Latest / What's Your Problem? / Faster, Cheaper Drugs with AI

Transcript

Faster, Cheaper Drugs with AI · What's Your Problem?

0:15
Speaker 1: Pushkin. Over the past few decades, it's become more and
0:23
Speaker 1: more expensive to develop new drugs. It now costs over
0:27
Speaker 1: a billion dollars on average to bring a new drug
0:30
Speaker 1: to market in the United States, and of course drug
0:33
Speaker 1: companies pass those high development costs onto us in the
0:36
Speaker 1: form of higher drug prices. This has been going on
0:40
Speaker 1: for so long that we have sort of gotten used
0:42
Speaker 1: to it. But when you zoom out, it's strange because,
0:47
Speaker 1: as I've said before on this show, and as I
0:49
Speaker 1: will say again on this show, one of the main
0:52
Speaker 1: things technology does is it makes things more efficient and
0:57
Speaker 1: therefore cheaper. Over the past few centuries, we've seen technologies
1:01
Speaker 1: make all kinds of things cheaper, everything from clothes to
1:05
Speaker 1: food to TVs. So why hasn't new technology made drugs cheaper? Two.
1:15
Speaker 1: I'm Jacob Goldstein and this is What's Your Problem, the
1:17
Speaker 1: show where I talk to people who are trying to
1:20
Speaker 1: make technological progress. My guest today is Alice Zang, co
1:24
Speaker 1: founder and CEO of verge Genomics. Alice's problem is this,
1:30
Speaker 1: how do you use artificial intelligence to drive down the
1:33
Speaker 1: price of discovering and developing new drugs? Why is it
1:38
Speaker 1: getting more expensive to develop drugs, despite the fact that
1:41
Speaker 1: we have better technology to do it. Yeah. Absolutely. One
1:44
Speaker 1: of the reasons is, you know, even though a lot
1:47
Speaker 1: of the new technologies we've developed have made us better
1:51
Speaker 1: at testing more drugs faster, but the fundamental problem is
1:56
Speaker 1: that even if we can get a drug all the
1:58
Speaker 1: way to clinical trials, which is the last step of
2:00
Speaker 1: drug development, ninety percent of those drugs still fail. So
2:04
Speaker 1: if you think about it, we're spending millions on each drug.
2:09
Speaker 1: Of those drugs are failing at the last and most
2:11
Speaker 1: expensive stage of drug development. And so really most of
2:15
Speaker 1: that billion plus dollar figure you hear is due to
2:19
Speaker 1: the cost of failure. Just to be clear, that figure
2:22
Speaker 1: more than a billion dollars. It's you've got to include
2:25
Speaker 1: the cost of all the drugs that don't work exactly,
2:28
Speaker 1: the ones that do right exactly. So the ones that
2:30
Speaker 1: do work have to pay for all the ones that fail.
2:32
Speaker 1: That's the fundamental problem, exactly, And you're setting out to
2:36
Speaker 1: fix that if you can. Absolutely, we think there's an
2:40
Speaker 1: opportunity for AI to fundamentally shift really the failure rate,
2:47
Speaker 1: and the most impactful time to do that really is
2:50
Speaker 1: the failure in clinical trials. So can we predict before
2:54
Speaker 1: we go in to these expensive clinical trials genes or
2:59
Speaker 1: targets or drugs that are more likely to work in humans,
3:02
Speaker 1: because even a ten percent decrease in that failure rate
3:07
Speaker 1: could have massive I saw a number of up to
3:11
Speaker 1: fifteen billion dollars annually in industry cost savings. You could
3:15
Speaker 1: still be in a universe where most of the drugs
3:18
Speaker 1: that go into clinical trials fail, but instead of ninety
3:20
Speaker 1: percent of them failing, seventy percent of them fail, and
3:23
Speaker 1: that would be a huge win. That would be a
3:25
Speaker 1: huge efficiency gain. It would save a ton of money, absolutely,
3:28
Speaker 1: And I think that's something that's underappreciated about AI and
3:31
Speaker 1: really any technology, is that oftentimes people have this expectation
3:36
Speaker 1: that this technology is going to absolutely transform a field overnight.
