrance of others.
But we don't do
only B cell biology,
but we try to understand how
we can translate this B cell
biology and knowing
in antibody responses
to try to inform
vaccine design and offer
novel therapeutic development.
And in this, I will give
two examples of those.
We have used this-- what we
call this mouse machine that I
mentioned to you, mouse
that produce antibodies
that look like humans.
And what we have shown through
immunizing or vaccinating
those mice that we can produce
much more potent antibodies
that they can now
protect us to malaria.
And this is of
particular function
because-- a particularly
important one
because if we have
an antibody that
works much better
than another, that
decrease the cost of the drug
by several orders of magnitude.
Similarly, we have been
involved in a large consortia
to try to use this humanized
mouse model to characterize
different antigens.
And this is the
surface of the HIV.
And these are different
antibodies bind
to different portions of HIV.
And in our collaboration,
what we have been using
is our humanized mouse models
to enable clinical trials
for the production
on an HIV vaccine.
And in that regards,
in particular,
we have taken these
immunogens to the clinic.
One clinical trial is going
to start in next January,
and this relates, in particular,
to this portion of the protein
and the production of
antibodies against that protein
or another ongoing clinical
trial together in collaboration
with Moderna that
is particularly
focusing in how we
can develop antibodies
that recognize this part of
the protein of the virus.
So with that in
mind, actually, I
would like to introduce
the next speaker that
is Arup Chakraborty, and
he will be digging more
into the B-cell response
and the biology of it.
[APPLAUSE]
Thank you, Facundo.
So as Facundo mentioned,
it is estimated
that vaccination has saved about
150 million lives in the past 50
years.
But we don't have
effective vaccines
against highly mutable viruses.
For example, we don't have a
universal vaccine for the flu.
So every year we guess which
strains may evolve and design
a new vaccine.
And the situation is similar
now for SARS-COV-2, which
causes the COVID-19 disease.
The way to better confront
these scourges this
is to design universal or
variant proof vaccines.
And a part of our
work is focused
on trying to address fundamental
scientific questions which,
if answered, would aid
the pursuit of this quest.
And toward this end,
we bring together
physics-based
computations and machine
learning along with trust
tests of predictions in animal
models and clinical data.
These latter efforts are
carried out in collaboration
with basic and
clinical immunologists.
And for the vignettes that
I shall highlight today,
the labs--
the collaborating labs were
led by these individuals.
Now, as Facundo mentioned,
upon infection or vaccination,
antibodies are generated by a
Darwinian evolutionary process
that occurs in germinal
centers, or GCs.
Herein, B cells evolve
by mutation and selection
to bind more strongly
to the antigen, which
is displayed in these GCs.
And that is why, as time ensues,
the memory cells and plasma
cells and antibodies that you
generate become more potent.
Now, upon re-exposure
to the antigen,
the existing memory cells
are first rapidly expanded
to create a wave
of antibodies that
is meant to confer protection.
But new GCs over formed--
also form, but over
much longer time
scales to generate more
memory B cells and antibodies.
Now, the first vignette
that I will describe
stemmed from our asking why
three shots of the COVID
vaccines protect better against
the highly mutated variant
like Omicron, while two shots do
not even though all three shots
encode for the Wuhan
original strain of the virus.
So this Darwinian
process is driven
by antigen displayed in GCs.
How does it get there?
Circulating antibodies
bind to the antigen
and deposit it there.
Now, when you get
your first shot,
all you have in you circulating
are generic, weakly binding
small numbers of antibodies
that bind to very little antigen
and deposit them there
before the antigen degrades.
And you can see this by
imaging lymph nodes of monkeys
after the first shot.
And here you see there's
hardly any antigen there.
So if you think of the antigen
as the prey and the B cells
as the predators in this
Darwinian process, when there's
very little prey, only the best
predators are going to survive.
And that is why,
after the first shot,
the antibodies and memory
cells that you generate
are those that
initially bind strongly
to the spike of
the virus and bind
to regions called
immunodominant.
And here I have depicted
these antibodies
binding to the
spike of SARS-COV-2.
Now, after the second shot,
the corresponding memory cells
are rapidly expanded
to generate antibodies.
Now, the regions
that are in red,
they are mutated in Omicron
compared to the Wuhan strain.
And you see they
are in the binding
footprints of these antibodies.
And so they're not very
effective against Omicron.
But after the second shot
you also form new GCs,
but now you've got
circulating antibodies
that are antigen specific.
And so they bind to and deposit
lots of antigen in the GCs,
and you can see this by
imaging the monkey lymph
nodes after the second dose.
You see there are
lots of antigen.
Now, when there's lots of
prey, they're not so good.
Predators can also
survive and evolve.
And so antibodies that bind
more weakly initially to regions
called subdominant that are not
mutated in Omicron can evolve.
And this effect is
further amplified
because these
antibodies that bind
to these immunodominant regions
can enter GCs, bind to them,
thereby masking
them and effectively
elevating the amount of these
subdominant regions that
are present.
So now the memory cells that
bind to these regions that
are not mutated in Omicron
that you've created
are rapidly expanded
and become antibodies
that are protected
against SARS-COV-2--
against the Omicron strain.
Similar mechanisms are at
play for our demonstration
that instead of giving
the vaccine, as just one
shot as a bolus, as
is normally done,
if you distribute the
same amount of antigen
over seven days in seven doses
over an escalating pattern, then
you get lots more antigen than
just the bolus, which leads
to a more potent antibody
response compared
to the bolus in mice.
And very recently, this
has been recapitulated
in a human clinical trial.
So from computational biophysics
to the clinic, if you like.
But seven days, seven
shots is not practical.
And so by better understanding
these feedback loops,
we have now designed
a two-shot regime that
puts a lot of antigen
on the surface in the GC
and you get potent responses,
just like the seven day one.
We are now trying to explore
how similar mechanisms can
be leveraged such that antibody
therapy can potentially
lead to a cure for HIV.
And finally, let me note
that we have computationally
designed nanoparticles
that display spikes
of diverse sarbecoviruses,
the virus family from which
SARS-COV-2 is derived.
And this is the potency
of the antibody response
in mice to several
sarbecoviruses
that may spill
over from animals.
And you see that our
computationally designed one
is more potent than the
state of the art nanoparticle
and much better than the
bivalent vaccine that many of us
have taken.
Thank you for listening.
And our next speaker will
be Professor Jessica Stark.
[APPLAUSE]
Thanks and good
morning, everyone.
It's a privilege to be here and
to share our group's perspective
on understanding and
engineering immune responses.
I want to start by first
affirming your choice
to come to this breakout
session because we really
are in a very exciting
time for immunology
in the midst of an immunological
revolution in medicine.
We now understand that there are
a wide variety of human diseases
and conditions that can
be effectively treated
by harnessing the immune system.
As we've heard from
Arup and Facundo,
we've been able to use
vaccination to prevent or even
eliminate infectious
disease by teaching
the immune system to recognize
and respond to pathogens.
In cancer, we've
been able to use
immunotherapies to elicit
decades long remission
from disease.
Really, the closest we've
come to a cancer cure.
And nevertheless,
we remain limited
in the kinds of conditions
that we can treat
and the number of patients that
we can help with immunotherapy.
And so we still
need new paradigms
to harness the immune
system if we're
going to realize the full
potential of immunology
for human health.
We see an opportunity to target
cell surface sugars called
glycans as an untapped axis of
regulation in the immune system.
Every cell in your body, and
in fact, every cell on Earth
is coated with a thick
layer of glycans,
and these glycans exert
influence in nearly
every immunological process.
Glycans can directly control
immune cell function.
They can activate or
inhibit immune cells
by binding to receptors
on immune cell surfaces.
They control how immune cells
move throughout the body.
They can help immune cells
exit the blood and home
to sites of inflammation
or infection in tissues,
helping your immune system
locate and eradicate disease
throughout the body.
Glycans play important
roles in innate immunity,
where they help your
immune system do
one of its most important jobs,
which is to distinguish self
and non-self, telling
the difference
between host and pathogen.
And they can further influence
adaptive immune responses,
for example, by
helping determine
whether an antibody will
elicit a pro-inflammatory
or an anti-inflammatory
immune program.
Despite all of these
examples, however,
of how glycans play important
roles in the immune system,
we have lacked the
tools to comprehensively
study this biology and to
target glycans therapeutically.
And these are the
gaps that my lab,
which opened here at MIT about
a year ago, seek to address.
We are building
platform technologies
to help us fully understand
how glycans regulate
immune responses,
and to translate
these biological
insights into the design
of new kinds of immunotherapy.
I want to share
an example of how
we've been able to target
glycans for cancer immunotherapy
that builds upon recent
fundamental and translational
advances in the field.
In order to grow and
spread throughout the body,
cancer cells upregulate
molecules on their cell surfaces
that bind to inhibitory
receptors on immune cells.
By engaging these so-called
immune checkpoints,
the cancer cells can put the
brakes on the immune response,
allowing them to evade
normal immune surveillance.
However, if you block
these immune checkpoints
with something like a
monoclonal antibody,
we can-- there's the potential
to reprogram and reactivate
the immune system to sense
and respond to cancer.
These checkpoint
blockade immunotherapies
were tested initially
in metastatic melanoma,
an incredibly aggressive
form of cancer.
And in one of these
trials, patients
who received a combination of
therapies targeting the PD1
and CTLA 4 immune
checkpoints, about 30% of them
mounted a robust immune
response to their cancer
that basically stopped
the disease in its tracks.
These patients diseases
had not progressed
in over a decade following
the start of treatment.
And in light of the
success in melanoma,
checkpoint blockade
immunotherapies
have now been tested in a
variety of other cancers
and are now approved to treat
about 50% of cancer patients.
However, across all cancer
patients, only about 20%
benefit from these therapies.
And so now the question is, why?
How are cancer cells able
to evade immune surveillance
and resist immunotherapy
in most cancer patients?
And that's where glycans
come back into the story.
It turns out that cancer cells
dramatically alter their cell
surface glycosylation.
This is a hallmark
of cancer that's
been observed in many
different tumor types
and across many different
patient cohorts.
And we now understand that
one of the reasons that cancer
cells remodel their
surfaces in this way
is to engage a class
of inhibitory receptors
on immune cells that
are called lectins.
So when lectins bind to these
glycans on cancer cell surfaces,
they trigger inhibitory
signaling pathways
that shut down immune responses
through the exact same mechanism
that occurs when the established
immune checkpoint receptor, PD1,
binds to its ligand, PL1.
And while we have a litany of
drugs that target the PD1/ PDL1
immune checkpoint, we have no
approved therapies to target
these glyco immune checkpoints.
We recently developed
a new kind of molecule
that makes it possible
to target glycans
for cancer immunotherapy.
These molecules we're calling
antibody lectin chimeras
or AbLecs, and
they're bispecific.
One half of the molecule
is an antibody domain
that can target specific
cells in the body,
for example, cancer cells.
The other half of
the molecule is
what we call a decoy receptor.
It's the binding domain
from the lectin receptor
we want to target, which
serves to bind and block
the same kinds of glycans
that would otherwise
inhibit immune responses.
And so our hypothesis
was that with AbLecs,
we could block these
glyco immune checkpoints
and elicit more potent
anti-cancer immune responses.
We tested these
therapies initially
in vitro with primary human
immune cells like macrophages
and NK cells.
These are two different
kinds of innate immune cells
that we could imagine activating
with our AbLec therapy.
And you can see that for both
macrophages and NK cells,
treatment with our AbLecs
elicited more potent cancer
cell killing compared to a
monoclonal antibody control.
And in vivo, in a humanized
model of metastatic cancer,
we observed that animals treated
with our AbLec immunotherapy
exhibited reduced tumor burden
and compared to animals treated
with the control molecule.
With this proof of concept, we
launched a new biotechnology
startup called Velora
Therapeutics that aims
to bring AbLecs to patients.
The exciting thing about
the AbLecs platform
is that it's modular.
You can imagine mixing and
matching different antibody
and lectin domains
to enable a variety
of different therapeutic
mechanisms of action.
And Velora is leveraging
this modularity
to develop therapies--
AbLec immunotherapies
for indications
in oncology and immunology.
But this is just the beginning.
Glycans play pivotal
roles in the immune system
and are really attractive
therapeutic targets.
My group is developing
the platform technologies
we'll need to explore this
emerging biological frontier
and enable both fundamental
and translational advances.
So with that, I'd like
to thank my lab who
made this work possible.
I look forward to
taking your questions,
and I'd like to invite our next
speaker, Professor Hari Wong,
to the stage.
[APPLAUSE]
It's great to be here.
I want to talk to you
a little bit about some
of the complexities
that exist in studying
the immune response,
why they exist,
and how our group has been
developing technologies
and implementing technologies
to try to overcome them.
We spend a lot of time
thinking about how
the immune system operates
in its native environment.
So not in a dish, not in
blood, but actually where it
needs to function in tissues.
And the reason we do this
is at least threefold.
One is that the immune
system is a system of cells.
It's not one cell.
It's multiple cells.
And these cells typically form
multicellular units of control.
They assemble in
space and in time
into intricate structures that
do things that no one cell
type can do on its own.
You get emergent dynamics,
emergent functions,
and it's very hard to
reconstruct these units
in vitro settings.
The second point
I want to make is
that the immune cells
within these units
are communicating dynamically.
They're engaged in physical
interactions with one another.
They secrete molecules
over short distances
and over long distances.
And the communication is always
evolving over time and in space.
And lastly, the
point I want to make
is that the most interesting
aspects of the immune response,
at least in my mind,
both in physiological and
pathophysiological
settings, are not mediated
by the bulk of the cells.
They're typically mediated
by rare outlier cells
on the ends of the distribution.
I'm not sure how to go back.
There we go.
So let me walk you
through technologies
that my lab has been
developing to try
to overcome each and every
one of these three challenges.
We typically take an
imaging-based approach
to really be able
to resolve and study
these cells in their
native environment.
So what you're looking
at here on the left
is a lymph node that lies
downstream of the kidney.
I want to direct your attention
to this inset over here.
We've developed multiplexed
imaging technologies
using antibodies that now
allow us to visualize anywhere
between 20 and 60 proteins
directly in tissues where
the immune response
is occurring.
What you're looking at
here on the zoomed in inset
is a multicellular unit that's
controlling T cell responses.
You don't have to worry about
what all the colors are.
The point I want to make is that
there are innate immune cells
here, like Stephanie introduced
earlier, that are initializing
T cell responses.
There are these
green cells here.
These are regulatory T cells.
There's not one of them,
there's many of them.
These are cells that are trying
to dampen the immune response.
And there are conventional
CD4 and CD8 T cells here.
These are cells that
are trying to promote
inflammation and promote
the immune response.
And if you try to study any
one of these cell types alone,
you can't really figure
out what's going on.
It is the integrated
behavior of all three
of these cell types
in space and in time
that is predictive of how the
immune response will behave.
So we can address this.
We can identify the
units of control.
We can identify the
relevant cellular players.
What about the
communication aspect?
Well, we've now developed
ways to preserve tissues
in a manner that
allows us to estimate
length scales of
communications between cells.
Let me walk you
through what I mean.
Here, outlined in white,
these are two endogenously
activated conventional T cells.
They are producing a
molecule called IL2.
This is a cytokine
that's shown in red.
This cytokine will get
secreted and diffused in space.
These surrounding
cells in magenta,
these are regulatory T cells.
Like I said earlier,
these are cells
that try to dampen
the immune response.
And in green, we're reading out
signaling downstream of the IL2
receptor by measuring the
phosphorylation of STAT5,
a transcription factor.
If we look to the right, what I
want you to take away from this
is that there is a
gradient of green,
and it's largely in the
regulatory cells here.
And by reading
out this gradient,
we can infer how
far the cytokine
must be diffusing in space.
And this is important
because cells
that are positioned here
close to the producing cells
are going to see a very
different concentration
of this cytokine than cells that
are positioned way back here.
And therefore, cells that
are positioned closest
to the producing cells are
going to adopt a different fate
than cells that are
positioned way back here.
So this is a source
of variation,
how the immune system generates
heterogeneity in its responses
during a dynamic
immune response.
This is all well and
good, but it's fixed.
These are static images, and
the immune system is not static.
It's dynamic.
As I told you earlier,
communication is dynamic.
So we've also invested in
technologies that allow us
to peer inside of living
organisms-- this is a mouse--
and actually watch the
immune system in real time.
What we've done here
is a small surgery
in a mouse under anesthetic.
And we're looking
behind the knee.
It's the popliteal lymph node.
And we're peering in and we're
watching the immune response
in real time.
It doesn't really matter
what's going on here.
But the cells in magenta, these
are the regulatory T cells
that are dampening
the immune response.
They're not fixed.
They're zipping all over
the place in real time.
And the cells in magenta,
these are conventional cells--
conventional T cells
that are trying
to initiate an immune response.
So there's a
complex choreography
that's always going on in our
bodies at any moment in time.
Now, what about the
last point I raised,
how can we identify
the rare cells
that are doing the most
interesting biology?
The technologies I showed
you so far are fantastic,
but they're limited in the depth
of tissue that we can image.
We can't look at enough
volume with those methods.
So we have to take a
different approach.
And the approach we use is
an optical clearing method
that allows us to treat
our organs of interest
with a chemical concoction
that homogenizes
its refractive index.
What that means, practically,
is that it turns our tissues
from completely opaque
to clear and transparent.
And from a microscopy
perspective,
this allows us to now
image the entire organ
in three dimensional space
without any light scattering.
Here's an example of that.
This is a mouse uterus
that's been cleared.
And we're going to
zoom in the cyan,
these are the uterine glands.
And what we can find are
these rare clusters of T cells
here in magenta that are
wrapped around the glands doing
some type of important function.
That's very distinct from
the rest of the cells
in the entire organ.
These are the rare cells in
the tail of the distribution,
and they're
functioning as a unit.
But again, it's very different
behavior than the rest--
the bulk of the distribution.
So I've shown you
the technologies that
allow us to overcome
the challenges,
But, it's not good enough
for us to just take
pretty pictures and movies.
We need to combine these
experimental methods
with computational
approaches that
allow us to extract as
much useful information
from our movies and
images as possible.
So for that, we've invested in
computational algorithms that
allow us to define the
boundaries of individual cells
and reconstruct our
organs of interest
with all of their features
and protein expressions.
We can do this in the computer.
So what you're
looking at here is
a computational
reconstruction of a lymph node
where every dot represents
an individual cell.
And we can analyze
specific regions
of space, which is important
because different reactions
occur in different places.
We can use these
measurements to then conduct
a variety of quantitative types
of analyses, similar to what
you would do in data science.
We can take these
measurements and parameterize
mechanistic models using
systems of equations
to simulate how different
aspects of the immune response
will proceed.
We can make predictions.
And then, we can go back
in to model organisms
and test those predictions
using genetic or environmental
manipulations.
And through these
types of technologies,
I think, this is how we'll
get a better understanding
of fundamental operations of the
immune system and understanding
how different
therapeutics actually
work in their sites of action--
in their tissue sites
of action, that is.
So with that, thank you,
and I'm pleased to introduce
our next speaker,
Professor Michael Birnbaum.
[APPLAUSE]
All right, thanks to Hari, and
it's good to see you all today.
So in my lab, we want to know--
we want to be able to
see what T cells see.
So to give you a
graphic I pulled
from a recent review, when a T
cell is looking for cells that
are infected or otherwise
transformed by cancer,
the way that this
actually works is
that every T cell has its own
protein sensor called a T cell
receptor, which
takes recognition
of what we call an HLA,
or an MHC molecule, which
presents a short peptide.
And these MHC peptide
complexes present
a census in terms of what the
cell is expressing at any time.
So when a cell is expressing
just the normal proteins that
make a hepatocyte a hepatocyte
or a lung cell a lung cell,
then T cells leave the cell
to go about its business.
When a cell is
otherwise altered,
so whether it has an
intracellular bacteria infected
by a virus or transformed in
terms of becoming a cancer,
then the peptides are
different and the T cell
is able to recognize it
through the T cell receptor
and then kill the cell,
among other immune functions.
So for me, one of the
fascinating things about this
is that this actually serves as
numerically an interesting thing
to think about.
So in each one of
us, and my math
could be off by a factor
of 10 in any direction,
but there's roughly, say, around
50 to 100 million unique T cell
specificities floating
around in each one of you
at any given time
out of a possible 10
to the 15 possibilities for T
cells for just combinatorially
for the receptor.
And these are recognizing
these peptide MHC complexes
where in terms of what you
theoretically could have in you,
again, math is back
of the envelope,
it's something around a
trillion possibilities in terms
of peptide and MHC
combinations in any one of us.
They're not all present
at any given time,
but there's no way
for the immune system
to know which one of these you
may encounter during your life
and which ones are
purely theoretical.
This is all happening
in a system that
has to be incredibly sensitive--
because if you're infected,
you have to be able to recognize
it-- at what we would consider
incredibly low affinities.
So interactions that occur
at body temperature that
lasts for maybe about a second.
And if you do the bookkeeping,
it turns out that each one
of these T cell specificities
would have to be able
to recognize on the order of
100,000 different antigens
in order for the
books to balance here.
Otherwise, you could
have, on one hand,
for example, a pathogen, which
is able to completely evade
detection.
And somehow, on
the other hand, you
have to be able to
do this in a way
where you still have
enough specificity
that the recognition of a
virus doesn't inadvertently
lead to the recognition of any
of your own healthy tissues.
So your immune system
is doing this every day,
all day, for each of us, very
nearly perfectly all the time.
And I really want to know
how this is able to work,
and able to work in a way
that is so specific in order
to allow this to happen.
Understanding this
specificity is going
to help us understand biology.
It's going to help us create a
next generation of therapeutics.
And increasingly, especially
with the explosion
of progress in AI, the
ability to make better
computational tools.
So for all of this, though,
what we remain limited in
is the number of
examples which we've
been able to really observe.
So this is a field,
understanding
how T cell receptors
recognize peptidomimetics
that has really been in progress
as long as I've been alive.
But we've studied, really,
relatively small numbers
of examples.
So in order to be
able to make progress,
we're going to be able to--
need to be able to
look at much more data.
So in my lab, we want to
develop better tools in order
to be able to understand this.
And so we think quite
broadly about how to do this.
We use a variety of approaches.
I want to tell you just
for a couple of minutes
about one of them that
we've done in our lab.
And what we've actually done is
repurposed a common laboratory
tool, which itself is
a repurposed virus.
So we work in our lab
with lentiviruses,
which is a heavily
engineered for laboratory use
safe version of a virus
that started as HIV,
of all things, where we
re-engineered it to say, OK,
what if we take these
lentiviruses, which are commonly
engineered to infect any
cell that you might want
into in the lab, and have them
infect only T cells by putting
peptide-MHC molecules
on their surface?
So in this way, you
create a virus that
is specific for a given T cell.
And lentiviruses are
quite unique because what
they are able to do as
part of their life cycle
is leave a permanent
record of their infection
into a cell's genome by
incorporating the viral genome
into the host genome.
So our thought was, if we put
a peptide-MHC on the virus,
we could have a system where
we have infection mediated
by the molecule on the
outside with a record kept
by the molecule on the inside.
And if we're able to build
large collections or libraries
of these as a tool, which
is common in biotechnology,
we could use this as a recorder
of this peptide-MHC T cell
receptor interaction.
So we set about doing
this a few years ago.
And we were able to show
that this actually works.
So you could create a
virus with a peptide MHC.
If that virus
encounters any cell that
doesn't have the right
T cell specificity,
even one that's close
but not exactly right,
you see little to no infection,
so this is a virus encoding
cells to turn green--
very few green.
Where if you have a virus
that does have that match,
you get robust
infection of cells.
So we created a
tool, a method which
we call RAPTR, Receptor
Antigen Pairing by Targeted
Retroviruses, we were
able to create libraries
of these viruses, mix
them with T cells,
and then, allow for the
mixtures to only infect
the right combination.
So the mixture of viruses only
infecting the correct T cell.
We're able to then isolate them
and use single cell sequencing
in order to get
the information out
of what the T cell receptor
is and what the antigen
is on a many versus many basis.
So something that we
are trying to approach
the scales of how the immune
system itself would work.
And we were able to show
that this is able to work.
That you're able
to identify, even
in a complex mixture of viruses
with different peptides,
once you sequenced the viruses
that are able to infect,
you're able to get the
right answer back out.
So this is a tool we're
very excited about.
But a little
belatedly, we realized
that if you're able to
infect a cell in order
to deliver a genetic
cargo for something
like a fluorescent
protein or a barcode,
that you should
also be able to do
it to deliver some type of
therapeutic cargo as well.
And so we're in the process
of exploring this, too.
And the way that we're
thinking about this
is in terms of different
types of cell therapies.
So it turns out that
your immune system
can mount a defense against
tumors all on its own.
We call these tumor
infiltrating lymphocytes.
This is something that often
is not effective on its own,
but is something
where just recently,
in the past few
months, there's been
a clinically-approved therapy
where these cells are taken out
of the tumor, grown
back up, and reinfused.
So this is great because it's
recognizing exactly what happens
to be in your tumor.
But is torturously
complex to make.
Correspondingly, we're
able to engineer T cells
through something called
a CAR-T therapy where
we're able to point T cells
in the right direction.
But this is also complex
to make and only works
for certain cancers.
So what we've done in our lab
is created the same viral system
where now, instead of
delivering a barcode,
we're able to deliver
therapeutic cargoes to the T
cells.
So we're able to, even in
vivo, infuse a virus which
is able to find
the right T cell,
reprogram it genetically to
become more immuno activated
in one way or the other, and
then able to extend lifespan
in mice.
So we're in the process
of exploring both
of these aspects in our lab.
And this is where the
collaborative and big picture
thinking of MIT
really comes in handy.
So on one hand, we
are a part of and I'm
the leader of a team funded
through a mechanism called
a Cancer Grand
Challenge in order
to truly try to
generalize the predictions
of peptide-MHC molecules
and T cell receptors
to design these receptors.
So this is a team of
immunologists, clinicians,
engineers all working together.
So several of which
at MIT, but also
for individuals around
the world, to really try
to address one of these
large problems in immunology,
how can we make
these predictions?
And then, we're also working
to take our targeted viruses
and turned them into a
drug in a company called
Colonial Therapeutics where
we were able to show really
promising data that's
been talked about
at meetings, both in
terms of mice and monkeys,
with the hopes of starting a
clinical trial in the next year
or so.
So thanks for your attention.
I believe we're moving
to a Q&A session.
[APPLAUSE]
Please join me in thanking
the speakers again,
and I would like to invite
all the speakers on stage.
[APPLAUSE]
And this will be
an open Q&A panel.
So if you have questions,
please raise your hand
and we will call on you.
But I would like to
start off this panel
by getting each one of your
takes on why immunology at MIT
is different to normal
immunology departments that
are typically housed
in medical schools
and are very embedded
into clinical practice.
Facundo.
Well, MIT is a unique in
ecological environment where
you get direct access to
engineering and computation that
doesn't occur, generally, in a
normal immunology department.
So I always train
in institutions
that they were kind of having a
large breadth of collaboration.
And particularly
for me, I can talk,
it's rather a unique opportunity
to be close people to do things
that I am not an expert on.
Well, as Harry mentioned,
the immune system
is a stochastic dynamical
system with many participating
components in a fluctuating
environment that work together
to ultimately give you
a coherent response,
either to pathogens
or to cancer,
or sometimes, when it's
[INAUDIBLE] regulated
to yourself.
And if we have to
understand that,
we have to bring together
tools from basic biology,
from engineering, computation,
and systems design
in order to really
make that happen.
And then, translate that
into therapies and vaccines,
et cetera.
And MIT has all
these components,
from basic biology to
systems engineering.
And it's an inherently
collaborative place
and a collegial place.
And so that brings all of
these together in a mix
that I really have not
seen in other places.
And I'm the old man
here, so I can tell you
I've seen more places.
I think immunology is a field
that really needs technology.
We've seen huge advances
in our understanding
of immunobiology through the
advent of genetics, machine
learning, computational
approaches.
And these are things that
MIT is really good at.
And especially in
my field, we really
need technology development
to enable new discoveries.
So I think that's a
unique feature of doing.
Immunology immunology at MIT.
We're at the cutting
edge of technology
that we can apply to
our biological problems.
I don't have too much to add
we haven't already heard.
I'll just say, I don't know if
my lab could operate in the way
it does at many other
places in the world..
Because we need to
be able to bridge
experimental immunologists with
computationalists and hybrids
that can do both,
and having access
to those types of trainees.
That's very uncommon that want
to do all three of those things
at most other institutions.
So this, I think, is especially
helpful for immunology
because it seems to be a
property of MIT in general.
It's pretty unique to combine
fearlessness and rigor
in a combination
that actually works.
Like the ability to try
what might seem like it just
definitely isn't going
to work, or something
that might seem a
little bit crazy.
But do it in a way where
substance is never lost.
Where you make
sure that we really
are going to do this
to make the impact,
not just to make the
flash in the pan.
And I think that, in immunology,
everything my colleague
said about technology
and connections
with each other, all
of that's important.
But I think that everybody
from the sophomores and juniors
in my lab who have
started projects
through to the graduate
students, the postdocs,
and my faculty colleagues
to combine fearlessness
and rigor in one setting
is a special thing.
Question over here.
OK, I have two questions.
You will get them.
I was a cheerleader.
I can do this.
Two questions.
The first, is in
the earlier session
they spoke about how difficult
it was to move from a mouse
model to a human model.
And I'm wondering if that is
less difficult than the work
that you do, because it
appears you're translating it
faster and more readily?
So that's question one.
Question two is, I think, for
professors Wong and Birnbaum is
my understanding is
that solid tumors are
harder to treat these days.
And is your work directed
towards solid tumor treatment?
So that's my question.
OK, maybe Facundo, you can
start on the first question.
Those are excellent questions.
And there is no perfect model.
And every model helps
to address a problem.
A human clinical trial
will take tens of millions
of dollars and
several years to plan.
What we can take the
advantage is in the mouse,
for example, is
to iterate things
extremely, extremely fast.
And then, narrow them down
to be then tested in humans.
And the technologies that
we do in the lab that
is trying to make this
mouse more similar to humans
are at the tip of this because
with all the predictions,
and there is an experiment
ongoing there, all the mouse
experiments that we and
many others have done,
they have now done the
human experiment and show
that the predictions
were absolutely right.
So again.
There is no perfect model.
We will not know-- everything
will not work the same,
but at least at a very
diminished cost and time,
we can operate in the
mouse as a credible model.
And I think maybe
to add specifically
because I was in the session.
The immune system,
I think, translates
better because we're looking
at systematic changes and not
a single molecule
in a pathway that
has to be conserved
between mouse and human.
And for the second
question, Michael,.
Sorry.
So for the second question,
in terms of translation,
it's a little bit of both.
So the slide I showed
towards the end
is in models of melanoma.
And so we're interested--
which is a place where these
tumor infiltrating lymphocytes
has been clinically approved
could stand to be improved.
And so we really--
even for mouse
versus human, it's
really every tool for a job.
So some of the questions we
ask in terms of solid tumors
where immunotherapy
has been successful,
some were working in
things like glioblastoma
where progress or solid
tumors has been slower.
Some of it's in liquid
tumors because that's where
some of the clinical data is.
And so we do our best
to match depending
on what we're working on.
Question over here
and then back there.
Thank you.
Thank you for the whole panel.
Fantastic talk.
So I have two
questions, actually,
one for the whole panel.
What do you think
about mRNA technology?
What are the advantages
you see currently,
or do you see any
limitations for your work
you are doing in immunology for
vaccines or maybe therapeutics?
Second question is for Jessica.
Love the talk.
Antibody lectins.
Curious to see how
you are going to apply
this technology from oncology
to autoimmune diseases.
Thank you.
So I'll start with the
question directed at me.
Thanks for that.
Yeah, so these applications
and indications
in autoimmunity and immunology
that we are going after.
In these indications,
there's biological rationale
for why you might want to target
a glyco immune checkpoint.
So for example, in
autoimmunity, while in cancer we
block glyco immune checkpoints
to potentiate immune responses,
in the context of
autoimmunity, you
could imagine engaging those
same immune checkpoints
to elicit tolerance
in the same way
that tumors tolerize the
immune system to themselves.
In the setting of
fibrosis, there's
a glyco immune checkpoint
that allows fibrosis
to continue unchecked that we
are thinking about targeting
to help slow that process down.
But all of this builds
upon fundamental biology,
fundamental life science
research that has shed light
on these biological mechanisms.
Let's move over there.
Good afternoon, everyone.
Stepping back a little
bit, I was impressed with--
across a number of
your presentations
about just how dynamic
the immune system is.
Just the scale, the quantities
that you're dealing with
are just astronomical.
And then, on top of
it, you have people.
And so I was also impressed
by the slide showing
if you could give seven doses
days with increasing amounts
and you get a really tremendous
response, but impractical.
And I really want to target or
push on that in practicality.
I work with a state
agency that really fights
infectious disease,
and I want to push back
on that in practicality.
What can we do to make
that practical or suggest?
That's it.
No, I'm happy to address that.
I mentioned that a bit too
quickly in a brief talk
like this.
But we have now, in a
very recent publication,
which I flashed briefly,
our understanding that
is a practical challenge.
We have now, by understanding
the mechanisms better,
been able to design a system
which has only two shots, which
takes it much closer to
practicality because two shots
you can do.
And so that's now slated
to go to clinical trial.
The seven shot was already
done, as I showed in humans.
But that's where we
are going with this
is to do the two shot
regimen because that does
just as well, at least in mice.
Question over there.
Thank you for the
exciting talks.
I'd like to follow-up a little
bit on that question on probably
a more fundamental question.
In the seven doses,
I can understand
that you're saying the
subdominant epitopes
will increase, but
viruses are equally smart.
I'm just wondering what
if the previous epitope
starts dominating again
because viruses can mutate?
You're trying to
decrease one epitope
and increase the epitope.
The second question following
up with Harry's talk,
he can actually now visualize
and quantify the timelines.
Is it possible to--
I don't know how different
B cell kinetics and T cell
kinetics are, but is it even
possible to get both of them
in the same timeline so
that you can actually
get both arms to
have a good response?
So your questions are excellent.
And to the part
that I will address
is the following that I
mentioned very briefly
that we are trying to see
how to leverage this thinking
into a potential cure for HIV.
And what we are finding
there is that when
people have suppressed
virus, but some amount there,
then this may be--
this is still very
early days, but it
may be happening that one by
one, the epitopes get masked
and then you get more and more
responses and so on and so forth
while the virus is dormant.
And so you can then create a
net of antibodies that target
different regions or epitopes.
And as a consequence,
when the virus starts
to come up, the escaping that
net becomes very difficult.
And so that's the part that--
I might also slightly
comment on the second part
of your question.
In the extended
dose delivery things
that we have done, in
order to optimize it
to two doses such that it
becomes closer to practicality,
one of the things
that happens is
that you sync the
T cell and antibody
response or B-cell
responses to come more
within symphony-- asynchrony.
And if you don't give
the adjuvant also
in an extended
way, then you won't
let that to happen
because that's
how you prime the innate immune
cells, the dendritic cells,
so that later on they are
more efficient with the T cell
dynamics.
So there has to be
synchrony between those two
get the best responses.
Now, how best to image them
and understand them better,
I'll leave to Harry.
I'll just say that we and
others have definitely
shown that you can--
similar to Arup's
work with B cells.
If you manipulate the dynamics
of what the T cells are seeing,
you can get better
T cell responses.
And we know that you can't
get great B cell responses
without great T cell responses.
And I think it's just--
we need to put it together
and actually--
Facundo and Arup should probably
work on something like that.
I mean, that is an
excellent question.
Antigen or virus is the
fuel of this reaction.
If that stays for a
long time, this reaction
goes on for a long time.
And it goes on from B
and T cell together.
Arup [INAUDIBLE] show that in
these lymph nodes, when you just
give one dose, you
don't get antigen there
for a prolonged period of time.
So answering to your
T cell question, that
is very important, if
you have the fuel that
is the antigen there, you
will keep the B cells and T
cells cooperating for
a prolonged period.
I made a comment
with regards to--
we know so little
about vaccines.
All we know is empirical.
And even when we are giving a
prime and a boost in humans,
that has never been
experimentally proved.
The time between the
prime and the boost
is something that has been
determined empirically.
What we are trying to
do and what people do
is really understand the
mechanisms for this in a way
that we may alter this
prime and boost to get
a more effective response.
Exactly.
Perfect.
I would have one
last question I'm
going to ask you guys
to answer really,
really quickly in
the interest of time
is, what is the next
big frontier each of you
want to tackle?
I'm going to start with
Michael for fairness.
[LAUGHTER]
It's going to be the
integration with AI.
Can you elaborate a little bit?
Yeah.
Stephanie said quick, and
there's a timer counting down.
We can mingle after [INAUDIBLE].
So in my world, it's being able
to make general predictions
and general designs
in the same way
that you can go to ChatGPT and
make a picture of anything.
Right now, the predictions
are good, but very narrow,
very, very narrow.
So would be the equivalent
of having a ChatGPT that's
really great at making
pictures of ducks,
and you ask it to make
you a picture of a tiger
and it gives you a duck.
So I think that getting to the
point where some of these tools
are going to be generally useful
is something that we're really
excited to work towards.
On my end, I think there are
certainly biological questions
we're very interested in.
But on the
technological side, you
might have noticed that a lot
of the dynamic live imaging
that we're able to
do, we're limited
in how many cell types
or readouts of function
that we can look at.
And there are
practical limitations.
So I think something we're
really excited about and trying
to think about is how can we
visualize many, many more cell
types and their states in
real time in living organisms?
And that's not a
trivial challenge,
but we have some ideas.
We are really focused on
developing technologies that
will allow us to fit
glycans into the broader
picture of immunology
and immune responses.
And we need to understand this
biology at a systems level
and how it connects to
all of the other aspects
of molecular and cellular
immunology that we understand.
I think those kinds of
fundamental advances
are going to be necessary
to translate this biology
and to get to clinical advances.
In addition to the
things I talked about,
one of the things that we are
launching in our group right
now is a program to
really understand
how the innate and adaptive
immune systems are integrated
because very little or much
less is understood about that
than anyone by themselves.
And I believe that will
get us to really start
understanding
questions like what
determines the memory
of an immune response.
Why is it longer for one
pathogen versus another?
Why is it longer, perhaps,
in protein antigens
versus mRNA vaccines?
And so that's one of the
new frontiers that in our--
new frontiers and our
group is to bring that
into innate and adaptive
immunity together.
Yeah, I think the connection
between quantitative immunology
that is very recent, and
then translating that
into computational
or AI models is
something that is
extremely exciting
and is going to be
transformational
over the next 10 to 20 years.
I mean, we still
know very little.
We know we can
generate antigens.
We know we can
generate antibodies.
We can produce structures, but
we cannot predict for better
molecule binding.
I think that the
next 10 to 20 years
are going to be transformational
in that respect.
Awesome.
With that, I would like to thank
the speakers and the panelists,
and particularly you
for your attention.
Thank you.
[APPLAUSE]

