Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
What’s Next for Personalized Medicine? | Sapient Bio CEO Dr. Jonathan Usuka
30 Plays12 days ago
Dr. Jonathan Usuka, CEO of Sapient Bio, is building high-dimensional proteomics platforms to help pharma understand disease at the molecular level and develop more targeted therapies.
With a career spanning early computational genomics during the Human Genome Project, to R&D roles at Roche and Celgene, and later pharmaceutical analytics at McKinsey, he has consistently worked at the intersection of biology and data. In this episode, we explore the shift from organ-based definitions of disease to biomarker-driven subtypes and ultimately toward truly personalized medicine. We discuss why proteins and metabolites—not just DNA—are key to understanding what’s actually happening inside patients in real time. We also dive into how AI is unlocking massive amounts of clinical data and why the biggest bottlenecks today lie in clinical trials, patient identification, and making sense of increasingly complex biological systems.
Recommended
Transcript
Introduction to Healthcare Theory Podcast
00:00:00
Speaker
Welcome to the Healthcare Theory Podcast. I'm your host, Nikhil Reddy, and every week we interview the entrepreneurs and thought leaders behind the future of healthcare care to see what's gone wrong with our system and how we can fix it.
00:00:15
Speaker
Today on today's episode, we're speaking
Dr. Asuka's Career Journey
00:00:17
Speaker
with Dr. Jonathan Asuka, the CEO of SapienBio, which is a biotechnology startup using large-scale proteomics and AI to better understand disease. and And he began his career way back at Stanford during the Human Genome Project, working on a PhD in Computational Chemistry, and then Since then he's spent a really interesting set of years pharma and consulting, including roles like leading R&D at Roche and Celgene.
00:00:41
Speaker
And in this episode we discuss the share from thinking about disease at the organ level to getting into personalized medicine where each patient's disease is biologically unique. And we explore how Sapient is tackling this complexity by first focusing on proteomics, arguing that proteins, not genes or organs, are where the disease actually happens.
Challenges in Biopharma
00:01:00
Speaker
And who does this drug work for, when, and which type of biology does it actually track with? And we get into the biggest barriers today, including working with some of the massive unstructured data sets we see in but and biopharma.
00:01:12
Speaker
Then we get into the bottlenecks and clinical trials. And finally, into how we can identify the right patients and better design studies in healthcare. care Hi, Dr. Yasuka. Thank you so much for coming on to Healthcare care Theory. We're super excited to have you. Yeah, great to be here, Nikhil. Thanks very much for making the time for me. And it's an awesome topic.
00:01:30
Speaker
Of course, yeah. And I want to get into your background a little bit. I mean, as you can see, you have a really interesting career spanning in many different points. But when you entered biopharma and chemistry at Stanford, I mean, it was kind of around the moment when the Human Genome Project hit that working draft milestone. And I assume back then that was a huge deal. I mean, this was like a huge push in biology. And it felt like the first time that biology was computable at scale.
00:01:54
Speaker
And so with that part of the hype cycle, I mean, looking back, What did you think genomics was going to fix in drug development?
Shift in Disease Understanding
00:02:00
Speaker
And what did you realize pretty early on it wasn't going to solve on its own? Yeah, great question, Akhil. I mean, i I was at Stanford Medical School really focusing on computational chemistry. So trying to simulate chemical reactions, simulate the the actions of proteins, for example, um which was a lot of physics, a lot of calculations. We were limited by computational power.
00:02:25
Speaker
ah But while I was at Stanford, um the lid really got ripped off of what was possible in terms of data and especially genomics data.
00:02:37
Speaker
So we started to be able to produce genomics data at an you know a scale that people really hadn't predicted. And there was this new field opening up of it was called bioinformatics.
00:02:49
Speaker
um and really very basic algorithms, the the the coding wasn't that complicated. What was just so exciting about the time was that the there was this availability of data to work on. So so You genomics enabled a lot of things in the pharmaceutical industry, a lot of drug targets that we didn't understand before then became pharmaceutically tractable.
00:03:18
Speaker
And then we also started to understand the differences in patient populations, gave rise to personalized medicine instead of kind of one size fits all therapies. um We started to define indications much more specifically.
