Stephen Quake on Safer Prenatal Genetic Testing, and Detecting Disease Earlier

In this episode of The Future of Medicine, we welcome Stephen Quake, a bioengineer, physicist, and serial entrepreneur whose innovations have transformed how we measure biology and deliver care.

Dr. Quake shares how his early fascination with building and experimentation led him from physics into biology, where he helped pioneer microfluidics, enabling the automation of complex biological experiments. He reflects on founding multiple companies to bring these technologies into real-world use, reshaping research and diagnostics.

The conversation explores the development of noninvasive prenatal testing (NIPT), a breakthrough that allows doctors to detect chromosomal conditions using a simple blood draw instead of invasive procedures. Dr. Quake explains the key insight behind this advance: counting DNA molecules, and how it has since impacted millions of pregnancies worldwide.

They also discuss the early days of genome sequencing, including Dr. Quake’s decision to sequence his own genome and what it revealed about the future of personalized medicine. From there, the conversation expands into liquid biopsies, transplant monitoring, and early cancer detection, highlighting how blood-based diagnostics are transforming how we detect and manage disease.

Looking ahead, Dr. Quake shares his perspective on the next frontier: using advanced molecular tools and AI to detect disease earlier, understand human biology more deeply, and ultimately reshape the practice of medicine.

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

0:00:00.000,0:00:04.440
Physics felt very mature. Biology was 
expanding in all directions. It felt like

0:00:04.440,0:00:08.120
there was more to discover in biology.
Dr. Stephen Quake is a pioneering

0:00:08.120,0:00:12.880
bioengineer and one of the founding fathers of 
microfluidics. He's also known for developing key

0:00:12.880,0:00:17.480
technology for non-invasive prenatal testing.
Huge red herring to try to separate maternal

0:00:17.480,0:00:21.440
from fetal DNA. You don't need to do that.
A serial entrepreneur and former Howard Hughes

0:00:21.440,0:00:25.760
investigator. He most recently served as the head 
of science at the Chan Zuckerberg Initiative.

0:00:25.760,0:00:29.720
When things get mature, I get bored.
In our conversation today,

0:00:29.720,0:00:33.760
we'll explore his journey in transforming the 
landscape of biotechnology, the challenges of

0:00:33.760,0:00:38.200
bridging science and industry, and his vision for 
the future of health and personalized medicine.

0:00:38.720,0:00:43.520
Welcome to Stanford Department of Medicine's 
inside look at the future of medicine.

0:00:43.520,0:00:48.840
Well, Steve Quake, hard to know where to start 
with you and your career and the impact that

0:00:48.840,0:00:54.960
you've had, but we always love with our guests on 
the show to talk about their background and what

0:00:54.960,0:00:58.520
first drew them into science and medicine. 
And I've talked to you about this before,

0:00:58.520,0:01:04.840
so I know one or two of the stories that relate 
to that, but tell us for the listening audience,

0:01:05.640,0:01:11.200
when did you first know you were going to be a 
mad scientist? And how did you move on from there?

0:01:11.200,0:01:16.000
The first experiment I might have done as 
a young boy was to try to make gunpowder.

0:01:18.360,0:01:22.440
And what led you to—how old were you first of all?
Oh my gosh. I don't know. Somewhere in elementary

0:01:22.440,0:01:28.840
school. My parents had given me this two volume 
set of books called The Way Things Work. And you

0:01:28.840,0:01:33.160
could read through it and explain all kinds of 
technological things and how this and that works.

0:01:33.160,0:01:38.080
And I liked shooting off fireworks.
Who doesn't? Exactly.

0:01:38.080,0:01:42.960
Unfortunately, I still have all my fingers.
Oh, I should try to make my own. And I read

0:01:42.960,0:01:47.560
in the book the formula for gunpowder and what 
goes in it. And so I went and tried to gather

0:01:47.560,0:01:54.080
the ingredients and mix them up. And I didn't 
quite appreciate the role of compression and all,

0:01:54.840,0:01:57.600
but so it was a failed experiment.
I see. That's good that it was a

0:01:57.600,0:02:00.000
failed experiment. You told your 
parents you were doing this.

0:02:00.000,0:02:04.520
Of course not. No, of course not.
In the backyard.

0:02:04.520,0:02:07.760
Exactly. Going to burn down 
the forest or something there.

0:02:08.560,0:02:14.280
So building things, inventing things, 
cooking up new things was something—sounds

0:02:14.280,0:02:18.200
like it was kind of almost in your 
DNA as it were from an early stage.

0:02:18.200,0:02:22.800
Absolutely. And I grew up in the early days 
of the personal computer revolution. And so

0:02:22.800,0:02:28.320
that was in the mix as well. I learned how to 
program the computer and use it to control things

0:02:28.320,0:02:33.600
and learn a little bit of electronics as well.
You used to get the magazine. I think I did the

0:02:33.600,0:02:38.920
same. We would get a magazine through and it would 
have software code written on it and you type it

0:02:38.920,0:02:45.760
in and then hack it a little. Was that also your—
For sure. All those trade rags come from this

0:02:45.760,0:02:50.360
far-off place called Silicon Valley.
For me, it was a very far-off. It felt

0:02:50.360,0:02:53.720
far off for me. I was upstate New York, 
so it was far off for me, too. When did

0:02:53.720,0:03:01.640
that transition then to an idea of a specialty 
called science or a profession called science?

0:03:01.640,0:03:08.520
When I first went to college as an undergraduate, 
I knew I was interested in physics. But I was also

0:03:08.520,0:03:12.840
interested in engineering, computers, and so I 
was sort of undifferentiated. It was probably

0:03:12.840,0:03:18.200
going to be one path or the other, but I had some 
very inspiring science, physics and math teachers.

0:03:18.200,0:03:22.120
And where was your undergrad?
Here at Stanford. And a couple

0:03:22.120,0:03:29.080
of very inspiring teachers set me on the 
path of science and kind of stuck with it.

0:03:29.080,0:03:36.880
It seems like it's gone pretty well. But 
computer science and engineering obviously today,

0:03:38.320,0:03:42.520
very obvious the impact that they're having.
Maybe I took the wrong path.

0:03:44.880,0:03:50.520
I don't know. But biology is obviously an 
area where you've had a huge impact. Biology

0:03:50.520,0:03:54.920
and medicine. We'll come to talk about that 
impact. But you started on the engineering side,

0:03:54.920,0:03:59.160
sort of making things work. How did 
you first start to pay more attention

0:03:59.160,0:04:03.240
to biology or what was it drew you into that?
So as an undergraduate, I trained in physics

0:04:03.240,0:04:09.280
and math and was very much in that mode. When I 
got to graduate school and I was trying to decide

0:04:09.280,0:04:15.360
what to specialize in for my PhD, I spent my whole 
first year trying to survey all of physics about

0:04:15.920,0:04:21.040
what would be the most interesting problems to 
work on. What I got to was that interface between

0:04:21.040,0:04:26.720
physics and biology would be the most interesting 
because physics felt very mature as a field and

0:04:26.720,0:04:32.120
biology was expanding in all directions and just 
felt like there was more to discover in biology

0:04:32.640,0:04:37.520
using the sort of approaches and philosophies 
of physics. And so I was at that interface of

0:04:37.520,0:04:42.520
biology and physics for a long time, but then 
brought in these ideas of measurement and

0:04:42.520,0:04:46.400
developing new measurement apparatuses and that's 
where the engineering started to fold into it.