---

### MIT HEALS Launch: Health plenary session
URL: https://www.youtube.com/watch?v=AxFQ9chLlVI

Idioma: en

Good afternoon, everyone.
On the afternoon of
Friday, July 30, 2021,
I stood outside
the emergency room
entrance of UT Southwestern
medical Center in Dallas, Texas.
I'd been there for a few hours--
not an unexpected amount of
time to spend at the ER--
while my sister and I waited
with our mother for a visit
from her surgeon.
A year earlier, she'd had a
bilateral mastectomy, performed
at the height of the COVID-19
pandemic and nearly 19 years
after her first breast
cancer diagnosis.
Our visit to the ER in
2021 was about eight weeks
after a second
eight-hour surgery
that was the direct result of
the prior year's mastectomy--
this time, for a quick check to
examine a troublesome incision.
Mom's surgeon released
her with instructions
to come back the following
Monday for a surgical cleanse,
and I set off to
retrieve the car.
As I stood waiting for
the car to arrive--
$5 valet parking for patients
is a nice side benefit
in a state with so much land
available for parking garages--
the valet attendant
started chatting with me.
After a few innocuous
comments, he suddenly asked,
did you take the COVID vaccine?
I replied, yes, absolutely, and
so did all of my family members.
And so began a
fascinating conversation
that included tales of
government conspiracies
and plots to track
citizens based
on devices included in the
vaccine, and even a connection
to monkeypox.
I decided not to tell him
that if he was worried
about being tracked
by the government,
he should put down
his cell phone.
[LAUGHTER]
At one point, the
attendant posed a question
that he undoubtedly thought
would carry weight with me.
Answer me this, he asked--
how did they manage to
develop a vaccine so fast?
My response--
because practically
the entire biomedical
world, from academic labs
to multinational
pharmaceutical companies,
started working on it.
Imagine the global problems
we could solve if only
we worked together to do it.
My name is Kristala
Jones-Prather.
I'm the Arthur D. Little
professor and department
head of chemical
engineering here at MIT,
and I am also a class of
1994 undergraduate alumna.
In my time as a student
and a faculty member,
I have come to see that
MIT is an institution that
is collaborative by design,
with our infinite corridor
representing endless
possibilities,
and our connected buildings
and hallways serving
as a physical reminder of
the lack of true boundaries
between our disciplines.
After all, the beneficiaries of
our scientific and technological
advancements,
including my mother,
who participated
in a clinical trial
after her initial cancer
diagnosis in 2001,
that, I learned 15 years later,
established the standard of care
for adjuvant hormonal therapy.
These patients are not concerned
with the fields of study
or the affiliations
of the inventors.
And so as we launched the
MIT Hills Collaborative,
we do so with the intention of
tackling big problems by asking,
what if?
And working together across
boundaries and barriers
and even institutions
to find the answers
and deliver solutions.
What if detection of
diseases like breast cancer
could be made easier, less
invasive, more accessible,
and more comprehensive?
Could this lead to earlier
detection, better health
outcomes, and greater
patient survival?
What if we really work to
understand the unique dimensions
of women's health, with women
being disproportionately
affected by autoimmune and
chronic inflammatory diseases?
Could we see new
human tissue models
giving way to novel
physiological insights
and new paths for
effective treatments?
What if we tackled ovarian
cancer in a fundamentally
new way, fusing knowledge
gained from basic immunological
sciences with novel
engineering approaches
to delivering immunotherapies?
Could we leverage
the partnerships
of academics and clinicians
to change the outlook
on this disease.
What if we thought
about education
in a radically different way
by working not just in research
labs but also in the communities
where epidemics take hold?
Could we incorporate
the social scientists
to foster a deeper understanding
of the challenges associated
with the adoption
of new medicines,
and develop new approaches
to more deeply understand
the many dimensions of health?
In this session,
you'll hear from four
of my amazing
colleagues, each of whom
is using interdisciplinary
approaches and collaborative
partnerships to ask and
answer these questions.
As you listen to
them, I hope you're
inspired to think about the
possibilities, the promise,
and the hope that
their work represents
to reflect on the patients,
perhaps some well-known to you,
and others who may remain
nameless, that would benefit
from the achievements yet to
come, and to consider if and how
you can contribute
to these missions.
Oh, and yes, my
mother is continuing
to live her best life, having
spent Thanksgiving week
traveling in Australia,
thanks, in no small part,
to a continuous community
of scientists, physicians,
and engineers who never
stop asking, what if?
[APPLAUSE]
[VIDEO PLAYBACK]
- Nobody knows your
body like you do.
Imagine spotting
dangerous diseases way
before symptoms arrive.
Protecting yourself by
detecting yourself is the goal.
- So this variable, non-invasive
ultrasound patch is specifically
for early breast
cancer detection.
- Professor Dagdeviren has
been working with her students
for six years on a
wearable breast monitor.
Because it uses ultrasound
waves instead of radiation
to create an image, the
monitor can be used regularly.
This invention
inspired her students
to create smaller, even
less-invasive silicone
patches that can scan
for other cancers.
- Wearable technologies will
grow rapidly in the near future.
But in the far
future, they will be
one of the most powerful
tools that we will
be seeing in our daily life.
[AUDIO LOGO]
[PLAYBACK ENDS]
So we create that far future
right here at the MIT Media Lab
by blending media,
arts, and sciences.
Specifically for my group, we
are focusing on women's health.
And we create technologies
for female technologies,
which I often called femtech.
And let me tell you why by
giving three dramatic facts.
80% of autoimmune
diseases are in women,
and we still don't
understand why.
Heart disease remains
the number one cause
of death in both genders,
yet women wait more in ER
to be treated.
Women were excluded from
clinical trials until 1993.
Hormones were
given as an excuse,
and which means there
is a massive data
gap in hormone health.
And since mice don't
menstruate, how can we catch up?
To catch it up, in my
group at the MIT Media Lab,
we create devices which
are organ-specific.
These devices can be thinner
than your hair fiber,
and implantable like a needle,
can go in the deep brain,
and do neurological
explorations and stimulus.
And at the same
time, these devices
can be directly worn on
your largest organ skin
to do hormonal drug delivery
all the way from face
to the abdominal skin
throughout the IVF process.
Or these devices
can be attachable
to your personal garments, like
your daily bra or underwear,
and they can detect your
breast cancer at an early stage
and/or monitor
your bladder volume
to better understand your
kidney health and wellness.
Let me give you an
example more in depth.
It's a wearable ultrasound
breast patch for early breast
cancer detection.
It's a technology that
was invented here at MIT,
and this device is non-invasive
and can be directly placed
in the internal
wall of your bra.
By using safe non-radiative
ultrasound technology,
this device can detect any
anomalies within subseconds,
literally by sitting
on your seats
and listening to
this presentation.
And this data is sent
to your medical doctors.
And your medical
doctors can tell
if you are supposed to come for
more sophisticated treatments
or screening at the hospital.
This is specifically
important for the women who
have barriers to this kind of
technologies in the rural areas,
and also for women
who are on high risk
and needs to do
periodic screening
to increase their
survival rate to up to 98%
And this is also important
because high-risk women often
are often diagnosed
between two mammographies
and decrease their
survival rate up to 22%
So with our technology,
what we are trying to do--
not only doing the science, also
creating stories behind that.
I'm not only doing
science at MIT.
It's a calling for me, and
it has a personal story.
I lost my aunt at the
age of 49.5 years old.
In this picture, she is Fatima.
She's very happy,
very beautiful,
and holds tons of
hopes for the future.
Whereas in this picture,
she is only 49.5 sick,
tired, and in pain.
The entire process was miserable
not only for her, for all of us
around her.
Despite the fact that she
had regular breast screening,
she was diagnosed
with breast cancer
and could survive
only six months
after this picture was taken.
So like my aunt,
many women suffer
from this devastating disease.
And the breast cancer
is the number one cancer
type among women.
And even the number is
higher than the combination
of other leading cancer types.
And more dramatically,
every eight women, one
will be diagnosed with breast
cancer during their lifespan.
And the current standard
screening tool is mammography.
It's painful, and
it's radioactive.
So you can't do it periodically.
You can't screen
your breast tissue.
And by the time you have
your second mammography,
by assuming the first one is
clear, 40% of high-risk women
develop incurable
cancer in between,
with the survival rate of 22%.
Imagine a world in which
your daily bra will
allow you to check your
breast tissue regularly
to increase your
survival rate up to 98%
That world is right here at MIT.
This is our product,
our technology.
It's a daily bra that you wear
with a wearable ultrasound
technology inside,
which can capture
the data within subseconds
with no skill operator needed.
And this data is sent to the
cloud and collect the data.
And by using AI and
machine learning,
your radiologist and doctor can
have an AI-supported report.
This is important because
current ultrasonography
technologies are handheld.
You need an operator.
You have to go to a doctor.
And oftentimes, as you can see
in the picture, they are bulky.
They have to press--
doctors needs to press against
your very personal part
of yourself.
And you need to use a
message [INAUDIBLE].
You need to clean up
afterwards, time consuming.
It cannot offer frequent
visiting, and it's expensive.
Whereas in our
technology, one scan
will cost less than
a cup of coffee.
And how it works?
We use-- anyhow, it doesn't go.
But anyway, we use inflexible
cables which can come all
together and print on a
biocompatible substrates,
which can go into your bra and
scan the tissue up to 260-degree
angle with 12-centimeter
information.
Why it matters?
Because our technology can
provide a resolution up
to 0.02 centimeters, which
is an order of magnitude
smaller than the
stage one tumor sizes.
With frequent screening, we
can enable early breast cancer
detection.
It matters because just in
US, we have 44,000 women lost
their life because
of this disease.
And the breast screening
is $30 billion,
and it increasing 30%
every single year.
With our technology, we can
increase the survival rate up
to 100% by decreasing the
medical costs by half.
So this was a dream
on a piece of paper
that I drew on the bedside
of my aunt just to relax her.
And even while she was in pain,
she was giving me feedback.
Now it's real and I can touch
people's bodies, skins, breasts
and their lives altogether
with my amazing students
at MIT Media Lab.
Thank you.
[APPLAUSE]
Good afternoon.
It's great to be here.
I'm Linda Griffith from
biological engineering,
and I direct the Center
for Gynepathology Research
here at MIT.
What I want to tell
you about today
is a new movement that
we're starting at MIT.
It's called the Move
Over, Mice movement.
And what inspired it is
the incredible challenges
we have creating
effective therapies
for diseases that
preferentially afflict women.
Now, they'll help
everybody, but they
are motivated by helping women.
Now, why do we want
to move over the mice?
Well, think back before
COVID, we had colds.
We had the flu, and then we
had that very special disease,
the man flu.
And you know what
I'm talking about.
Husband and wife get a cold.
The husband's in bed
dying while the wife's
taking the kids to soccer.
OK, everybody knows that story.
And if you look at the x and Y
chromosome, it's not surprising.
So many things missing from the
Y, including some immune genes.
[LAUGHTER]
OK, now so it's
biologically true
that men will get
sick-- this is proven.
They'll get more acutely sick.
But women stay sick
longer, and they
develop more chronic
autoimmune and
chronic inflammatory diseases.
They're the result of
multiple gene environment
interactions in humans--
very, very difficult to
study in mice-- and that
have provocative
links to the way
we reproduce and menstruation.
And so we can't really hope
to make a lot of progress
if we continue to study animals.
We really need to
study the patients
and replace our animals
with living patient avatars
and studies on patients.
Now, I'm going to tell one story
to illustrate the Move Over,
Mice principles.
About 15 years ago,
we started working
in a chronic systemic
inflammatory disease called
endometriosis.
And probably most people
have heard of this,
but it's when you
find bits of tissue
similar to the uterine
lining, the endometrium
growing, other places
like the abdominal wall,
invading into the bowel.
It can invade through the
diaphragm and get to the lungs
and so on and so on.
It's incredibly
debilitating, causes
pain, infertility,
vomiting, systemic sickness.
And we don't know what causes
it, but we do about 10% of women
after the age of 12 or so
will have this disease,
and it's very hard to diagnose.
Now, I wouldn't be here
talking to you today
if we had good
treatments for it.
Some patients respond
to birth control pills
or menopause drugs,
but a lot of them
have repeated,
repeated surgeries.
Now, I got involved in this when
a local surgeon, Keith Isaacson,
who's a fantastic
surgeon and clinician,
called me one day and said, I
operated on a 23-year-old woman.
It was her 10th surgery
for endometriosis.
And she got operated
on in that OR
there on the right
at Newton-Wellesley,
where we now collect patient
samples because I got together
a group of people, and we
started to study endometriosis.
OK, so what do we do?
Well, I'm not a-- we're not
the kind of engineers who
make devices or surgical tools.
We're biological engineers.
We study how systems work.
So we reasoned
that we could maybe
start to classify patients by
studying their immune networks
in the peritoneal fluid.
So we studied
about 100 patients,
did some targeted
proteomics, and we
were able to identify in
about a third of the patients
a particular inflammatory
network associated
with an intracellular enzyme and
immune cells called Jun kinase.
And so Jun kinase--
this is very exciting--
maybe if we inhibit it,
we could have a new
non-hormonal drug target.
And so it turns out, we go and
find out that there was a--
find these two papers--
that there was a drug
company, Merck Serono,
that had been
developing Jun kinase
inhibitors for other diseases.
And one of their amazing
scientists, Steve Palmer,
convinced them to do preclinical
models of endometriosis,
and that their
drug worked there.
Unfortunately, it
didn't work when
they licensed it out and there
went into clinical trials.
It didn't have the right potency
or patient heterogeneity,
whatever.
But the Steve Palmer
there turns out
to stay involved in the field.
So I got invited to give a
keynote talk at the American
Society for Reproductive
Medicine meeting in Hawaii
in 2014.
There was a hurricane, over
a foot of rain that weekend.
And who did I run
into but Steve Palmer?
So we sat inside, and we
talked about the future
of endometriosis drugs.
And he said, I'm leaving Merck.
I'm going into academia.
I'm going to develop new drugs.
But what we need to
get into the clinic
is we need human
efficacy models.
We need the patient on a chip.
We need a lesion connected
to blood vessels,
circulating immune
cells and all of that.
So you go do that.
Well, then, OK.
So the first thing we had to
do-- we had plenty of access
to patient biopsies.
And other people in the
field had figured out,
how do you make organoids
from these patient biopsies.
You can see the glands from
the endometrium on the left.
But what they were doing is
using some terrible reagents
to make 3D cultures of these.
They had something derived from
a mouse tumor to make the 3D
environment.
It degraded.
It was really not
great for the biology.
So we used our polymer science
and biological engineering
skills to design a
completely synthetic hydrogel
to support all the different
cell types in a lesion.
So you see here the
beautiful epithelial organoid
blue with green in the middle.
And the red cells are the
supporting fibroblasts
growing in this gel.
And labs around the
world have used this,
and it's now being
commercialized.
But of course, the lesion
has to be in contact
with blood vessels.
And we were fortunate that
our colleague Roger Kamm
had invented a way
to grow a vascular
bed in between two microfluidic
channels, as shown here.
And our student in the
movies not playing--
but if it were playing--
why is it not playing?
Our student Ellen
Khan had shown how
to flow immune cells
through that vascular bed
from one side to the other.
But what we had to do was
adapt this to accommodate
endometriosis lesions.
And so what we did is designed
a new chip, bigger with pumps.
And we put the vascular cells
and the lesions together.
And what you see, over a
six-day time lapse here,
you can see the lesions.
They're in the
green boxes growing.
And you can see
the blood vessels
forming from the vascular cells
to build a network of blood
vessels around them.
And there's immune
cells in there also.
OK.
So now we're ready to go.
We did our part.
What did Steve Palmer do?
He went to Baylor Drug
Discovery Institute,
developed a whole new set
of Jun kinase inhibitors
with NIH funding.
And then he licensed them to
a company called Celmatix.
He left Baylor.
He's now at Celmatix.
And here we are a few weeks ago,
starting the launch of a project
so that we will do
preclinical efficacy
testing with the Celmatix team,
Steve Palmer, and the CEO Piraye
Beim, and the MIT team.
Of course, Laura Bahlmann
actually does all the work.
She's a postdoc in the lab.
But this EndoChip,
there's still a lot
of things evolving about
how regulatory agencies view
this kind of efficacy data.
So there's great
conversation going on.
But no matter what, you have
to prove your drugs are safe.
And so we come into this project
with decades of experience
working with drug
companies on building
these kind of patient avatars
for drug safety, particularly
liver.
So I invented something
called the liver chip.
It's widely used.
Novo Nordisk has built a lab
close to MIT to interact with
our team on how to use this
3D liver model of human liver
for cardiometabolic disease.
We've done interacting organ
systems as shown on the right.
And smaller versions of them
are commercially available.
So we know we have
to get these out
into the wild in collaboration
with lots and lots of users.
And we have to also
build many other organ
systems similar to what
we have for the chip
and connect them together.
So our vision and it's
under construction
is a living patient avatar
core facility at MIT.
And Kelly Pate, the
head of the division
of comparative medicine,
is very on board with this
because she would like to
move some of MIT's mice over.
In order to do this, we had to
develop a whole new platform
technology.
What I showed you on
the previous slide
is about 10 years old.
So together with Dave Trumper
in partnership with Novo Nordisk
and a lot of other
partners, we're
developing all of these
for pain, for liver,
for chronic infection like
Lyme disease, gut microbiome.
And these are real
things that we
want to have on campus for
people to be able to use.
So I'll close there
and invite you
to come to the poster session.
If you want to hear more
about this and other diseases
we study, I encourage you
to join the MIT Move Over,
Mice movement and
be happy to talk
to you about it later today.
Thank you.
[APPLAUSE]
Good afternoon, everyone.
So I am a chemical engineer
who works in biology,
and I'm going to be talking
about how we can use extremely
small lasers to address cancer.
In our work, we
actually use alternation
of positive and negative
charge to build up
extremely thin nanoscale films.
And in this work, we've been
studying how we could take
an existing small
drug nanoparticle,
like a liposome or a
polymeric particle--
maybe it has a net
negative charge--
and positive and
negatively charged layers.
In this way, we can actually
incorporate a chemotherapy drug
in the core and wrap around
this nanoparticle other agents.
For example, after
positively charged layer,
we can put something
negatively charged,
like siRNA that can silence a
gene that enables the cancer
cells to survive that
chemotherapy drug, thus creating
a way to address
chemo resistance.
And one of the most important
parts of these layers
is that final layer,
negatively charged,
that enables us to have
a nanoparticle that
doesn't engage with other cells
but might be designed to engage
specifically with cancer cells.
So these are squishy materials.
I'm a polymer
materials scientist,
and we enjoy making
these systems.
But the idea is that we
can actually coat them
in a way that allows us
to direct therapeutics,
specifically to cancer
cells in chemotherapy.
And I like to compare
this to a Wonka Gobstopper
for anyone who remembers
what those things are--
[LAUGHTER]
--that have layers of
candy that dissolve
at different times, which
gives us a chance to regulate
how the drug is introduced.
However, when we became
really interested
in addressing ovarian
cancer, we found
that this approach
was not necessarily
the best approach using siRNA.
Ovarian cancer has been a very
difficult cancer to address.
It affects a large
number of women,
and it is detected late stage
because of the very subtle
symptoms that it provides us.
For that reason, although
a large number of patients
are able to recover after the
first chemotherapy treatment,
about 80% to 85% of them
will suffer a recurrence,
and that recurrence will
be a highly resistant form
of the tensor.
For that reason,
it would be really
interesting to have
an immunotherapy that
would allow this
cancer to be attacked
by the human immune
system when it comes back.
Unfortunately, there's
been little progress
in doing this
because there are not
a large number of immune cells
that reside in ovarian cancers.
So when we began
looking at this problem,
we started thinking about
making a nanoparticle that
would be extremely sticky
to ovarian cancer cells
but not to other cells.
And we did this by looking
at a library in which we
started with the liposome
and wrapped positive charge,
and then a range of
different negatively
charged stealth layers,
polysaccharides, polypeptides,
and a range of other
biomacromolecules, that
might be able to help
us with targeting,
as well as with getting
through the ovarian
space, this
intraperitoneal space.
And in doing so, we found
that there were a few that
were very interesting to us.
And those few actually
showed a very high affinity
for ovarian cancer,
whether it appeared
in the primary tumor-- here
we see a tumor lit up--
and the infrared
particle showing up
and a very nice
overlay, but also
in metastases within that
intraperitoneal or abdominal
space like this gut system.
Now, we found that
there were two that
were really interesting to us--
one of them, hyaluronic acid
actually binds to cancer cells
and gets taken up
into the interior
and releases things like
siRNA in chemotherapy drugs.
That's really interesting.
But we found
another one of them,
poly glutamic acid, a
very simple polymer,
that causes the
nanoparticle to stick
on the outside of the
ovarian cancer cell--
not great for delivering siRNA,
which has to-- on the inside.
However, it does allow
us to present something
to neighboring cells.
And we began to think about how
we might be able to use this.
I began to talk to my
downstairs neighbor
at the Koch Institute, Darrell
Irvin, who knows everything
about immunology and cancer.
And he was able to
inform about how
we might be able to use this
to deliver a cytokine, known
as interleukin 12.
IL-12 actually can
induce large amounts
of infiltration of immune
cells, in particular T
cells and NK cells that are
responsible for killing cancer.
So we decided to then
take our liposome,
attach that IL-12 the liposome
directly, wrap it in plus charge
and then minus charge.
And that minus charge will take
us to the ovarian cancer cell.
Deliver this into the
intraperitoneal cavity
and let this get presented
to neighboring cells, which
will eventually lead to the
generation of interferons
and the infiltration of T cells.
And we did indeed find
that we had enriched
T cell infiltration
in these layer
by layer nanoparticle systems.
In fact, when we looked in
mouse ovarian cancer model,
we found that there were
something like 10,000 or more
infiltrated T cells
compared to untreated mice.
And what was really
exciting is that we
found we could combine
this IL-12 treatment
with checkpoint inhibitors.
Now, it's worth mentioning
that the reason IL-12 isn't
used in the clinic today is
because it is extremely toxic.
It can also excite the immune
system in the bloodstream.
In our work, attaching
it to a nanoparticle that
is very specific to
the cancer allows
us to rescue that toxicity and
avoid any kind of side effect.
Checkpoint inhibitors
are typically
not effective in
ovarian cancer at all.
So what we did was we combined
IL-12 with checkpoint inhibitor.
And in this case, we changed
the immune microenvironment.
So now there are
immune cells that
can be impacted by the
checkpoint inhibitor.
And what we found remarkably,
was in this red line as you see
here, 100% of the
treated mice survived,
compared to all of
the other treatments--
checkpoint inhibitor,
which does barely nothing--
and any of the other IL-12
combinations that we looked at.
So we're very excited
about this work
and looking at how we can
take this on to clinic.
Now, I mentioned
that little story
about getting inside the cell.
With other collaborators,
Joan Brugge,
who is at Harvard Medical, and
Ursula Matulonis at Dana-Farber,
we've been looking at how we can
encapsulate inhibitor molecules,
again, so toxic that,
although they are together
very effective against
ovarian cancer,
they're also very effective
at harming the liver.
And in this case,
we encapsulated
these in a polymeric
nanoparticle
and used our hyaluronic
acid targeting system
to get this directly inside
ovarian cancer cells.
And again, we saw a
great deal of improvement
in ovarian cancer treatment.
So this theme of collaboration
is really important to us.
I usually show a slide of
all of the students involved,
and we have some incredible
students and postdocs involved
in this work, but I
wanted to point out
collaborators who have
impacted us in our journey
to address ovarian cancer.
I mentioned Darryl Ervin,
my wonderful neighbor,
who has informed us about
every aspect of the design
of these systems, and we have
worked together to develop it,
but we also have had this
wonderful collaboration
with Harvard Medical
School in Dana-Farber.
We're involved in breakthrough
cancer, which brings us together
with one of our
younger colleagues
at the Koch Institute,
Stephanie Springer, also
an expert in immunology
and tumor biology,
and Sohrab Shah at Memorial
Sloan Kettering, who is looking
at patient cells that are being
treated and determining changes
in the genetic
signatures of those cells
before and after chemotherapy.
If we can understand what the
outside markers of those cells
are, we can find
nanoparticles that can then
be targeted directly to
minimal residual disease
and treat them and excite
the immune response in them
using the studies that
Stephanie is doing.
And finally, we have a
wonderful collaboration
with my MIT colleagues
and colleagues
at Johns Hopkins with Angela
Belcher, Sangeeta Bhatia, Kripa
Varanasi, and Rebecca
Stone at Johns Hopkins.
And in this case, we're
reaching much earlier.
Intervention is really critical.
Ovarian cancer tends to be
detected at very late stages.
Angie has developed
nanoparticles
that are able to allow us
to see inside deep tissue.
And we are using our
sticky nanoparticles
to address being able to
detect those tumors and even
early tumors that appear
on the fallopian tube.
So with that, thank you.
And I'm really excited
about collaboration.
And I think that
this new initiative
is going to allow even
more of these opportunities
to take place.
[APPLAUSE]
Hello, my name's Bruce Walker.
I'm a physician scientist.
I've spent my career taking care
of patients and studying HIV.
And it's my great pleasure
to tell you about a course
that I teach to
MIT undergraduates.
Each January in
the intersession,
we select 20 students
from all majors
and take them to experience
an ongoing epidemic
to understand the many
dimensions of an epidemic.
We go to South Africa.
You can see from this
slide the staggering,
staggering prevalence of HIV
infection there in the areas
that we work.
40% of men and 62% of
women in the ages 25 to 44
are infected and will
need lifelong therapy.
It's in that context
that the students
get to see what it's
fueling, that epidemic.
We land first in
Johannesburg, where
we begin the didactic
portion of the course.
We incorporate a number
of African students
with the MIT
students, and we begin
to teach them where
HIV came from,
how it causes disease,
how the body fights back,
and how therapies are
developed, and why
four and a half decades
after discovery of HIV,
there's still no vaccine,
but we don't stop there.
We also teach them about how
policy and advocacy influence
the course of an epidemic.
Here, they're learning
from Edwin Cameron, who
was the chief justice of the
South African Supreme Court,
who presided over the trial
that led to the forced--
to force the South
African government
to provide treatment
to its citizens.
He's also a gay man, and
he tells the students
about the problems of
stigma and discrimination
that he himself
encountered and that
are so widespread
in that population.
We then go to Durban,
where we had participated
in collaborative research for 25
years with African scientists.
We helped catalyze the
construction of two research
institutes there by funneling
money from the United States
to support that construction.
The students get a chance
to see African scientists
and talk to them about the
opportunities and challenges
of doing research
there, but they also
get a chance to see ongoing
projects that are underway.
One of those projects is
called FRESH, probably
the most exciting project I've
ever worked on as a scientist.
Stands for Females Rising
Through Education, Support,
and Health.
And it has two goals.
One is it's an HIV prevention
program and poverty reduction
program geared at young,
high-risk women 18 to 23 years
old who are living in poverty
and because of that, are
at high risk for HIV infection.
But it's also a study
of acute HIV infection.
And essentially, we see
the women twice a week
for nine months, teach them
empowerment, life skills,
job readiness training.
And each time they come in,
we do a finger stick blood
draw to identify anybody, who
despite our efforts to prevent
infection, has become infected.
The students get a chance
to talk to these women
and understand what
it is that's fueling
the 10% per year
incidence rate of new HIV
infection in that population.
They also get a chance to see
part of the empowerment program,
which is teaching these
women, most of whom
have never graduated
from high school,
teaching them how
to use a computer.
Most of them have never
touched a computer before.
And we conduct the
study at a shopping mall
to avoid stigma that
would be associated
with coming to a clinic
or to a or to a hospital.
And the course really
completely changes
the lives of these young women.
They come in with no real
sense of hope about the future.
And they leave having
found their voices
and feeling like they really
can make a difference.
Of the 3,000 or more women that
have gone through this course,
80% of them we've been
able to get employed.
And the unemployment
rate is so high
that these women
would really never
have had an opportunity
to become employed,
which moves them out
of poverty and reduces
their risk of HIV infection.
We also take the students
to local hospitals
to understand
delivery of care in
a resource-constrained setting.
Here, talking to
a young man who's
infected not just with
HIV, but also with TB,
and the challenges in
a resource constrained
setting of delivering care with
a highly contagious disease
like tuberculosis.
We go deeper into the rural
areas of KwaZulu-Natal
to teach them about delivery
of health care there.
This is a pharmacy.
And out on the table, you
see traditional medicines.
80% of South Africans will go
to see a traditional healer
before they go to
see Western medicine.
We've been collaborating
with traditional healers
for the past seven years.
And through a process
of building mutual trust
and respect, we've been able
to implement with them an HIV
counseling and testing program
so that when they see patients,
they now not just rely on
their traditional medicines,
but they test them
for HIV and refer them
to us if they need therapy.
So the students see
those traditional healers
and hear from them how they
were called to that profession.
We also introduced them
to infected mothers
and their infected
babies so that they
can understand how those
mothers became infected,
what the factors were
that led to that,
and understand how
difficult it is to raise
a child who's HIV infected.
We also go deep into the
rural areas of KwaZulu-Natal.
Here are some of the students
with a mobile clinic.
This is a catchment area where
the Africa Health Research
Institute has been collecting
data on the whole population
of about 160,000 patients once
a year on every person in that
region.
And the students go
and make house calls
with the RE investigators
and then come back
to the Research
Institute and use
that data to answer questions
that they have generated
through the research that--
or through the visits
that they've made.
At that point, we're so
far north in KwaZulu-Natal,
that there are no local
hotels to stay in.
So we take advantage of that
and take them to a game park
where there is a hotel.
And there they do problem sets
in traditional MIT fashion,
comparing and contrasting
different epidemics, cetera.
And they then present those data
to one another as a wrap up.
In between the problem
sets, they go on game drives
to see another dimension
of the African landscape.
And finally, when we
get back to Durban
to the airport to
head back home,
I think it's fair to say
that their lives have really
been changed by the experience.
They realize that epidemics
are multifactorial.
They're not just the realm of
medical doctors and scientists.
But really, every
discipline has something
to offer to help
come to a solution.
They also learn that
it's really important
to learn from the
patients themselves
and to understand
what the barriers are
to care that prevent
some of the things
that we develop from ever
being actually utilized.
And what I learned year after
year from teaching this course
is that MIT students are
incredibly motivated, incredibly
bright, and
incredibly passionate
about making a
difference in the world.
And I think we as educators
have to show them those worlds
and facilitate their
involvement in that.
I'd like to finish
just by acknowledging
MISTI and IMES and the Reagan
Institute for their support,
as well as
philanthropic support.
We set up this
course specifically
so that it would not
cost any of the students
any funds of their
own because we didn't
want economic considerations,
to prevent anybody
from taking the course.
And so I thank
you for listening.
[APPLAUSE]

