Priscilla Chan on AI, Rare Disease, and the “Virtual Cell”

In this episode of The Future of Medicine, we welcome Priscilla Chan, MD, pediatrician and co-founder of Biohub, a first-of-its-kind research initiative combining frontier AI with frontier biology to accelerate progress toward curing or preventing all disease.

Dr. Chan shares how her experience caring for children with rare and undiagnosed conditions shaped her commitment to transforming how science is done. She discusses how patient-led research communities are driving breakthroughs, and how new approaches to data sharing and collaboration are reshaping the pace of discovery.

The conversation explores Biohub’s work to build a “virtual cell”—a computational model designed to simulate human biology—and how advances in artificial intelligence, large-scale datasets, and imaging technologies could allow scientists to better understand disease at its most fundamental level. These tools may one day make it possible to predict disease risk earlier, test interventions virtually, and personalize treatment based on an individual’s biology.

Dr. Chan also reflects on the future of medicine, where the boundaries between research and clinical care continue to blur, and where physicians increasingly engage with data, biology, and technology to guide patient care in real time.

Looking ahead, she shares her vision for a more proactive and precise healthcare system—one that moves beyond treating illness to anticipating and preventing it.

Thank you for listening!

Call to action: If you enjoy The Future of Medicine, subscribe for more conversations with leading scientists shaping the next era of healthcare. Please rate and review the podcast to help others discover these important discussions. Share with friends and colleagues who are curious about how science becomes medicine.

Transcript

0:00:00.160,0:00:06.400
It's not some distant far away land where papers 
get published and then 20 years down the line we

0:00:06.400,0:00:12.320
get a drug. Dr. Priscilla Chan is co-founder of 
Biohub. 10 years ago, Dr. Chan and her husband

0:00:12.320,0:00:17.120
Mark Zuckerberg launched their audacious 
mission to cure or prevent all disease.

0:00:17.120,0:00:22.640
The science has moved incredibly quickly because 
the patients are bringing assets to the table. In

0:00:22.640,0:00:28.000
this fireside chat, Dr. Chan speaks about how her 
time as a pediatrician shaped her vision today,

0:00:28.000,0:00:32.880
what Biohub is doing to try to compress 
decades of scientific discovery into months,

0:00:32.880,0:00:37.920
and how technologies like the virtual cell 
and AI might transform both research and

0:00:37.920,0:00:42.800
clinical practice. We are fully committed 
to building tools faster, more effective,

0:00:42.800,0:00:47.200
and more efficient. Her vision for the future 
of medicine may sound like science fiction,

0:00:47.200,0:00:53.200
but with the power of AI, their team believes that 
soon enough it will be very real daily practice.

0:00:53.200,0:00:58.080
Welcome to Stanford Department of Medicine's 
inside look at the future of medicine. We

0:00:58.080,0:01:02.400
are very excited to welcome you to Stanford. 
Thanks for being here. Thanks for having me.

0:01:02.400,0:01:06.480
When we think about the philanthropy and 
the work you've done through CZI and Biohub,

0:01:06.480,0:01:13.280
it seems, I think, to flow from your identity as a 
physician. I mentioned your pediatrics residency.

0:01:13.840,0:01:18.640
You've talked in the past about how you remember 
times when people would show up with really very

0:01:18.640,0:01:22.720
little knowledge about the rare condition 
that their kid would have, and wanting to

0:01:22.720,0:01:28.240
really try and help those kids. Talk a little 
bit about your inspiration for the work that

0:01:28.240,0:01:33.680
you're doing with CZI, and we'll get on to some 
of the newer things relating to Biohub shortly.

0:01:33.680,0:01:41.120
In pediatrics at UCSF, we saw kids from all over 
come. We just didn't know what they had. And

0:01:41.120,0:01:50.640
oftentimes, you get a printed PDF or you're trying 
to search on PubMed, a name of a gene, or maybe a

0:01:50.640,0:01:56.320
bunch of symptoms, and there's just not very much 
information available, and there's definitely not

0:01:56.320,0:02:00.000
a clear way to translate that to what you're 
supposed to do when they're hospitalized with

0:02:00.000,0:02:06.320
you. That was my first taste of, oh my goodness, 
this is where basic science needs to come in and

0:02:06.320,0:02:12.320
is the source of hope. Because I will tell you, 
when I was applying to residency, a beloved mentor

0:02:12.320,0:02:18.160
of mine was like, you're not going to go down 
the basic science route. And I was like, okay,

0:02:18.160,0:02:23.280
whatever it takes to get into residency, right? 
And so I sort of went down a different route

0:02:23.280,0:02:29.520
around leadership and service. But it was those 
moments when I was like, this is all we have,

0:02:29.520,0:02:35.600
and this is all that this family is clinging 
to and in search of answers for their child.