0:04:47.200,0:04:51.680
I remember also one of our prior conversations, 
you talked about the impact of reading Feynman's

0:04:51.680,0:04:56.360
work. Talk about that a little bit.
Oh, when I was in high school,

0:04:56.360,0:05:04.040
I had his autobiography and stumbled across it and 
read it and just loved his fearlessness and his

0:05:05.040,0:05:10.400
irreverence. That was very inspiring to 
me. I mean, I reread it again last year.

0:05:11.080,0:05:15.720
Certainly didn't age well.
No, that's for sure. It didn't age so well,

0:05:15.720,0:05:20.880
but for the moment and of the time it was—
60s when would he have written?

0:05:22.280,0:05:24.880
I think he wrote it in the 80s.
Oh, I see. Okay.

0:05:24.880,0:05:27.920
But definitely about things from 
World War II era and beyond.

0:05:30.120,0:05:34.120
Very interesting. And of course your own 
path took you geographically at least

0:05:35.040,0:05:39.560
down to Southern California of course, but you 
were at graduate school and graduate school was

0:05:40.080,0:05:45.280
started here, right? But you ended up in Oxford.
I did my doctorate in theoretical physics at

0:05:45.280,0:05:48.800
Oxford. But I came and spent part of 
it here visiting to do experiments.

0:05:49.760,0:05:53.400
And why Oxford?
For the adventure of it basically.

0:05:53.400,0:05:58.200
I had not had the opportunity to study abroad 
as an undergraduate and so I thought, well,

0:05:58.200,0:06:04.760
I'm either going to travel abroad and work as a 
ski bum or something, or if I get a fellowship,

0:06:04.760,0:06:10.440
maybe I'll go study abroad. I ended up getting 
a fellowship and so that paid for my experience.

0:06:10.440,0:06:15.120
Nice. So that was a Marshall scholarship. And 
you went to Oxford. Pretty prestigious thing.

0:06:16.320,0:06:18.320
Remind me what college you were at?
Merton.

0:06:18.320,0:06:22.160
Oh, Merton. One of the older ones.
The oldest.

0:06:22.160,0:06:29.080
Okay. So pretty amazing. But very 
different place from Stanford and New York.

0:06:29.080,0:06:34.960
Oh, yeah. I struggled with that. There 
were formal dinners every night. You

0:06:34.960,0:06:39.200
had to wear a tie and a jacket, academic gown.
Don't think I've ever seen you wearing a tie.

0:06:39.880,0:06:44.040
I see you wearing a jacket.
So I did all kinds of experiments to

0:06:44.040,0:06:50.240
understand what the true limits of the dress code 
were. And you definitely needed to wear the tie.

0:06:50.240,0:06:55.640
You had to have a collar on your shirt, but I got 
through with a rugby shirt. Socks ended up being

0:06:55.640,0:07:01.440
important. I didn't wear socks one day. But the 
big discovery was that they could regulate the

0:07:01.960,0:07:06.200
dress code, but they couldn't regulate taste.
I see. So I had the most hideous polyester

0:07:06.200,0:07:10.480
ties you could imagine. And those were fine. 
They didn't give me a hard time about them.

0:07:10.480,0:07:14.960
That was color, any fabric. I 
mean they glow. It was wild.

0:07:15.960,0:07:18.240
Physics professor for sure in the making.

0:07:19.320,0:07:23.360
And then you came back and 
then you were at Caltech.

0:07:24.040,0:07:28.760
I started my faculty career at Caltech. Spent 
almost a decade there. Started as assistant

0:07:28.760,0:07:34.920
professor and moved up the ranks and it was a 
wonderful, wonderful place and experience. Just

0:07:36.240,0:07:41.360
the history, the focus on what it does. 
They really just do great science and

0:07:42.200,0:07:46.720
the community there was just terrific for me 
because I was able to really go across these

0:07:46.720,0:07:51.200
boundaries. I collaborated with a ton 
of people, knew almost everybody, and

0:07:51.960,0:07:57.000
it was just a great spirit of working together.
Did you have a specific mentor down there that

0:07:57.000,0:08:01.320
led you in the direction of measurement and 
sort of helped underline your interest in

0:08:01.320,0:08:05.320
biology and intersection with physics, 
or was it really around collaboration?

0:08:06.240,0:08:12.000
There were so many things that were going on 
over that period. I worked very closely with

0:08:12.000,0:08:15.960
a fellow named Axel Scherer. That's where we 
got the microfluidics off the ground—sort of EB

0:08:15.960,0:08:21.600
lithography person, double E background and just 
a lovely person, and we mentored several students

0:08:21.600,0:08:27.840
and postdocs together and launched a whole bunch 
of the microfluidics work there. Frances Arnold

0:08:27.840,0:08:32.680
was a mentor for me and we worked together on 
using those microfluidic tools for screening and

0:08:32.680,0:08:38.440
molecular evolution, and she's remained a good 
friend to this day. And those are probably the

0:08:38.440,0:08:43.560
two closest collaborators, but I worked with a 
bunch of the biologists—Henry Lester, Bruce Hay,

0:08:43.560,0:08:48.680
Mel Simon. It just kind of went on and on.
Well, I think many people who have been exposed

0:08:48.680,0:08:55.040
to and been impacted by your inventions that came 
in your Stanford time and beyond—some of them,

0:08:55.040,0:09:01.040
especially the doctors, might not know that really 
your early career was around essentially inventing

0:09:01.040,0:09:05.840
microfluidics. The principles were there, 
but really in terms of making it practical

0:09:05.840,0:09:10.040
and making applications. That's the first time—I'm 
not even sure if we met—but the first time I was

0:09:10.040,0:09:16.160
in a talk that you gave was in Seattle at the 
Institute of Systems Biology and it was all about

0:09:16.160,0:09:21.760
microfluidics. So that's how we all knew you, and 
of course Fluidigm was a huge success and really

0:09:21.760,0:09:29.760
has impacted, continues to impact the field. Maybe 
just for the audience, just remind us what that

0:09:29.760,0:09:35.160
is. There are valves that—you don't call them 
these, but other people call them Quake valves.

0:09:36.160,0:09:37.880
They do.
Tell us about that.

0:09:38.640,0:09:46.760
So when I started at Caltech, I was interested in 
trying to do automation of biology because I knew

0:09:46.760,0:09:50.400
I wanted to work on the interface of physics 
and biology. I knew that biologists tended to

0:09:50.400,0:09:55.440
have labs with lots of people and there's a lot of 
manual labor involved and that's not what I wanted

0:09:55.440,0:10:00.520
to mentor. And so I thought, all right, let's 
build tools that automate biology. And I had

0:10:00.520,0:10:06.720
been following these ideas about trying to make 
the integrated circuit of biology—miniaturized

0:10:06.720,0:10:11.560
plumbing with valves and pumps and pipes and all 
that—and had started to dabble in it a little bit

0:10:11.560,0:10:18.000
when I was a postdoc and really then leaned 
into it as an assistant professor. We tried a

0:10:18.000,0:10:22.720
bunch of different things and eventually one of 
the things we landed on was these valves where

0:10:22.720,0:10:27.080
we figured out how to batch fabricate tens of 
thousands of valves on a single chip, sort of

0:10:27.080,0:10:32.560
arbitrary plumbing complexity. And then we went 
about trying to figure out what to use that for.

0:10:33.960,0:10:37.000
One of the first things we decided to 
use it for was protein crystallization.