---

### MIT HEALS Launch: Entrepreneurship panel session
URL: https://www.youtube.com/watch?v=iy7Hry92Xzw

Idioma: en

Good afternoon.
I'm Ernest Fraenkel, a Professor
in the Department of Biological
Engineering and Co-Director of
the interdisciplinary program
in Computational
and Systems Biology.
Any of you who've walked
around Kendall Square
over the last few
years will have
noticed that the landscape
is being transformed
by an influx of companies in the
life sciences and health care.
And that's no coincidence
because MIT is an engine
of innovation in those areas.
At MIT, the Life Sciences
and Health Sciences
span many departments--
obviously, biology, chemistry,
physics, and many of the
engineering departments--
my own, biological engineering,
chemical, civil, mechanical,
materials, and of
course, computer science.
And we live in an
amazing ecosystem
where we also have the Broad
Institute, the Whitehead,
the many hospitals around us,
and of course, many venture
capitalists.
And that's led to,
really, a profusion
of companies emerging from MIT.
You can see some of them
that have biological themes
on the slide above us.
And you've seen similar
slides earlier today.
And I think what strikes
me about these companies
is their diversity.
Everything from bionics, bionic
limbs, better diagnostics,
the platforms that accelerate
the delivery and discovery
of therapeutics.
So today, we have
the opportunity
to hear from two
leaders in developing
entrepreneurial companies
that are, really, while early,
still already having
an amazing impact.
And I'll ask them each to
introduce themselves and tell us
a little bit about
their company.
So let's start with you, Cullen.
All right.
Happy to be here.
Cullen Buie, I'm on the faculty
in Mechanical Engineering
and Biological Engineering.
The company that's spun out
of my lab is called Kytopen.
It's a company which is
developing a technology that
allows you to do
scalable gene delivery
into human cells for things
like CAR-T therapy and gene
therapies, ex-vivo gene therapy,
so therapies where they want
to take your cells, engineer
them, and then give them
back to you to serve as
a living therapeutic.
The beauty of the technology is
that it scales from the bench
to the clinic.
So a challenge in a
lot of these fields
is that the technologies
that you use for discovery
don't always scale
up to manufacturing,
but we have a technology
where the same thing that you
use at the bench can quickly
and easily scale up to what
you can use at the clinic.
So to put some
numbers to it, it can
do enough for a human
patient in about one minute.
So you could imagine easily
doing this for a large patient
population.
Fantastic.
Tasuku.
Hi, my name is Tasuku Kitada,
and I'm Co-Founder Head
of R&D at Strand Therapeutics.
Strand is a next-generation
messenger RNA therapeutics
company focused on
developing cancer drugs.
Our drugs are based
on messenger RNA.
We actually programmed the
mRNA using molecular sensors
and switches so that
we can express protein
in a cell-type specific manner.
And what that does
is that now you
can take really potent
proteins like cytokines
that might be very good
for counteracting cancers
but might be very systemically
toxic to the patient.
So now we can take that, express
it specifically in the tumor
so that you can take a
difficult-to-develop mechanism
and make it developable.
So I developed this technology
while I was a postdoc at MIT
at the Synthetic
Biology Center, together
with my business co-founder
Jake Becraft, who
was a student at MIT,
biological engineering,
and our scientific co-founders,
Ron Weiss at the Synthetic
Biology Center and
Darrell Irvine,
who was at the Koch
Institute at the time.
Fantastic.
So I'd like to hear
from each of you.
What inspired you
to start companies?
Was there a eureka
moment, or was it
more like Edison, one part
inspiration and 99 parts
perspiration?
Let's start with you, Tasuku.
Yeah.
Entrepreneurship for me
was very serendipitous.
So I didn't come to MIT
necessarily thinking
that I wanted to
become an entrepreneur.
But I feel like maybe I
came at the right place
at the right time.
A few years before I started
my postdoc, Katalin Karikó,
Drew Weissman at Penn had
developed this technology
that showed that you
can take messenger RNA
and express protein in
a very efficient manner.
And then Moderna
also had just became
a company, a startup company.
And they were working
on the first generation
of messenger RNA therapeutics.
And so we were in academia.
And while we thought
that most people were
skeptical about messenger
RNA therapeutics at the time,
maybe it was only
like a 1% chance
that it would become
useful for society.
But as academic researchers,
we can think boldly.
And we thought, let's imagine
a world in which this messenger
RNA became useful for society.
And in that world,
what would be missing?
And what we realized
was that the ability
to control protein
expression precisely
was missing for the
first-generation technologies
that companies like Moderna,
BioNTech were developing.
And so that's why
we decided to apply
the principles of synthetic
biology and program
the messenger RNA so that we can
express protein in cancer cells
specifically or immune
cells specifically.
And this was more
than 10 years ago,
way before anyone
believed in messenger RNA.
And so when we
presented this work,
most people thought it
was kind of like sci-fi.
And so I think the world
wasn't really ready yet
for this kind of technology.
So we didn't immediately
think we were going
to spin it off as a company.
It wasn't until
several years later,
when we saw that companies
like Moderna, BioNTech,
they were gaining financial
traction, at least,
thinking about IPOing.
And that's when we
thought that maybe,
maybe with these resources
behind these technologies, maybe
now there's a 10% chance of
success that mRNA could achieve.
And that's when we
spun out the company,
and we started in early 2019.
Thank you.
Cullen.
Yeah, I wish I could say
that I was two years old
and I wanted to start a company
on gene delivery technology.
And this is the fulfillment
of a lifelong dream.
That's not the way it happened.
It started for us, really,
with the grant from DARPA.
So I was actually excited
to hear about the new grant
opportunities.
DARPA kind of took a risk on
some technologies from my lab,
and this was 2013.
I hadn't worked in the space
of gene delivery at all.
I just had a crazy
idea, and they
were willing to put
some money towards it.
And I had a great post-doc
at the time, Paulo Garcia,
who became my co-founder.
And we developed this
technology for gene delivery.
And around 2016, he would
present at conferences,
and people would come up
to him after the talk,
say, this was awesome.
That was amazing.
Where do I buy it?
Where can I buy this thing?
And we're like,
oh, I don't know.
We weren't selling it.
It was very much we
started the company
to satisfy a market need.
We didn't get into
the space thinking
we were going to
start a company,
and then we're going
to be rich and famous.
That still hasn't happened yet.
[LAUGHTER]
We got into it because
there was a market need,
and we felt it was really clear
that for some research at MIT
and for some research, the
end of your progression
is publishing a paper
or a student's thesis.
It was really clear that in this
case, the end of this research
was in the marketplace.
It had to get out there where it
could potentially help patients.
And I would say
there was another--
there was a lot of
serendipity also in our story.
We founded the company
right around the same time
as the engine was being founded.
And so we submitted
an interest statement
saying we were interested in
learning more about the engine.
And they reached out to us.
I'll never forget
meeting with Katie Rae.
They didn't even have
offices at the time.
We met in a random
conference room at MIT
to talk about our technology.
And she said, I really want
to put gasoline on this
and accelerate it forward.
And they took a risk on us,
and we started the company
based on that kind
of market pool
and the interest
from the engine.
A fantastic story.
Tasuku, what
challenges did you face
taking your academic project
and turning it into a startup?
Yeah.
So the objective
of a company is so
different from
academic research.
As a company, we need
to be really laser
focused on value creation.
And so to me, it comes down to
company culture and the team.
And company culture, it seems
like such a nebulous concept,
but it goes hand in hand
with company strategy.
So what is it that
you choose to do?
What is it that you choose not
to do and how do you do it?
And so that goes hand in hand
with the company's mission.
And our mission
is to cure cancer.
And so every day, we
ask our employees,
what is it that you're
doing specifically
that creates value towards
that goal and mission of curing
cancer?
And if it doesn't really
contribute to that,
let's drop it.
And then really focus
on the activities
that lead to value creation.
And then on just setting
the company culture,
and the strategy isn't enough.
The team needs to
execute on that.
And so from a team
perspective, of course,
you need the really talented
people, the intelligent people.
But that's not enough.
You really need grit.
So as a startup, you're facing
crises all the time, right?
And it's so easy to give up.
So you need not just
intelligence but that grit
to hang in there and figure out
how to solve these problems when
they come.
But that isn't enough either.
You need to be a real team
player because drug development
is a team sport.
And so in the early
days, we brought
in a lot of talented
research scientists,
a lot coming from MIT as well.
And once we were successful
with the drug discovery
phase of the
company, then it was
time to really bring in a lot of
industry experience folks, that
can work on the development.
Our SVP has decades of
experience in industry
developing gene therapies.
Prashant Nambiar.
And then we also brought in
some folks that can really
manufacture the drug.
That's super important,
or else your drug
isn't going to be successful.
So Joe Barberio is
our vice president
of technical operations
that is really
experienced with lipid
nanoparticles, messenger RNA
drug development.
So yeah, with that, now we're
a clinical stage company.
And I think we're
proving out our concept
of taking these really
potent mechanisms
and curing cancer patients.
Cullen, as a professor,
as an academic,
do you encourage your students
to do entrepreneurship.
And if so what message
do you give them?
Yeah, that's a great question.
These days, you almost don't
need to encourage them.
Probably half the
students that come to you
are already interested in it.
So I actually spent a lot of
time discouraging students.
[LAUGHTER]
I think there are
two poor reasons
to want to start a company.
Then I'll give a good one.
The two bad ones that
you see, unfortunately,
being an entrepreneur has
become a bit glamorized,
and you can tell
that there are some
that are interested in
being an entrepreneur
because it's going
to sound really
cool to tell their friends
they're raising money.
That's a terrible reason
to be an entrepreneur.
There's a second one,
and you see this a lot.
And it's a hard one.
But people get really excited
about their technology.
They love their science.
They love their
technology. so much.
They just think everyone else in
the world is going to love it.
And the truth is that
the world doesn't
care about your technology.
They care about what it can do.
They care about the solution.
And so that gets me to the good
reason is, you need a good, why.
You need a good, who
are you going to help,
whose life is going to change by
the success of your technology
or of your company.
If you have a
compelling why, that's
a good reason to be an
entrepreneur because that's
the thing that's going
to keep you going
those late nights, those long
days, when you're raising money,
which you'll always
be raising money.
That's the thing that's
going to keep you going.
So I usually tell students
like, find a good why,
find a good problem.
That's the thing that you
want to be motivated by, not
by the technology.
The technology is nice.
The technology is
nice to get a degree,
but not nice for
starting a company.
That's a wonderful lesson
for our students here.
Tasuku, I understand
that you took
a bit of a detour between your
time at MIT and starting Strand.
Can you tell us a
little bit about it
and how it may have
influenced your path?
Yeah, and a lot of this was
really serendipitous, right?
It wasn't like this was
my grand plan in life.
So I met my wife while
I was here at MIT,
and she was a visiting
student from Belgium.
And so after I
finished my postdoc
so that I can be with
her, I moved to Europe.
And when I moved to
Europe, I decided
that getting outside of the
MIT environment, it's so hard.
You get spoiled with the
resources that you have here.
And I thought, maybe
I'll try something else.
And I joined an asset management
company, a biotech investment
fund, in Brussels.
And so I was a so-called
buy-side analyst.
And so we had
billions of dollars
of investor money invested
in public biotech and pharma
companies.
We had more than 100
companies in our portfolio.
And my job as an analyst was
to follow obsessively every
single move that these
public biotech and pharma
companies make.
And so I analyzed the platform
technologies, the pipeline
drugs, and all the
diseases that are
being pursued by the biotech
and pharma companies.
And so I didn't necessarily
really love that job, per se.
But it was so educational in
learning about what people are
doing in the biopharma space.
And as Cullen
mentioned, that helped
us think about what are
the unmet medical needs
that this next-generation
messenger RNA
technology can address?
And that led us to
focusing on cancer
and immuno-oncology using
Strand's messenger RNA
technology.
Wonderful.
Well, maybe this is
the last question,
given how much time we have.
Are there any lessons
that you learned,
advice that you'd give
to your former self
when you were just starting out?
Yeah, I'd probably
give myself the advice
I just gave all of you, which
is like, don't forget your why.
I think as you get
into it, you can
start getting distracted by
all kinds of things which
seem like success or seem like
milestones along the way--
hiring people,
finding new space,
raising money,
finding customers.
Those things are all great.
But don't forget
the core thing that
motivated you to get
into it in the beginning.
Think of the faces
of the people that
are going to be saved and
helped by what you're doing.
And maybe try to keep a
good connection with them.
In the technology that
we are working on,
it's a gene delivery technology.
And so we interact more
with customers who then
will interact with patients.
And so you can feel
a little bit removed.
And early on, I
think it might have
been helpful to just continue
to remember why you're doing it
and who you're doing it for.
Oh, that's a great message.
And so I want to thank
you both for, really,
a fascinating conversation.
I think what came out to
me in these conversations
with you is really that that
exemplifies what's really
best about MIT, people working
across disciplines, really
focused on addressing some of
the really challenging problems
for society, especially
in health care.
Let's give them a big
round of applause.
[APPLAUSE]