3:40
Speaker 1: And I think what people don't appreciate is that most
3:42
Speaker 1: of the time that doesn't happen. It's always step by
3:45
Speaker 1: step incremental. But even a ten percent change would have
3:49
Speaker 1: billions of dollars of cost savings and would be a
3:52
Speaker 1: huge win for patients in the industry worldwide. I like
3:56
Speaker 1: that frame, actually, I like that frame of maybe AI
3:59
Speaker 1: can have drugs fail most of the time, but not
4:03
Speaker 1: as much of the time as they fail. Now, like,
4:05
Speaker 1: it seems very credible, It seems very plausible. Would you
4:08
Speaker 1: put it that way? Yeah, it's all life is nothing
4:10
Speaker 1: but a learning process, Yes, getting less bad at everything.
4:16
Speaker 1: So I know you were studying to be a doctor
4:20
Speaker 1: and a researcher not that long ago, a few years
4:22
Speaker 1: ago before you started your company. Like, tell me how
4:25
Speaker 1: you went from an mdphd program to starting the company. Well,
4:30
Speaker 1: my PhD research was actually in using genomic analysis and
4:36
Speaker 1: computational biology to analyze large scale data sets and find
4:41
Speaker 1: new drugs that could improve drug development. And we found
4:46
Speaker 1: that from our very first drug that was predicted from
4:50
Speaker 1: our algorithms when we put it in mice after they've
4:53
Speaker 1: been injured, help them walk and recover from that injury,
4:58
Speaker 1: that nerve injury about four times faster than the leading standard.
5:01
Speaker 1: And that was just the first drug that was predicted.
5:03
Speaker 1: And I looked at this technology in this approach and
5:06
Speaker 1: I thought, Wow, there's so much promise here. You know,
5:09
Speaker 1: am I really going to just publish this and let
5:12
Speaker 1: it sit on a bookshelf somewhere, or if I'm not
5:16
Speaker 1: going to be the one to really develop this to patients,
5:19
Speaker 1: you know who will, And when I looked out off
5:21
Speaker 1: the field, I did not see a ton of biotech
5:25
Speaker 1: or farmer companies that were truly computationally driven. Usually within
5:29
Speaker 1: pharma companies they might bring in computational biologists to support.
5:34
Speaker 1: There are scientists or their biologists, but there wasn't really
5:38
Speaker 1: a genomics computationally driven company at that time. Now there
5:41
Speaker 1: are many, but at the time there are very few.
5:44
Speaker 1: And so I actually, you know, it wasn't a binary decision.
5:47
Speaker 1: People always ask me, how did you make the courageous
5:50
Speaker 1: decision to leap? It wasn't really like that. I think
5:55
Speaker 1: what we did first is that we just took three
5:57
Speaker 1: months three month leave of absence. We joined a program,
6:00
Speaker 1: an incubator called a y combinator. We as you and
6:04
Speaker 1: you and well me and my co founder Jason. And
6:09
Speaker 1: the first question really was, you know, can we even
6:13
Speaker 1: generate some data that validates that computational biology can predict
6:17
Speaker 1: targets that work? And then when we saw some data,
6:21
Speaker 1: the next question was can we even hire people that
6:25
Speaker 1: want to come on? And the next question was can
6:27
Speaker 1: we even raise money from people that will care? And
6:31
Speaker 1: I think that is so such an important lesson because
6:34
Speaker 1: I think people oftentimes get caught up in just the destination,
6:38
Speaker 1: you know, is where I want to be? Is this
6:39
Speaker 1: the career I want to have that they don't take
6:42
Speaker 1: the first step, And really it's the first step that's
6:45
Speaker 1: needed to actually get the data to even decide if
6:48
Speaker 1: it's the appropriate track for you. And did you really
6:50
Speaker 1: just keep thinking, well, this might not work, but let's
6:53
Speaker 1: do the next thing. Were you in a place where
6:55
Speaker 1: you could have gone back to the MD PhD program
6:57
Speaker 1: for a while. Yeah. I took a leave of a
7:00
Speaker 1: continuous leave of absence for probably over five years, probably
7:05
Speaker 1: more than I should have, until the point where a
7:07
Speaker 1: lot of my friends are like, are you really, are
7:09
Speaker 1: you really gonna go back? And finally the medical school
7:13
Speaker 1: is like, you're not really going to come back, let's
7:15
Speaker 1: just terminate your leave of absence. But it was in
7:18
Speaker 1: the first few years a really important safety net for