0:02:35.600,0:02:43.440
So when we had the chance at CZI to work with rare 
patient groups, I remembered exactly what this was

0:02:43.440,0:02:50.720
like. And in 2019, we started the program where we 
gave some seed funding to patient advocacy groups.

0:02:50.720,0:02:55.840
At first I thought — my imagination had not 
been open to what they were going to do. I was

0:02:55.840,0:03:02.640
imagining sort of patient support, walking each 
other down the route of diagnosis and treatment

0:03:02.640,0:03:09.520
options awareness, but what actually happened was 
phenomenal. And now we've done multiple cycles of

0:03:09.520,0:03:18.000
the rarest ones, where we give funding to rare 
disease groups and they have since taken that

0:03:18.000,0:03:26.240
funding to convert it into research agendas for 
their disease. 50 groups working on rare diseases,

0:03:26.240,0:03:33.120
about 20,000 researchers now plugged into 
those research agendas, and the science has

0:03:33.120,0:03:38.880
moved incredibly quickly because the patients 
are bringing assets to the table — incredibly

0:03:38.880,0:03:45.680
important assets without which the research 
would not happen. They've built disease models,

0:03:45.680,0:03:54.800
they've helped cultivate cell lines — hundreds 
of cell lines, hundreds of disease models — then

0:03:54.800,0:04:00.640
clinical registries, biobanks, clinical 
studies, natural history studies. All of

0:04:00.640,0:04:07.040
that makes it easy for folks who then have 
the skills in basic science to contribute

0:04:07.040,0:04:12.000
and actually move their science forward in a way 
that will make a difference for their disease.

0:04:12.000,0:04:22.960
One that I'm particularly excited about this week 
is around the FOXG1 mutation, which causes a rare

0:04:22.960,0:04:33.040
neurodevelopmental disorder in children. The group 
just got FDA approval for their clinical study for

0:04:33.040,0:04:40.480
gene therapy. And I can't even say this without 
getting a little teary. The mom who started the

0:04:41.520,0:04:47.200
group promised her daughter they'd have a 
clinical study by the time the child was 10,

0:04:47.200,0:04:56.800
and they got it approved on Amara's 10th birthday. 
That is incredible. It's also incredibly fast.

0:04:56.800,0:05:02.320
So that work has been incredible and inspiring 
to see how you can do science differently.

0:05:03.520,0:05:09.040
It's really just incredible to see what these 
rare disease communities can do. As you know,

0:05:09.040,0:05:13.360
we've been involved here through our Undiagnosed 
Diseases Network, which you have supported — and

0:05:13.360,0:05:19.120
thank you for that — for many years with the rare 
disease community. But I think until you really

0:05:19.120,0:05:26.000
look straight in the eye a rare disease parent, a 
rare disease family, and understand where they're

0:05:26.000,0:05:31.280
coming from — they talk about being on an island 
often before they're diagnosed, and then joining

0:05:31.280,0:05:35.280
a community when they find that diagnosis, 
and how that's so empowering. When I think

0:05:35.280,0:05:39.440
of that and compare it to what we often get from 
reviewers, which is, where's the actionability

0:05:39.440,0:05:44.880
in your diagnosis? — the parents of these kids 
never ask where's the actionability. To them,

0:05:44.880,0:05:49.120
it's incredibly empowering. That's the start 
of their mission, and then they come together,

0:05:49.120,0:05:52.960
they bring together groups like the FOXG1 
group that you mentioned, and they are focused

0:05:52.960,0:05:59.360
on a cure. There is no one with more energy and 
motivation and focus than a rare disease parent.

0:05:59.360,0:06:06.880
Totally. The parents are so powerful and often 
selfless. Because oftentimes this is not going

0:06:06.880,0:06:12.000
to help their child. And it's about making 
it better for everyone else. Because it can

0:06:12.000,0:06:18.560
sometimes be too late by the time their action 
has led to big changes. But I think the rare

0:06:18.560,0:06:25.280
disease community — we're so grateful that you in 
particular, and obviously CZI itself, backed them

0:06:25.280,0:06:30.960
in this unusual way, which was to combine support 
for bringing the parents together and the families

0:06:30.960,0:06:37.440
together with connecting them to the right 
scientists, because that's the magic. The families

0:06:37.440,0:06:42.880
on their own can support each other, but the cures 
are going to come from the connection to science.