00:03:33
Speaker
And there are, you know, a couple different data trends that are behind that. Yeah. And I think that we're going to see it like a thread throughout this episode that we started off with thinking of things in terms of like lung cancer, for example, and then specific subtypes of that. And now we're eventually going to eventually get to the point where every person's form or version of lung cancer is different. So we have far more indications that are personalized. And I think that's going to be interesting. But part of that comes from a deep knowledge of bioinformatics and artificial intelligence. And I think that you've also worked on that quite a bit at CellGina, for example. I know you were
00:04:05
Speaker
working with their team there, working in R&D and also at McKinsey, all across getting phar pharmaceutical analytics and taking this real-world data and bringing insights to that. So i'd love to hear a little bit more about these other opportunities that you've had to work on um after getting your PhD. What were those like? And what were kind of the questions that you were trying to answer, the things you were doing day to day? And um mean what was that kind of goal you're pushing for and that thread that you were kind of throughout all of those experiences?
00:04:32
Speaker
Yeah, I'd love to dig into that with you, Nikhil. You know, it's interesting that you bring up lung cancer because that is such a good example of how data has more tightly defined exactly what the indication is for a therapy or what the, you know, our definition of disease. Because traditionally, lung cancer, you know, there was adenocarcinomas, there was squamous lung cancer, and then there was large cell.
00:05:00
Speaker
And then sort of in the mid 2000s, we got KRAS mutations, EGFR mutations and therapies specific to that. Now, if we fast forward to today, you still have KRAS and EGFR, but then you have tons of other highly specific therapies, you know, Ntrek, Ros1, RET, ALK.
00:05:23
Speaker
So they're just like, literally a dozen different molecular definitions of the disease. And so people, you know, fundamentally, you don't have a cancer of an organ anymore. You have specific biomarker defined types of cancer.
Role of AI in Drug Development
00:05:41
Speaker
Now, I know we wanted to talk about my background, but it really just parallels that development of the personalization of disease, taking something from an organ and some some rudimentary biological understanding of what was going on in the organ, and then going now towards sequencing of aspects of the disease. I mean, in oncology, you're directly sequencing aspects of the the cancer cells themselves.
00:06:09
Speaker
But in rare diseases or in cardiometabolic diseases, there are lots of other applications to omics in the development and the use of the therapies.
00:06:20
Speaker
So, you know, I'd love to talk to you about my background against that backdrop of the personalization of disease. Yeah, I think I would be curious, I guess, at what point in your background or through these experiences, at what point did you realize or notice that we were going from a drug works for this like disease to we know why truly why it works and what patients under what conditions? i feel like that's a much harder barrier to hit, and we haven't really gotten there exactly yet. But how did that transition? I mean, through your career, these different experiences, were we slowly getting towards that point? And what did that look like for you on the ground?
00:06:54
Speaker
Yeah, I mean, I started... a bioinformatics company around 2000, then that got acquired by Roche and Genentech. And so I started to work with and at Roche and Genentech, Genentech really leader in the understanding of disease.
00:07:12
Speaker
Not necessarily at that time did we understand fully why therapeutics worked, but we started to get much deeper into the molecular characterization of the disease.
00:07:22
Speaker
I moved to Celgene, which is now BMS, and there, that that's where I really got exposure to the use of what we call real-world data, or how patients are responding outside of clinical trials to therapies.
00:07:39
Speaker
And you know the very rich world of physicians making treatment choices with therapies that you know where the indication is not exactly matching what they see in their patients, but they want to try that therapy to help their patient.
00:07:58
Speaker
And a lot of the success that Celgene had around 2010 or so was in bringing forward therapies like Revlimid, know, that were first approved for other indications or even other diseases. I mean, we're talking about therapies that are coming from aids and HIV research moving into cancer.
00:08:23
Speaker
And then as we understood how physicians were prescribing the therapy, we were really able to expand the label for Revlimid based on what we were seeing from the real world.
00:08:36
Speaker
And the label expansions then gave us indications of what we should try to understand at a molecular level.
00:08:47
Speaker
So in the case of Revlimid, for example, we were first approved for patients that had very large deletions in their DNA. um We didn't fully understand why, but we could see that from some of the fluorescent imaging technologies. As we got into the genomic sequencing, we could see and identify very specific genes or even single point mutations that would make the drug particularly effective for certain cancer patients.