7:21
Speaker 1: me that gave me the psychological safety to really take
7:25
Speaker 1: a risk and really pursue a new idea that I
7:28
Speaker 1: don't know if I would have otherwise. And I think
7:30
Speaker 1: that's so important. I think for universities to provide is
7:34
Speaker 1: that to recognize there can be more than one track
7:37
Speaker 1: for people to do really excellent science and make an
7:40
Speaker 1: impact more than just becoming a professor. And sometimes that
7:44
Speaker 1: psychological safety is what's needed to help people find their
7:49
Speaker 1: ultimate calling too. By the ways, so far, By the way,
7:55
Speaker 1: what's a very brief definition of computational biology. It's really,
8:02
Speaker 1: at the end of the day, in my view, just
8:04
Speaker 1: the use of computers and data sets to understand and
8:09
Speaker 1: biology better. By the way, what happened to that molecule
8:14
Speaker 1: that you were testing in mice in grad school? That
8:17
Speaker 1: seemed useful? I don't know. It's a good question. Actually,
8:26
Speaker 1: I think the project was taken on by someone else,
8:30
Speaker 1: but I'm not actually completely sure. So, Okay, you leave
8:33
Speaker 1: grad school, you start a company you in fact now
8:37
Speaker 1: have taken You have a bunch of molecules that you're
8:41
Speaker 1: working on, and that seemed promising. But there's one that
8:45
Speaker 1: is in clinical trials now right to treat als Luke
8:49
Speaker 1: Gary's disease, a terrible disease that is very poorly treated.
8:53
Speaker 1: And I thought that we could talk about the story
8:57
Speaker 1: of that molecule of that drug as a way to
9:00
Speaker 1: understand the way your company works. Can you just sort
9:03
Speaker 1: of take me through the life of that drug? So far? Yeah, absolutely.
9:08
Speaker 1: I'll start off just by talking about als and why
9:13
Speaker 1: it's been so hard to discover the right therapy, and
9:16
Speaker 1: then you know why how we did that differently. So,
9:19
Speaker 1: as you might know, LS Luke Garrig's disease is a
9:22
Speaker 1: really horrible disease. What happens is that these neurons called
9:27
Speaker 1: motor neurons start dying, and most patients experience paralysis and
9:31
Speaker 1: then death, usually within three to five years of diagnosis.
9:35
Speaker 1: A very fast progressing disease, and there really aren't any
9:40
Speaker 1: meaningfully effective treatments that really slow or stop the disease today.
9:44
Speaker 1: So a very horrible disease with a horrible prognosis and
9:48
Speaker 1: no available treatments, and why it's been so hard I
9:52
Speaker 1: think to discover really effective treatments is really just the
9:55
Speaker 1: complexity of the disease, and really any disease of the brain,
9:59
Speaker 1: the brain is the most complex organ in the body.
10:02
Speaker 1: So you end up having a lot of drugs brought
10:05
Speaker 1: into clinical trials that worked in mice. I always like
10:07
Speaker 1: to say we've cured LS or can There are many
10:09
Speaker 1: diseases in mice a thousand times, but none of them
10:13
Speaker 1: have really worked in humans. So what we did differently
10:16
Speaker 1: was we started from day one by collecting data from
10:19
Speaker 1: over a thousand ALS patients as well as controls, and specifically,
10:24
Speaker 1: we collected samples of brain tissue as well as spinal
10:28
Speaker 1: cords from these patients that actually passed away from ALS.
10:32
Speaker 1: So you got samples from a thousand patients who had
10:37
Speaker 1: died of ALS. How did you do that? So what
10:40
Speaker 1: we've done over the last seven years is we've signed
10:43
Speaker 1: partnerships with over twenty one different brain banks, hospitals, labs,
10:49
Speaker 1: academic centers worldwide that collect these brain tissues. They're usually
10:53
Speaker 1: donated from patients that have passed away from the disease
10:57
Speaker 1: and whose families want to contribute to research. Could So
11:00
Speaker 1: step one basically is get tissue samples from real patients.
11:06
Speaker 1: And you said controls as well, right, So tissue samples
11:08
Speaker 1: from healthy people as well, so that you can use
11:10
Speaker 1: them as a basis of comparison. You have the samples, Now,
11:14
Speaker 1: what's step two? So step two is that we put
11:18
Speaker 1: an enormous amount of effort into quality controlling these, So
11:22
Speaker 1: that's a big underappreciated step. They can be very noisy samples.