---

### MIT HEALS Launch: AI plenary session
URL: https://www.youtube.com/watch?v=WCkj484Lw7Y

Idioma: en

Hello, everyone.
My name is Asu Ozdaglar.
I'm the head of
electrical engineering,
computer science
department, also
the Deputy Dean of the
Schwarzman College of Computing.
I'm delighted to introduce the
session on AI and the health
and life sciences.
It's a very exciting
time for this initiative
and its intersection with AI.
Advances in AI and
machine learning
are transforming not just
every aspect of our lives,
but also the fundamentals
of academic inquiry.
And life and
medical sciences are
very much at the
forefront of this
with the promise of accelerated
scientific discovery
and personalized
treatments in medicine.
While the opportunity
is clear and present,
realizing it is
absolutely not trivial,
it really requires a two way
exchange between biologists,
chemists, chemical engineers
and health care professionals
on one hand, and those at the
forefront of AI and machine
learning on the other.
And this session aims to
contribute to this exchange
by highlighting early
innovative advances from MIT
and how we can make this
aspiration a reality.
I'm delighted to introduce
our four amazing speakers.
First is Joey Davis, Associate
Professor of Biology.
Joy will focus on
structural dynamics
of multiple molecular
machines that
are essential for performing
critical cellular functions,
including synthesis and
degradation of proteins.
He will present his
recent groundbreaking work
that develops a novel
computational method called
cryo DRGN and leveraging
deep learning to analyze data
from cryogenic electronic
microscopy, which enables
to image millions of
single molecules rapidly,
but also extending this
method to determine
dynamic structures of protein
complexes directly in cells.
The second speaker is Caroline
Uhler, Andrew and Erna Viterbi,
professor of Engineering
in the Department
of Electrical Engineering
and Computer Science,
also core member of
Institute for Data Systems
and Society and Broad Institute,
and the director of Eric
and Wendy Schmidt Center.
Caroline will highlight
the big opportunity
coming from rapidly growing data
sets in biomedical sciences.
Of course, there's great
promise in understanding
programs of life, but also a
phenomenal opportunity for AI
and ML in creating
inspirations and problems
for the foundational techniques.
Of course, I think the
interesting part about her work
is that she's going to
highlight that, AI applied
to biomedical data
will require focusing
on causal mechanisms rather
than simply correlations
and forecasts, where-- which is
the focus of machine learning
in several other domains.
And she will actually talk
about the roadmap that
combines causal inference
and machine learning
to understand various key
questions in cell biology.
Our third speaker
is Marzyeh Ghassemi,
Germeshausen Development
Professor in MIT Electrical
Engineering and Computer
Science Department,
as well as Institute for
Medical Engineering and Science.
Marzyeh focuses on decision
making in health care.
And she's going to highlight
the importance of evaluating
not only model
performance, but also
the heterogeneous
performance of these models,
including biases when
a model is applied
to specific demographic groups.
She will investigate whether
the AI models currently in use,
use demographic
information as shortcuts,
and whether biased decisions
follow from such shortcuts.
But of course, with
a positive spin,
she will also tell us
how to actually de-bias
these outcomes in
a way that sort
of calibrated to each
distinct input and query.
And last but not least,
our fourth speaker
is Dimitris Bertsimas, the
Boeing Leaders for Global
Operations, Professor of
Management and Operations
Research at Sloan School,
as well as the Vice Provost
for Open Learning.
Dimitris will
conclude the session
focusing on precision medicine,
providing personalized care.
How do we provide personalized
care to every patient?
Of course, the key
here is to understand
the variation of
treatment effects
across patients, meaning how
the same treatment will affect
different patients differently.
Gold standard approach is to use
randomized controlled trials.
But of course, this
is very costly.
And we have tons of
observational data.
So he's going to present
a terrific approach that
enables the use of observational
data by removing unobserved
confounding as if
they were randomized,
thus opening really the road
to true personalized medicine.
With that, I'm a very
delighted to invite you.
[APPLAUSE]
All right.
Thank you for that
wonderful introduction.
So I thought I would
start with some slides.
Oh, do I need to advance them?
I wanted to start with a
historical perspective.
This is a quote from renowned
physicist Richard Feynman.
And he's sort of describing the
interesting biological problems
of the day.
Things like how do cells
deal with DNA mutations,
how do they perform
translation, that
is, how do they
synthesize proteins,
how do things like
photosynthesis work.
And he says, "It's
very easy to answer
many of these fundamental
biological questions.
You just look at the thing."
And I really love this quote.
It sort of captures the
promise of structural biology.
That is, if we could
determine structures,
we could inspect
those structures
and use them to infer sort of
chemical mechanisms of action
to really understand
how these machine works.
He goes on to say, and
lament, unfortunately,
"The present microscopes," and
here he's talking about electron
microscopes, "they see that a
scale that is just a bit too
crude.
If we could make
them more powerful,
many of these problems
would become easier."
And so when we think about what
he was describing at the time,
we can go and look at a
micrograph from around that era,
this is a human cell.
And it turns out all
these little black dots
you see are ribosomes.
These are the machines
that are performing
synthesis in the cell.
And when you look at
them, it's quite obviously
it's very difficult to figure
out chemically how are they
working.
What are the atomic
contacts that allow
them to perform translation?
Because the microscope
was so crude.
If we flash forward roughly 20
years, microscopes got better.
People were able to infer sort
of low resolution structures
of ribosomes shown here.
They sort of look
like clay pottery.
We got a general sense of
the shape of the molecules,
but we really still couldn't
see the important details.
So the subsequent 40
or so years of research
led to huge advancements, both
in hardware and importantly,
in software, that now allow
us to routinely determine
structures to high resolution of
important biological molecules.
And just sort of capstoning
Feynman's predictions,
I've pulled a couple
of recent papers.
Here's a complex that was solved
determining how this protein
complex repairs DNA mutations.
Our group and many others have
used cryo electron microscopy
to look at the
process of translation
and understand how these
ribosomes work in atomic detail.
And there's a
recent study showing
how one of the
photosystems involved
in photosynthesis works.
So how does this actually occur?
When we perform cryo
electron microscopy,
we're going to take a sample
of interest, typically highly
purified in one state
that you're interested in.
You'll apply that to an electron
microscope grid shown here.
And then you're
rapidly going to freeze
the sample in a thin
layer of vitreous ice
that sort of freezes the
molecules in whatever
conformation they were in.
And then we're
going to image it.
So this grid is about
3 millimeters across.
An electron microscope allows
us to magnify it up to 100,000
fold.
Here I'll show you a
30,000 fold magnification.
This is a copper grid.
You can see the grid bar sort
of separate these into squares,
and in these squares there's a
thin layer of carbon into which
we've etched small holes.
Suspended across these holes as
this thin layer of vitreous ice.
And as you actually
look into the holes,
you start to resolve,
miraculously,
individual molecules.
So these are single protein
molecules directly imaged.
How do we use them?
We're going to use very
typical machine learning tools,
such as convolutional
neural networks to isolate
the particles of interest.
We can extract those
from the micrographs.
And now we have this
sort of ensemble
of data, which is
tens of thousands
to millions of particle images.
We're going to use
computational approaches
to then align all these images.
And once we align
them, we can average
across them, which reduces
the sort of noise increasing
the signal to noise ratio.
And that allows us to now
solve these high resolution
structures.
Now inherent in this
notion of averaging
is the basic assumption
that these molecules are,
in fact homogeneous.
That is that they're all
in the same conformation.
They all have the
same composition.
Anyone that has kids realizes
realize that biology is not,
in fact, static.
It is highly dynamic.
And so we start to think
about, say, drug development,
trying to understand
what types of pockets
we might want to dock a ligand.
We can think about kids.
So here is a small
molecule docking problem.
The task is to get the
ligand into the binding site.
If you look at a
static structure,
this looks like it
should be an easy task.
And when you look at the
dynamic version of it,
there are all sorts of motions.
There are all sorts of dynamics
that complicate the problem.
All right.
So what we really want to do is
look at the dynamics of this.
[APPLAUSE]
So how are we going to do that?
We began to think about
other ways to use this data.
So typically you would
take your set of images
and infer a single structure.
We thought instead we could use
these images to learn a mapping
function.
So this mapping function
at its base level
is going to take
these images and learn
to map from some position
in Cartesian coordinates
to some density at
that same position.
We first showed that
one was able to do this.
And then later,
we realized, well,
we could modify this
function to now take
as a second input
some condition.
Say, where is this molecule
in a low dimensional energy
landscape?
How do we do this?
So this is the work [INAUDIBLE]
that also introduced.
This was done in collaboration
with Bonnie Berger's group
in CSAIL, and led by my
first student, Ellen Zhang.
We built a variational
autoencoder,
consists of two parts.
There's a decoder network
that does exactly what I just
showed you.
It's going to predict the
EM density at some position.
The way it learns to do that is
it generates these model slices,
compares them to
the input images.
If they're the same, the neural
networks stay left alone.
And if they're
different, that's a loss
that we can backpropagate
to that decoder network.
We couple that to
an encoder network.
That will map these particles
into a low dimensional space,
and you can think about
them as sorting them
based on their structures.
So the particles
get sorted based
on what they're similar
to, and the decoder can
generate structures from those.
So how do we use this?
Here's an exemplar
data set we analyzed.
This is a spliceosome.
So this is a machine involved
in splicing mRNA together.
If you look at the
static structure,
there's regions of it that are
highly resolved shown in blue.
And there's other regions that
we weren't able to resolve.
Those are shown in red.
We hypothesize that those
are likely in motion, which
made it hard to resolve them.
When we pass this
through Cryo Dragon,
we can now start to see
this molecule in motion.
You see these
dancing motions that
help you understand the
dynamics of the molecule.
So with this in hand,
we began to think, well,
if we can resolve sort of highly
purified molecules like this,
perhaps we can push
this further and begin
to resolve sort of
heterogeneous structures out
of say, cell lysates.
And we turn to translation
as a way to test this.
It was well known
that ribosomes go
through all sorts of
conformational and compositional
changes as they perform
their functions.
And about 20 years of work,
using primarily cryo-EM
has resolved key
states within this.
And we said, well, if
these states are important,
they should be
populated in cells.
And if they're
populated in cells,
then we should be able to
just put sort of cell lysates
directly onto grids,
image them by cryo-EM,
sort the particles with
something like Cryo Dragon,
to resolve each one of
these individual states.
And so I had a series of really
wonderful students and postdocs
that worked on the
technology to do this,
and I'm excited to say this
actually worked quite well.
So we can go through that
whole set of translation states
that I just mentioned.
We can see initiation
sort of building up
more and more factors
joining to the particle.
We can look at translation.
We can see tRNAs in different
states within the ribosome.
We can see these
ratcheting motions.
And again the remarkable
thing about this
is that there's no
purification upstream.
We're just putting lysates
directly onto a grid.
And we can fill out this
diagram and basically
one or two weeks of work,
versus these decades of work.
What do we really want to do?
We want to understand how
these machines work in cells.
So how do we go about that?
There's a related technology
to electron microscopy called
electron tomography that
allows us to look at structures
directly in cells.
And so Barrett Powell,
a student in my lab,
applied some of the
similar approaches
to Cryo Dragon and
particle picking
using convolutional
neural networks to try
to apply this to tomography.
And so what I'm going to
show you here is a movie.
This is a bacterial cell
imaged by tomography.
As we slice through that cell,
you'll see the sort of membranes
on the right side of the image.
All these white
sort of speckles.
Those are ribosomes.
We can then segment all of those
out to populate the tomogram.
We can then extract
all of these particles
and use Tomo Dragon
to try to find those
that are in similar states.
And there was a
little trick here.
We had actually
treated the cells
with a small molecule
named chloramphenicol
that binds to the
ribosome and inhibits it.
And what we were
really excited to see
is that the resolution of
the maps that came back
were high enough for actually
to see the small molecule
drug bound.
Oops, jumped ahead there.
So this is a small
molecule therapeutic
bound to its functional target
image directly in cells.
With that, thank you
so much for your time.
[APPLAUSE]
I'm super excited to be here
and see this establishment
of this HEALS Collaborative.
I think it's really exciting to
see all this vibrant community
come together here and see
all the collaborations that
will actually come out of it.
So I'm here to talk to you
about the unique opportunities
that I see at building this
two way street between machine
learning, AI, and the
biomedical sciences,
and why I believe that the
life sciences are not only
uniquely suited for
being one of the greatest
beneficiaries of research
in AI, but actually also one
of the greatest
sources of inspiration
for foundational
developments in AI.
So in the life sciences and
the biomedical sciences,
we've seen a huge
data explosion where
we're now able to
measure the same system
in many different modalities.
Think of single cells where
we can, in one experiment,
measure the activity
of all genes
at the single cell level in
a million cells at a time.
Or all of the
imaging data where we
can image at subcellular
resolution, cellular resolution,
tissue scale, organ scale,
even whole body scale.
And we have all
these biobanks where
we have access to
health data of millions
of individuals combined
with genetics data
of these individuals.
So I think the real
power here comes
from thinking about
how to integrate
these different views
of the same system
into some joint space.
And so that's really where
representation learning,
the field of representation
learning in AI
comes in, where you're trying
to identify information
about all of these
different modalities
and combine them into
one view that is most
informative about the system.
And so that has
unique opportunities.
You can think of multimodal
biomarkers that better capture
a disease, because
it can combine
all these different views.
Or you can think about
the opportunities
for experimental
design to think about,
what is information
that is actually
captured by different
types of measurements?
Say a cheaper measurement,
easier to obtain measurement
like a heart ECG, or a
much harder to obtain
and more expensive
measurement like a heart MRI.
If you have shared information,
I could just translate from one
to the other.
Or what is information that
is really only contained
in a particular measurement,
like a particular protein
stain, where you really need to
get that measurement in order
to get that information out?
And maybe you can
go even further.
In the life sciences,
generally, we
care about getting at the
underlying mechanisms,
moving from just
associations and biomarkers,
actually moving on to
mechanisms and causality.
So maybe multimodality can
actually help us there.
And so that's the question
I want to talk to you about.
But before-- let's
make the case.
And I think I
probably don't even
have to make it
here-- for why we
need mechanisms and causality.
So let's take a simple example
of fibrosis, where it's still
a disease that actually has
a huge effect on mortality.
And there is still no drug
that can revert fibrosis.
Now, how do we think about
identifying drug targets
in this setting?
Well, we need to identify--
and I should say, fibrosis
is when the tissue stiffens.
And so what you want to
identify are genes proteins
that are upstream.
So causal of tissue stiffening.
And not genes,
proteins, that are
associated with
tissue stiffening,
but are downstream of it.
Because these would be
very bad drug targets.
So it's really critical to
understand what is cause
and what is effect.
For example, for identifying
a drug candidate or protein
candidates in this case.
However, if we're thinking
about the current AI models,
they generally fail
miserable in causal tasks.
And so here I just want to give
you a very, very simple example.
So we all know that good
warm weather and sunshine
is a cause of increased
ice cream consumption.
Right?
But if you now go out and
you have a gloomy cold day
and you give out a whole
lot of free ice cream,
well, you're increasing ice
cream consumption artificially.
That doesn't change the
weather, because weather
is upstream of it.
However, if you ask Dall-E so
OpenAI's model for generating
images from text, and you
should all do this experiment,
you ask it to draw an
image of an ice cream stand
where they're giving
out ice cream for free.
It always puts good
weather in the background,
because there is a really
strong association between ice
cream and good weather.
And these models are trained on
learning based on associations.
They are not trained on
trying to identify what
are the causal relationships.
But that's critical,
as I just showed you
in the previous example.
We need AI models that can
actually reason causally.
And that's exactly
the types of questions
that we're working on in my lab.
So how do we?
What does it mean to fully
understand the mechanisms
in a biological system?
So I guess
traditionally we would
love to get to these pictures,
these amazing pictures here
by Davidson, for example, of
really understanding fully
how one gene protein turns
on another gene protein,
maybe even knowing all the
differential equations above it,
and really having this
full causal picture.
Now, the framework for thinking
about these regulatory networks
causally has been already
developed in 1920s by Sewall
Wright.
And so that has been like--
the field of causality has been
a very, very vibrant field
basically since the 1920s.
Now, in causality, actually,
until now, basically, there
has always been one very
important assumption.
That we know what the
nodes of the network are.
So say we assume they are genes,
and we have data on the genes,
like the activity of all genes.
And we would like to infer
the regulatory relationships,
these causal edges
between the genes.
But I'd like to argue that in
all modern problems in the life
sciences, we actually
don't even know what
the nodes of the network are.
Like let's take the example
of cholesterol, for example.
Where for a long time we
thought that total cholesterol
was one causal variable.
Well, now we know
that's not the case.
We actually have two causal
variables, LDL and HDL.
And they have different
causal effects.
So we don't actually know.
We have to learn what
the causal variables are.
Similarly in images, pixels
are not causal variables.
We have to learn what
the causal variables are
in any of the imaging data
sets that we're collecting.
Is it the shape of
the cell nucleus?
Is it a particular protein,
and where it is localized?
Is it how much of it is
and where it is localized?
So we have a really more
fundamental or really new
foundational
question in AI, where
we have to first learn what
even the causal variables are,
and then from there go
on and actually learn
the causal networks among them.
So over the years, we've
taken on this problem
and really thought
about how, or identified
that actually having
access to multiple views
of the same system, so
different measurements
or different data modalities
on the same system
can really help us understand
the underlying latent
causal structures.
So why is that?
Well, so let's think about what
is hard about understanding what
are causal relationships
is you need
to sieve out all kinds
of spurious associations.
So multimodality can
help you do that,
because we're looking for--
or a causal relationship
in the system
as an intrinsic relationship
should be there,
whether you're
looking at the system
from the left, meaning
with measurement type 1,
or from the right meaning
with measurement type 2,
or from the top or
from the bottom.
So in order to
really identify what
are these causal relationships
or causal features,
you should be looking for
what is invariant to how
you're looking at the system.
So trying to understand which
information or what information
is really shared between
different modalities
can actually help
you get at causality.
And so that's really
exciting because we
have all these
different measurements
of the same biological system.
And so we're always thinking
about, well, how can
I identify what
is actually shared
between these
different modalities?
And then actually
learn from there.
And so I just want to end with
one example that we're currently
working on in ovarian cancer,
where we can, for example, think
about, well, what is shared
among abundant pathology images
that we have access to, and
newer and much more expensive
spatial transcriptomics
data, and actually
see that there is a lot
of shared information.
It's actually possible to
translate between them pretty
well.
And you can not
only identify what
is shared among
different modalities,
but also what is shared
or specific to particular
individuals or patients.
Now, that's a really
important problem.
Anyone here who has worked
with, say, transcriptomic data
or basically any
patient data knows
that if you look
at these data sets,
you have such big batch effects
that basically every patient
always looks its own Data Cloud.
So it's very hard to see
what is actually shared
and what is specific to
a particular patient.
But if you just use these
multi-modal causal features,
you can directly identify, in
this case, cancer niches which
are specific to a
particular patient,
and other cancer niches
that are actually shared
between different individuals.
And identifying these
cancer niches, as we know,
is really critical for
informing therapeutic strategies
down the road.
So I hope I was able to give
you an example of showing why
I think it is so important
to build this two way
street between AI and
the life sciences.
I think that like RA
Fisher, was motivated
to develop modern statistics
based on applications
to agriculture and to breeding,
I think here we have the life,
the modern life
sciences is really
going to be a driver for new
foundational questions in AI
that will really inspire maybe
even deeper types of questions
that we have seen so far.
Certainly that's the case
in the field of causality.
So thank you very much.
[APPLAUSE]
Hi, everyone.
I'm Marzyeh Ghassemi,
and I'm going
to talk to you today about the
pulse of ethical AI and health.
My lab is the Healthy
Machine Learning Lab at MIT,
and we focus on
evaluating whether there's
a need for machine learning
in a health care setting.
Developing a robust private
and fair model if there is,
and then potentially
deploying that model,
understanding how people
might use that model for good.
What I'd like to
start out with--
and there will be a
quiz later, so remember.
You're all going to--
its final season at MIT.
What I'm going to start out with
is the first learning, which
is at this point, maybe based on
the presentations you've already
seen, you should know
that with some caveats,
we can predict or
generate almost anything
if we have the right data
and the right models.
These are two really
fantastic examples
of AlphaFold for
protein folding,
and robust prediction
of breast cancer risk.
And in both of
these cases, these
were tasks that we
weren't very good at.
And that machine
learning really improved,
transformed the practice in.
That's really exciting.
How do we do that?
Well, the way that
we do this now
is by following this pipeline.
This is pretty well established.
We choose a problem, collect
some data, define an outcome,
develop an algorithm, and then
potentially deploy that model.
And if I walk you through
this for another prospective
application, let's
say that I want to do
X-ray triage at a hospital.
So if you're healthy I
want to send you home.
How would I do that?
Well, I already
picked the problem.
And so I would go out
and get some data.
There's three really large,
publicly available chest X-ray
data sets in the US.
Over 700,000 images.
I would train a model.
Here I'm training a type of
convolutional neural network
to predict no finding,
which means you're healthy.
And then I would look at what
the performance of that model
was.
And here I get an AUC of 0.859.
That's really high.
Random is 0.5, 1 is perfect.
So this is a really good score.
And many people
don't just stop here.
They go all the way
to FDA approval.
And in fact, there's now many
hundreds of FDA approved devices
with AI/ML in them
that are being
deployed across a wide range
of clinical subspecialties.
This is really exciting.
But the second thing
to keep in mind
is that predicting or
generating something,
well, for the average,
especially in health,
is not good enough.
Why?
Let's go back and
pick on my own work.
That performance
right there, 0.859.
That's really good.
But that's one number.
What happens when I see
the individual performance
numbers for different
subgroups of patients?
When I do that audit, I find
that this model underperforms
in female patients, young
patients, Black patients
and patients on
Medicaid insurance.
And actually, if you exist in
an intersectional identity,
if you're a Black or
Hispanic female patient,
then you do significantly
worse than if you
were a part of a larger
aggregated identity.
And this is happening in part
because medical data is really
complex, and you can
infer just by looking
at somebody's medical note
or their medical image what
their demographics are.
What their biological sex is,
what their self-reported race
would be.
And this is even when you
don't give this information
to the model.
This is also in situations
where we've tested doctors
and they can't tell based
on the medical image
or the medical note.
And so what this means is that
it's actually pretty complex
to make a model optimal
in a high dimensional,
multimodal setting.
How hard could this be?
Well, let's take a look.
Medical imaging is one of the
spaces where AI is applied
to most frequently in health.
And so we looked at
a couple of settings
where it's frequently applied.
This would be for a chest X-ray
based diagnosis for dermatology
and for ophthalmology.
And in each of
these cases, we're
trying to predict an
important clinical outcome.
And we want to look at
how poorly these models do
in different subgroups depending
on how we optimize the model.
The first thing we found is that
the more that a model secretly
learns in the layers to encode
your demographics, for example,
that you are
female, the worse it
does at predicting the clinical
target for that subgroup
that is minoritized, in
this case, female patients.
Put another way, if
you're training a model
to predict pneumonia,
and it learns who is male
and who is female, it will
predict pneumonia worse
for female patients.
We then tried to
fix this, and we
found that if you take data
from hospitals in Massachusetts,
train the model and
force it to be optimal,
meaning all the way on
the bottom right hand
part of that plot, very,
very performant on average,
and on the bottom,
very low fairness gap.
Then when you test the model
on patients from Massachusetts,
it does really well.
And so that means modern
machine learning strategies
can be locally optimal,
if you deploy them
in the settings at
their trained in.
But that's not how we train
machine learning models
right now, even when
they're approved by the FDA.
We train them in Massachusetts.
They get approved, and then we
invite the world to use them.
And when we take this model and
we throw it over to California
and ask them to
test on their data,
we actually find that optimal
fairness doesn't transfer.
And this has important
policy findings for the way
that we audit models
and evaluate them
before they're approved in a
local setting, for example,
state to state.
And this gets even
harder when we
think not just about
deep learning models,
but vision language models.
Everybody's used a
vision language model,
I would assume at
this point, for fun.
As Carolyn said, they can give
you really interesting results.
And they're really good, not
just at image generation,
but they have been
used in medicine
and clinically adjacent tasks
for zero shot classification
and zero shot retrieval.
And they're very
good at these tasks.
But it's also been
pretty well established
that they have these
problematic human biases in each
of these settings.
For example, they always tend
to generate male doctors.
And one of the reasons
it's really hard
to de-bias these
models is because we
have a really simplistic
approach to thinking
about their mind space.
And so here, if I wanted to
de-bias a vision language model,
much of the existing
work assumes
there's a direction for male,
there's a direction for female.
Draw a linear projection
and collapse that down.
It's fair.
But actually these
models are reasonably
complex and non-linear.
And so maybe the direction
for male and female doctors
is very different than the
direction for male and female
pilots or managers.
And so in order
to get past this,
we actually have to
think about doing a more
complex, non-linear approach.
And so in recent
work, what we did
is looked at a zero
shot retrieval example
where we want to get
a picture of a doctor,
and we want that to be unbiased
between male and female image
generation.
So when I ask for a
doctor, maybe my query
gets embedded right
there at the star.
And if I did a
traditional de-biasing,
I would just project a line,
find this middle point between
these two spaces, and
say, "Now I'm fair."
But you can see here, because
it's complex and non-linear,
I'm still closer to all of
these pictures of male doctors
than female doctors.
So I would still just generate
a picture of a male doctor.
Instead, we propose an
approach to bend your debiasing
in the direction of the
complex non-linear space.
And so don't stop here.
Move along this
surface until you're
at a place where you're
equidistant from male and female
images, in this case.
And this is nice because
approaches like this
require no fine tuning of
your vision language models.
So you can just use a
vision language model
as it comes to you.
And they can be
tailored to de-bias
for existing or new
queries that come in.
The last thing I
want you to remember
is that good models,
even if we de-bias them,
they don't necessarily guarantee
that we'll have good outcomes.
And so we have a lot
of work in the past
that's shown that the
losses from poor AI
can be larger than the
gains from good AI.
And so we have to be really
careful that these models are
trained well.
We know that doctors are
susceptible to poor advice,
even when they have expertise.
Here we showed radiologists
incorrect advice
for X-ray images, and they
were still fooled by it.
And finally, we
have work showing
that the way that you
deliver bias content
is more convincing sometimes
than the bias content itself.
So if you tell somebody to call
the police on minority patients,
they're much more likely
to listen to it than if you
say there's a risk of violence
at an equal rate for minority
patients.
So we need to pay
careful attention
to how we deliver
advice from models,
not just the advice we give.
And so the four things you need
to keep in mind as you build
models along this pipeline
are, we can predict,
or generate anything.
But doing well on average
is not good enough.
We need to know that making
models optimal as complex,
otherwise they already
would be that way.
And good models won't
improve outcomes.
Instead, we need to move
forward with creating
actionable insights
in human health.
Thank you.
[APPLAUSE]
Thank you and good afternoon.
Very glad to be here.
What I would like
to talk to you about
is an area that occupied
my thinking for many years.
And recently with
two of my colleagues,
we have made some progress
that recently appeared,
meaning recently in the
last month in nature
digital medicine.
So what is the story?
As many of you
know, of course, is
that the golden standard in
medicine and many other fields,
of course, is to do
randomized clinical trials.
For more than 100 years, this
has been the name of the game.
However, RCTs are very expensive
and often last very long.
Also, while RCTs,
the conclusions
are suitable for
the average patient,
they may be not appropriate
for not the average patient.
So there is a need
to identify a way
to understand what is
called heterogeneity
of treatment effects,
namely more personalization.
In contrast, observational
data are very widely available
and inexpensive.
One would argue that the cost
of developing a new drug,
which currently is of the order
of between one or two billion,
is definitely affected by the
fact that we have to do RCTs.
So what is the
challenge on this idea?
The presence of
confounding bias.
Confounding exists, especially
in observational data.
This is exactly the topic
that RCTs try to address.
At the same time,
RCTs, while strong
in identifying whether a
treatment works or does not
work, they do not define
effectively patient subgroups.
In other words,
they don't typically
personalize treatments.
There has been some
progress of what
is called one
variable at a time,
but I would say quite limited.
So what is the purpose
of this effort?
That we propose and
demonstrate using
one clinical example,
and later on many,
but I only have time for one.
A framework to reach
clinical conclusions
based on observational data.
And this is data that
we used in collaboration
with Memorial Sloan Kettering
that one of my collaborators
is from.
So in other words, we
can now have an approach
that can lead to medically
valid conclusions
based on observational data.
The second is that even
if we insist to use RCTs,
we have a way of
personalizing the RCT.
In other words, from an
RCT, you can actually
get personalized information
of how treatment should
be appropriately personalized.
So in my view, this
leads to a way of--
which is, in my
opinion, something
I have tried for a long time
to utilize observational data
than RCTs for
precision medicine.
So let me tell you, get
you a sense of the story.
We have some patients that
have a particular cancer,
of a particular type
of tumor, what's
called GIST,
Gastrointestinal Stromal
Tumor, that MSK specializes
in the observational data set.
So these patients have
already undergone surgery.
And the question
is, are we going
to give them a particular
chemotherapy called imatinib?
By and large today
in the world, we
offer imatinib as
a way of doing it,
even though there has been some
studies that suggest that there
are patients who don't need
it, and therefore would
have benefit by avoiding it.
So what is the data
that we do have?
So for each patient we
have some covariates,
or what is called
[INAUDIBLE] count
the tumor site and the tumor
size potentially others.
And for every one
of these patients,
these are observational data.
We also know whether the
treatment was applied,
whether imatinib
was given or not,
and whether or not
the patient recurred.
So we have the
data, the treatment,
whether the treatment was
given, and whether reoccurrence
happened.
Of course, there is confounding.
What is the notion
of confounding here?
That the treatment was
not randomly assigned.
So patients who were predicted
to have worse outcomes typically
tend to get the treatment.
So these are not--
this is not a randomized trial.
So the very act of
having the treatment
has a higher baseline
risk of recurrence.
So which is the key difficulty.
So what we try to
do, I will briefly
tell you, the key idea
of how to overcome that.
There are two components.
I will only have
time for the first.
But the key here
is to take a model,
take a machine learning model.
Take only people who did
not receive the treatment,
and build a classification
model of whether they
are recurring or not.
Notice that this is
only based on people
who did not get the treatment.
Then apply the model for those
who did take the treatment.
So now for every one
of these patients,
we have a probability
of recurrence
for both types of patients.
But the patient was only
developed for those patients
who did not--
I'm sorry-- did not
get the treatment.
Put them in buckets.
So now let's say we put them
in 10 buckets between 10%.
So within these buckets,
we have patients
that have similar
probability of recurrence.
Some of them got the treatment.
Some of them did not.
And then at the same
time, we try now to
in each of these buckets,
there are say 10 of them,
we try to make them in
each particular bucket
so that they are as
if it were randomized.
And how do we achieve that?
We try to use optimization.
We try to match the
characteristics from the data
that we have so that
patients within every bucket,
that includes both
treated and untreated,
are similar in these
characteristics.
And we use
optimization for that.
The specifics that, you
would not follow it.
I would not have followed
it if I were you.
So key here is that we
have a methodology now
that we, in each
of these buckets,
we are able to match
their characteristics.
Just to demonstrate the
process of confounding,
the process of the left.
The data on the left
shows the survival
among those who took
the treatment and those
who did not get the treatment.
And you observe that people
who got the treatment
are faring worse than the
people who did not get,
because they are healthier.
After the matching, after
this optimization process,
there is still confounding
involved, but it has decreased.
So there is remaining
confounding even
after this optimization.
So what we do as
a second step is--
and to demonstrate
the effect is that,
so this is the graph on the left
shows the percentages of people
in each of these
buckets who have
similar probability
of recurrence,
but who took the treatment and
who did not take the treatment.
What the optimization does
is on the right, in which we
have the same number of people.
So we actually delete patients
from the observational data,
but we don't delete
random patients.
We delete those patients so
that the remaining patients
are very similar in their
characteristics, as well as
the probability of recurrence.
And then as a final step, we
also, because as I mentioned,
there is a confounding
left, we actually
train a model where
we artificially
increase the weight
for those patients who
received the treatment who
did not have recurrence.
So the model tries to
make better predictions
for those patients
with a larger weight,
since this yields
greater improvement.
And when we are-- so in a way,
we clean the data in such a way
that we try to subtract
confounding from the data.
And then as a final step, we
apply machine learning technique
called optimal policy trees,
that was developed in my group,
to be able to personalize
the treatment.
And here's the effect
for patients from MSK.
So you can see that
although you might not
see the exact
personalization, you
observe that when you
look at a particular,
let's say the mitotic count,
if it's above a certain level,
you prescribe the treatment.
If it's is below a certain
level, you do other things.
So this is definitely
a personalization story
on the observational data.
So how do we know that
it has some benefits?
We actually took completely new
data, unseen data from Poland.
And in that particular
data we have observed
that while the sensitivity
of the trial data and ours
is similar, we were able
to actually find 7%--
the personalization
helped us spare
7% of the patients who
would not have benefit
from unnecessary treatment.
While the result is
relatively modest,
it shows that there is a
benefit by the personalization.
This is just on-- and
this is unseen data.
I mean, a completely
different study.
My final point is how to
now personalize trials.
So this has been so the benefit
of radiotherapy for patients
with what is called
truncal sarcomas
has been demonstrated
in two major trials.
So MSK, as many
other institutions,
utilize a radiation
therapy after surgery.
So that's more or less
the key therapy today.
However, it has been
reported that many patients--
some patients might
not benefit for that.
So applying a
similar story-- this
is now not observational data.
This is randomized
data for which we
don't know how to personalize.
So this data does not have
the effect of confounding
because it's randomized trials.
So we do the optimization
and this other part.
And we also find a
personalized treatment
on the randomization data.
So this is a general purpose
randomized clinical trial
that allows personalization.
And then we try this
approach for data from MSK
that are different from
the clinical trial.
The original-- the methodology
was trained by the randomized
clinical trial data in that
happened in the early 1980s.
And then we utilized the data
from MSK observational data
over 16, 17 years.
We have found that the
approach is equally
effective in
protecting patients,
but spares 15% of the
patients that need not--
they didn't need radiotherapy.
And therefore, the
personalization
helped in improving that effect.
So of course, since then
this was the original paper.
We have tried it in about
10 different studies,
from cardiology to oncology,
to other areas of medicine.
And we have found
across the board
that a benefit by the
personalization in studies
outside of the data we have.
So in summary, what I would
like you to remember from
this is that the area
of trying to find
an approach for personalization
just on observational data
is, in my opinion, a very
important aspect in medicine.
We hope that the approach is
definitely-- would definitely
help how to change
medical research,
and also perhaps improve and
decrease the cost of developing
therapies around the world
in a way that provides
a road to precision medicine.
Thank you.
[APPLAUSE]