0:06:42.880,0:06:47.600
And science has really been obviously at the 
center of everything that you've done and the

0:06:47.600,0:06:55.840
huge ambition that you have. I wanted 
to jump up to some of the newer things,

0:06:55.840,0:07:00.480
because you've announced some really exciting new 
things just this year. I was lucky enough to be

0:07:00.480,0:07:07.920
at your launch for Biohub and with some of the 
new focus. You've talked about the virtual cell,

0:07:07.920,0:07:13.520
which may be the hottest thing in biology right 
now, and there are many groups focused on it.

0:07:13.520,0:07:18.800
Give us a sense of the scale of that project and 
why you think that could be the answer to many

0:07:18.800,0:07:22.960
of the questions that we're trying to answer 
in basic science and translational medicine.

0:07:22.960,0:07:28.160
First of all, the virtual cell is super exciting, 
and there's not even a true consensus of what

0:07:28.160,0:07:33.200
we're all talking about. So maybe I'll say 
what it means for us at CZI. We want to be

0:07:33.200,0:07:41.520
able to build a virtual model that helps us 
understand the underlying biology that powers

0:07:41.520,0:07:50.400
the human cell. We want to have a computational 
model that allows us to quickly and more cheaply

0:07:50.400,0:07:56.720
look at cause and effect in the cell — look 
at various disease states, perturbations,

0:07:56.720,0:08:02.560
how various genes contribute to changes within 
the cell that may later cause downstream disease.

0:08:02.560,0:08:09.760
That work is super exciting. I think for medicine 
in general — and you work in precision medicine,

0:08:09.760,0:08:15.680
we call it frontier medicine, I think they're very 
similar things — to cure, prevent, and manage all

0:08:15.680,0:08:22.320
disease. That is a big mandate that we don't think 
we're going to achieve alone. And the reason why

0:08:22.320,0:08:29.440
we think we have a good chance at it is because at 
CZI, we are fully committed to building tools that

0:08:29.440,0:08:35.760
make every other scientist faster, more effective, 
and more efficient at their work and research.

0:08:35.760,0:08:42.320
If we can speed everyone up, then science just 
moves faster and we can have new discoveries,

0:08:42.320,0:08:47.280
new knowledge that impacts patients' lives.
And to clarify for everyone, maybe some who

0:08:47.280,0:08:52.640
hadn't realized — that is your mission, 
to cure and prevent all disease?

0:08:52.640,0:08:56.960
Well, we dropped "manage." We're just 
going to cure. Okay, that seems fine

0:08:56.960,0:09:01.200
too. That's not a small ambition. We 
love that ambition in this audience.

0:09:01.200,0:09:06.880
It was very — I will tell you, most reasonable 
scientists could not look at us in the face

0:09:06.880,0:09:11.520
with a straight face and say yeah, that 
sounds like a good idea, 10 years ago.

0:09:12.400,0:09:17.360
But we've gotten lucky because right now we're 
at a moment where I think it actually is feasible

0:09:17.360,0:09:22.400
and possible. And the virtual cell model is 
going to play a large role in this. Right now,

0:09:22.400,0:09:32.960
a lot of basic science is done through someone 
having a clever idea, happy accidents, years of

0:09:32.960,0:09:39.200
dedicated trial and error. And what we're hoping 
to do with the virtual cell model is actually

0:09:39.200,0:09:45.520
de-risk a lot of that — have a computational model 
that allows scientists to test a lot of ideas,

0:09:45.520,0:09:53.440
especially riskier ideas, and then say, 
okay, these are the few that we think are

0:09:53.440,0:09:59.920
most important and highest yield. And then we'll 
test it in the wet lab or in a model or whatnot.

0:09:59.920,0:10:07.280
But what is the holy grail for me is developing a 
model that allows us to understand an individual's

0:10:07.280,0:10:13.120
biology. You talked about the diagnostic journey, 
which is terrible, and you've walked through the

0:10:13.120,0:10:19.520
diagnostic journey with patients many times. 
We have fewer answers than questions when we're

0:10:19.520,0:10:25.120
walking patients through that journey. And 
it's actually sometimes a relief and lucky

0:10:25.120,0:10:32.880
when someone gets a rare disease diagnosis. More 
often than not, you say, I don't know, we looked

0:10:32.880,0:10:37.360
at your whole genome and we think you have 
these three unusual things and we don't really

0:10:37.360,0:10:45.680
know what that means and we don't know what risk 
profile to put you into. That is such a tricky

0:10:45.680,0:10:52.000
part of the patient journey and your journey 
walking through it with them as the physician.