11:27
Speaker 1: And then step three is that we sequence them, so
11:31
Speaker 1: we profile, what is the expression of all twenty thousand
11:35
Speaker 1: genes in the genome, and we also sometimes do DNA sequencing,
11:40
Speaker 1: we look at genetic mutations. We also have a clinical
11:44
Speaker 1: information about that patient, how long did they have the disease,
11:47
Speaker 1: when did they die? And that makes for a very rich,
11:52
Speaker 1: multidimensional data set, and that gives us essentially a global
11:56
Speaker 1: snapshot of what happened in that patient. H okay, and
12:01
Speaker 1: you and presumably the sequencing that you're doing on the
12:05
Speaker 1: patient's tissue samples, you're doing the same sequencing on the controls,
12:09
Speaker 1: the samples from healthy people. So now you have this
12:12
Speaker 1: very large data set. What's the next step. So then
12:17
Speaker 1: you have this snapshot of what happened, and the tricky
12:20
Speaker 1: part is to figure out what caused it. I often
12:22
Speaker 1: liken it to a plane has crashed, right, You're looking
12:26
Speaker 1: through the rubble and you want to figure out how
12:28
Speaker 1: the plane crashed and how that information can be used
12:31
Speaker 1: to prevent further planes from crashing. So that's when our
12:36
Speaker 1: software engineers and data scientists as well as machine learning
12:39
Speaker 1: scientists come in and we have algorithms essentially to integrate
12:43
Speaker 1: multiple data types all the way from the RNA, so
12:46
Speaker 1: how the genes were expressed to genetic mutations to essentially
12:50
Speaker 1: create a map of disease biology, and within the map
12:54
Speaker 1: our networks of genes that are all interconnected that we
12:58
Speaker 1: believe cause disease. And so I like to think about
13:02
Speaker 1: it like when you're looking through a plane crash the rubble,
13:04
Speaker 1: you want to find the black box, which I'll help
13:07
Speaker 1: you figure out the cause of the disease. And by
13:10
Speaker 1: having all the information, we essentially locate the black boxes
13:14
Speaker 1: of disease, the targets that are really at the center
13:17
Speaker 1: of those networks, and then we design drugs against those
13:20
Speaker 1: targets that we believe can reverse disease. It seems like
13:24
Speaker 1: differentiating between correlation and causality in this particular setting would
13:29
Speaker 1: be really hard, right, Like to use the plane metaphor,
13:31
Speaker 1: if you had a bunch of planes that crash and
13:32
Speaker 1: a bunch that hadn't crashed, you might say, oh, like
13:35
Speaker 1: the wings were off all the ones that crashed, and
13:38
Speaker 1: that's why they crashed. But actually the wings came off
13:40
Speaker 1: because they crashed, right, and it was something else that
13:42
Speaker 1: caused the crash. I feel like that would be I mean,
13:45
Speaker 1: an obvious problem. Yeah, that might be hard. To solve Absolutely,
13:49
Speaker 1: you hit the nail on the head, and actually the
13:51
Speaker 1: plane metaphor is a really great one here. For one
13:53
Speaker 1: of the biggest challenges with looking at tissue from a
13:57
Speaker 1: patient that already died is that you're getting the crash right.
13:59
Speaker 1: You're not seeing video of before the crash. You're really
14:03
Speaker 1: getting the crash. And the challenge is how do you
14:06
Speaker 1: figure what caused the crash versus as well was just
14:10
Speaker 1: the effect of the crash, like a burned wing, etc.
14:15
Speaker 1: And one of the ways we do that is we
14:18
Speaker 1: combine different data types. So we found that looking at
14:21
Speaker 1: one type of data, for example, just RNA data is
14:25
Speaker 1: in particularly helpful, but it's actually looking at where do
14:28
Speaker 1: you get convergence signal that pulls through multiple types of
14:31
Speaker 1: data to start revealing more compelling signal. So as an example,
14:35
Speaker 1: we look at genetic data as well. So genetic data
14:39
Speaker 1: is useful for looking at cause versus effect because it
14:42
Speaker 1: contains information about genetic mutations that you were born with
14:46
Speaker 1: as a baby that then lead to increased risk later
14:50
Speaker 1: in life for a disease. And that's kind of nature's
14:53
Speaker 1: human experiment for really cause and effect. And when we
14:56
Speaker 1: layer that on that information on with the RNA data.
15:00
Speaker 1: It actually gives us information about how the genetic drivers
15:03
Speaker 1: are acting in these functional pathways, which is a big
15:07
Speaker 1: issue actually with just looking at genetic data on its own.