---

### MIT HEALS Launch: Translation plenary session
URL: https://www.youtube.com/watch?v=kVmqwAnp7Oo

Idioma: en

Good afternoon, all.
Thank you for attending this
exciting session on translation.
This session is all about how
we get our discoveries out
of the lab and to the clinic,
into the hands of patients
and doctors and into companies.
My name is Ellen Roach.
I'm an associate professor
in mechanical engineering
and the Institute for Medical
Engineering and Science.
And I have the privilege of
chairing today's session.
We are so lucky to have four
distinguished speakers today
who will speak to us about
different aspects of translating
their science and technologies.
I could listen to these
speakers for hours,
but we have a short
amount of time,
so I'll introduce
everyone at a high level
now and keep the talks
moving afterwards.
So first, we have
Professor Andrew Lo.
Professor Lo is the Charles
and Susan T Harris Professor
of finance at Sloan and the
director of the Laboratory
for Financial Engineering.
Professor Lo will speak with
us about financing translation,
describing the inflection
point of biomedicine
and proposing new business
models to translate technology.
Next, we'll hear
about translation
through multiple startups
in the area of nanomedicine.
Professor Sangeeta Bhatia is
the John and Dorothy Wilson
Professor of Health
Sciences and Technology
and electrical engineering
and computer science.
She is a Howard Hughes
Medical institute investigator
and directs the marble center
for cancer and nanomedicine.
Professor Bhatia
will speak to us
about translating nanomedicine
technologies for cancer
detection and describe her
prolific co-founder timeline
in translating technology
out of her lab.
She will then describe how
she supports other faculty
in translating their technology
and the incredible impact
and success of the
Faculty Founder Initiative
she has created.
Professor Hugh Herr
will speak next.
Professor Herr is a professor
of media arts and sciences
at MIT Media Lab and the
co-director of the K. Lisa Yang
Center for Bionics.
Professor Herr's work
melds human physiology
and electromechanics
to produce systems
that emulate human biomechanics,
improving mobility for those
with paralysis and limb loss.
He will speak about how he
has translated technology
from the Biomechatronic
Mechatronics Lab
to enhance human
physical capability
through for-profit and
non-profit mechanisms.
Finally, Professor Alex
Shalek will speak to us
about engineering our biomedical
ecosystem for global impact.
Professor Shalek is the
J.W. Kiekhaefer Professor
in the Institute for Medical
Engineering and Science
and the Department of Chemistry.
He is also a member of the Broad
Institute, the Koch Institute,
and the Ragon Institute,
and the director
of the Institute for Medical
Engineering and Science.
Alex will describe how his
group develop and apply tools
to understand balance
in tissue ecosystems.
He'll describe the
opportunities at MIT
for translation, how we can
engineer the right solutions
to key problems, leveraging
global partners to scale
and addressing
systemic challenges.
He will conclude the
session describing
how this is achieved through
broad collaboration, which
will create a natural segue
into the next session.
I do hope you enjoy these talks
and you are inspired by them.
I thank the speakers in
advance for sharing examples
of their work and for imparting
their wisdom and philosophies
on biomedical translation.
Thank you.
[APPLAUSE]
So I'd like to start
by thanking Ellen
for that generous
introduction and for chairing
the session and
President Kornbluth
for putting together this event.
Thank you all for
being here as well.
I'm a financial economist
by training and trade.
And you might be wondering,
who ordered that?
What am I doing here in an event
of otherwise serious scientists,
clinicians, and
biopharma leaders?
Well, no matter what
field you're in,
no matter what you're
doing, no matter
what your goals are, at
some point or another,
if you're looking to
affect patient lives,
you will need money,
and that means
you'll be speaking my language.
[LAUGHTER]
So what I'm going
to be talking about
is how we engage
in the financing
of translational medicine.
And to begin, clearly we are at
a biomedical inflection point.
And I'll give you two
illustrations of that.
The first comes from
my world of finance.
This is a chart of the 30
best-selling drugs in the year
2000.
I'm sure you can't see
the specific drugs.
But highlighted in
blue are those drugs
that came out of
academia or biotech
four 26 out of the top 30
came out of big pharma.
And in the good old days,
that's how translation occurred,
really through big
pharma efforts.
Fast forward 18 years
to 2018, and now here
are the 30 best-selling
drugs in that year.
Highlighted in blue
are those that came out
of biotech and academia--
24 of the top 30, nine
out of the top 10.
Medicine has changed.
The second way.
We know that
biomedical advances are
at an inflection point is
because my MIT colleagues told
me so.
[LAUGHTER]
In particular, Susan Hockfield,
Tyler Jackson, Phil Sharp
in 2016 published a report
titled Convergence--
The Convergence of the Life
Sciences, Physical Sciences,
and Engineering.
All of the different
fields are coming together
to produce remarkable
advances for medicine.
And the insiders
in this industry
call it the omics revolution.
Every one of these omics has
experienced tremendous advances
over the last even five
years, never mind 10 or 20,
with the exception of one.
There's been one omics that's
been a bottleneck to progress,
and that is economics--
[LAUGHTER]
--because we got to
pay for this stuff.
And the business
models that we're using
are still tied to the
past and haven't really
taken into account all of the
innovation that's going on.
So I want to talk to you
about that omics economics
and, in particular, describe
how translation occurs
at a high level at a
caricature from my perspective
as an economist.
So what is the life cycle
of biomedical innovation?
Well, it begins here
in the laboratory
with a researcher and an idea.
And that idea sparks,
usually, a collaboration
of additional experts that come
together to take that idea,
round it out, put it into
various different tests,
experiments we call them.
And ultimately, when
all is said and done,
and you collect the data,
and you've got the results,
and it looks like
you've got something
that might one day help a
patient, what do you do?
You publish a paper, right?
[BUZZER]
Well, wrong.
[LAUGHTER]
Do not do that.
What you are supposed to do--
and it's remarkable how many
life scientists
aren't aware of this--
you file a patent disclosure,
after which you are totally
free to publish our paper.
And this is important because,
while many people think
that filing a patent is
somehow dirty and too
commercial and
pecuniary, and you're
getting involved in the world
of money, what you're doing
is reserving the right for
somebody else, typically
an investor, to basically
benefit, get a rate of return
for deploying capital
to take that idea
and translate it into medicine.
So after you publish a paper,
do some more experiments.
You're able to now
demonstrate, perhaps
in animals, that maybe this
actually might turn into a drug.
And so at that point, you're
able to get your patent,
be able to create a
company, perhaps rent
some space at Lab Central,
and then ultimately start
going through the very
long and important process
of clinical testing, at
the end of which you might,
if you're lucky--
[CHA-CHING SOUND EFFECT]
--get a drug.
But more often than not--
[EXPLOSION SOUND EFFECT]
--you will fail.
And that, in a nutshell, is
what translation is all about.
What makes this process
very challenging
is that you need money
every step of the way.
And if that money is
interrupted for any reason,
then the whole thing
pretty much ends.
Biomedical innovation is
binary, and that kind of risk
is something that most
investors are simply not
able to withstand, so we
need better ways of financing
translational medicine.
And that's what I would like
to talk to you about today.
Here's an idea-- instead of
developing drugs one at a time,
why not pool these
various different drugs
into a portfolio, maybe
even across universities?
Imagine, for example,
a business model
that looks like this, where
on the left hand side,
you've got multiple
universities pooling
their intellectual property to
create a vehicle that investors
can participate in.
Why multiple universities?
Well, because MIT
can't do everything,
and so we have to leave
room for some other folks.
The bottom line is that we don't
know where the next breakthrough
is going to come from because
of the complexity of the life
sciences, and so we
want to be able to have
multiple shots on goal even
beyond a single university.
But if we can construct
that portfolio,
manage it, create programs
around the most promising ideas,
not necessarily requiring
faculty to give up tenure
and become CEOs when they
really don't want to do that,
and they want to
do their research
and let other folks who are
ideally suited to manage
the businesses
grow that business,
that's really what
we're talking about.
With this kind of
business model,
all of the various
different stakeholders
will get what they
want, and they
will be able to
contribute what they can.
What do we call
this business model?
Well, I call it
tech transferase.
[LAUGHTER]
And for those of you
who are laughing,
you are critical for the
success of this business model.
Thank you.
[APPLAUSE]
Hi, I'm Sangeeta.
I'm trained as a
physician and an engineer.
And I'm going to
tell you a story--
a case study, really,
that came out of my lab
in the field of
nanomedicine-- and use it
to illustrate how you can
translate through startups,
in the way Andrew described, and
how it helps to advance things
to patients, but also how it
can drive the next generation
of innovation here at MIT.
And then I'll close
with some ideas
about how we hope to
enable more faculty to do
in this next chapter
that MIT is embarking on.
So to begin, I want to tell
you what nanomedicine is.
It's nanotechnology
applied to medicine.
And nanotechnology
itself is a field
of materials that are
smaller than 100 manometers.
Human hair is 100 microns, so
this is 1,000 times smaller.
Now, this is a field
with two origin stories.
On the left, you see microchips,
semiconductor fabrication.
These drive all of
your smartphones.
And you may know the
new NVIDIA chip has
feature sizes of 4 manometers.
So using what we call
top down fabrication,
we can make very
teeny, tiny things.
The other origin story comes
from the world of chemistry.
On the right, what you
see are quantum dots.
These are when you
assemble nanomaterials
by collecting atoms and
creating emergent properties.
And this is an example of
quantum dots, something
that Moungi Bawendi shared
the Nobel Prize for chemistry
last year here at MIT.
And these materials
are really cool.
They're exactly the
same composition,
but they glow different colors
simply based on their size.
So there's a whole world of
material science and physics
happening at the nanoscale
where properties-- electronic,
magnetic, optical-- emerge
when you make them small.
And what's really
exciting in this moment
is that if you think about
what that does for medicine,
it turns out there's also
biological properties
at this length scale.
And if you think
hard about it, it's
not surprising that
this is the case,
because if you look at this
molecular ruler, actually,
the nanoscale is the scale at
which biology communicates.
DNA is 2 manometers in width.
A receptor is 10
manometers in size.
A virus is 100 manometers.
So that is the scale that
all of the interesting things
are happening.
And in fact, the field has
made many, many nanomaterials
that have been in clinical
trials and in humans.
In fact, many of you got
one of these nanomaterials.
The lipid nanoparticle
on the top
there was the encapsulation
for the mRNA vaccine,
both of the mRNA
vaccines for COVID.
So these materials have now
been in millions of people
and been proven to be safe.
And as a community, over about
20 years now around the world,
people have been
working in nanomedicine
and discovering what is the
biology of the nanoscale.
And I'm not going to go through
everything on the right,
but just to say that we're
finding out all kinds of things
about the size and shape of
a material, dictating how it
interacts with the human body.
So in my lab, what
we've been doing
is trying to think about what
could this do for cancer.
Now, cancer is a
disease that you
know could be
really best treated
if we could find it early.
And some cancers, unfortunately,
can grow for eight to 10 years
before being detected even by
the best screening technologies
that we have today.
So we had the idea, what if
we can make a machine that
can travel through the
bloodstream, look for cancer,
and when it finds it, report
it to the outside world?
This is how the machine works.
So here you see
100-nanometer particle.
It's circulating
in the bloodstream.
And it's going to leak out into
the surrounding tissue, which
is a tumor.
Once it's inside the tumor,
it's designed to detect enzymes
that tumors make.
Those of you who are
biologists, these are proteases.
They're depicted here in yellow.
And these enzymes can
activate these materials.
Because it's a
catalytic reaction,
each enzyme can release
1,000 reporters.
So this is an amplification.
And those reporters are now even
tinier than the first material.
These are about 3
manometers in size.
So when they get back
in the bloodstream,
now they are small enough to
be filtered and concentrated
twenty-fold by the
kidney into the urine.
So this starts out as a shot.
And an hour later, you collect
something in the urine that is
20,000-fold amplified for
every copy of an enzyme.
So we've been working
on this in the lab.
The first paper was
published in 2013.
We started a company
called Glympse Bio, like,
get a glimpse inside the body.
And we did exactly what
Andrew directed us to do.
We first filed a patent,
then raised the capital.
I co-founded it with
my postdoc, Gabe Kwong,
who is now a full
professor at Georgia Tech.
And that company did things at
a scale with dedicated teams
that we couldn't do here at MIT.
We did scale-up manufacturing.
We went to the FDA.
[INAUDIBLE] two
species toxicology.
We did human clinical trials.
And now they're on their
way in liver cancer,
trying to show that
this could work
to find liver tumors
in patients that
have advanced fibrotic disease.
So that's been very exciting.
That's been happening
outside of campus since 2015.
But what's exciting about this
is that the students in the lab
now are unlocked.
They can see that
this can happen.
There's a path to patience,
and they get more innovative.
And so what did they
start thinking about?
Well, what if you could make
a urine test like a pregnancy
test?
What if it could be on paper,
and you could read out a cancer
test at the point of care?
So we invented that.
What if you were
interested in lung cancer?
You could breathe out like a
breathalyzer in 10 minutes.
So we invented that.
What if you could find
out where the tumor was
and if it had spread?
So we invented a
version like that.
And what if you could use AI to
design these materials with--
by design instead of creating
libraries like we used to?
And so we're about to publish
a paper with Microsoft on that.
So there's all kinds of
things that the translation
into the company then
fed back into MIT.
Finally, we have physicians in
the labs, not just engineers.
And they decided that this would
be good, not just for cancer,
but for something that may
be surprise all of you--
pediatric pneumonia.
When you have a child in
the ER that has pneumonia,
the best thing to know is do
they need antibiotics or not?
Is it bacterial or is it viral?
So we're working
on a breath test
to be able to detect that in 10
minutes with foundation support.
And this whole
field now has a lot
of investment coming
in from places
like ARPA-H. So
the innovation is
keeping going by
virtue of having
done the start up company.
So Glympse here I
started in 2015.
I want to just say that
MIT is an incredible place
to start companies.
And as an academic entrepreneur,
you learn as you go.
This is a muscle.
You can see that my
entrepreneurial journey
accelerated over time.
And I feel very lucky to be
here, not just geographically,
but with colleagues like
Bob Langer and Phil Sharp,
who gave their time and
their wisdom and advice
about how to stay a professor
and be an entrepreneur.
Now, I will say that in
this next chapter of MIT,
I hope we can do even more.
And that's because,
as I did this,
I found that very
few of my women
colleagues had a
foot in both worlds.
And in fact, MIT is
an incredible place
to make that observation
because we really
have a playbook for how to
handle gender disparities, which
are documented in this amazing
book if you haven't read it,
the playbook is--
make a team of people
to address the problem,
gather the data, which we did,
find institutional champions
like Anantha Chandrakasan and
Maria Zuber and now Sally,
and talk about what addressing
this gap would address.
So we've been doing that.
Finally, do experiments.
So we have the Faculty
Founder Initiative started.
It's just celebrated
its three-year birthday.
And I'm happy to tell any of you
who want to learn more about it
how we've been helping these
21 amazing faculty unlock
their own inventions.
19 of them are now
building startups,
and we hope that all of them
become serial entrepreneurs
in this next chapter.
Thank you.
[APPLAUSE]
Good afternoon.
My name is Hugh Herr, and here
you see my beautiful legs.
From the knee down, I'm a
bunch of nuts and bolts--
three computers the size of your
thumbnail, six sensors, a muscle
tendon-like actuator.
These devices
propel me as I walk,
and I'm actually able to
walk in a normal fashion.
Years ago, I founded a
company called Bionics Medical
Technologies that translated
the prosthesis outside of MIT.
Today, they've been fit
to tens of thousands
of people, many wounded
US soldiers as well.
So at the New K. Lisa Yang
Center for Bionics here at MIT,
we're taking the next
step, pun intended.
The legs that I'm wearing
are intrinsically controlled.
They're not connected
to my brain.
So there's algorithms
running on chip.
So what we want
to do is actually
link the next
generation of prostheses
to the brain, to the spinal
cord, and central brain
via muscles and tendons.
So we're actually
not only manipulating
bone and soft tissues
and surgical designs,
but we're actually putting
hardware inside the body.
This is an
osseointegrated prosthesis
where there's a titanium shaft
going through the skin membrane
into the residual limb bone.
It's a mechanical
anchoring point.
And as you can see, you
see an electrical connector
where we can pass 16 fine
wire leads from computers
on the external limb
to, for example,
electrodes inside the body
and bidirectionally singles.
So this is not a
wireless approach.
It's a wired approach.
We recently got a paper
published in Nature Medicine
a few months ago.
This is the first
bionic limb that's
completely controlled
by the brain that
restores natural movement gait.
So all these movements that you
see, all the nuance of movements
that occur biomechanically
when a person goes up and down,
steps are being mediated
through the electromechanics.
So we were also
applying our techniques
to the upper extremity.
Sorry, let me go back.
Tricky.
Well, let me try once more.
Doesn't quite work.
There we go.
This is a blocks and
blocks demonstration.
The task is how many
blocks can this person
put in the box in 60 seconds?
So we currently have the record
with our neural interfacing
technologies.
We're not only doing
brain controlled limbs,
but we're also closing the loop
where we reflect proprioception,
the sense of position, movement,
and load, and cutaneous signals,
the sense touch,
from the prosthesis
back into the nervous system.
This is a first
human subject that
received a cutaneous
interface design put forth
by my group called the
cutaneous mechano interface.
Basically, surgically, we took--
when her limb was
amputated, we took skin
from the bottom of
the foot, and we
put that skin inside
her amputated residuum,
wrapped a muscle around it.
And then we're actuating
that muscle via computer,
via the 16 fine wire leads.
This video is shot
immediately after we
stimulated the constructs,
and this is her response.
[VIDEO PLAYBACK]
- So fascinating.
[MAN CHUCKLES]
This is just so
different from everything
else I've been feeling
the last day and a half.
- OK, so this is definitely
not residual limb--
- No.
- --or muscle stimulation?
This is pure phantom
limb sensation.
- Yeah.
I'm not feeling anything
in my limb at all.
[END PLAYBACK]
That was wonderful.
Obviously, the constructs are
physically in her residuum.
She didn't feel anything there.
She actually felt the bottom of
her foot that's no longer there.
So obviously, we're going to
put one of these devices on her,
put pressure sensing that
senses pressures as she walks
and then reflect that
into her nervous system
so she can feel the ground.
So that's the neural interfacing
part, the prosthetic part.
There's a startup company
called Muscle Metrics that's
being launched outside of MIT.
This is a technology--
so that's an exoskeleton.
It's a bionic shoe,
kind of like an e-bike
for legs, that augments
walking, running, and jumping.
This is a company I
co-founded, called
Defy, that's not only
focused on rehabilitation,
but human augmentation,
extending capability
beyond nature.
So we're using
this platform today
to also address medical needs.
So using small magnetic
spheres into weak muscle.
In the calf, for
example, after a stroke
and using magnetic
arrays on the skin,
we can track weak
muscle dynamics
and then reflect those to
control a motorized exoskeleton
to augment strength.
So I think we'd be very remiss
if the only thing we did here
at MIT is expand the
boundaries of technology.
Sadly, there's so many
areas of the world
that do not have access to
prosthetics and orthotics.
Arguably, the area of the
world with the greatest need
is in western Africa
and Sierra Leone.
As many of you know,
there's a civil war there
that ended in 2002, where
rebel forces hacked off
limbs with machetes.
Now there's tens of thousands
of people with limb loss.
So as part of the K. Lisa Yangs
Center for Bionics here at MIT,
we're actually building a
prosthetic and orthotic sector
in this war-torn nation.
What is our goal?
Sustainable care and delivery
of orthotics and prosthetics
services in Sierra Leone.
We want to increase annual
production of prosthesis
by an order of magnitude.
Our approach that
we call SCITT--
an acronym that stands for
Supply Chain Infrastructure,
Technology, and
Education-- we feel only
by having results
in all four pillars
can we introduce
sustainable change.
We're three years
into our program,
and we're experiencing
tremendous success.
We have many glorious stories.
I'll end with one such story.
This is Mohamed [INAUDIBLE].
Like tens of thousands of
other citizens of Sierra Leone,
he lost his limb
in the civil war.
As a young man, the rebel forces
were attacking his village,
and he ran, and he stepped on
a land mine and lost his limb.
We just fit a prosthesis to him.
Now he can walk freely
for the first time
without crutches or canes.
He's passionate about
soccer, and he can freely
be seen on the soccer field
enjoying his favorite game.
So just to finish up, restoring
quality of life and independence
of living for people with
physical disability is,
I believe, a moral imperative.
Disability experienced by
a friend or family member
is so common that
we just view it
as part of the human
condition, but it
doesn't have to be that way.
A great narrative in this 21st
century will be the emergence
of humans with technology,
resulting in the mitigation
of disability and unleashing
humanity's greatest potential.
Thank you so much.
[APPLAUSE]
Hi, everyone.
I'm Alex.
Thank you very much for
the opportunity to present,
and I definitely don't have any
videos that are quite that cool.
What I was hoping
to talk about today
was what I really think today's
event is all about, which
is the opportunity to come
together to really engineer
our entire ecosystem to have
global impact, as was very
beautifully illustrated by
the talks that just occurred
from Sangeeta and Hugh but
also from many of the ones that
came earlier this morning.
So before I begin, I
thought I would just
tell you a little bit about
what we do within our lab.
And fundamentally, what
our lab is interested
in is this idea of
homeostasis and tissues.
Now at MIT, I
teach 560, which is
thermodynamics for those of you
that haven't had to take it.
And as a chemistry
professor, if you gave me
a pressure and a
temperature, I could tell you
what equilibrium is
for any given reaction.
But if you ask me what
balance is in a tissue,
I wouldn't be able to tell you.
I can't tell you what that
means molecularly or cellularly.
But I know it's
important because if I
get too much of an
immune response,
I can get tissue
damage and death.
Like, I might see in an
autoimmune condition.
Or if I have too
little, I can become
susceptible to
opportunistic infections
or the outgrowth of tumors.
So a lot of what we try
and do within the lab
is to understand,
through studying
different human contexts, what
is it that defines health?
What are the deviations
that are induced by exposure
to environmental perturbations
like pathogens or pollutants?
And how could we rebalance
the system therapeutically
or prophylactically buffer
it through something
like a vaccine in
the first place?
And the reason that
it's such a good time
to think about these
sorts of questions
is that we can now
develop tools that
will let us study these
cellular ecosystems
at incredible resolution.
So just to give you a sense
of what is now possible,
because it seems relatively
science fiction-like we can take
a very small amount of cells--
so 50,000 cells, 20,000 cells,
the kinds of things that you
would have in fluids that
you might not even be aware.
So imagine scraping the
inside of your cheek.
Now rather than
saying that's too few,
we can actually load them into
innovative technologies-- many
of the ones that
you've seen earlier
that rely upon things like
microfluidics and new chemical
capture reagents.
And with these, we can actually
measure not one thing or two
things, but look at 20,000
plus variables at a time.
So really get a comprehensive
view of everything
that's happening in every cell.
And this gives us
the kinds of big data
that Caroline was talking about
earlier where we can actually
get a trillion data points just
from one individual sample.
And so just to make this a
little bit more real, what
you can do is you
can take a tissue
like the retina, where it has
this beautiful multilayered
structure.
And you can see there's
multiple different cell types,
but we don't really know
what the components are.
We don't know what
their markers are.
We can load it into
one of these devices.
And for each cell,
we can measure
thousands of different genes.
And here I'm showing
you expression
of a gene being high as white.
Black means off.
What we can then
do is we can take
a set of cells that express
the same kind of programming.
And this, we think, might
be a specific cell type.
We can take one
of those markers,
go back and label it using
standard molecular biology
approaches, and
characterize it in situ.
And so what this
does is it takes us
from a place where it would have
taken us 30 years to identify
a cell type, its markers,
and characterize it
but now do it in a week.
But what's so impressive is that
we don't just do this one cell
type at a time.
Instead, we can look at every
single cell type that's up here.
So we're not compressing 30
years of research into a week.
We're compressing hundreds
of years of research
into a single week.
So when you look at
something like this,
you can easily imagine why I
had to be at MIT as a professor
and was lucky when I
had the opportunity.
It's because I need incredible
multidisciplinary scientists
that are chemists, that are
engineers that really understand
how to apply this to
different contexts, that
can analyze the resulting data.
But it wasn't just MIT.
It was really that
I needed to be
embedded in this
incredible ecosystem
that you're all part of.
As you probably all know from
Phil's talk this morning,
MIT is surrounded by
this incredible community
of biotechs, of other
institutions, of hospitals,
of academic medical centers.
And what this does is it
creates an incredible ecosystem
in which to innovate.
And that's because,
like Colin said earlier,
the most important question
when you engineer is why?
And I interpret that a
little bit differently.
I think about that as exploring.
And what I mean by that
is really taking advantage
of this incredible ecosystem
and its willingness
to collaborate, to
go talk to people,
to identify problems
that matter.
So for example, by
talking to Sara Fortune
over at the Harvard
School of Public Health,
she said, well, that's great.
Have you ever thought about
applying it to something
like tuberculosis? which
you heard about from Brian
this morning.
Brian said 4,000 people a day.
It's 1.6 million people a year.
So that's more than any
other infectious disease
except for one year in COVID.
And this is literally
happening every year
and has been happening
for hundreds of years.
And about a quarter of
the world's population
is latently infected.
Now, the reason we don't
think about it quite so often
is because it happens that
the majority of infections
are in sub-Saharan Africa
and Southeast Asia,
as you heard about a
little from Bruce today.
So when you think about
the kinds of technologies
that you need, if you want
to tackle these problems,
well, you need things
that are deployable,
that can go out in the field.
You need them be scalable
so you can run them
across lots of patients.
You need them to be
low input, and you
need them to be
cheap so that they
can be used in all the different
contexts around the world.
And so once you have
the design constraints,
it's easy to go find incredible
partners like Chris Love over
at the Koch Institute
in chemical engineering
to take these really
cool machines that
look nice in our labs
at MIT, but could never
go out to where the
problem sites are
and re-engineer them into
really simple technologies that
accomplish the
goals that we have.
So we can take the
same kinds of things
that enable this
incredible discovery
and adapt it into really
simple portable chips that
can literally go
anywhere in the world.
What that does is
it then puts us
in a position where we
can go hop on a plane
and fly down to places
like South Africa--
and this is actually in
Durban, right at the site
that you heard about from Bruce
Walker earlier this morning--
and look at what's different
between individuals
that have more
mild or more severe
tuberculosis to find new
cellular targets that we
can use for therapies
to improve outcomes.
You don't just have
to go to South Africa.
You can actually start
applying these sorts of things
to all kinds of problems,
like pediatric stunting
due to malnutrition and
environmental infections
like environmental enteropathy.
And you find really
interesting things--
like, in the gut
right here we're
looking at the small
intestine-- and we're
finding that there are cells
that normally only show up
in the stomach, arising
in the small intestine
and preventing
absorption of nutrients.
And it gives us a
new idea of what
we might do to counteract it.
We can similarly go hop
on a plane to Thailand
with Sangetta's team,
who you just heard from,
and go and identify
new targets that we
could use to try and combat
specific stages of malaria
infection.
And what's really
cool is that when
you can do this, flying
all around the world
that your students get to
go-- and as I heard somebody
behind me say earlier,
it looks like that course
that Bruce teaches is really
fun, and it is really fun.
But you can also
improve everything
that we do here in Boston
because you make it simpler
to do things like partner with
people over at the Dana-Farber
to identify distinct
states of pancreatic cancer
and therapeutic treatment
regimes that will work better
for each.
And so engaging
people is important.
But a lot of what our roles are
as educators and in connecting
with people, as you
heard about from Andrew,
is to really empower.
And so what I like to think a
lot about is how we can go out
into the world inspired
by models like the one
you heard from
Bruce Walker today
and train people in
these technologies
so that they can apply
them to their own science
so that we can do more than
we could do independently
in our own lab.
And so I've been lucky enough
to go to a lot of places
with the team to train
individuals in places like Ghana
and Senegal in some
of these technologies
and to share what we've
learned along the way
so that others
that are interested
feel like they have at least a
place to start as they begin.
And another part of
really empowering people
is not just going out
and teaching them.
It's also working with the
innovation ecosystem around here
to take the widgets
that we build,
harden them and then sell them.
And I think that
commercialization
isn't the dirty secret.
And patenting isn't
the dirty secret.
There are things that
we do really well here.
We do that zero to one and
sometimes that one to 10.
But academia is a place that's
hard to go sometimes from 10
to 1,000.
And that's where partnerships
with local biotechs,
local industry,
with venture firms
becomes really critical to
having the kinds of impact
that we want to have
and, most importantly,
figuring out what
the right targets
are to accelerate good
outcomes for patients.
So I think that, to me,
what's so powerful about being
MIT is this
opportunity to explore,
to identify good problems.
So what are you going to do?
The opportunity to engineer
with incredible colleagues
solutions that will work.
And then to engage
others in fun science,
and then to empower others to do
the same in their own research.
And I think that
this model needs
to be extended to a place
where we can tackle all
of the emerging problems that
are showing up in the world
today.
You heard about
some of them earlier
from people like Linda,
issues with women's health.
Unfortunately, this is
probably not the last pandemic
we're going to see.
We need to think about
children and their thriving.
We have issues at
the intersection
of global of human
health and global warming
that we have to work
our way through.
And we have places
where we need to think
about how to care
for the elderly
and deal with
chronic conditions.
But I think that this model
that we can have here at MIT
taking advantage of this
incredible ecosystem
of biotechs, hospitals, and
surrounding academic centers
like HMS, puts us in a position
to actually really go after this
and to have transformative
impact for everybody everywhere.
And so with that, I just want
to thank some of our funders
and some of our collaborators.
There's too many to
throw on the screen.
But it's incredible
to be in a place
where this kind of
cooperation and partnership
is something that people love.
It's not should we do something?
It's can we do it
together and how?
And so thank you very much.
[APPLAUSE]