0:10:52.000,0:10:57.360
I want us to be able to have virtual models that 
look at an individual's biology and say, based on

0:10:57.360,0:11:03.920
these three mutations, we see that it impacts 
this protein downstream, and that protein is

0:11:03.920,0:11:09.600
integral in this process. And we can look at the 
risk profile, and we can also look at treatment,

0:11:09.600,0:11:14.960
and we can predict natural history for something 
that we have no word for right now — we just have

0:11:14.960,0:11:22.080
a constellation of symptoms and a few genes that 
we don't know how they actually impact. That model

0:11:22.080,0:11:29.440
will truly get us to frontier biology, precision 
medicine in a way that allows us to treat the

0:11:29.440,0:11:36.240
individual that walks into the clinic and really 
close the gap between basic science, clinical

0:11:36.240,0:11:43.520
research, and treatment. That is super exciting.
No, I mean, clearly the cell is the individual

0:11:43.520,0:11:49.280
unit of biology. It's the singular unit, and we 
have gotten better and better understanding. We

0:11:49.280,0:11:53.920
have single cell genomics, single cell 
proteomics. We're increasingly able to

0:11:53.920,0:12:00.080
measure things at a single cell resolution. But 
I think the scope and ambition of modeling an

0:12:00.080,0:12:04.480
entire cell — especially a human cell; I mean, 
perhaps some recent papers are starting to get

0:12:04.480,0:12:12.240
close with much simpler cells — but to model an 
entire cell means multi-dimensional, and then

0:12:12.240,0:12:16.720
I think part of it as well — and I think you've 
embraced this already — is also the perturbations.

0:12:16.720,0:12:22.320
Because it's one thing thinking about a quiescent 
cell just sitting there and trying to predict its

0:12:22.320,0:12:26.160
gene expression or something similar, but really 
what you want, as you've been talking, is to be

0:12:26.160,0:12:30.240
able to predict what happens when you mutate this 
particular gene in this particular way. And that's

0:12:30.240,0:12:36.960
another scale again. And then there's the spatial 
aspect, right? Things aren't just soup within a

0:12:36.960,0:12:42.080
cell — you're looking at how things are arranged 
within the structure of an individual cell.

0:12:42.080,0:12:47.760
The really exciting thing is that maybe 24 months 
ago we were asking, are the models powerful

0:12:47.760,0:12:55.760
enough? Is there enough compute? We spent a lot 
of time building up a very robust GPU cluster

0:12:55.760,0:13:01.520
to allow us to do this type of AI research. But 
right now the key constraint is data and looking

0:13:01.520,0:13:09.280
at different data formats. There is great data in 
biology, but most of the data available — whether

0:13:09.280,0:13:18.720
it's single cell or spatial work or protein 
work — is often in proprietary data sets, and

0:13:18.720,0:13:25.600
often configured to answer specific questions that 
someone had gone into the work looking to answer,

0:13:25.600,0:13:31.120
which is reasonable. But what we really need 
in biology right now are foundational data sets

0:13:31.120,0:13:37.120
that are open and standardized and that everyone 
has access to in order to do their research and

0:13:37.120,0:13:43.840
build upon. And we're very focused on that at the 
Biohub. We've had some lucky history. You know,

0:13:43.840,0:13:48.720
10 years ago when people told us we were stamp 
collecting, we started doing single cell and

0:13:48.720,0:13:54.960
we built out the Human Cell Atlas Tabula 
Sapiens right here at Stanford. And then we

0:13:54.960,0:14:01.200
built a data set called CellxGene, which 
is one of the largest single cell data sets

0:14:01.760,0:14:10.000
available. And the incredible thing about those 
data sets is that they were built for general use

0:14:10.000,0:14:18.560
and general knowledge. We learned that no one 
group is going to build all these data sets. A

0:14:18.560,0:14:26.320
very happy accident happened with CellxGene. 
We were trying to get a single cell data set

0:14:26.320,0:14:32.720
off the ground and we realized that annotating 
the data was really hard. So CellxGene was not

0:14:32.720,0:14:38.880
originally meant to be a very large data set. It 
was an annotation tool that allowed researchers to

0:14:38.880,0:14:44.880
identify the cells that they wanted to look 
at. This little tool became very useful for

0:14:44.880,0:14:50.400
scientists, and because they were all using 
the same tool, they happened to use the same