00:09:17
Speaker
so you know That's an example of where we observe things in the real world, how physicians are using the therapy off-label or beyond the label, That guides where you want to go with a business strategy for your therapy to expand the opportunities to serve more patients and really help more patients.
00:09:35
Speaker
And then ultimately, what further experiments you need to do all in your therapy so you can really understand at a molecular level how it's working in the patients. Yeah, I think that's really interesting. I mean, Revlimid, Linaldomide, think it's like those are just they're interesting because you had a single therapy that could span across different patient contexts and different indications over time, which is something we're seeing more and more of today. But it wasn't necessarily the most common paradigm back then. And I'd love to know, I mean, with...
00:10:06
Speaker
what what What exactly were the major roadblocks going on back then? I know one problem that I've heard a lot about is that when you have so much data and new types new ways of understanding biology, for example, fluorescence, you have so much data, it's hard for people to analyze. And you almost need like artificial intelligence or just a larger staff to even handle all that data. So it's yes, data is always good, but you need to be able ingest all of that also. i mean, what were the major roadblocks that you had seen there that were kind of changing today and in a new world with AI?
00:10:36
Speaker
Yeah, I mean, great question. There have been a number of changes. um Some of them have been regulatory in nature. I think the FDA became, you know, during the decades that we're talking about, like from 2000 to like the last years, let's say from to now,
00:10:52
Speaker
became much more accommodating to the use of real world data, but also came up with new regulatory paths. um You know, they addressed some of the orphan drug challenges.
00:11:05
Speaker
They came up with accelerated approvals, the ability to move um some of the bottlenecks in a clinical trial enrollment.
00:11:17
Speaker
So you're in phase two, you're having trouble getting patients, but you see a lot of efficacy. You identify patient populations where the therapy is particularly good and you move on in this in the span of one trial into phase three and approval.
00:11:32
Speaker
Now that was a completely unknown mechanism that is now really common. The label expansions that we were talking about, it's very common now to get your first approval in something that is easier to secure or a more straightforward pathway, um something like skin cancer, and then move on to other treatment contexts for your therapy and you know try to understand it more.
00:12:01
Speaker
A lot of these are very data-driven though. Like you know that the data is there, you arrange a range of biological story around it and a regulatory package.
00:12:14
Speaker
But as you're saying, a lot of the problem was the data existed, but needed to be either cleaned, combined, analyzed, um or just packaged in a way that actually moved a therapy forward towards the clinic.
AI's Impact on Drug Approvals
00:12:32
Speaker
And how is that changing now? I'm curious because I know with LLMs and transformers, those are like really important architectures, but you also have so many different other types of models that it seems to be um i mean like making a huge difference. And then there's also different types of storing datas like data like Databricks and warehouses and things like that. so there's a lot of new transformations, but which of them have you actually seen?
00:12:53
Speaker
come to be more valuable in drug discovery and drug development and which are you think more applicable to like more commercial size, like helping doctors be clinicians and things like that? Where do you see the divide between what works in drug discovery, which is what's more something involved in the clinic or personal usage, like often like chat GPT, for example?
00:13:13
Speaker
Yeah, I mean, that's ah that's a tremendous question. um First of all, the impact that we've seen has been sustained and is longstanding. So it's not like we're hitting a particular step function with AI.
00:13:30
Speaker
What we've seen, they if you just take new drug approvals, going back, say, I don't know, like we're talking about 20 years. In the mid 2000s, like around 2006 to 2010, you would see 20 to maybe 30 new drug approvals each year. These are new molecular entities. So fundamentally new therapy, not the kind of label expansions we were just talking about. These are these are truly new drugs.
00:13:58
Speaker
Around the middle 2010s, 2014, 2015, 2016, and there, it stepped up and what started to consistently move into 40 new drug approvals.
00:14:11
Speaker
And then really since 2020, we've been consistently above 50. So you clearly see a progression in the purest form of patient impact, which is new drugs being approved.
00:14:26
Speaker
Those have almost doubled or or more than doubled in the last 20 years. I think the question is, what is behind that and what kind of technologies like your you're identifying really are going to continue that trend um and also cut down the time required from these ideas and discovery to clinical trials and actual approvals.