---

### MIT HEALS Launch: Collaboration panel session
URL: https://www.youtube.com/watch?v=1GFQHdI_qsQ

Idioma: en

Good afternoon.
I'm Anantha Chandrakasan, MIT's
Chief Innovation and Strategy
Officer and the
Dean of Engineering.
It is my honor to introduce
our panelists and moderator
for the session
on collaboration.
Our panel brings together global
luminaries in biotechnology
and health care.
And they work with us right here
in the Boston Cambridge area.
They've each received numerous
prestigious awards and honors,
and you can review them in their
detailed biographies online.
During today's
discussion, our panelists
will explore the rapidly
changing landscape
of life sciences research
and how we can collectively
embrace and shape the new
challenges and opportunities
that lie ahead.
Moderating our discussion
today is Dr. Noubar Afeyan.
Dr. Afeyan is the
Founder and CEO
of Flagship Pioneering
and co-founder
and Board Chairman of Moderna.
Because of COVID,
virtually everyone
knows about Moderna, a world
leader in mRNA technology.
Some of you might not know as
much about Flagship Pioneering.
Flagship Pioneering empowers
entrepreneurially-minded
scientists to invent solutions
to health and sustainability
challenges that seem unsolvable.
The result has been the
creation of more than 100
first in category
bioplatform companies
that are having significant
real world impact.
During his career as an
inventor, an entrepreneur,
and a CEO, Noubar has
co-founded and helped
build over 70 life science
and technology startups.
Terrific to have you
moderate the panel today.
Noubar will be joined by our
four distinguished panelists.
Dr. Anne Klibanski is
the President and CEO
of Mass General Brigham,
or MGB, a world class,
nonprofit, Boston-based
integrated health care system.
It is the largest private
employer in Massachusetts.
We all know that MGB delivers
world class health care.
However, it is also a
leading hub for innovation
in health care, helping
to bring academic research
from the lab to the patient.
Anne and her team have overseen
substantial investments
in leading edge research
as well as the creation
of more than 300 companies.
With Anne's leadership,
MGB is building
the integrated academic health
care system of the future,
and it's a system
that puts patients
at the center of it all.
Anne has been instrumental
in shaping the MIT MGB
collaborations.
It is truly a pleasure
to work with you, Anne,
to work on advancing
health care.
Robert Langer is the
David H. Koch Institute
Professor, one of only nine
Institute professors at MIT.
Bob is a world renowned
biomedical and biochemical
engineer who has pioneered the
development of new materials
for controlled drug
delivery systems,
particularly for continuous
controlled delivery
of genetically
engineered proteins.
He has written over 1,600
articles and 16 books and holds
more than 1,500 issued
and pending patents.
His patents have been licensed
to over 400 companies.
As an academic, my favorite
metric is that he is the most
cited engineer in history with
an H index of 327 with 437,533
citations, at least
as of yesterday.
[LAUGHTER]
It is estimated that
Bob's research--
[APPLAUSE]
It is estimated that Bob's
research has improved
the lives of over two billion--
with a B-- people and counting.
I'm very grateful to Bob
for all of his advice
in the creation of MIT.
Christopher Viehbacher
is the president and CEO
of Biogen, one of Cambridge's
leading global biotechnology
companies.
Chris has extensive
international experience
in large pharmaceutical
companies,
including GlaxoSmithKline
and Sanofi,
where he was the global CEO.
As the co-founder of
Gurnet Point Capital,
a Cambridge-based health
care investment fund,
he also values the
role of startups
in moving good science
out into the world.
As CEO of Biogen, Chris
is creating a new path
by broadening the
company's investments
in advancing innovations in
immunology and important--
increasingly important
field of research.
We're delighted to have a
long-standing collaboration
with Biogen, as we've
heard this morning,
and delighted to have
you join us, Chris.
Finally, Dr. Reshma Kewalramani
is the CEO and President
of Vertex Pharmaceuticals, a
global biotech company that
was founded here in Cambridge.
Now it's headquartered
in the Boston seaport.
Reshma is literally on her way.
She'll be here in a
couple of minutes.
With training in internal
medicine and nephrology
and extensive
business experience,
Reshma's leadership bridges
the medical, biopharmaceutical,
and the drug
manufacturing domains.
Using this expertise,
she has helped
vertex from increasingly
effective cystic fibrosis
treatments to a new
CRISPR-based gene editing
therapy for sickle cell
disease and transfusion
dependent beta
thalassemia that FDA
approved within the past year.
It's been a pleasure to work
with Reshma, who's here,
and brainstorm how academia
and industry can work together
in this space.
We're very delighted to have
this incredibly distinguished
group join us today.
And now I will turn
it over to Noubar,
who's going to make some opening
remarks and lead the discussion.
[APPLAUSE]
Well, thank you, and I feel
like introducing Anantha
after that very elaborate
introduction you
made for all of us.
Thanks for your
leadership, Anantha,
for originating this
initiative and for allowing
me to participate in
this very special event.
I just want to make a
couple of framing comments
and then we'll get right
into the discussion.
I came to MIT in 1983 to
pursue a doctorate thesis
at what was then the beginning
of this awkward marriage
between engineering and biology.
You heard about that from
Phil Sharp this morning.
It was called biochemical
engineering at the time,
and I had the great
honor of working
with one of the founders
of that field, an Institute
professor, Professor Danny Wong,
who unfortunately passed away
a few years ago.
I came here as an immigrant
fleeing a Civil War in Lebanon
after a few years
in Montreal, where
I did an undergraduate degree.
But 40 years later,
I had the honor
of delivering the commencement
address just this spring,
and just as great an honor
to be here with you today
on this very special occasion.
As the saying goes,
what a country.
And even more
appropriate today, what
an Institute MIT has become.
As a visiting committee member
in biological engineering,
chemical engineering,
and biology,
those of you who are
on MIT know that this
is a very important part
of MIT'S governance,
I've had a front row seat on
the two decade long discussion
about how MIT should best go
about becoming more relevant
in the biotech ecosystem.
It is already extremely
relevant, but how could
it do more?
How could it be more
impactful, not only here,
but in the whole world?
And looking at that today and
participating in the program,
I think that HEALS is a
very appropriate response
to that challenge.
As was announced by President
Sally Kornbluth this morning,
my wife Anna and I were
pleased to contribute
in a small way to
this initiative
through supporting fellowships,
graduate fellowships.
This is our way
of paying forward.
I did a degree here.
I have two daughters who have
done three degrees between them
here.
And I think that
everybody here who
has the means to
contribute through HEALS,
whether it's through
involvement or otherwise,
I would urge
everybody to do that.
This is the beginning and
it's got a long way to go.
And Anantha didn't ask
me to make this plea,
but I want to make it
because I think everything
you saw today is really--
it's hard to fathom,
but I get to see in many,
many other institutions
similar presentations.
And as I was saying
during one of the breaks,
most basketball teams have
one seven foot player.
None of them has 12
seven feet players.
And what you saw today
were the equivalent
of seven foot tall players
in every single area.
So we couldn't be more pleased
with our ability to support.
Now, as a segue to
this session, we're
going to have a
very important topic
to discuss to wrap up the day.
And I wanted to share
with you one last thought,
which is a curious, coincidental
connection between the words
that start with the
letter C. In addition
to curious, coincidental,
and connection, bear with me.
I am a bit of a
geek, so I went here.
So I call these the C's
that drive the biotech--
world leading biotech
position of Cambridge.
Now, let me try
these out on you.
Concentration of capital,
culture, connections,
creativity, community,
critical mass.
We have the critical
mass in this field,
in my view, bar none.
Competition and collaboration.
And it's that last C that this
panel will be talking about.
We're going to do this by my
asking them a few questions,
and then we'll have a
discussion, which I hope and I'm
sure you'll enjoy.
Thank you.
[APPLAUSE]
So we're going to
start-- we're going
to go down this group here.
But let me start, Anne, with
you to talk about the research
mission of Mass General.
Obviously, Mass General is
a huge research institution
in addition to being among the
best hospitals on the planet.
How do you think about the
differences in this mission
when that's being
done in a hospital
setting versus academia,
and how is it different,
how is it similar?
Let's start with that.
Yeah, so I love starting
with Mass General Brigham
because we are an integrated
health care system.
And we start and
end with patients.
So when I think about
health care innovation, when
I think about
collaboration, when
I think about all of the
things that we are all here
to think about, it
has to, in the end,
begin and end with patience.
So starting with the
health care system,
thank you for that
because I think
that is a fundamental thing.
And just as an anecdote,
I've spent my life
before doing this job as a
researcher, as a clinician.
Just today, and I
do this sometimes,
I like to walk
around the hospital.
I can walk around Mass General,
or the Brigham, or Mass Eye
and Ear, or McLean,
or Spaulding,
any of the institutions
in the system.
And it's always wonderful
to just go and walk
on one of the floors,
talk to a few Fellows.
I talked to a few
oncology Fellows today.
I talked to a few nurses who
work in labor and delivery.
What does that do?
It grounds me, and it should
ground all of us in, ultimately,
these phenomenal people who
are 100% focused on patients,
whether it's through
the research they do,
the training they do, the
clinical care they deliver.
So I just want to
emphasize that part of it
because I think that
is just so critical.
To get to your question
about research,
when we think about
the missions, certainly
the missions of
Mass General Brigham
and this is the missions
of so many, particularly
in the academic
health care system,
it's really providing the
highest level of, really,
integrated quality care
and an academic system.
You have to go to that
mission and tie it
to the next mission, which
is research and innovation.
And what distinguishes us and
other academic health care
system is that tie.
It's bringing the best of
research and innovation
and infusing it, instilling
it into clinical care,
that highest level
of clinical care.
And the other missions,
when you think about it,
when you think
about training, when
you think about
servicing community,
they all come together.
Research is the fundamental
driver of the system.
And we are the largest
academic health care
system in the country
in terms of NIH funding.
We have an enormous amount
of funding, and most of it
is federal funding.
I'm going to take a
pass on where that's
going to go in the future.
That's an important topic
very much on our minds.
But that's not
really your question.
Your question is, what does it
look like in an academic health
care system?
So I would say that research
and the, again, infusion
of that research
into clinical care
is what that looks like in an
academic health care system.
Now, in contrast to many
academic health care systems
around the country, we
are not under a university
or a medical school.
We are affiliated with Harvard.
But the grants,
the research, that
is all done on our campuses.
That's something that I think
is very special because it ties
together that research
mission, the researchers
with the clinicians,
they sit together.
And in fact, when you look
at a lot of our clinicians,
particularly those in
the subspecialties,
they spend a lot of
their time in research.
So this translation of
research into those things that
will really affect
people, the new therapies,
the clinical trials, the new
discoveries, the new devices,
the diagnostics, it's
being lived in real time
at the hospitals.
So if I think about what this
looks like in a university basis
or in a hospital basis, I'm
putting that all together
and saying, the phenomenal
work that gets done here at MIT
is absolutely ripe
for that collaboration
to come to our hospitals.
But every day these things
are a real life laboratory.
So taking all of the
work that goes on
and bringing it to
another laboratory,
that is the clinical laboratory.
And I think, sometimes,
the challenge,
when we think about this,
is they're still in pieces.
And some of it is a bit
of linguistic dissonance
because the language
of the basic scientists
may not be the
language of clinicians.
We hear a lot about that.
These are things that
you can overcome.
But living all of
that together means
that every day, top
of mind for the people
who are delivering care, who are
trying to solve those problems,
are deeply embedded,
living with, connected to.
And it can be a
geographical collaboration,
a conceptual collaboration.
There are many, many different
forms of collaboration.
I'm looking at that
word collaboration
and thinking of the many,
many things that means.
Some begin with a C, some don't.
And just thinking
about what we can
create with a university
like MIT, and part of it
is the people.
You bring the people together.
Part of it is the ideas.
But I have found,
certainly in our system,
the way to get the best
and the brightest people
working together is to give
them a problem to solve.
You give them a
problem to solve,
and that's where you get
the energy, the passion,
and the talent working together.
So basically, we're going to
try to fill in the Charles
River between MIT
and MGB and increase
the lanes on the Longfellow.
That's basically what we
have to do with HEALS.
We can get the urban
planning folks involved in--
They will solve that for us.
That's right.
So the other thing we
heard on the introduction
was that MGB itself is heavily
involved in a lot of translation
and a lot of companies.
How is MGB going about
thinking about another form
of collaboration, which
is between industry
startups and the kind of
science that's done there?
It's interesting.
I was the Chief Academic Officer
before taking on this role,
and that was back in 2012.
And when I started,
I had spent my life
as a researcher
and most of my work
was NIH funded, although
because I was in rare diseases,
I was very interested in
partnering with industry.
So I had a lot of collaborations
going with industry because I
knew that for rare
diseases, you really
weren't going to get anywhere
unless they had an industry
collaboration.
That was sort of a given.
And I think when I
think now about MGB
and how it's thinking
about it, the reason
I'm going back to 2012
is because I really
think there has been a cultural
transformation in the thinking
of a lot of our scientists,
and that's not unique to MGB.
And there are a lot of ideas
about why that is taking place.
But let me start by
what is taking place.
The cultural transformation for
many, many of our investigators,
and we, again, have thousands
of investigators, many of them
have spent their lives focused
on the research in a singularly
focused way, which if you
think about where it leads to--
and this is what I was
taught when I was first
doing my training after clinical
fellowship-- it leads to grants.
That's good.
Ideas lead to grants.
And that leads to papers and
that leads to promotions.
That leads to an
academic reputation.
That leads to making
important discoveries
in a field, whether basic,
whether they're translational,
no matter what they are.
So the end was that piece of
scholarly work and its impact.
The shift is to broaden out
what impact really means.
And I think the cultural
shift among many--
and it's interesting,
some of the surgeons
were way ahead of this.
They would come up with
ideas and they were more
keen on seeing that instrument.
What's that new instrument?
How do I operate better?
What is that invention that
will make things happen?
There were always a bit
more focused on that.
I don't want to
overly generalize,
but that's what I was saying.
But it was really getting
all of those investigators
to understand the full
life cycle of an idea
and how to have the
biggest impact on patients.
And that, I think, was
a very different way
of thinking about
industry, thinking
about pharmaceutical
companies, startups,
how do you actually take that
idea and bring it to patients?
You can't do that by yourself.
You can't do that
only with a grant.
What you have to do is have
those critical partnerships
and go through the
life cycle of an idea.
And it has to go to a company
or start a company or a partner
with a company.
You need to get a
product in the end,
whether it's a drug or
a device, whether it's
an algorithm,
something that then you
can bring back to patients.
And I think for our
investigators, particularly
those who are more
translational, but also,
some of the more basic
science investigators, that
understanding that if you really
wanted to impact human health,
you had to think more
broadly, and I would just say,
complete the circle.
So that's the cultural
change that took place.
And I think at Mass
General Brigham,
it is just now fundamental and
very much core to who we are.
I would also say that if you
look at a lot of our trainees,
they're already there.
They don't need it.
They're already there.
They come in and they're
already thinking,
how do I take this idea?
Where does it go?
What's the result of it?
How do you partner?
And some will actually start
companies and go there.
Some will join industry.
People have very
different career paths.
But I think it speaks to
the broadening of what
that circle looks like and
the interdependencies of all
these different sectors.
So the concept is
academics are not
an endpoint in and of itself.
Super.
Bob, you cover a lot of these
same topics from the MIT angle.
You run a lab that does basic
research, applied research,
engineering, translation,
and a lot more.
But you didn't
start out that way.
So how and when did unmet
clinical need and translation
get into the way you think
about research directions?
And also, how would
you-- because we
saw a lot of younger
academics who
don't come in with connections
to the clinical world,
how have you forged these
kinds of relationships
with the clinical world?
Yeah, well, it's
a great question.
So for me, I mean,
Noubar knows this,
but I got my degree in chemical
engineering in 1974 at MIT.
And then, when I got it,
almost all my colleagues
went to work for oil companies.
They had a lot of jobs
that paid a lot of money.
Anyhow, I ended up doing
something very different.
I wasn't that excited about the
oil companies and I had lots--
and I ended up getting a job as
a postdoc at Children's Hospital
with a surgeon
named Judah Folkman.
And by the way, for chemical
engineers at that time,
nobody did anything like that
because it paid so much less
and it also didn't look like
a very smart career move
according to everybody
that I talked to.
But I was fascinated by it.
And so I went there
and I was actually
the only engineer in the
hospital at the time,
and certainly, in
the surgery lab.
And the way I got
started, he had this idea
that if you could
stop blood vessels,
that maybe that would
be a whole new way
to think about treating cancer.
people.
Thought he was crazy.
They didn't think
that made any sense.
But at any rate, there
was no way to study it.
And so what I--
so we had to develop
what's called a bioassay.
And to do that,
we had to create,
we thought, tiny
little particles
that could deliver
a large molecules.
And Dr. Folkman would
talk to different experts.
At that point, drug delivery
was hardly done at all.
There was only one company
in the world working on it
and he went-- he was
an advisor to them
and he asked them
if they could help.
And they just said this
delivering a large molecule
would be impossible.
Anyway, I spent several
years working on it,
failed hundreds of times.
But finally, got a
way to get it to work.
And then, we published
articles in Nature in 1976
showing that you could
for the first time
deliver large molecules,
including it was the first paper
to deliver nucleic acids.
And also, we published a
paper the same year in Science
showing that you could use
it to stop blood vessels.
So that was kind of my
first taste of everything.
And of course, it took many,
many years before those--
I mean, the first one took
Genentech doing great work
building on some
of the things we
did before that would lead to
Avastin and many other drugs.
Second one, of
course, that's where
I got involved in starting
some companies, including
a number with you, Noubar,
using drug delivery
techniques to protect and
deliver different molecules,
including messenger RNA.
And Noubar and I
have shared that.
But the thing is to
try to figure out
what the applications are,
That's, a team effort.
I mean, all these things have
very broad potential uses
and figuring out what the ones
that are going to be important
you don't necessarily know.
Certainly, when we publish
those papers in 1976,
we had, obviously, no
idea that messenger RNA
would be important, that
COVID would even exist.
And so, really, it ends up being
a team effort over the years.
And also, to go over part
of your question, what
was great for me personally
was being in a hospital
with a lot of clinicians
working next door to them,
and it would lead to
more and more ideas.
And some of them are
my best friends, even
today, 50 years later.
So another one was
Jay Vacanti, who
started at Children's
and went to Mass General.
He came to see me one
day and said, Bob,
could we make new
tissues and organs?
And we came up with
this idea that would
lead to tissue engineering.
And again, we didn't
the right applications.
He was talking
about liver, but it
would lead to a lot
of other things,
including organs and
tissues on a chip.
And of course, the other
great thing about what's
happened in terms
of clinical things,
and this is, I think,
the beauty of Boston,
you have MIT and all
of the hospitals.
You get all these
incredible Fellows
who come to your lab for
several years from Mass General,
from the Brigham, from
Children's Hospital and Beth
Israel.
Jeez, I mean, I've probably
had about 20 or 30 summer,
[INAUDIBLE] Traverso,
professors at MIT now
and some have started companies.
So it's just been-- so
you just get ingrained
in that because of all
the great people around.
So on the company
side, then, you also
have been a prolific founder
of a number of companies.
And I'd say probably
so much so that it
must affect how you train your
students and the postdocs,
or at least what
exposure they get.
Because if they
want to be like you,
most academics who
don't do startups
have their graduate
students and postdocs
being trained to be like them.
Well, in your case,
they also need
to learn how to
start companies, how
to influence large companies.
I ask it seriously.
How do you think you
have incorporated or have
you some of the things you have
learned in the startup world
into the way you actually train
the students and run your lab?
Yeah, well, I mean,
part of it to me is--
I don't know that
I do that much,
but I think probably--
as my wife would say,
I'm probably the opposite
of a micromanager.
I try to come up with
big ideas and give people
a lot of freedom, but with
clear objectives in mind
and try to get big
papers and big patents.
And that does happen,
to your point.
But people see,
it's not just me.
I mean, Noubar, again,
knows a lot of this.
I started a couple
companies back in the '80s.
And then, like Jay Vacanti,
he said, well, can we do one?
And then, I had this
postdoc, David Edwards,
who you know and he said,
well, Bob, I want to--
people-- so it kind
of gets ingrained.
And so it's not just me.
People see the older graduate
students and postdocs
and colleagues doing it,
and it kind of just happens.
I think, a lot of times with
companies, people don't--
like if you go to some
other universities,
they think it's like,
how could you ever do it?
It just seems almost impossible.
But when they see people a few
years older than them do it,
they think, well, yeah,
that could be nice.
And when we started,
I think it was
AIR, which was David started.
Jeez, I don't know how
many people from the lab
when they graduated went
there and they ended up
doing very well.
And so I think it just--
they'd see that by example.
And then, finally, we
have great seminars
as we had Noubar once give us a
lecture and I still remember it.
He starts out about
entrepreneurship
and how everything, really,
with entrepreneurship, almost
every scientific-- every
scientific discipline
is really embedded in that.
And some of the examples
of the things that you did.
And sometimes, with me,
sometimes with many others.
And I think people
just see that.
And I think that--
so having role models, I think,
ends up being a really big deal.
Super, Chris.
You've been in large
companies, you're
in a slightly less
large company,
but it's still really
large compared to startups,
and you've done a lot in
the various hats you've worn
partnerships with academia.
Talk about what you've
seen works in that arena,
the value of such partnerships.
We're in the collaboration
section of this day.
It'd be great to get your
insights on that, plus maybe
some things that you've found
over the years doesn't work.
Yeah, I mean, Andrew Lowe
gave a really nice synthesis
about how drugs get developed.
And there is
certainly, OK, you have
to have some idea of
what causes a disease.
And, then you're going
to find a target.
You find a molecule and
you do a clinical trial.
But in actual fact,
there's just a never ending
series of puzzles that
present themselves to which
you have to find an answer.
And pretty much developing any--
particularly first
in class drug,
there is no pathway to follow.
And even if you find
a molecule, well,
how do I deliver the molecule?
If it's a small molecule,
how do I make it soluble?
I mean, do I need a device?
What's an endpoint?
Even if I have a
molecule, how am I
going to measure whether
this is working or not?
Even if I know that, then I
might have another problem here
to solve on scaling up in
manufacturing, for example.
We heard the presentation
earlier about nanoparticles
and what a role that played
in getting mRNA, actually,
to where it needs to go
to make the mRNA vaccine.
So even after we figured
out mRNA and what it does,
we still had to deliver it.
And so when you're
in the company,
you're really focused
on value creation,
as we heard from one of the
entrepreneurs this afternoon.
So your job is really
trying to take that molecule
and move it as fast,
but you don't really