15:10
Speaker 1: So I wish I had a better I wish I
15:11
Speaker 1: had a way to actually string that through to the
15:13
Speaker 1: plane metaphor. But and there's a time for leaving metaphors behind.
15:20
Speaker 1: Your company uses AI in drug discovery. I appreciate in
15:24
Speaker 1: a certain way that you haven't said AI yet, but
15:27
Speaker 1: also I don't want to not talk about it. I
15:29
Speaker 1: mean in the sort of figuring out what's going on
15:32
Speaker 1: in this step? Is that well, is that the first
15:36
Speaker 1: instance in this process where you're using AI? Is it?
15:38
Speaker 1: We're talking about that here? Yeah, I mean, I think
15:41
Speaker 1: AI is a really broad term for any kind of
15:45
Speaker 1: process where the computer is learning from something. So there
15:48
Speaker 1: are all sorts of applications of AI in this entire process,
15:53
Speaker 1: for example, how we're integrating the data sets together, how
15:56
Speaker 1: we're inferring what are the central nodes or the key targets.
16:04
Speaker 1: I would say the most classical use of A on
16:07
Speaker 1: the way that most people think of it is then
16:09
Speaker 1: once we have this network of say one hundred genes,
16:12
Speaker 1: how do we actually find what the cause is? How
16:15
Speaker 1: do we find what is the hub or the right
16:18
Speaker 1: target to hit to turn off or on all hundred
16:22
Speaker 1: of those genes. And that's where machine learning and AI
16:24
Speaker 1: comes in handy. In a minute, Alice explains how this
16:31
Speaker 1: actually works in the case of the ALS drug verges
16:35
Speaker 1: working on. Now now back to the show. So okay,
16:46
Speaker 1: Alice and her colleagues at Verge have collected all these
16:49
Speaker 1: tissue samples from ALS patients. They've used the samples to
16:52
Speaker 1: generate this huge data set that shows genetic variation and
16:56
Speaker 1: changes in how genes are expressed, along with lots of
17:00
Speaker 1: clinical data about the patients, and then they build these
17:03
Speaker 1: basically these AI models to try to figure out where
17:08
Speaker 1: in this complicated biological process that's happening in this disease,
17:12
Speaker 1: where they should try to intervene with a drug, Basically
17:16
Speaker 1: where they should try and target a drug. I think
17:19
Speaker 1: of this oftentimes, like if you think of a map
17:22
Speaker 1: of all the airports in the US, you want to
17:24
Speaker 1: figure out how to go after the hubs like Chicago
17:29
Speaker 1: or New York. You don't want to go an airport
17:32
Speaker 1: in Kansas or I will wouldn't be very effective at
17:34
Speaker 1: stopping airplane travel in the country. So there's a lot
17:38
Speaker 1: of different pieces of information that we collect to then
17:42
Speaker 1: infer what are the best genes that are not only
17:45
Speaker 1: central within this network, but also there's independent evidence of
17:49
Speaker 1: a disease causal effect or a relationship to disease. And
17:54
Speaker 1: so you do all that in this instance, and what
17:59
Speaker 1: do you figure out? So what the algorithms spit out
18:03
Speaker 1: is essentially a ranked list of targets, all right, So
18:06
Speaker 1: these are ranked list of targets that are predicted if
18:08
Speaker 1: we could dry them, would restore that network back to
18:13
Speaker 1: levels of healthy people and potentially slow or stop the disease.
18:19
Speaker 1: And then what we do is we take those targets
18:20
Speaker 1: and we start testing them in the lab, all right,
18:23
Speaker 1: So we actually what is kind of cool about the
18:25
Speaker 1: platform is we get all these targets from human brain
18:27
Speaker 1: tissue and we also can test them in human brain
18:30
Speaker 1: cells in the lab. So you get a list it's
18:34
Speaker 1: basically genes to target. You either it says upregulate or
18:37
Speaker 1: make this gene express more or make this gene express less.
18:40
Speaker 1: Is that basically what the AI is out putting exactly, Like,
18:44
Speaker 1: so how long in the instance of this ALS drug.