00:14:49
Speaker
um yeah And one of the most fundamental ones has just been the availability of the different data assets. We talked about real-world data, yeah the ability not only to get at electron electronic medical records, but deeper into the physician notes where you get a lot of the patient response therapy, patient response to therapy information.
00:15:14
Speaker
encoded beyond the um sort of structured data an EMR. And that's an area where ai is absolutely perfect to go from unstructured notes to structured use for analytics.
00:15:30
Speaker
And so AI is both used for cleaning up the medical records and putting them in a structured form, but also really just dealing with the unstructured data directly in a way that you know is beyond traditional statistics.
00:15:46
Speaker
Yeah, I think that's interesting, especially because a lot of the data that's most insightful isn't really quantitative necessarily. I mean, it is to an extent, it's oftentimes like clinical notes or patient notes or just notes in the EHR by a clinician that does a lot of things that are deep, but you have to dig through a lot of it to actually find out real insights, which is really difficult.
00:16:06
Speaker
And I'd love to hear a little bit more about where do you see that come into play now? so I mean, it seems like in the drug discovery or it seems like also in understanding the patients through the EHR, it's very important. But in terms of drug discovery and drug development, even in the me market commercial access, like where do you see it coming to play most? I know it's a huge question. and We don't have a huge answer to that yet.
00:16:27
Speaker
um But through your work at SapientBio, what have you seen in terms of AI being used to actually accelerate drug discovery? like Where tangibly does it come into play and getting that 15 drug approvals to 50 and hopefully further beyond the future?
00:16:41
Speaker
Yeah, i mean, some of the really surprising areas um that are so encouraging are outside of oncology. So oncology continues to be an extremely important part of new drug approvals and the success that we've had in treating really difficult unmet medical needs in patients.
00:17:07
Speaker
And it still accounts for like 25, 35% of annual approvals, but that's significantly down from even 10 years ago when therapeutic development was essentially oncology development. It was over 50%, like some years, 60% of new approvals.
00:17:25
Speaker
of new approvals So one of the big stories you see is the rise of other therapy areas, infectious disease, neurology. Rare disease has been an amazing place for putting together disparate data sets, real world data, patient data with you know different types of omics data.
00:17:47
Speaker
But then also cardiometabolic disease, immunology, these are all really complicated biological areas. So, you know, try to get disease understanding and specific disease mechanisms in these diverse therapy areas um that has become tractable now by biotech and pharmaceutical companies.
00:18:08
Speaker
And you don't need to have the way you used to, a very specific and well understood disease hypothesis before you would invest the kind of money that was required for clinical trials.
00:18:23
Speaker
Now, based on you know just the enormous quantities of data in these therapy areas and in specific indications, you're getting a lot more comfort with investing behind therapeutics where the disease biology is not well understood, but the data is clearly indicative of ah of a therapeutic
Proteomics vs. Genomics
00:18:45
Speaker
opportunity. And you see this again and again in immunology and in rare disease and in neurology.
00:18:51
Speaker
Yeah, I think in neuroscience where I have the most time in, you see so many examples of that. like Even something like depression, it sounds simple, but there's so many different like things that we just don't understand yet. We're constantly seeing new therapies like psychedelics and other new treatment paradigms that are a little bit unique.
00:19:08
Speaker
And I'd love to get your idea on a little bit as to if AI is a huge lever pushing this forward. I think a lot of it's coming with the biology too. um dna DNA, of course, was like incredibly informative back then. it's It still is, but it's mostly static data. And I think you've focused on before protein metabolites are dynamic and really reflect what's actually going on right now.
00:19:28
Speaker
And I think a lot of that is... I'd love to really hear a boardroom version of like why these proteins and metabolites are often like a more decisive layer for precision medicine. like Where do they come into play versus genomics? If people don't understand biology, how would you explain what they do and what they can do versus what simple genomics and other paradigms, even though they are important, what they do instead?
00:19:52
Speaker
Yeah, that that's a great question to kill. I mean, most people have heard of what they call the central dogma biology, which is you have DNA.