have the capacity to do an
awful lot of basic research
and a lot of discovery research.
But in some ways, a
lot of these things
you have to discover
just to solve them.
And that's where you
need these partnerships.
How do you solve these problems?
And you're not going to
always have the capability
inside your own company.
Now, a lot of companies
do think they need that,
and that sometimes is a problem.
But generally, I mean, you
look at all the cool stuff
we heard today and all the
different technologies.
And you think about,
Bob, you haven't just
developed molecules,
you've developed
how to get things to places.
And that's where you
got to figure out,
what's the right combination of
people who are trying to drive
something to an endpoint and
people who are really having
the time and maybe the risk
capital or the government
capital to actually solve some
of these basic problems that
need to be done?
Now, it happens to happen is
some level of trust has to exist
and a sense of common purpose.
And you were saying it starts
and ends with the patient.
And that has to be for
everybody in the chain.
You can get hung up on things
like intellectual property,
for example, can get in the way
of some of these collaborations.
And that's where a
certain amount of trust
is going to be needed
to try to figure out,
how do we share some
of these things?
Because if different people
and different organizations
are solving different
parts of the problem,
somewhere there has to
be return on all of that.
Otherwise, they
can't survive either.
And so, sometimes, that might
be an advisor at Mass General
and they're doing
clinical trials for us
and they can advise us how
to-- where's the unmet need?
How should we design
this clinical trial?
It might be coming to MIT to
solve some of these delivery
problems.
So there's sort
of a never ending
series of these collaborations.
It's not just one project.
It is creating a porosity almost
between the organizations.
Now, you also find that all
of this depends on people.
And I always say,
running companies
would be easy if it weren't
for all the people inside them.
[LAUGHTER]
And people really create
the complexities in this.
And you get egos involved.
You get different incentives
that are not aligned,
and that can cause problems.
Personalities.
And so there has to be, in some
ways, some sort of alliance
management, generally, to try
to smooth those problems out.
One of the things that
you don't want to create
is a sense of one is a
subcontractor to another.
There has to be a respect for
the intellect, the talent,
the capability on both
sides of a collaboration
to say, OK, we are not going
to succeed without the other.
So you have to create a lot
of the right people dynamics
to make this as well.
Sometimes we can
get carried away
with the science or the
problem, but the number
of times I've seen
collaborations
that don't work that just come
back to basic people problems
is a real issue.
And so, certainly,
whenever we start one,
create the relationship and
build the relationship over time
and build it on the respect
and capability of what
the other brings.
But as I say, if
you don't do that,
you're never going
to get to the end
because these
collaborations just don't
start at the very beginning.
Because in Andrew's
model, it was sort of,
hey, there's the idea,
you create the patent,
and then, it goes
to the companies.
Well, it's not quite so simple.
It's these constant
collaborations.
And that's why I think it's so
helpful to be in this ecosystem.
Biogen is across
the street from MIT.
We're across the river from MGH.
And when you talk about
those relationships,
that makes it a whole lot easier
when people know each other
and see each other all the time.
And I do think that is
this proximity of all
of the talent and
people that we have
make those people issues a
whole lot easier to deal with.
And I think that's what--
since collaboration
is so critical
and people issues
are so critical,
I think that's what makes
this work here super.
And Reshma, you're are
leading a native kind
of Cambridge-born
company, Vertex,
one of the great
standout successes
of the biotech industry,
really, in the biotech era.
You're not literally
in Cambridge,
but you're in the
greater Cambridge
area is kind of how we view it.
And so with that
in mind, give us
also your view on this topic,
if you don't mind, on industry,
or at least Vertex's
collaborations with academia.
And in particular,
obviously, you
have David Altshuler
from MIT, you
have Doug Milton from Harvard,
and many leading academics
who have gone into industry.
So give us your sense of
this dynamic of collaboration
between biotech and university--
research universities.
Yeah, I think a lot has
been said about this,
so maybe I'll point out two
or three additional points.
The work that we
do in biopharma,
biotech, the work
of drug development
is a very particular
kind of work
and, honestly, doesn't
happen in academia.
I left academia because after
my own circuitous journey,
I realized that when I started
as a physician scientist,
I thought what I wanted to be
was a triple threat, run my lab,
see patients, teach students.
I really thought that's
what I wanted to do.
And I fashioned my
entire life to do that.
Until I actually did
it and I realized
I was doing research in
areas that didn't actually
interest me.
But that's where
the funding was.
That's where the
grant money was.
And I just decided one day
I was not going to do that.
And then, I realized when I said
to myself all those years ago
that I wanted to do
research, I actually
meant I wanted to
make medicines.
I didn't know how to
say it prospectively.
But when I was there, I realized
that's what I actually meant.
And then, once I decided that's
what I really wanted to do,
it turned out to be extremely
difficult, if not impossible,
to do in academia.
You need to have all of
the basic science discovery
infrastructure, which I
think academia has in spades.
Then you need clinical
development, regulatory affairs,
manufacturing, and a
whole lot of capital.
That was not
available in spades.
And then you need
another group of people
who could take that medicine
and actually bring it out
to patients around the globe.
You need access, you
need health economics,
you need a sales and
marketing infrastructure.
And I actually left
for that reason.
When I went to my
chair and I told them
all this exciting
stuff I wanted to do,
they said, OK, you are at--
I was at the Brigham
at that time.
You have 20 hours of
statisticians time.
20 hours.
I wouldn't even be able to
write one protocol in 20 hours.
And that's what drove
me to go to biotech.
And I went to a company
on the West Coast.
And the collaborations
that I saw there
and the collaborations
we do at Vertex
are along these lines of
the very best innovation,
and we have
fantastic scientists.
But given the vast amount
of science that can be done,
it cannot all be done
within the walls of vertex.
So we partner often
with academics,
with small biotech companies,
and with others who
are just doing basic
science research.
We also collaborate
in-- then in-license
and/or acquire companies because
that company, that technology,
their asset works perfectly
for what we want to do.
And then, the third
way we partner is there
are some unique
problems that come up
that you don't expect to
come up along the lines.
We might be making a
medicine for cystic fibrosis
or we might be making a medicine
for sickle cell disease.
But along the way,
you realize you
need some data set, you
need some algorithms,
you need to understand how to
do certain off-targeting work
that we don't know
how to do inside.
We just know that it's
a new problem that's
come up along the way.
And working with
academia, institutions
like MIT and Harvard, has
been incredible for us.
One of our medicines that
was just recently approved
came from a partnership
with CRISPR Therapeutics,
a small biotech
company right here.
Another one of our
programs that's
just in phase III development
as of the last couple of weeks
came to us by purchasing
Doug Melton's company.
So if you said, look,
you've got to do work
without collaborating, I'd
say patients would suffer.
It is absolutely
essential to what we do.
Super.
So staying with you.
The other thing
that MIT contributes
and as such
universities contribute
to Vertex and to Biogen and
others is human capital,
is the students.
And with HEALS
today and what we're
kind of celebrating
and launching,
I know that MIT very much is
interested in what more it
can do to better prepare
students, graduate students,
postdocs, everyone to be even
more impactful in biopharma
industry.
So with a lens towards the
future, what types of things
would you expect
that MIT could maybe
add to its arsenal of how
it's preparing students?
I think President Kornbluth's
focus at MIT on life sciences
is really critical.
I happen to be
married to an engineer
and it's not that awkward.
We do OK.
[LAUGHTER]
But I realized that we actually
think completely differently.
To be a fantastic
practicing physician,
you need to connect dots.
And you need to
repeatedly connect
the dots the same exact way.
Photophobia, neck, pain,
fever, you think meningitis.
And you think that
every single time.
Otherwise, a patient will die.
So it's a lot of
pattern recognition.
It's not that good to
be all that creative.
As an engineer, my
husband sees every problem
and he sees the
most, I don't know,
this-- a small, obscure
piece of evidence
that he wants that to be proven
to him based on first principles
that it's true.
And all I say is,
photophobia, neck pain,
fever, that is meningitis.
[LAUGHTER]
I think there is a lot of
benefit to bringing engineering
into the life sciences,
and the best element of it
is something that Anne
said, it helps patients.
You can use engineering to make
tires and pencils, I suppose,
but there's no better cause
than to do this in health care.
What I think that does
mean is introducing
physicians like myself
to engineers, to math,
to technology.
In our medical school existence,
we don't learn a whole lot
about that.
And ensuring that
engineers, even if they're
mechanical engineers and
not biomedical engineers,
have a good, strong
foundation in biology.
I find that a lot of
engineers are turned off
by biology because they
think it's rote memorization.
And as a physician,
I have to admit
I'm a little turned
off in my earlier days
from engineering
because I thought it was
about oil and building bridges.
[LAUGHTER]
So I think that
there is a lot more
we can do to make sure that
engineers know mechanical,
computer science, electrical
that there's a lot of great work
that they can do in biopharma.
And the only way we can do that
is to make sure that, one, we
talk about it, two, we
have seminars like this,
but three, internships
and fellowships,
there needs to be a lot
more movement of engineers
into biopharma.
There already is
quite a bit of that.
If you're a PhD in
the life sciences,
there's actually an
entire Vertex program
that welcomes you and you have
the opportunity to do that.
We need to do more to make
that opportunity for engineers.
Chris, same question to you.
Yeah, I think I've been
a trustee, actually,
of Northeastern
University for 10 years.
And I think one of the
interesting things they do
is they've got this
experiential learning, which
used to be called interns.
It goes back to something
I was saying earlier.
A lot of the problems in life
are really people problems.
And if you can get the
people to work together,
you can solve an
awful lot of things.
But the higher you go
up in an organization,
the more you're going to be
focused on getting people
to work together effectively.
And I think younger
students, often,
it's a question of
how much do I know?
What's my technical knowledge?
What's my capability?
But at some point, if you're
creating these collaborations,
or at some point, you're
going to lead a team,
you have to develop some
of these softer skills.
And I think that's important
to start to introduce even
in engineering and sciences.
I don't think that
often gets enough.
Some of that can be through
the fellowship program
that you've
generously sponsored.
We've got fellowship programs.
We have an intake
of MBA students
and we bring them through.
But there is an element, once
you get inside a company,
you are in an organization.
And once you're in
an organization,
you have to learn
about culture, you
have to learn about leadership,
and the other element
I would say is business acumen.
Andrew's omics were fabulous.
I mean, economics
is fundamental.
I mean, you have to raise money.
You've had to raise money.
You deal with omic economics
a lot at the hospital.
We do it as company CEOs.
And the reality is nothing is
going to happen without capital.
And so I remember I had hired
one guy when I was at Sanofi
and he was one of the first
MDs to do an MBA as well.
And at Harvard, they
almost said, well,
why are you doing that?
Now, today, that is
a lot more common,
and you see these
combination degrees.
I think Wharton,
for instance, you
can get a combined science
and finance degree.
But people certainly who have
the scientific capability, who
bring these other human
capabilities as well
as the business acumen, I think,
will really go a long way.
So continuing the
discussion, we've
gone 30 minutes without
using the AI expression.
So we see we're disciplined.
But now I do want
to turn to that.
Because we heard today that
this was the 50th anniversary
of the Cancer Center here and
how much molecular biology
influenced all that.
And then, we heard about how
engineering got involved.
And now, we're all kind
of-- not only do we
have to combine
biology, engineering,
patients, clinical
care, but now we
have to get into AI or AI omics
if Andrew's expression works.
But that's kind of unfolding
in front of our eyes.
So taking out AI as it
relates to productivity gains
in our regular life, how are
each of you thinking about this
as--
or are you-- as a
pillar of what you're
building on in the future?
Obviously, at MGB, you go from
the data generation, data access
set of issues that
AI raises all the way
to many, many
other applications.
So maybe, if you
don't mind, I also
wouldn't mind because
this is being filmed
that we take a little
bit of the opportunity
to just project a bit over
the next 10, 15 years.
And so when this group HEALS
gets together and celebrates
their anniversary,
we can look back
and they can make fun
of what we projected.
And with that, tell us
the current reality,
how you're thinking about it,
and how you see the future.
So you brought up a
lot of different things
that come together.
When you think about a patient
from a purely clinical point
of view, you're always
thinking about bringing
in multidisciplinary
strengths to solve a problem.
I'll just get to
the problem-solving.
So you could have a
patient and no one
really knows what's going on
and there's an internist there.
And someone will say, yeah, I
think I know what's going on,
but I'm wondering if maybe
this relates to endocrinology.
Then the endocrinologist will
come in and say, well, no,
this is not endocrinology at
all, but I think it's this.
So at the end of
it all, you might
have five different
medical subspecialists
all with their own point of
view, their own expertise,
their own view of the world, and
their own learnings who are all
focused around the patient.
So then, if you broaden
it out, you get more
into these themes of, it's not
just multidisciplinary in terms
of medicine, what else
are you bringing in?
And then, you get into the
technologies, the engineering.
Now you're bringing
in multidisciplines
of a different point or a
different part of it that's
all focused in.
So when you talk about this
in terms of the clinical care,
the AI, the engineering,
et cetera, I just
want to put the
patient at the center
and think about
these things that
are going to be wraparounds
in terms of how you ultimately
solve the problem of
what's the diagnosis, OK?
What's the treatment?
How do you find the
right treatment?
How do you get the
treatment to the patient?
All are wrapped
around the patient.
So when I think of AI right now,
and I'll try and summarize this
maybe in three
points, number one,
it is, what we doing
in terms of AI that
is helping the clinician?
That's the easiest thing to do.
So that is, how do you
take and deal with burnout?
How do you look at
different data sets
and put them together
that are already existing
that will help the
clinician do what?
Write a better note?
OK, well, that's
great, actually.
But that is here and
we want to go here.
So if I look at the parts
of this, number one is,
how do we broaden out
the knowledge base
beyond the physician,
their own experience,
to the five physicians
in the group,
to the medical literature,
to what we all based
on these different articles,
different observations,
and be able to have a massive
amount of data all compiled
together to put it together
in the most objective form?
So number one is, how will this
help us in terms of diagnosis?
And there are a lot
of issues about this.
I mean, there's a radiology
conference many years ago that
basically said,
will there be a job
for radiologists in five years?
That's not really the question.
The question is, how do we put
together lots of different data
sets in--
that are huge, that have been
applied in a gold standard way
that are validated.
What does that mean?
That are equitable so
that you can actually
make a real diagnosis.
So I think the diagnostic
possibilities--
because we're already seeing
it-- are extraordinary.
And it's going to be hard
because of all the issues
that we have with AI that we
could spend another hour on.
So I think the second thing is,
how do you personalize therapy?
How do you match
therapy to the person?
That's been one of the
single most pressing
therapeutic problems.
We look at drug trials.
We have broad answers.
But then, there's that subset
that could have done this.
It's the personalized medicine
that we keep talking about
that we're not there.
AI can definitely
help us get there.
So I think the
predictive properties
of this in terms of
diagnostic accuracy,
in terms of looking at targeted
therapies, drug discovery,
I could go on and on.
But honestly, I think it's going
to revolutionize health care.
I was at a Microsoft
meeting about two years ago
and they had a presentation
about AI in the car industry.
And I learned more about
the application of AI
to health care in
that presentation
than I'd ever thought of before.
And you could say,
how is that possible?
It's the car industry.
Because the concepts that they
were talking about in terms
of data and how you view
it and what it means
and how you can envision the
future needs were extraordinary.
Super.
Bob, maybe put on the kind of
science if you look forward--
maybe just take the
forward question.
If everybody had access to all
the knowledge you have gathered
and your imagination
in the form of ability
to create new patterns,
how do you compete?
And then, what does
that look like?
Yeah, well, one other
thing I was going to say,
and because I wasn't--
I was talking to students and
thesis committees all day,
so I didn't hear some
of the earlier sessions.
But I hope it was
said at some point,
if it was the 50th anniversary
of the Cancer Center, jeez,
I think MIT has been doing AI
research for way more than 50
years and really deserves a lot
of credit for pioneering it,
people like Seymour
Papert and others.
And we ourselves
have used it for--
I mean, again, in science,
things-- actually,
for Chris's point,
like about solubility.
We came up with ways of
doing high throughput
systems for solubility
with Mike Simon,
looking at thousands and
thousands of combinations
and analyzing what you could--
and keep iterating
until you get something
that does make the
drug soluble enough
with FDA approved excipients.
But I think, going forward, the
way I think about it is, I mean,
I don't know if
it's what I know,
but I think it's like to me, it
does go to the science like--
to data, where can large
amounts of data be very useful?
One of the things I do
feel, like when I mentioned
this idea of tissue
engineering before,
is you could have-- to the
extent that you can make organs
and tissues on a
chip that really do
simulate the human
model, which are doing
quite a bit of that at MIT.
I could envision that
as the years go by
could just screen thousands,
maybe millions of molecules
and find new targets, find new
drugs using things like that
and use AI to predict
what's the next iteration
to the next iteration till you
maybe would have someday help
Alzheimer's or other things.
Because right now, if you have
to do everything in humans,
that's pretty tough.
But we're working with
Li-Huei Tsai, for example,
with the postdoc, Alice Stanton.
So we have a brain on a
chip where we could take,
actually, anybody's
cells here, and actually,
with Merit Cudkowicz,
we're going
to be talking to her about--
I was her pre-med
advisor and now she's
head of neurology at
MGH about maybe using
it to find new targets for ALS.
And so I just
think, again, where
you could take these things
and get thousands and thousands
of molecules, maybe
millions, and then
use it to find new targets
and new molecules that
could treat disease.
Super.
Maybe, Chris, to continue with
a view towards maybe a little
bit the present and a
lot more of the future.
And it's OK to speculate.
On the area of drug
discovery, drug development,
and the delivery of
clinical products,
how do you see this play out/
Well, first, I would go
back to the economics
before we go anywhere.
Because some of us were
at a dinner last night.
Sally and Susan,
various 50 members
of the broader ecosystem here.
And there was a lot
of concern raised
about the threat on innovation.
Certainly, we have a
current new administration,
but I would argue that
innovation was somewhat
under threat even before.
But now we're talking about
overhead recovery rates
and grants.
We're looking at pricing
implications for our industry.
And I think when you look
at the federal deficit,
you look at the
percentage of people
who are going to be over
the age of 65 by 2030,
it's going to be significantly
higher than where it is now.
And I'm sure, Anne, you would
agree that, actually, it's
those people that are pretty
heavy users of health care.
And we see already that Medicare
is potentially not solvent
just when I'm about to join it.
[LAUGHTER]
There is a massive amount
of economic pressure.
And when you go back to the
presentation by I think it was
Professor Lauffenburger this
morning where he started-- well,
we start with-- once we
actually get it out of academia,
we're still looking at
10,000 molecules to get one
that's approved.
And somewhere, I think, we have
to be changing fundamentally
the model of how we're doing
research and development.
Because society is telling
us that pricing and cost
is important, and we're going to
have to do this a whole lot more
efficiently.
If I look at clinical
trials, as much innovation
as we do on the molecular
level, we have done very little
on the actual process level.
And so what we're doing
at Biogen, for instance,
we have dry labs.
If you're in small
molecule chemistry,
you can actually now use
AI to, actually, develop
a more optimized molecule.
We used AI to create
virtual waiting rooms
for hospitals to actually find
patients for clinical trials.
Today, if I'm going to do
a study with 400 patients,
we're going to have to go to
400 or 500 clinical trial sites.
I mean, that's crazy.
Who does that?
If we could do 400
patients in 100 sites,
you have a massive
benefit on productivity.
So I think AI is not
just a cool tool,
I think we have to figure out
how do we reverse engineer now
how we're doing
research and development
because I don't think that
this pressure economically
on us as universities, as
biopharmaceutical companies is
going to go away any time soon.
And AI strikes me as a
way we can revolutionize
how we're doing things.
Reshma.
Maybe I'll give you
some practical examples
of what we do.
I think that there's
a lot of scope today.
We don't have to envision too
far into the future for AI
in imaging.
So for us, part of what we do
when we do our type 1 diabetes
cell therapy, our scientists
will describe the cells
as fluffy, cloud-like, plump.
And some of those are good and
some of those are not good.
And right now, you have a
scientist looking at a tube
to see if it's fluffy,
cloudy, or plump.
That we are
converting reasonably
with ease into an
imaging AI model that
will tell us these are the good
cells, these are the bad cells.
I think that is
already here and now.
We try to do that with studies
in clinical development
when you have to take biopsies.
I think that's a good place to
go and that's easy to do now.
I think the other place
that's really ripe,
it's not exactly now, but it's
maybe the day after tomorrow,
in manufacturing,
now that we're going
from chronic small molecules
or biologics molecules
that you take by
mouth every day,
biologics that you may take
once a month to these one
and done curative therapies,
the manufacturing of those
and the quality release of those
is an enormous human task, far
better done by computers.
But it's actually manual.
We still write batch
records in a log sheet
by one human being who is
then checking the same thing
to other human beings because
that's all we have right now.
So we don't need fanciful AI.
We just need not to be faxing
things back together or PDFing.
We just need to automate
simpler things like that,
and that'll get us a
huge bang for our buck.
In the distant future, right
now, we do what Bob describes.
We have really
smart people taking
30, 40 years of
experience and recognizing
that that carbon
moiety or that oxygen
atom in that spot,
that's where it usually
triggers a signal for maybe
a drug-induced liver injury.
It would be great to
catalog all of that
and to have that be a
little bit more automated.
It's not as bad as
I portray it to be.
Even today, we have
software and we
have algorithms that tip you
off to this is a red flag.
When you have this
moiety in this corner,
it's not going to work out.
We have ways of making things
more soluble or less soluble.
But I would say we are 70%,
80% human, 10%, 20% computer.
And there is a lot more
ability to do that.
But if we just tackle the simple
stuff, like in manufacturing,
like in batch records,
like in imaging,
and I agree with Chris
on clinical trials.
Today, when you do clinical
trials and now we're doing
clinical trials of
30,000, 40,000 people,
because the effect
size is smaller,
because we've made health
care better, there are people,
doctors, physicians, scientists
who are reading through reams
of safety data.
That's called line listing
review because we literally
review lines of data.
That is far better
done by a computer.
It's not there yet, but
that's the low lying fruit.
Super.
Well, look, we're out of time.
I think this has just
been a glimpse of what
interdisciplinary collaborative
discussions will look like.
Hopefully, it will happen at
every classroom and every corner
of MIT going forward.
And please join me in
thanking our panelists
for this discussion.
Thank you, guys.
[APPLAUSE]