0:14:50.400,0:14:54.480
data format, and then we had data that was 
building upon itself, and it became useful,

0:14:54.480,0:15:02.320
and people started giving their data back. We 
seeded the original data — we paid for and seeded

0:15:02.320,0:15:10.000
the original data — and now about three quarters 
of the data in CellxGene is not something that we

0:15:11.680,0:15:17.040
put into the world ourselves. People are just 
building this data set together. And because

0:15:17.040,0:15:25.120
it's standardized, it is incredibly useful and is 
used by folks across academia and industry. And so

0:15:25.120,0:15:30.560
we're following that playbook. We are looking 
to work with academic partners and industry

0:15:30.560,0:15:36.640
partners to build the next data sets in our 
billion cell project. And we are very careful

0:15:36.640,0:15:44.800
to look across species, across ancestry and 
human data, age, and disease states. The key

0:15:44.800,0:15:52.480
will be to get multiple parties to be excited 
and willing to put it in the same data format

0:15:52.480,0:15:58.240
for us to be able to build a foundational 
asset for everyone to be able to use.

0:15:58.240,0:16:01.760
So let me just be clear — you 
just said a billion cells.

0:16:01.760,0:16:07.280
I know, it got big real fast. We were talking 
about one, and clearly, as you mentioned,

0:16:07.280,0:16:13.120
over the last decade you and others — with a 
very big focus for CZI on single cell sequencing,

0:16:13.120,0:16:17.600
as you mentioned the Human Cell Atlas and all 
that work, the mouse one as well and multiple

0:16:17.600,0:16:22.240
organisms — that was in the range of hundreds 
of millions of cells. But you've raised the bar

0:16:22.240,0:16:27.360
a little bit more. The technique has been 
established; it's really about getting the

0:16:27.360,0:16:34.480
right folks involved, doing the science, and 
bringing it back together. We're not limited

0:16:34.480,0:16:40.640
by the number of cells anymore. It's really about 
whether we have the right views on the data.

0:16:42.960,0:16:48.880
The world moves so fast — it's so hopeful.
I love also that you mentioned the spatial

0:16:48.880,0:16:52.800
element, because you've also had this 
interest in imaging. And I feel like,

0:16:52.800,0:17:00.480
talking about perturbing the cell, time is a kind 
of lost dimension. We stop cells or we kill them

0:17:00.480,0:17:10.000
and we measure them. Whereas in life, time is part 
of our lives, and certainly we go to see doctors

0:17:10.000,0:17:15.200
at different points in time. So ultimately, if we 
go all the way to patients, time is an inevitable

0:17:15.200,0:17:20.400
dimension. But with cells we haven't traditionally 
thought about the spatial element. You have unique

0:17:20.400,0:17:28.880
tools from your imaging Biohub that — I hope no 
one here ever needs to become a cryo-EM slide.

0:17:28.880,0:17:34.800
It's much better to be dynamic, warm, and 
alive. Both have cryo abilities and our

0:17:34.800,0:17:38.800
electron microscopes, and we're very proud 
of the work we've done there. It has been

0:17:38.800,0:17:46.080
pretty high ROI investing in a few things like a 
laser phase plate in partnership with a group up

0:17:46.080,0:17:51.440
in Berkeley to really improve our contrast 
and resolution. So we can see down to the

0:17:51.440,0:17:57.760
atomic level of what's happening in various 
cells and preparations. That's really cool.

0:17:57.760,0:18:07.280
We can actually say, with this mutation, this 
protein changes in this configuration, and we

0:18:07.280,0:18:13.520
can see it clearly with this imaging ability. Next 
is going to be looking at video — we want video,

0:18:13.520,0:18:20.160
not static slides. So what are the different 
labeling methods, without labeling, that allow us

0:18:20.160,0:18:28.320
to look at living, dynamic cells at a resolution 
that helps us understand how the gene connects

0:18:28.320,0:18:34.000
to the protein, connects to the living cell.
No, it's amazing. And it's remarkable how often

0:18:34.000,0:18:37.520
the scientists who do that kind of work aren't 
really talking to the scientists who do the

0:18:37.520,0:18:41.600
other work. So being able to bring them together 
— that convening power becomes really important.

0:18:41.600,0:18:48.160
It's been awesome, and that's actually the whole 
premise of our research groups in the Biohubs. We

0:18:48.160,0:18:55.680
put out a grand challenge — whether it's tissue 
engineering or cell engineering or single cell

0:18:55.680,0:19:02.480
work — and we say, this is what we want to do, 
this is our big pie in the sky, do you want to do

0:19:02.480,0:19:20.560
this with us? We invite folks from whatever seat 
they may come from — AI researchers, specialists

0:19:20.560,0:19:29.440
in the physics of lasers, physicians, engineers — 
coming together and building together as a team.