0:10:38.160,0:10:44.760
And I knew we were on to something when I sent 
my postdoc—my student up to James Berger's lab

0:10:44.760,0:10:51.480
for a week at Berkeley and he did in a week more 
experiments than James' best postdoc had done in

0:10:51.480,0:10:59.920
a year. So we were like, okay, this is—we're on.
Exactly. And then it just went from there. And

0:10:59.920,0:11:02.720
then Fluidigm. Was that your first company?
Yeah.

0:11:02.720,0:11:07.200
So tell us about that. Was it just a very obvious 
thing to do? Was it something you'd always wanted

0:11:07.200,0:11:13.400
to do or did it come out of somewhere else?
Well, it was a little bit maybe

0:11:14.760,0:11:18.600
in the family history. My dad was an early 
software entrepreneur. So I'd seen over his

0:11:18.600,0:11:23.760
career him founding companies and getting them 
off the ground. And it was also a bit that it

0:11:23.760,0:11:28.960
was pretty clear that making these chips 
was complicated enough that no biologist

0:11:28.960,0:11:32.360
would be able to do it. It required a bunch 
of specialized knowledge. And so to see these

0:11:32.960,0:11:37.840
tools have impact in biology would require 
somebody making them in some kind of commercial

0:11:37.840,0:11:42.800
availability. And so I didn't want to stop with 
the work just being published as a paper. I

0:11:42.800,0:11:48.640
wanted to have very broad impact. And it seemed 
like starting a company was the way to do that.

0:11:48.640,0:11:58.000
Did you find it easy, or—I mean, it's never easy.
It's never easy, but raising money is never easy.

0:11:58.000,0:12:02.520
One of the things that places like Caltech, 
Stanford stand out is that there's quite a

0:12:02.520,0:12:07.920
few people around who've done it, so you can 
sort of round a corner and find somebody who

0:12:07.920,0:12:13.000
can give you advice. Was that what you found at 
Caltech, or were you really trying to break out on

0:12:13.000,0:12:17.160
your own and had to learn it by yourself?
Much different environment at Caltech.

0:12:17.960,0:12:23.200
And Caltech had a tech transfer office that was 
run by a fellow named Larry Gilbert in those

0:12:23.200,0:12:30.160
days and he was terrific, and he mentored me and 
many other faculty on how to be an entrepreneur. I

0:12:30.160,0:12:36.560
eventually found the CEO Gajus Worthington who was 
up here in the Bay Area and we founded Fluidigm,

0:12:36.560,0:12:41.360
and he quit his day job and he convinced two other 
guys to quit their day job, very confident they'd

0:12:41.360,0:12:45.840
be able to raise money and launch this thing, 
and almost a year later still hadn't been able

0:12:45.840,0:12:51.000
to raise money. They're living on ramen. One 
guy was getting ready to sell his motorcycle.

0:12:51.000,0:12:57.600
It was much harder than any of us realized it was 
going to be. And biotech was in kind of a down

0:12:58.640,0:13:06.080
cycle then. And Larry—one of the things he 
did was introduce faculty entrepreneurs to

0:13:06.080,0:13:10.880
potential investors. He eventually introduced 
us to a fellow named Bruce Burroughs who became

0:13:10.880,0:13:17.000
the first investor in the company. And Bruce wrote 
us our first check and that got the whole thing

0:13:17.000,0:13:24.040
going, and then it was off and running.
What was—do you remember how long it was

0:13:24.040,0:13:26.720
between when you started the company 
and when the first kind of product

0:13:26.720,0:13:32.480
really shipped to the first customer?
Oh my gosh. Let me think about that.

0:13:34.440,0:13:39.280
It must have been three or four years. Because 
we had to figure out how to do manufacturing.

0:13:39.280,0:13:42.600
And that in itself was a challenge 
because—and you spoke about this

0:13:42.600,0:13:47.200
during your grand rounds—but universities are 
great at some things and are great environments

0:13:47.200,0:13:50.760
for collaboration and for pushing the 
envelope, demonstrating the end of one,

0:13:51.280,0:13:57.440
but then to scale manufacturing or to even think 
about selling something is a whole different

0:13:57.960,0:14:02.200
set of skills. Last time I checked, which 
was last night, but correct me if I'm wrong,

0:14:02.200,0:14:05.680
you have started 12 companies.
Give or take.

0:14:06.480,0:14:10.880
So we'll definitely touch on one or 
two of them as we move forward. But

0:14:10.880,0:14:16.720
there are—probably when I first contacted the 
Office of Technology Licensing here at Stanford,

0:14:17.800,0:14:24.640
we were talking about a variety of different 
faculty who'd started companies also, and I think

0:14:25.480,0:14:30.280
they refer to the Quake factory, because in 
the sense that you were a factory of company

0:14:30.280,0:14:35.680
formation, but each of those has a very specific 
idea and is often based around a very specific

0:14:35.680,0:14:41.760
technology that is new. It's an amazing venture. 
Maybe we can come to talk more about that sort of

0:14:41.760,0:14:48.720
duality as a faculty member. But I really wanted 
to get to some of what I would say in terms of

0:14:48.720,0:14:54.600
lives affected, millions of lives affected in this 
case, patients, and many of the listeners here

0:14:54.600,0:15:01.920
will be familiar with non-invasive prenatal 
testing, and there really—I think there are

0:15:01.920,0:15:07.680
several people who contributed to early work, but 
I think from my perspective there's really two

0:15:08.360,0:15:13.560
big names in the world and you both worked both 
together and maybe even a little competitively

0:15:13.560,0:15:19.920
for a while—Dennis Lo of course in Hong Kong, 
and your work here at Stanford. That all started

0:15:19.920,0:15:27.720
though with something quite close to home.
I never worked on anything clinical before.

0:15:27.720,0:15:34.360
It was all very basic science oriented. And when 
I first became a father, we had to interact with

0:15:34.360,0:15:42.080
doctors and have amnio and this sort of thing, and 
you realized this was not a good state of affairs,

0:15:42.080,0:15:45.760
to have to take risks about your baby 
just to get a diagnostic measurement.

0:15:45.760,0:15:51.480
And so that sensitized me that there's an 
important problem to be solved there. And

0:15:52.800,0:15:57.000
it was in the back of my mind for a number of 
years and then eventually I stumbled across the

0:15:57.000,0:16:03.400
literature on cell-free DNA. And of course in 
the late 90s Dennis and Jim Wainscoat had been

0:16:03.400,0:16:10.240
first to show with molecular methods that some of 
the cell-free DNA in the blood is from the baby.

0:16:10.240,0:16:11.940
And in their case they were 
focused on the Y chromosome.

0:16:11.940,0:16:16.560
Focus on the Y chromosome. Exactly.
And so picking up in the context of a

0:16:16.560,0:16:22.000
male baby—woman of course, female mother—and 
the Y chromosomes must be coming from—

0:16:22.000,0:16:26.400
Exactly, the fetus. And that work had set 
off a decade-long effort to try to build

0:16:26.400,0:16:32.040
a practical diagnostic, and many things were 
tried and failed, and there's no biochemical

0:16:32.040,0:16:36.560
difference between fetal and maternal 
DNA and so forth. No way to enrich, and—

0:16:36.560,0:16:40.800
Sorry to—but Dennis was in Oxford around 
that kind of time. Did you overlap with him?

0:16:42.000,0:16:43.960
Have you ever seen the picture?
No, I have not.

0:16:43.960,0:16:49.320
I'll show you the picture sometime. So it turns 
out we were in Oxford exactly the same time.

0:16:49.320,0:16:51.680
We were.
Wow. He was getting his medical training.