00:20:03
Speaker
That DNA encodes for specific RNA messengers and those RNA messengers get translated into a new type of molecule called proteins.
00:20:19
Speaker
yeah DNA is very durable. you know it can can It exists for you know literally millions of years, very durable. yeah RNA and proteins are not, and they're always being degraded and replaced in the body.
00:20:34
Speaker
So you get a pretty complicated picture um as you get into the proteins, but the proteins are where the action is. The proteins are what catalyze reactions in the cells.
00:20:48
Speaker
They're almost entirely the targets of our therapeutics to interact with a protein in some way. And most diseases are caused by disruptions in or lack of presence of a particular protein in a particular type of cell.
00:21:09
Speaker
And so fundamentally in therapeutics, you're treating a protein. You're not treating DNA. You're largely not treating RNA. And so the action for therapeutics is around interacting with proteins.
00:21:23
Speaker
um But it gets much more complicated than that. Through the Human Genome Project and the other genomics work that has been done, we know pretty much the number of genes and kind of what they encode for in humans. There are about 20,000 of them.
00:21:41
Speaker
Those will then encode for 20,000 RNA. those RNA will be translated into roughly 20,000 different types of proteins.
00:21:55
Speaker
From here, it gets extremely complicated because the proteins themselves can exist in many different forms they're called you know proteoisoforms, just different forms of the proteins.
00:22:06
Speaker
They can be modified in a lot of different ways after they're translated initially as a protein. So any one um Gene in DNA can code for many different of the same type of protein that have similar functions, but are are really important, the differences in them when you get into non-function the cause of disease.
00:22:36
Speaker
where you have 20,000 genes, you're looking at like hundreds of thousands of different forms of proteins. So you get a combinatorial problem right there.
00:22:47
Speaker
And then those proteins themselves are interacting with a number of smaller molecules like metabolites and lipids and cytokines inside of the cell.
00:23:01
Speaker
um And those can go up into the, we don't even know how many, up into the millions. So there's a significant area of, you know, what we call the dark metabolome or the dark proteome, you know, things that we just fundamentally don't understand and have barely detected now that are really the cause of disease and cause of response to therapy.
00:23:22
Speaker
Yeah. And so would you say, i i think like, There's a few difficulties here. Of course, first of all, it's hard. Not all proteins have like easy binding pockets you can just bind to and create a new therapy. And also, I think we've seen like a lot of different drugs. think 1,600 drugs, but only around 700 or through these type of proteins. i mean What do you think is the main difficulty here? Is it because the interactions are so nuanced and and complex and there's basically...
00:23:54
Speaker
hundreds of thousands of possible combinatorial like ah things that could be going on here? Or is it just like you don't have the data? We don't know exactly which matter for what? What would you say is like the main difficulty? If this is the next step that biology needs to take, what's what's in the way of that? Yeah, the main difficulty is part also the thing that is most exciting about where science is going.
00:24:15
Speaker
You know traditionally we have thought about human biology and especially disease as a as a pathway. You know, you would see a set of proteins that would catalyze reactions and it was very linear, right? You'd have a protein, a reaction arrow, another protein in a reaction and another arrow and you'd have disease or healthy as the output, right? It was extremely linear A to B to C.
00:24:44
Speaker
However, you know, biology doesn't really, it's good for visualizing, good for trying to get, you know, a human understanding of it. But what's really going on is we we're talking about a high dimensional system and we're talking about combinatorial problems.
00:24:59
Speaker
So just to cite some, you know, very quick numbers about what we're talking about. yeah there are trillions, an estimate of about 30 trillion different cells in ah in an individual human person.
00:25:15
Speaker
And then each cell has a trillion molecules floating in it, you know, in the in the cells. So, you know, you're talking about at that point, 30 trillion cells times another trillion molecules per. You're obviously in a combinatorial problem.
00:25:36
Speaker
You, you know, you start to address this only by measuring the different analytes, you know, these different trillion molecules in the cells and try to understand, uh, um, sort of networks, not just pathways.
SapienBio's Proteomic Innovations
00:25:55
Speaker
And as things, that's actually really interesting. I think, of course, if we things get more complex, it's like we will, yeah, there's more potential, but it's much harder to get there. And a lot of what you guys are trying to do at Sapien they're trying to move from this surface level pathway a to B to C to D and just knocking out one of those points. Now we're having...