0:19:29.440,0:19:36.880
These multidisciplinary teams are able to do the 
work towards our grand challenges, answer these

0:19:36.880,0:19:42.480
questions while building important data sets in 
partnership. We have nine universities — Stanford

0:19:42.480,0:19:48.640
is one of them — that partner with us deeply. And 
at the same time we build something that's open

0:19:48.640,0:19:54.560
and available for everyone far beyond the nine 
universities to build upon. That's our big goal.

0:19:54.560,0:19:58.400
It's amazing. One of the things that happens 
when you bring people from very different

0:19:58.400,0:20:02.640
parts of science together is they have slightly 
different views on how long things are going to

0:20:02.640,0:20:11.120
take. As you were relaunching Biohub for the next 
10 years, the AI people were on one side saying,

0:20:11.120,0:20:17.440
biology? We can solve that in two or three 
years. Maybe 18 months. And the biologists

0:20:17.440,0:20:22.160
are on the other side going, what? We've been 
doing this for decades. How do you parse that?

0:20:22.160,0:20:28.080
Who do you believe? How do you resolve that? Or do 
you just throw them together and see what happens?

0:20:28.080,0:20:34.000
Well, this is not a new problem for us, 
because we have always had a mix of very

0:20:34.000,0:20:40.640
different people coming together. Cori Bargmann, 
our founding scientist leading our work at CZI,

0:20:40.640,0:20:47.760
joined us 10 years ago. I adore her, love her. 
And she — we wrote down one day in a team meeting,

0:20:47.760,0:20:56.480
we have to sequence our priorities. And 
she writes down "sequence the priorities"

0:20:56.480,0:21:02.240
and then looks around and goes, "Where will 
we get the samples from?" And I was like,

0:21:02.240,0:21:06.480
that's not what we're talking about, we 
just need to write down what we want to do,

0:21:06.480,0:21:17.680
Cori. So we've always had the excitement and the 
additional work of getting people from different

0:21:17.680,0:21:22.800
trainings and different perspectives to come 
together. Who do I believe in this? Well,

0:21:22.800,0:21:33.200
I believe no one. Like a healthy skeptic. 
And the cool part is having people say, well,

0:21:33.200,0:21:37.120
this is why I think it's impossible. My favorites 
are the people who think it's impossible, because

0:21:37.120,0:21:42.160
the people who think everything is possible 
just haven't gotten down far enough. But the

0:21:42.160,0:21:47.120
people who think things are impossible — you ask 
them to enumerate why they think it's impossible,

0:21:47.120,0:21:54.320
what are the blockers — and then there 
are people in the room who can say, oh,

0:21:54.320,0:22:01.120
I actually from a different experience know how 
this works. And that's when the problem solving,

0:22:01.120,0:22:06.480
or the resource allocating, and figuring out 
how to solve the problems together happens.

0:22:06.480,0:22:15.440
I would say science already looks different. 
Therapeutics are being identified and built

0:22:15.440,0:22:22.720
using AI right now, and right now it's anecdotal 
— you're like, oh, this cool thing happened. But

0:22:22.720,0:22:30.720
I think maybe in five years we are systematically 
working in a new way, and the breakthroughs aren't

0:22:30.720,0:22:37.280
anecdotal. We're really looking at marching 
through a deep understanding of human biology

0:22:37.280,0:22:42.800
and helping patients on a regular basis.
I was wondering a little if Mark was on one

0:22:42.800,0:22:47.280
side of that and you were on the other, and you 
were balancing the two. The Mark and Priscilla

0:22:47.280,0:22:53.840
collaboration — what does that look like?
A little bit. Where you see Mark is really

0:22:53.840,0:23:02.800
through the engineer's perspective — the idea of 
doing single cell transcriptomics and building a

0:23:02.800,0:23:08.160
virtual cell model where you understand how the 
system works. That very much appeals to him,

0:23:08.160,0:23:13.680
because he wants to be able to understand and 
instrument the system, understand where the bugs

0:23:13.680,0:23:22.640
are, look at cause and effect, and tinker and say, 
here's the system, if we understand something at

0:23:22.640,0:23:35.760
the fundamental level then we can understand human 
biology. For me, it's always the question of, are

0:23:35.760,0:23:43.920
we answering important questions for patients and 
for clinical use? That's the bit where I'm always

0:23:43.920,0:23:53.120
asking, is this interesting? Tell me why building 
a series of mirrors around the cryo-EM is going to

0:23:53.120,0:24:00.000
help patient care later down the line. And I think 
the work is so exciting because we have a strategy

0:24:00.000,0:24:07.680
that really closes the gap between the physics in 
electron microscopy and the impact in a patient.