0:16:51.680,0:16:57.400
I was doing my DPhil. His wife was a physicist 
in my department in the physics department and

0:16:57.400,0:17:06.720
we graduated at the same time. We never met. But 
at Oxford as students, at the graduation, someone

0:17:06.720,0:17:09.320
snapped a picture of him and her and I'm in it.
Wow.

0:17:09.320,0:17:12.680
We were standing near each other. Didn't 
realize it. And the first time I visited

0:17:12.680,0:17:17.480
Dennis in Hong Kong, he says, "I'm going to 
show you something." He pulls out this picture

0:17:17.480,0:17:23.640
and it's three of us in the same picture wearing 
graduation robes. I have a copy of it somewhere.

0:17:23.640,0:17:27.880
Okay. Amazing.
So anyways, there was this long

0:17:27.880,0:17:33.600
effort to try to figure out how to take advantage 
of this phenomenon. And I think the key discovery

0:17:33.600,0:17:39.440
was this counting principle. If you can count 
the molecules and measure over-representation,

0:17:39.440,0:17:43.640
that was the breakthrough that enabled all of 
NIPT. And that's what we did here at Stanford.

0:17:43.640,0:17:49.040
And this is counting specific molecules of 
DNA that then map to specific chromosomes.

0:17:49.560,0:17:53.640
So you could look for over-representation 
of chromosomes and therefore see—

0:17:53.640,0:17:58.280
Yeah, the whole—it was a huge red herring to try 
to separate maternal from fetal DNA. You don't

0:17:58.280,0:18:03.000
need to do that. All you need to do is look at 
the over-representation, and it has this amazing

0:18:03.000,0:18:08.000
property that it can be as sensitive and precise 
as you want if you just count high enough. The

0:18:08.000,0:18:11.640
higher you count, the better it works. It goes 
like one over square root of n. So it's a very

0:18:11.640,0:18:16.840
unusual property for a diagnostic test.
And what technology were you using?

0:18:18.320,0:18:24.200
So we started trying to do it with digital PCR.
And remind us what that—

0:18:24.960,0:18:31.280
Okay. So digital PCR is a way to get around 
the nonlinearities of PCR amplification

0:18:31.800,0:18:36.440
by trying to amplify single molecules in discrete 
wells. And it's been around for a long time as a

0:18:36.440,0:18:43.400
concept. Goes back to someone named Sykes and 
Bert Vogelstein and others had been trying to

0:18:43.400,0:18:48.800
do it with emulsions and had published some 
papers for cancer, and we had seen that.

0:18:48.800,0:18:53.720
And I had a rotation student in the lab who 
became my PhD student, Christina Fan. I said,

0:18:53.720,0:18:58.040
"For your rotation project, why don't you try 
to replicate this Vogelstein paper, and let's

0:18:58.040,0:19:03.640
try and make digital PCR happen." And she couldn't 
do it. It's a very fickle thing. The chemistry is

0:19:03.640,0:19:11.880
challenging, the emulsions are hard to work with. 
And so then Fluidigm had just then made their

0:19:12.880,0:19:17.320
digital PCR chip and they made the first 
commercial digital PCR product. So then we said,

0:19:17.320,0:19:23.560
"All right, let's use that." And that worked right 
away. And so we were able to count molecules. We

0:19:23.560,0:19:28.960
could do these kind of dummy-up experiments and 
show that it would work in principle. And then

0:19:28.960,0:19:34.160
we tried to do the real clinical samples. And you 
couldn't count high enough on a single chip for

0:19:34.160,0:19:38.520
those. And I said, "I don't care. Use as many 
chips as you need. I'll spread the sample over

0:19:38.520,0:19:45.040
them. It doesn't matter what it costs." She didn't 
listen to me. She was very conscious of my budget,

0:19:45.040,0:19:50.800
which I guess I appreciated that, but I wanted 
her to do this experiment. And so eventually

0:19:50.800,0:19:56.160
the first sequencers had just come out then. We 
had ours at Helicos, Illumina had theirs. And I

0:19:56.160,0:20:00.000
said, "All right, fine. If you're not going to do 
it with digital PCR, let's do it with sequencing

0:20:00.000,0:20:04.440
and try that. That's another way of counting." 
And for some reason, she agreed to do that.

0:20:05.400,0:20:09.280
And that's how we ended up doing the counting, 
and that's mostly how it's done today.

0:20:10.600,0:20:15.440
Remarkable. And then as a test, you've obviously 
talked about amniocentesis, which was the

0:20:16.360,0:20:24.680
principal method for getting a sample of fetal DNA 
in the past, or cells, in order to do karyotyping.

0:20:25.960,0:20:30.520
But in terms of impact, in terms of 
the number—millions of lives affected—I

0:20:30.520,0:20:34.720
think you made this point well in your talk. 
If you're giving an individual a choice between

0:20:34.720,0:20:40.680
a large needle that would be approached towards 
the head of their unborn child or a blood test,

0:20:41.360,0:20:43.440
most people are going to choose the blood test.

0:20:43.440,0:20:49.160
Absolutely. And of course it's a different 
test in the sense that it's for screening,

0:20:49.160,0:20:55.880
but it essentially takes you to the same 
place, which is an answer to a question.

0:20:56.400,0:21:00.520
It's interesting, this question of 
screening versus diagnostic. And my

0:21:02.080,0:21:06.080
perspective on that—when a diagnostic test 
doesn't work well, it gets used as a screen.

0:21:06.680,0:21:12.680
If it works well as a diagnostic—and in the NIPT 
world, there are some that don't work very well

0:21:12.680,0:21:18.040
because they're optimized for cost or something 
like that, or just weren't designed well. And so

0:21:18.960,0:21:24.640
they all market them as screens, but the best 
ones are as good as amnio in fact. And so

0:21:26.160,0:21:30.280
I think it is just a matter of time and 
accumulation of data and experience before

0:21:30.280,0:21:34.360
that actually becomes the gold standard.
I think the test characteristic is what matters,

0:21:34.360,0:21:41.280
not what you label the test as. You also showed 
the paper—I remember when it came out—the way you

0:21:41.280,0:21:47.360
sequenced essentially an entire baby's genome from 
the blood of, with some knowledge of the family,

0:21:47.360,0:21:51.800
but otherwise from the blood of the mother 
and cell-free DNA, which is remarkable.

0:21:53.880,0:21:59.400
So this counting principle works not just for 
chromosomes, but for whatever, however you want to

0:21:59.400,0:22:03.440
bin the genome. You could do structural variation. 
You made the bins so small they're just a single

0:22:03.440,0:22:08.400
base pair, that let you read out essentially the 
whole genome. And in that paper you're talking

0:22:08.400,0:22:16.880
about, we did it two ways. So one way was to 
get haplotype information from the parents and

0:22:16.880,0:22:21.360
then you could reconstruct the fetal genome 
from that. The other was to do it completely

0:22:24.000,0:22:30.920
naive to the parents. And so you could just draw 
the blood sample without having worked out the

0:22:30.920,0:22:37.520
genetics of the mom or the dad and do a pull-down 
to get the exosomes and do the counting principle

0:22:37.520,0:22:43.640
on the exosomes. And that worked as well.
And it essentially has taken over for all the

0:22:43.640,0:22:47.640
obvious reasons. The test characteristics have 
improved over time. More than just chromosome

0:22:47.640,0:22:53.400
counting has happened over time. Certainly routine 
sequencing of fetal genome isn't something that

0:22:53.400,0:23:00.680
happens yet, but obviously the technology moves 
in that direction. And it's become an everyday

0:23:00.680,0:23:07.120
routine clinical test that everyone really 
expects to have. It's not always reimbursed. It

0:23:07.120,0:23:11.960
seems to be reimbursed more often thankfully than 
it used to be. Things take a while it turns out in

0:23:11.960,0:23:20.840
diagnostics. And then the company you started 
was taken over by Illumina in that case, and

0:23:20.840,0:23:25.920
of course now this is just a routine medical 
test that happens millions of times every year.