00:26:14
Speaker
much more complex interactions. And when you're moving from that surface level view to something more fundamental and getting a more complex view, how not just how these proteins interact, but how they interact for each person at a given time in any given state. I mean, what would you say Sapiens is trying to do in this space and and why is it so important? I think whether it's ah a pharma partner or a smart generalist, it's really interesting, but maybe, well, how would you describe that in the simplest terms? Yeah. Yeah. a Great question. So,
00:26:44
Speaker
We have this problem of a trillion human cells and a trillion different molecules in those cells. We need to understand what's going on at the molecular level.
00:26:54
Speaker
So what Sapient does is we are able to characterize thousands of proteins simultaneously from a sample. or thousands of metabolites simultaneously from a sample and really get a much deeper understanding of what is occurring throughout this information network in a diseased cell or a cell that has been treated by a therapeutic.
00:27:20
Speaker
um The problem can be reduced a little bit of those trillion by trillion because there are only about 200 different cell types in a human. As I said, there are about 20,000 proteins and all drugs work on proteins largely.
00:27:37
Speaker
So what we focused on at Sapien is being able to apply proteomics to as many of those different sample types, those different cell types, 200 cell types, or the different types of sample you can get, solid tumor or tissue samples, blood samples, plasma, um different um types of samples that would be important for neurology and be able to characterize those samples using our methods to really understand what levels um and what abundance of molecules are in these different ah in these different disease tissue.
00:28:23
Speaker
And i think, i of course, it's a difficult problem. And I'd love to hear a little bit more from like where you guys are, what angle you guys are starting from, because I know that you're not really just trying to collect more data.
00:28:34
Speaker
One interesting thing that I saw is that you're collecting completely different type of data and doing it longitudinally. You're building ah like a snapshot, um almost like a time series of like human biology where the patient becomes this like data cube of, proteins, metabolites, and all these changes over time. And can you walk through what that actually looks like and looks like in practice and why that's so important? There's the mass spectrometry layer, the AI layer, the longitudinal sampling layer, and how do they all come together and what makes that fundamentally different from how data is typically generated at in Algarve?
00:29:05
Speaker
Yeah, yeah. and And very good characterization of what we do because it is like single snapshots and then putting them together into something that looks more like a video of a patient state evolving.
00:29:18
Speaker
So what we do is we employ mass spec, which is pretty well-established technology. I mean, it's been around for 70-some years. What is different about mass spec now, though, is the throughput and just the scale that it can be applied at. It used to be a very artisanal custom type workflow that was very laborious, and very manual.
00:29:45
Speaker
We've ridden a wave of automation and data analysis, you know lot of largely driven by the instrument makers and some of the technologies that you were citing earlier, enabling the data analysis and the production of the data at scale.
00:30:03
Speaker
So now what we do is we take um single patients, we will observe them for multiple years. And so get samples from multiple years and see how the disease state or the background metabolites in that patient change over time.
00:30:22
Speaker
And so you can't do this kind of network analysis if you only look at one patient in isolation, or if you just look at a group of patients at one time points, we have to watch how they their disease state evolves over time.
00:30:38
Speaker
So we're able to observe the patients from pre-diagnosis, diagnosis, the um start of a treatment course, and then that patient's response to therapy, all at the metabolic and protein level. So looking at thousands of proteins, thousands of metabolites as that patient state evolves.
00:30:59
Speaker
And I'd love to hear from a business model perspective a little bit because that's obviously a super transformational technology. In mass spectrometry, we've had a few guests, even guests that we've had a year and a half ago when we first started. The language around the technology has already changed since then. So I think it'll keep accelerating. Yeah.
00:31:17
Speaker
from a business model perspective you guys aren't trying to be a traditional diagnosis company you're not going to fda approval yourself you're also not really in academia per se and i think what it seems to be is you're partnering with pharma at these specific points in their pipeline especially're on discovery translational work but where do you think sapient generally plugs into the pharma value chain today and like what's the And why is that the right place to start rather than trying to commercialize something yourself? Why was that the decision you made?