0:24:09.840,0:24:16.720
Bringing together basic science and clinical work 
side by side — that is the future of medicine.

0:24:16.720,0:24:22.640
That's — we couldn't agree more with that. And 
having computation next to — and by the way,

0:24:22.640,0:24:28.000
you mentioned your cluster — it's the largest 
nonprofit GPU cluster basically outside of tech.

0:24:28.000,0:24:33.120
It's the largest GPU cluster. Yes.
So you're ready to do this. But having

0:24:33.120,0:24:36.960
that next door to these wet labs, as you're 
talking about, and the cryo-EM, and then having

0:24:36.960,0:24:43.760
on top of that your philosophy of what does 
this mean for a patient — that seems like the

0:24:43.760,0:24:49.360
north star of what you're trying to do. It's an 
incredibly exciting time for biology right now.

0:24:50.240,0:24:54.480
We talked about cells. You also have a 
project called the virtual immune system,

0:24:54.480,0:24:58.080
which is really interesting. I think 
some of the researchers are thinking

0:24:58.080,0:25:05.120
about essentially immune cells floating around 
the body, surveying and looking for disease,

0:25:05.120,0:25:11.120
extinguishing disease wherever they find it.
I do not underestimate immune cells. T cells

0:25:11.120,0:25:14.960
are exquisite. We all have our 
favorite cell. What's yours?

0:25:16.000,0:25:20.800
Maybe an NK cell. Oh, I have a friend who does 
a lot of work on that. She's kind of persuaded

0:25:20.800,0:25:24.160
me over the years. What about you?
Oh, what is my favorite? I have

0:25:24.160,0:25:26.800
to say a heart cell, actually.
I thought you were going to say

0:25:26.800,0:25:31.911
heart cell. I should have said cardiomyocyte. So 
that's obviously the answer. Just forget I said

0:25:31.911,0:25:38.480
that — your favorite cell is the cardiomyocyte.
Let me rephrase my answer. I don't know if

0:25:38.480,0:25:45.360
I have a favorite cell. Maybe my 
favorite cell doesn't exist yet.

0:25:45.360,0:25:50.560
That's a good answer. That is a very good answer.
So you're talking about our work at our New York

0:25:50.560,0:25:57.520
BioHealth where we're doing cellular engineering. 
Our goal is to be able to build an underlying

0:25:57.520,0:26:05.200
platform — and I use "technology platform," not 
the software kind — where we can help cells be

0:26:05.200,0:26:13.760
targeted to a specific destination, read out a 
certain state that we're looking for, encode it

0:26:13.760,0:26:20.960
back into the cell's own DNA, and lyse itself, so 
we can pull out the cell-free DNA and read out the

0:26:20.960,0:26:28.560
data that it has gathered within the human body. 
And then the next step will be, can you get it to

0:26:28.560,0:26:36.560
take an action? That would be incredibly cool. 
The example I like to give is about the heart:

0:26:38.880,0:26:50.320
can you use this technology to send a cell to read 
the plaque burden of a coronary artery, encode the

0:26:50.320,0:26:55.520
data, lyse, and read it out — so that to get 
that information you're injecting cells into

0:26:55.520,0:27:01.760
a patient and then drawing blood and sequencing 
it? Obviously that is still science fiction,

0:27:01.760,0:27:06.720
hopefully not for long, because we have a lot 
of unanswered questions about the inflammation

0:27:06.720,0:27:11.440
that you would trigger, what would happen, can 
we actually read with fidelity. But that's the

0:27:11.440,0:27:17.280
type of technology and platform we think about 
building, because that's a general purpose

0:27:17.920,0:27:23.920
technique. Then you can do a lot of other things 
— you can go into very privileged places in the

0:27:23.920,0:27:32.000
human body and read out data that's useful for us 
to know, in a minimally invasive way. But that's

0:27:32.000,0:27:36.480
where partnership with the community comes in. 
We see ourselves as building the platform and

0:27:36.480,0:27:43.680
technology, in partnership with others. We need 
scientists, physician-scientists, to be thinking

0:27:43.680,0:27:51.760
about the applications of that technology and 
how they would use it in their daily work.