18:47
Speaker 1: How long was the list? More or less, our initial
18:50
Speaker 1: set of targets was twenty two high confidence targets, and
18:55
Speaker 1: then we actually then generated another chut choosing updated data
19:00
Speaker 1: of about thirty more targets as well. And what was
19:04
Speaker 1: really striking when we tested these targets is that when
19:07
Speaker 1: we tested them in the lab, we found that on
19:10
Speaker 1: average over sixty percent of them, though more recently actually
19:14
Speaker 1: around eighty percent of them actually validated in the lab,
19:17
Speaker 1: so they actually protected ALS patient cells from dying, which
19:22
Speaker 1: is very high. So we're really excited that we're actually
19:25
Speaker 1: seeing very robust validation of the computational predictions, at least
19:29
Speaker 1: in the lab. Okay, so you have this list, you're
19:33
Speaker 1: testing it, something like half of them seem promising, you said,
19:37
Speaker 1: sixty percent seem promising. What happens next? Okay, So what
19:42
Speaker 1: happens next is that we so we test them in
19:44
Speaker 1: these human brain cells. We understand the mechanism. One of
19:48
Speaker 1: the really interesting findings from this ALS program and specific
19:52
Speaker 1: is that when we looked at the network that we
19:54
Speaker 1: found in these patient spinal cords, we found a new
19:57
Speaker 1: cause of disease that was previously unknown so most of
20:02
Speaker 1: the hypotheses in ALS, where many of them to date,
20:04
Speaker 1: have really been focused around these protein aggregates, these clumps
20:08
Speaker 1: of proteins that we can easily observe by ie that
20:11
Speaker 1: you see in ALS patients. Right, A lot of them
20:13
Speaker 1: are observational hypotheses. But what we found by looking at
20:17
Speaker 1: a deeper cut of the data is actually, at baseline,
20:21
Speaker 1: most of these patients actually had a baseline dysfunction in
20:25
Speaker 1: their life csomal pathway, which I like to call the
20:28
Speaker 1: garbage disposal pathway. It's what is critical to clear out
20:32
Speaker 1: junk from the cell. And because patients were at baseline
20:36
Speaker 1: vulnerable to these toxic insults, it wasn't so much the
20:40
Speaker 1: protein clumps that were directly causing it. It was because
20:42
Speaker 1: they're already vulnerable to these clumps of proteins that their
20:47
Speaker 1: cells started dying. And is the idea that the gene
20:51
Speaker 1: you're targeting is causing the cell's garbage disposal to not work, right,
20:57
Speaker 1: Like you're trying to fix the garbage disposal by targeting
20:59
Speaker 1: this particular gene. Yeah, it's a central regulator of that pathway.
21:04
Speaker 1: And it was also a target that was ranked I
21:06
Speaker 1: think it was ranked number one or number two on
21:08
Speaker 1: the list. So just to be clear, how how do
21:12
Speaker 1: you get from you know, so you have fifty or
21:16
Speaker 1: so things to test, fifty or so targets, something like
21:23
Speaker 1: thirty of them seem promising. How do you decide which
21:26
Speaker 1: of those thirty to proceed with? Yeah, so that's a
21:30
Speaker 1: great question. We get asked that a lot. I think
21:33
Speaker 1: at that point it's a strategic decision. Right, you were
21:36
Speaker 1: a startup, Right, we have to be able to develop
21:39
Speaker 1: things quickly and capital efficiently. So we were lucky in
21:43
Speaker 1: that sense that one of the top targets was also
21:47
Speaker 1: a target that already had where the path to developing
21:51
Speaker 1: a drug was relatively smooth, A lot was known about
21:54
Speaker 1: that target. We could start doing chemistry and designing molecules
21:57
Speaker 1: relatively easily, and the target itself had actually been tested
22:02
Speaker 1: in the clinic for other diseases, not als, but things
22:07
Speaker 1: like Crohn's disease and surrounds, so we did know there
22:10
Speaker 1: was some safety data around hitting that target. We do
22:15
Speaker 1: then for targets where we can't develop all of the targets, right,
22:20
Speaker 1: we can only take focused bets for targets where there's
22:23
Speaker 1: a bit more technical risk, Right, It might be a
22:25
Speaker 1: bit more exotic. People don't really understand how it works.
22:29
Speaker 1: There's not a lot of tools out there to really
22:33
Speaker 1: develop drugs against it. That's where we might partner with
22:37
Speaker 1: a pharma company to develop those targets. And we have
22:42
Speaker 1: such a collaboration with Eli Lily where we developed our
22:45
Speaker 1: als target, but actually Lily has the opportunity to essentially
22:48
Speaker 1: take you targets number three through twenty two plus and
22:53
Speaker 1: choose four of them to develop themselves. Oh interesting. So
22:57
Speaker 1: in that way, you're essentially laying off the risk to
23:00
Speaker 1: this giant pharma company that can afford to make more bets.