0:23:27.280,0:23:32.760
It's been an amazing journey. It's an 
incredible journey, a real transition in

0:23:32.760,0:23:37.480
the diagnostic world, I think, in that from the 
business perspective, people used to look down

0:23:37.480,0:23:42.760
on it. Venture capitalists didn't like it because 
diagnostic readers, antibody tests, 10, 20 bucks,

0:23:42.760,0:23:50.320
something like that. And so hard to build a 
big business around it. But invasive biopsies

0:23:50.320,0:23:54.480
have been expensive. Those are thousands of 
dollars a test. So if you can replace that

0:23:54.480,0:23:59.240
with something that's safer and better, you 
can charge less than what the invasive cost

0:23:59.240,0:24:03.520
but still enough that the business starts to 
look very attractive and the numbers add up.

0:24:05.440,0:24:09.840
Well, when you think about the current state 
of non-invasive testing, or the prior state of

0:24:09.840,0:24:16.920
non-invasive testing, it was based on one protein 
marker, ultrasound of the neck of the baby.

0:24:18.360,0:24:24.160
That speaks to a world that now sounds 
like it's 200 years ago when we think about

0:24:24.160,0:24:30.360
the molecular world we currently live in.
We talked about briefly the first time I think

0:24:30.360,0:24:35.600
I heard you talk, but the first time we really 
interacted, we were meeting in the Clark building,

0:24:36.440,0:24:40.680
planning—I think Mike Snyder, who's now 
the ex-chair of genetics here at Stanford,

0:24:40.680,0:24:47.640
but at the time was about to be the next chair of 
genetics—and I remember well actually wandering

0:24:47.640,0:24:52.240
into your office after I found it, and 
there you are pecking on your keyboard.

0:24:52.240,0:24:57.440
I still think you don't do full typing, right? 
Of your many skills, that's not one. But anyway,

0:24:57.440,0:25:02.280
I found you behind piles of journals. I still 
remember this. And we were supposed to talk

0:25:02.280,0:25:07.280
about the seminar that we were putting together 
for Steve, but instead you had up on your screen

0:25:07.800,0:25:13.520
your genome, which was just mind-blowing 
to me because there were only, I think,

0:25:13.520,0:25:17.480
four people in the world who'd had their genome 
sequenced. They were essentially mostly unknown

0:25:17.480,0:25:23.560
other than Jim Watson. And you had sequenced 
your own genome with two people in your lab

0:25:23.560,0:25:29.440
for—if I recall—this $50,000 or so on the machine 
that you invented, Helicos, and the one prior,

0:25:29.440,0:25:37.160
I think, was done on Illumina with a group of 
like 30 people over 6 months. It cost $250,000. So

0:25:37.160,0:25:42.320
you'd already made headlines for sequencing 
your genome, but nobody had really spent a lot

0:25:42.320,0:25:47.000
of time yet thinking about what medicine might 
do with that genetic information. And that was

0:25:47.000,0:25:52.880
the beginning of an adventure. A little bit of 
an adventure for both of us, certainly for me,

0:25:52.880,0:25:58.360
and thinking about trying to build for ourselves 
first, for you definitely as patient zero

0:25:59.120,0:26:05.040
and then for the world of—what does it look like 
if you have an entire genome of an individual?

0:26:05.040,0:26:10.320
And I tell that story a lot because it had a big 
impact on the way my career went. But I wonder

0:26:10.840,0:26:16.520
how you fit that into the overall picture of your 
career where you've invented multiple things.

0:26:16.520,0:26:21.360
Well, there's a bunch to unpack there, I suppose. 
When you go back to the early days of the Human

0:26:21.360,0:26:26.520
Genome Project when it was just getting underway, 
people appreciated the massive cost that it was

0:26:26.520,0:26:32.040
going to take with existing technologies and 
the need for newer technologies. And so many

0:26:32.040,0:26:37.680
of us in the physics community were aware of that 
and thought, "Oh, we're physicists. We can invent

0:26:37.680,0:26:43.280
measurement machines." And so a bunch of 
us were involved in that. I was one, and

0:26:44.480,0:26:48.720
this is still at Caltech. We invented the first 
single molecule sequencing and published a paper

0:26:48.720,0:26:56.240
on that. That turned into a company called 
Helicos and they scaled it up massively, and

0:26:58.400,0:27:03.520
their product was the first sequencer capable of 
doing an entire human genome with just one machine

0:27:04.400,0:27:09.920
in principle, but nobody was actually doing it. So 
I'm like, "All right, we're going to try it out.

0:27:09.920,0:27:17.040
See if we can really make it happen." And so we 
did it. We sequenced my genome and published it.

0:27:17.560,0:27:21.320
The editor at Nature Biotech wrote a really 
nice piece about how I was a genome altruist

0:27:21.320,0:27:27.720
by giving my genome out there, because at that 
point every individual genome was revealing many

0:27:27.720,0:27:32.880
new loci—hundreds of thousands of loci of human 
variation. And so it was really starting to

0:27:32.880,0:27:38.920
fill out our view of what human variation looked 
like. And then we began to think about, well,

0:27:38.920,0:27:43.680
can we interpret it? And you came in and we 
put a whole team together here at Stanford to

0:27:44.280,0:27:47.800
try to answer the question of what happens when 
a patient walks into his doctor's office with a

0:27:47.800,0:27:54.280
genome and says, "Help me." And that was 
fun. And I love being part of it because,

0:27:55.440,0:27:59.960
as you may recall, I was the one who 
had done the sequencing. So I knew the

0:27:59.960,0:28:04.520
strengths and weaknesses, potential errors, 
and how much trust you should have on any given

0:28:05.160,0:28:08.520
base call. And so I was in the room while all 
the discussions were happening about what or

0:28:08.520,0:28:13.120
what might not be in my genome. And the poor 
genetic counselor had her head in her hands.

0:28:13.120,0:28:18.480
This was not the way it was supposed to happen 
in the textbooks. But it all turned out okay.

0:28:18.480,0:28:23.040
I think it did. Pretty interesting times. 
Like you said, we were challenging our

0:28:23.040,0:28:27.000
colleagues, genetic counselors, to suddenly 
deal with whole genomes and to think about

0:28:27.000,0:28:30.360
how this might happen. You, as the inventor 
of the technology, were also the patient,

0:28:30.360,0:28:36.520
and you were in the room as we discussed it, as 
you said. In the end, pretty clear at that point,

0:28:36.520,0:28:41.200
really no one had received medical advice on 
the basis of having their whole genome data,

0:28:41.880,0:28:45.600
but we discovered a bunch of things with that big 
team, and it really took a team. Remember Russ

0:28:45.600,0:28:50.440
Altman helped, of course, on the pharmacogenomic 
side. Atul Butte's team really put together a

0:28:51.920,0:28:56.280
whole engine for a common variant—and it was 
like an early version of a polygenic risk score

0:28:56.280,0:29:02.160
what he did, right? Very much ahead of its time. 
Pulling together, built one of the first databases

0:29:02.160,0:29:07.600
of genetic variation in relation to human 
disease and then built polygenic risk scores.