---

### MIT HEALS Launch: Connections plenary session
URL: https://www.youtube.com/watch?v=aJDpsCJgx8Y

Idioma: en

So we've reached the end
of our plenary program.
And [? Katerina ?]
and I would like
to take just a couple
of minutes in closing
to express our gratitude
to the people who made
today's symposium possible.
So this morning, Sally
thanked a lot of people.
We would, of course,
like to thank her.
Thanks, Sally--
[APPLAUSE]
--and also Dean Anantha
Chandrakasan and Dean Nergis
Mavalvala.
The commitment of
Sally and Anantha
and Nergis to advancing our
understanding of biology
and human health through
interdisciplinary collaboration
is the driving force behind
the HEALS collaborative
that we're launching today.
So today's symposium would
not have been possible
without our multitalented
planning committee.
Their names are listed here.
You saw many of them
on the stage here today
or running breakout sessions.
This is the group that took
the charge of the deans
and turned it into a day of
exciting MIT research highlights
from all across campus.
So we're very
grateful for the ideas
and the enthusiasm
of the committee that
brought the event to life.
[APPLAUSE]
And, finally, a
huge thank you to
the amazing administrative,
technical, communication,
safety, transportation,
facilities staff
who worked tirelessly to
ensure a successful event.
[? Katerina ?]
and I particularly
want to thank Marsha Warren,
who's lurking here somewhere.
Thank you, Marsha.
She's been keeping us
organized from day one.
We never could have
gotten here without you,
so thank you so much.
So while the formal part of
our day is coming to an end,
the discoveries
do not stop here.
We invite you to
continue the conversation
and explore the amazing work
of our students and postdocs
that really drive the innovation
at the poster session that
is set up right next
door in the tent.
This is your chance to connect
with the brilliant minds
directly.
And I promise it will feel like
a treasure trove of discovery
and exciting research.
A special thank you
for the poster session
goes to Iain Cheeseman, a key
member of the committee also,
and Katey Provost
and Tia Giurleo.
It is incredible how much
work they put into it.
I really underestimated
how many hours.
Very, very, very big thank
you, their dedication
made this exciting
opportunity possible.
And thank you all
again for being
part of this vibrant community
and bringing your curiosity
to this wonderful day.
Let's celebrate the future
of health and life sciences
and continue to push the
boundaries of what's possible.
[APPLAUSE]

---