23:04
Speaker 1: I'd say we're distributing the risk and we're allowing us
23:07
Speaker 1: to really capitalize on the entire opportunity all of the targets,
23:12
Speaker 1: because it's impossible for any small startup to do, you know,
23:15
Speaker 1: thirty different programs. And it's actually in line with what
23:18
Speaker 1: a lot of pharma companies are looking for. A lot
23:20
Speaker 1: of pharma companies are looking for. What is that novel
23:24
Speaker 1: target that no one else is working on that's kind
23:26
Speaker 1: of unexpected, Where if we could really get a competitive
23:29
Speaker 1: edge in here, this would be really meaningful for a
23:32
Speaker 1: position within within drug development in the next ten years. Well,
23:37
Speaker 1: and I mean it also seems compelling because even though
23:43
Speaker 1: this seems like a more promising way to do drug development,
23:46
Speaker 1: drug development is hard enough that anyone candidate drug is
23:51
Speaker 1: probably not going to work, right. Yeah, An any biotechniqus
23:55
Speaker 1: to be able to have a pipeline and the ability
23:57
Speaker 1: to withstand I think some failures because I think it's
24:01
Speaker 1: unrealistic to expect one hundred percent of what you try
24:03
Speaker 1: will work. But that doesn't reflect on the technology itself,
24:08
Speaker 1: and that can be something unfortunate in biotech, where you know,
24:11
Speaker 1: if the first thing fails, everyone's all can be. It
24:16
Speaker 1: can be tempted to say, oh, the technology didn't work,
24:18
Speaker 1: but in reality, you think about how many different drugs
24:21
Speaker 1: that pharmac companies test all the time. Right, So I
24:24
Speaker 1: think really promising technologies need to be afforded that runway
24:28
Speaker 1: and that ability to really take multiple shots on goal
24:31
Speaker 1: before you can get the end to really see if
24:33
Speaker 1: it's working. Right. Well, I mean, if nine of traditionally
24:36
Speaker 1: developed drugs fail once they get to clinical trials, you
24:41
Speaker 1: could be way better but still likely to fail on
24:44
Speaker 1: anyone drug. Yeah, Yeah, even a fifty percent would be huge, right,
24:48
Speaker 1: but still that means one out of two drugs will fail.
24:54
Speaker 1: Relative to the world we live in now, a world
24:57
Speaker 1: where one out of two drugs fail could be a
25:00
Speaker 1: world where we get more new drugs for less money.
25:05
Speaker 1: In a minute, the Lightning Round including the worst thing
25:09
Speaker 1: out being named to the Forbes thirty Under thirty, and
25:12
Speaker 1: the best thing about accepting that your company might sail.
25:22
Speaker 1: That's the end of the ads. Now we're going back
25:24
Speaker 1: to the show. Let's let's close with the Lightning Round.
25:28
Speaker 1: You personally interviewed over a thousand people when you were
25:32
Speaker 1: starting your company, as I understand it, which seems very intense.
25:37
Speaker 1: And I'm sure as if there's anything in your life
25:39
Speaker 1: outside of work where you've been that intense. Oh, everything
25:44
Speaker 1: that is a core to my being. If you ask
25:48
Speaker 1: my spouse, you would say any new game that we
25:51
Speaker 1: start playing. And I'm very competitive and it's just part
25:54
Speaker 1: of my being. I iterate, I get a lot of
25:56
Speaker 1: reps in He always likes to make fun of me
25:58
Speaker 1: that I have an AI in my head. I'm constantly
26:02
Speaker 1: learning and improving the model until eventually I become a
26:07
Speaker 1: lean mean. We've been saying a lot of Katan recently,
26:10
Speaker 1: and I think if we him fifteen times in a row,
26:14
Speaker 1: So yeah, I am very intense and thorough in my life.
26:20
Speaker 1: Is chat GPT overrated or underrated? Both? Actually? I think
26:26
Speaker 1: it's both over and underrated. It's overrated for some applications
26:30
Speaker 1: and underrated for others. I think it's overrated for things
26:34
Speaker 1: where there aren't a lot of information available already on
26:38
Speaker 1: that thing. I think it's underrated for applications at coding,
26:43
Speaker 1: where there's already a large body of literature out there.