0:29:08.200,0:29:13.000
And then our team, we thought about rare disease, 
and we tried to put all of that together. And then

0:29:13.000,0:29:18.560
also we had you in the clinic. Finally your family 
had been on at you to go see a cardiologist.

0:29:19.240,0:29:24.320
There was one in front of you. So you ended up 
coming into our clinic and we tested you on a

0:29:24.320,0:29:27.120
bike. You took very good care of it.
Oh, good. I'm glad to hear it.

0:29:27.920,0:29:32.400
Checked out your heart. Thankfully, we didn't find 
anything there, but we did find a few things that

0:29:32.400,0:29:37.880
were of relevance for your disease and for your 
risk of disease, particularly heart disease. And

0:29:38.480,0:29:42.920
I ended up making some of the first kind of 
recommendations that were genomically informed.

0:29:42.920,0:29:49.600
I think famously—at least in my mind—we had an NPR 
interview together, kind of like this as I recall,

0:29:49.600,0:29:53.600
might have been the last time you and I were next 
to each other in a microphone, and they asked if

0:29:53.600,0:29:58.280
you had actually taken the advice that we'd given 
you. And I think at the time your answer was no.

0:29:59.240,0:30:05.240
You got me over the hump eventually.
Right. So glad to hear that you bought into

0:30:05.240,0:30:12.000
the genomic medical revolution that you helped 
ignite. And now sequencing genomes for medicine is

0:30:12.000,0:30:16.840
routine. It literally is something that we do 
every day and it happens around the world. So

0:30:16.840,0:30:22.600
it's exciting, I think, to see that. I still think 
we're far short of realizing the full potential of

0:30:22.600,0:30:27.920
genomic information for medicine, but especially 
for rare disease, I think we've seen a lot of

0:30:27.920,0:30:33.160
lives impacted. And hopefully, I think that in 
addition to the work you did with cell-free DNA,

0:30:33.160,0:30:38.600
that's a whole other area where I hope 
you would gain some sense of pride and

0:30:38.600,0:30:44.840
fulfillment from being there at the beginning.
Oh, it was a ton of fun. And maybe to get back

0:30:44.840,0:30:51.960
to your earlier question, once we did my genome, 
I felt like I had to make a decision. Did I want

0:30:51.960,0:30:59.320
to go full in to the human genomics field or do 
something else? And it was clear that the big wigs

0:30:59.320,0:31:03.480
in the field were just raising enormous amounts 
of money. And it was like for them all about scale

0:31:03.480,0:31:09.360
and resources and how many could you do. And that 
didn't seem intellectually interesting to me. And

0:31:09.360,0:31:16.680
so I kind of made a very conscious decision. All 
right. We did my genome. We did one tumor genome.

0:31:16.680,0:31:24.400
And I'm going to find something else to do. And it 
was because of that decision and creating kind of

0:31:24.400,0:31:32.760
the open space that I turned into liquid biopsies. 
And that ended up being a very good decision.

0:31:33.640,0:31:38.240
Well, and I remember at the time talking to you 
as we were thinking, we've done one patient,

0:31:38.240,0:31:42.680
we've done 10. The next we have to do a hundred 
or a thousand and really think about how this

0:31:42.680,0:31:46.080
is going to impact medicine more broadly. 
But at the time, it was around then that

0:31:46.080,0:31:50.880
you had another conversation at Peet's Coffee, 
I think, with one of my friends and colleagues,

0:31:50.880,0:31:57.400
Hannah Valantine. Just tell us about that.
So we had published the first NIPT paper, and

0:31:58.440,0:32:02.920
Hannah called me up and said, "Hey, is this 
going to work for heart transplant? Because

0:32:02.920,0:32:07.440
we have a similar problem. We have to do invasive 
biopsies and we've gone through all this trouble

0:32:07.440,0:32:11.800
to give these patients a new heart. Then what 
do we do? We go start pulling pieces of it out

0:32:12.360,0:32:16.120
every couple of months to see if it's rejecting 
or not. And wouldn't it be so much better if

0:32:16.120,0:32:21.240
there was a blood test?" And so we put our heads 
together and figured out that we could monitor the

0:32:21.240,0:32:26.600
specific cell-free DNA from that transplanted 
heart because the donor's genome was different

0:32:26.600,0:32:30.360
than the recipient's, and the amount of donor DNA 
in the blood, by actually sequencing and looking

0:32:30.360,0:32:35.320
at polymorphisms, would reflect how much rejection 
is happening. We called it a genome transplant.

0:32:35.320,0:32:40.760
Yeah, exactly. We called it a genome transplant. 
Well, and even more than that—I've done the

0:32:40.760,0:32:44.880
cardiac biopsies, not for many years now, 
but I'm glad thankfully we're all doing

0:32:44.880,0:32:48.600
many fewer of them. But the catheter that's 
used is actually called a Stanford catheter.

0:32:48.600,0:32:53.280
But that doesn't mean it's great. In this case, 
you're passing it through one of the valves,

0:32:53.280,0:32:57.440
the tricuspid valve in the heart, and then just 
sort of poking it towards the middle of the heart,

0:32:57.440,0:33:01.040
which first of all isn't necessarily the right 
place to go, but it's the place you can access

0:33:01.800,0:33:07.160
and then you just kind of grab and pull. You think 
of a biopsy, you hope that there's millimeter

0:33:07.160,0:33:13.080
precision. This is not one of those things. It's 
rather a barbaric procedure actually. So the same

0:33:13.080,0:33:17.360
way that amniocentesis is something that you 
would never invent if there was an alternative,

0:33:17.880,0:33:21.560
I think you could say the same for cardiac 
biopsy. So but this was another story—and

0:33:22.840,0:33:27.880
you alluded to this earlier—it didn't immediately 
work like that. There were challenges with signal,

0:33:28.440,0:33:35.640
but eventually it worked very successfully.
That's right. The first pilot experiment worked

0:33:35.640,0:33:40.880
very well and then the first time we tried to 
scale it up to a larger study, it stopped working.

0:33:40.880,0:33:50.400
And it took a long time to get to the bottom 
of it. One of the interesting things about the

0:33:50.400,0:33:54.720
project is it's a precision measurement. We're 
trying to measure things at the 1% level, and

0:33:56.040,0:34:00.760
small sources of noise and error throughout 
the whole thing can really mess that up. And

0:34:00.760,0:34:05.200
this is not like with NIPT where you're 
counting chromosomes and doesn't matter,

0:34:05.200,0:34:09.480
as you mentioned, whether you'd separate the 
mother and the fetus—here you really have to—

0:34:10.000,0:34:12.440
The point is separating. Here 
the point is to separate.

0:34:12.440,0:34:16.600
And in fact, so we would genotype the donor 
and the recipient and some of the errors in the

0:34:16.600,0:34:21.200
genotyping arrays were propagating their way 
through the whole thing and we had to figure

0:34:21.200,0:34:26.040
those out, compensate for them, and then it all 
started working. But it took a while to get there.

0:34:26.760,0:34:31.400
None of these things happen overnight. But then 
a pivotal clinical trial and then a company,

0:34:32.360,0:34:36.400
and now thankfully there are 
many, many fewer cardiac biopsies.