0:27:51.760,0:27:55.120
That creative process — you used the word 
"science fiction." I was going to use that myself,

0:27:55.120,0:27:58.480
because it seems so futuristic. 
But I've been thinking about this,

0:27:58.480,0:28:02.320
even just over the last few days, about how so 
much of the world we look around us right now,

0:28:02.320,0:28:07.360
particularly where AI is concerned, looks like 
science fiction from just a few years ago. If you

0:28:07.360,0:28:12.160
do your own version of science fiction and look 
forward 10 or 15 years — and we were talking about

0:28:12.160,0:28:16.720
the movie in theaters right now, Project Hail 
Mary, which sounds like it's really excellent.

0:28:16.720,0:28:21.680
It's so good. Yeah.
Rocky — that's the name of the alien,

0:28:21.680,0:28:26.400
right? And they name the planet — I'm not 
going to say anything. No spoilers. But it's

0:28:26.400,0:28:32.240
a good movie. Science fiction is creative 
people, many of whom understand science,

0:28:32.240,0:28:37.280
thinking about what the future could be like. In 
a way, that's kind of what you're doing, and also

0:28:37.280,0:28:42.640
trying to help build it. If you have succeeded and 
you recently looked forward to the next 10 years,

0:28:42.640,0:28:49.520
given the tools we have today as well as the 
foundation you built over the last 10 years — how

0:28:49.520,0:28:55.040
do you see medicine having changed? Because we 
are the ones practicing, and we all love science,

0:28:55.040,0:29:00.480
and lots of us have labs as well, but we are 
hoping, as you are I think with your vision,

0:29:00.480,0:29:07.360
that we will change how we do medicine.
I talked a bit about being able to have that

0:29:07.920,0:29:13.520
virtual model of individuals. I think that 
is super powerful — to be able to look at

0:29:13.520,0:29:19.520
an individual's genetics and actually translate 
those genetics in a meaningful way to clinical

0:29:19.520,0:29:25.120
practice. I hope very much that we will start 
looking at individuals that way in 10 to 15

0:29:25.120,0:29:32.080
years from now. But also — this is Stanford 
University, and God bless physician-scientists,

0:29:32.080,0:29:38.960
especially those who've gotten an MD/PhD, that is 
so long. I think if I stayed in medicine, I would

0:29:38.960,0:29:42.720
still be finishing if I were on that route.
And you know that you're always welcome back.

0:29:44.160,0:29:50.080
That is actually my secret retirement plan. 
I wanted to be a pediatric cardiologist. Oh,

0:29:50.080,0:29:57.040
well now we're talking. And I actually did a 
rotation here. It was one of those things where

0:29:57.040,0:30:03.120
maybe Stanford had some advantages over 
UCSF. So I spent some time here. It was

0:30:03.120,0:30:08.720
really something I wanted to do and I haven't 
closed the door on that. But that's a separate

0:30:08.720,0:30:12.960
conversation — kids have to go to college and 
I'll be the oldest fellow that's ever been.

0:30:12.960,0:30:17.760
When you're ready, we know some people. I keep 
my board status active because I plan on this.

0:30:17.760,0:30:24.560
But the part that I think will hopefully change — 
the way physicians can evolve their practice and

0:30:24.560,0:30:29.760
the way we evolve training — is that everyone 
actually becomes a physician-scientist. It's

0:30:29.760,0:30:35.680
not some distant far away land where papers get 
published and then 20 years down the line we get

0:30:35.680,0:30:42.080
a drug. I don't want us to operate that way 
anymore. It is slow and ineffective. We have

0:30:42.080,0:30:49.440
better tools than that. How can we actually close 
the gap so that everyone is thinking deeply about

0:30:49.440,0:30:56.400
the biology, the protein structure, the mutations 
that are happening within individuals and the

0:30:56.400,0:31:03.680
factors that put them at risk, and use that in 
our clinical practice every day? That is how

0:31:03.680,0:31:10.320
we get to the individualized care that we're all 
looking for. And that is how we can very quickly

0:31:10.320,0:31:18.080
translate what's being learned in a lab to impact 
patients. That has to look different in 10 years.

0:31:18.080,0:31:20.960
Thank you, Dr. Ashley, as always. 
And please join me in thanking

0:31:20.960,0:31:25.520
Dr. Priscilla Chan. Thank you everyone.
The preceding program is copyrighted by

0:31:25.520,0:31:33.920
the Board of Trustees of the Leland Stanford Jr. 
University. Please visit us at med.stanford.edu.