26:45
Speaker 1: So it's really good at replicating things that exist, less
26:48
Speaker 1: good at discovering new things that don't exist. I read
26:53
Speaker 1: an interview where you said one of the things you've
26:56
Speaker 1: learned as in running your company is you learn to
26:59
Speaker 1: be okay with your company dying with your company not
27:02
Speaker 1: making it, which I found like very surprising and interesting.
27:05
Speaker 1: Can you just tell me a little bit about that. Yeah,
27:08
Speaker 1: I mean, I think it gets to really the core
27:11
Speaker 1: of how we drive our culture, which is I think
27:13
Speaker 1: that soul for so long companies have been driven through
27:17
Speaker 1: fear and bravado of you know, we're crushing it, We're
27:20
Speaker 1: pounding on our talking about how we're crushing it, and
27:23
Speaker 1: less about emotional vulnerably and introspection and self awareness, and
27:27
Speaker 1: ultimately I found the thing that really transformed my leadership
27:30
Speaker 1: style was learning what I had grips over of where
27:34
Speaker 1: I was really attached to outcomes, And ultimately, I think
27:37
Speaker 1: for all CEOs, a lot of that is tying meaning
27:41
Speaker 1: to what happens with the company. If the company fails,
27:43
Speaker 1: this means something about me as a person, and I
27:46
Speaker 1: think that stifles a ton of innovation and curiosity and
27:50
Speaker 1: tends to drive those cultures of fear. So ultimately, the thing,
27:54
Speaker 1: for example, that got me to stop micromanaging was really
27:57
Speaker 1: being okay with the company dying, because ultimately, what is
28:00
Speaker 1: micromanaging if not just fear right or fear or control.
28:04
Speaker 1: And once you let go of that fear and you
28:06
Speaker 1: recognize you're just open to learning. You can still really
28:08
Speaker 1: want the company to succeed, and you can be passionate
28:10
Speaker 1: about it, but you're no longer thinking, oh, I'm screwed,
28:15
Speaker 1: or like I'm a failure if this fails, and that
28:18
Speaker 1: just opens a whole new level of levity and lightness. Nice.
28:23
Speaker 1: What's the worst thing about being named to the Forbes
28:25
Speaker 1: thirty Under thirty list? I think they did a photo
28:31
Speaker 1: shoot where there was a there was a very revealing
28:35
Speaker 1: split on the dress, and I still get constantly made
28:37
Speaker 1: fun of by my close friends for that. What's one
28:43
Speaker 1: example of a thing that went wrong as you were
28:46
Speaker 1: building the company? Something bad that happened? Oh so many things.
28:50
Speaker 1: We had a whole period where there was a ton
28:52
Speaker 1: of attrition and people leaving, and you know, the first
28:55
Speaker 1: time that happens to a founder can I took it personally,
28:59
Speaker 1: It's like someone leaving your baby, and you wonder why.
29:02
Speaker 1: That was actually a huge growth moment for me because
29:05
Speaker 1: I was for so long trying to put for the
29:09
Speaker 1: strong face. If it's okay, it's okay. And finally, at
29:12
Speaker 1: the end of like a month of this, I just
29:14
Speaker 1: sat in front of the company at an all hands
29:16
Speaker 1: and I honestly I just broke down in tears. I said,
29:19
Speaker 1: I feel like I failed you guys. You know I'm
29:23
Speaker 1: still grieving this. I really don't know what to do.
29:25
Speaker 1: And it was paradoxically in that moment, most of the
29:29
Speaker 1: team really rose up to the occasion and I found
29:31
Speaker 1: support in ways I didn't even know where possible from
29:33
Speaker 1: the team. Alice saying, is the CEO and co founder
29:42
Speaker 1: of verge Genomics. Today's show was produced by Edith Russolo.
29:47
Speaker 1: It was edited by Sarah Nix and Lydia Geancott and
29:51
Speaker 1: engineered by Amanda ka Wong. You're always looking for more
29:56
Speaker 1: guests for the show. If there's someone out there working
29:58
Speaker 1: on an interesting technical problem with big stakes, tell us
30:01
Speaker 1: about that person. You can email us at problem at
30:06
Speaker 1: Pushkin dot fm, or you can find me on Twitter
30:09
Speaker 1: at Jacob Goldstein. I'm Jacob Goldstein and we'll be back
30:12
Speaker 1: next week with another episode of What's Your Problem.