0:34:36.400,0:34:41.800
It's not just the heart, I guess, as well. Lung.
It's a principle that works on any solid organ,

0:34:41.800,0:34:47.080
which is super powerful and flexible.
Amazing. And of course there's a whole

0:34:47.760,0:34:52.960
revolution also happening. You talked a little 
bit about it today within the cancer world,

0:34:52.960,0:34:57.600
thinking about both early detection and 
potentially multi-cancer early detection,

0:34:57.600,0:35:04.120
and I think what is more—I think sort of embedded 
now clearly in mainstream medicine is the idea of

0:35:04.120,0:35:09.080
minimal residual disease testing for cancer 
recurrence, let's say after surgery, which is

0:35:09.080,0:35:15.040
very powerful. So this move towards molecular 
tests rather than relying on imaging—why wait

0:35:15.040,0:35:19.480
till a cancer is big enough that you can image 
it if you can pick up small numbers of molecules

0:35:19.480,0:35:27.680
in the circulation? So it's really, I think, been 
extraordinarily impactful across all of medicine.

0:35:28.320,0:35:35.440
Mostly been DNA, but RNA is out there, too. 
And you have at least one company now that's

0:35:35.440,0:35:40.200
focused on RNA and a lot of really interesting 
science. Just tell us a little about that.

0:35:40.960,0:35:45.000
So cell-free RNA ends up being a very 
powerful way to measure phenotype. Not

0:35:45.000,0:35:48.160
everything is genetic and you don't always 
have a different genome in your body.

0:35:48.720,0:35:57.080
And so the RNA ends up being a very good 
complement to DNA. And we initially went back to

0:35:58.520,0:36:04.040
maternal health to try to solve the problem 
of preterm birth, which—the genetic component

0:36:04.040,0:36:08.200
contributor is very small, but there's clearly 
a phenotype thing. So we thought maybe RNA would

0:36:08.200,0:36:13.400
provide a signal for that, and it does.
And the RNA from—is it specific cell

0:36:13.400,0:36:17.480
types that it's reading out? What 
is the signal actually coming from?

0:36:17.480,0:36:20.880
Some of it comes from placenta, some comes 
from mom's immune system. Those are the two

0:36:20.880,0:36:25.800
main contributors for preeclampsia anyways 
and for preterm birth. You can also measure

0:36:25.800,0:36:31.320
fetal transcripts, which is very interesting 
biologically, sort of scientifically. But we

0:36:31.320,0:36:36.320
haven't seen a clinical application of that yet.
And so that now led to another adventure, Mirvie,

0:36:36.320,0:36:41.720
launched the Encompass test, 9,000 women 
prospective clinical trial. And so last year

0:36:41.720,0:36:50.000
that's now kind of out there and it's been sort 
of amazing to see. For preterm birth, which is

0:36:50.000,0:36:55.280
almost 10% of pregnancies affected by it 
with no meaningful diagnostic before this,

0:36:55.840,0:37:01.880
we think that's going to be just another 
major, major impact on helping to save lives.

0:37:02.880,0:37:08.000
Any of us who've worked at the bench understand 
that DNA is fairly robust. RNA falls apart just by

0:37:08.000,0:37:15.160
looking at it. It's pretty stable circulating. 
It's a sort of counterintuitive thing.

0:37:16.280,0:37:23.320
The cell-free RNA that survives in your blood 
has been either trapped in exosomes or wrapped

0:37:23.320,0:37:28.720
on protein particles. So it's got some kind of 
protection that has kept it intact, and it's

0:37:28.720,0:37:34.160
remarkably stable, much more stable than 
completely purified RNA you get on the benchtop,

0:37:34.160,0:37:38.480
which is very labile. And we've 
got amazing luck going back into

0:37:39.160,0:37:47.600
bank plasma really and serum from years ago 
really and lots of cell-free RNA, because it's not

0:37:47.600,0:37:54.520
completely purified, it's stabilized.
I remember one of the early insights,

0:37:54.520,0:37:59.280
I think from you or Dennis, was just how 
short the half-life of DNA was in a pregnant—

0:37:59.280,0:38:03.600
Dennis did beautiful work on that. Absolutely. 
Looking at Y chromosome after delivery of a male

0:38:03.600,0:38:12.120
baby, and it goes—turnover time is half an hour.
But the RNA is stable enough that even in bank

0:38:12.120,0:38:17.560
samples you can find it. That's remarkable.
You've invented so many things in the last few

0:38:17.560,0:38:23.360
years. You've spent quite a lot of your time not 
taking grant money and inventing things with it,

0:38:23.360,0:38:27.880
but rather giving out money with a big 
vision as part of the roles. You've

0:38:27.880,0:38:33.880
played multiple roles within the Chan Zuckerberg 
Initiative. How did that first come about, and

0:38:34.800,0:38:38.400
tell us about your time there.
That was an amazing adventure.

0:38:40.360,0:38:43.480
When Mark and Priscilla decided to 
launch their science philanthropy,

0:38:43.480,0:38:49.160
they reached out to the leaders of Stanford and 
UCSF and asked for advice, and universities put

0:38:49.160,0:38:54.360
together committees to help propose ways to design 
an institute and what to work on. I was part of

0:38:54.360,0:39:00.280
that. And then eventually got asked along with Joe 
DeRisi to be the founding leader of the institute,

0:39:00.280,0:39:06.720
the first Biohub. And so we brought Berkeley 
into the fold. So we had all three major Bay

0:39:06.720,0:39:11.680
Area institutions to take on a couple of big 
problems. One in infectious disease, which was

0:39:11.680,0:39:16.520
Joe's interest area, and another in building 
cell atlases, which is what I wanted to do.

0:39:17.280,0:39:24.840
And as we were talking about this earlier, as 
powerful and as much as we know about the genome,

0:39:25.720,0:39:29.360
you still can't predict the cell types of an 
organism from the genome. That's an unsolved

0:39:29.360,0:39:37.120
problem. And we had developed single cell 
techniques to characterize gene expression

0:39:37.120,0:39:42.720
of single cells and thereby in principle create 
an atlas of all the cell types in the human body,

0:39:42.720,0:39:50.600
and we set out to do that at the Biohub and 
succeeded. It was just great team science. There's

0:39:50.600,0:39:56.760
like 150 authors on each paper involved, dozens of 
groups from all three institutions, and figuring

0:39:56.760,0:40:02.080
out how to coordinate that was a new thing for me.
You were getting every human tissue and then

0:40:02.080,0:40:09.520
sending it out to labs and then getting level—
Right. And then that is available to the world.

0:40:09.520,0:40:11.040
We just gave the data away.
Absolutely.

0:40:11.040,0:40:18.640
Even before we published it. So we had a very 
strong commitment to open science. And we were

0:40:18.640,0:40:23.520
probably the first organization to require the 
use of preprints. So everyone we funded, we said,

0:40:23.520,0:40:27.480
"You've got to put a preprint of your work up 
when you submit it to a peer-review journal."

0:40:28.280,0:40:31.920
And for our big projects, we not only put 
the preprint up, we put the data out there.

0:40:32.800,0:40:37.200
And we gave people access to the data before 
the papers were published in the name of trying

0:40:37.200,0:40:43.400
to accelerate science and lift boats everywhere.
I think there isn't an institute or an initiative

0:40:43.400,0:40:50.240
with a more bold aim. This is obviously something 
that people focused on managing or curing all

0:40:50.240,0:40:56.160
disease within—I think initially a 10-year 
period, maybe a little bit longer—100 year period?

0:40:56.160,0:40:58.840
100-year period.
100-year period, okay. And

0:40:58.840,0:41:05.040
I think that the word "manage" is key, but we 
live in an amazing moment for discovery science

0:41:05.040,0:41:12.240
and its translation to medicine, and I think this 
was your initial work in the context of Biohub.

0:41:12.240,0:41:16.040
The last few years you were also helping 
build other Biohubs. There's multiple of them.

0:41:17.240,0:41:20.880
Yes, exactly. Well, I moved on to manage all of 
Mark and Priscilla's science philanthropy as the

0:41:20.880,0:41:25.640
head of science, and part of that was creating 
a Biohub network which we established in Chicago

0:41:25.640,0:41:29.520
and New York, as well as in collaboration with 
the Allen Institute and outpost in Seattle.

0:41:30.680,0:41:35.520
And more generally funding science globally. 
We funded grants in something like 30 different

0:41:35.520,0:41:43.440
countries, and that was a lot of fun to be part of 
that and to really try to get that launched. And

0:41:43.440,0:41:50.120
part of what I got to do was lead the pivot 
towards AI. That was not the obvious thing

0:41:50.120,0:41:57.280
to do at the time, but there was enough 
early kind of examples that it felt like

0:41:58.040,0:42:03.280
the perfect thing for CZI to be investing. Still 
risky. Wasn't clear it was going to work out,

0:42:03.880,0:42:07.160
but a good thing for philanthropy to do.
What struck me with that was just how

0:42:07.960,0:42:14.480
elegant the foundational work for the cell atlas 
is for the pivot towards AI, the idea of AI

0:42:16.560,0:42:19.800
and biology. It couldn't have been predetermined 
because so much of what we understand about the

0:42:19.800,0:42:26.880
power of the new AI in this third revolution is 
only been two or three years old. So in a sense,

0:42:27.440,0:42:31.360
you might have had other reasons for 
originally initiating the cell atlas work,

0:42:31.360,0:42:35.960
but of course it was just good for the world, 
and then it's perfectly poised thinking about

0:42:35.960,0:42:41.040
now a sort of in silico immune system or some of 
these other ideas that are coming out as potential

0:42:41.040,0:42:48.680
priorities for the Biohub network going forward. 
I think Priscilla's actually coming on our

0:42:48.680,0:42:54.040
podcast—32 grand rounds in 32. So 
we'll be able to follow up and ask her.

0:42:54.040,0:42:57.440
Is she going to give grand rounds herself?
I think she might. We may do it in this format

0:42:57.440,0:43:01.640
though. We may talk about it. She being 
a physician, she's certainly qualified.

0:43:03.200,0:43:06.680
No, we're very excited that 
they have such a big vision

0:43:06.680,0:43:12.040
that goes all the way from very basic discovery 
all the way through to disease. And obviously

0:43:12.040,0:43:16.440
we're really excited about the impact of AI at 
multiple levels—molecular level, cellular levels,

0:43:16.440,0:43:22.160
all the way up to patient interaction levels.
So we have missed out a whole bunch of stuff

0:43:22.160,0:43:26.880
that you've invented over the years and impact 
that you've had. But I wanted just to spend the

0:43:26.880,0:43:35.120
last few minutes sort of thinking about more 
generally—in anyone's early career, physicians,

0:43:35.120,0:43:40.200
scientists, physician-scientists, they're 
trying to build, hoping for maybe one big hit,

0:43:40.200,0:43:44.240
something that has impact maybe all the way 
through to a patient and maybe starting a

0:43:44.240,0:43:51.280
company that has some impact. You've done this 
tenfold more than most people. And I wonder,

0:43:52.040,0:44:00.720
what lessons do you take, or what do you ascribe 
that sort of disproportionate success to?

0:44:00.720,0:44:02.960
It's hard to know because I 
don't get to replay the tape.

0:44:04.280,0:44:11.360
Part of it is being willing to jump into the 
unknown and work on emerging areas rather than

0:44:11.360,0:44:16.440
mature ones. And that enriches your chances, 
I think, of doing something like that. And

0:44:18.000,0:44:24.080
so I've transitioned in my career from one area to 
another, and as you know, when things get mature,

0:44:24.080,0:44:28.360
I get bored. And I look for sort of 
what's the next emerging thing to work on.

0:44:29.200,0:44:31.600
The gunpowder.
Exactly. Also,

0:44:31.600,0:44:35.400
it's just working with terrific people. I've 
been so fortunate to have great students and

0:44:35.400,0:44:40.400
postdocs and collaborators over the years, 
and that's driven so many things as well.

0:44:40.400,0:44:45.000
We've talked a little about in the past about 
the sort of secret sauce of Stanford and the

0:44:45.000,0:44:48.720
ready collaboration, but also just 
the fact we're all here in one place.

0:44:48.720,0:44:53.840
Is that, do you think, an important part?
100%. When I moved from Caltech to Stanford,

0:44:54.760,0:44:59.200
I knew that there were a bunch of cell biologists 
and such that I wanted to collaborate with,

0:44:59.200,0:45:05.320
the stem cell folks, and that all came to 
pass. I had no idea how much fun I'd have

0:45:05.320,0:45:10.000
with the doctors and physicians. That opened up 
a whole new world to me and we've done things,

0:45:10.000,0:45:15.320
as we talked about in the grand round—that's 
the whole story of that over the last 20 years

0:45:16.320,0:45:20.160
that I never thought was going to happen, and 
it would have been hard to do it anywhere else.

0:45:20.840,0:45:26.000
It is a great, great environment for that. Last 
question then really is—this podcast is called The

0:45:26.000,0:45:31.320
Future of Medicine. There are few people better 
positioned. We've certainly had a lot of guests

0:45:31.320,0:45:36.080
on who've given us our view. Eric Topol was on 
recently and he's very well positioned as well.

0:45:37.200,0:45:44.000
But you're someone who invents the future. So what 
are you most excited about, the next few years,

0:45:44.000,0:45:51.760
either for yourself or just more generally?
So I am—on the liquid biopsy world, very

0:45:51.760,0:45:58.000
bullish about the early cancer detection work, 
multi-cancer detection. I feel like there's been

0:45:58.000,0:46:02.640
a bunch of early attempts, a bunch of marketing 
and things. It hasn't quite come together yet,

0:46:02.640,0:46:07.320
but it's poised to do it now. I think we know 
enough collectively that that's going to really—

0:46:07.320,0:46:11.680
And do you think that will be 
DNA methylation plus RNA? Do you

0:46:12.760,0:46:18.920
have a sense of what the measurement will be of?
I'm most optimistic about hydroxymethylation.

0:46:18.920,0:46:22.480
That's like your best compromise between 
methylation and RNA. It's right in between.

0:46:22.480,0:46:27.520
It's the genes that are being turned on, but it's 
a DNA measurement. It's remarkably predictive.

0:46:28.520,0:46:35.240
The NCI is doing this Vanguard trial of early 
cancer detection. Only two companies in it, and

0:46:35.240,0:46:39.760
one's methylation, the other's hydroxymethylation. 
But the other one I'm super enthusiastic about is

0:46:39.760,0:46:45.480
dementia and neurodegeneration. I think there's 
going to be tests for that. That whole field is

0:46:45.480,0:46:49.400
ripe for disruption. There's been a lot 
of orthodoxy and received wisdom, and

0:46:50.200,0:46:52.920
I think there's going to be some 
amazing things happening there.

0:46:52.920,0:46:57.600
We certainly need it. Well, Steve, we could 
talk for much longer and you've certainly

0:46:57.600,0:47:01.800
done many more things, but I'm glad to have 
a chance to talk about your entire life so

0:47:01.800,0:47:06.120
far. Many more things still to happen. Thank 
you for joining us on The Future of Medicine.

0:47:06.120,0:47:10.120
Thank you. It's been great.
The preceding program is copyrighted by

0:47:10.120,0:47:18.400
the Board of Trustees of the Leland Stanford Jr. 
University. Please visit us at med.stanford.edu.