00:31:43
Speaker
And how does it kind of fit into the drug discovery and the drug approval process today? Right. Yeah. Great question. I mean, we are... a We work in the pharmaceutical industry supporting pharma and biotech R&D operations.
00:32:01
Speaker
And so while we are discovering and developing therapeutics, we're doing it for our clients um that are you know probably some of the most established names ah in the industry.
00:32:15
Speaker
We specifically help them to understand what is going on with their therapy or what is going on with their patient populations. So we talked earlier in this call about how disease is becoming much more personalized and the therapeutics themselves are interacting with those fragmented, highly personalized disease states in new ways.
00:32:42
Speaker
What Sapient does is we partner with pharma and biotech companies to really understand what is going on with the pharma's patient population or with their therapy.
00:32:55
Speaker
By being able to identify and define and quantify the number of proteins and types of proteins in a sample, especially after the proteins have been modified,
00:33:09
Speaker
that gives the pharma companies a much better understanding for use cases like target identification, the identification of new drug opportunities.
00:33:20
Speaker
the understanding of the disease, the underlying biological mechanism of the disease, and then in particular, the ability to stratify patients well so that drugs are only given to patients that will respond to the therapy in a positive, beneficial, your efficacy way and not put patients at risk in putting patients in trials where they won't benefit or worse, they'll have adverse events.
Future of Personalized Medicine
00:33:48
Speaker
So Sapien really partners with pharma companies to produce this kind of data and then help them interpret it. Because as as you understand, it's ah it's a huge volume of data that pharma companies are not necessarily used to interacting with.
00:34:04
Speaker
Yeah, especially because you guys have your own proprietary data. But one thing I've noticed recently is that a lot of the drug, I wouldn't say failures, but the lack of approvals in the market isn't really just from the biology. It's often like a patient identification problem or trial design problem. and I think a lot of that stems from the fact that you might know the biology of a certain indication or a subset of the indication, but knowing which patients it correlates with is difficult to find.
00:34:30
Speaker
I know an example within neuroscience is that for depression, not everyone has depression because they have reduced serotonin levels. But if you go with some serotonin reuptic inhibitor, which we know works to the patients that... um aren't really affected by serotonin, then your drug won't get approved, even though it could work for many others. So a lot of it's getting the right data to find out who your patient population is, how to build good trials. So it's, yeah, I think it's it's interesting because I think a lot of the scientists are doing great biological biological work, but there's a step beyond that needs to be taken. and
00:35:02
Speaker
As we're kind of seeing biology get more science driven and personalized, and we're eventually going to get this, I think, true personalization, maybe not today, but in ah in a few decades or something where we can really understand personally what does each disease look like. What do you think are are the still the major roadblocks and things you're excited about in the future?
00:35:20
Speaker
as you see biology and sapiens and even this whole industry get more personalized and data driven and eventually it become, i guess, the unit of measure for these diseases gets lower and lower from the organ to the cell to hopefully even lower than that. um What do you see as a major roadblocks of things you're excited about here?
00:35:37
Speaker
Yeah, I mean, Such a good question. but First, let's look at where we've come. In terms of discovery, discovery times from just a biological hypothesis to a ah molecule in the clinic have dropped precipitously. I mean, that that whole...
00:35:55
Speaker
area has had a revolution in terms of acceleration. Where we've seen particular success is in AI designed molecules entering the clinic, especially since 2015. So the clinical trial process, there are three phases, phase one, phase two, and phase three.
00:36:14
Speaker
Phase one is really focused on is the molecule safe and can it be you know does it have indications of efficacy? You know will it be successful in patients? Phase two and phase three really prove that out on larger populations.
00:36:29
Speaker
Traditional phase one success rates are like 40%, maybe 50%. If you look at those molecules that are designed with AI, we've had about 75 of them over the last 10 years, and they have success rates over 80%, pushing 90%.
00:36:49
Speaker
So number one, you're seeing major impact just in the design of the therapeutic. Then you put that together with some of the advances in understanding the biology, which are things that Sapient does. You really understand the disease state and how the therapeutic is going to work in a particular patient population, identify those patient populations through a segmentation.
00:37:13
Speaker
um The next revolution is really going to be in finding those patients for the clinical trial. Probably the biggest bottleneck right now is you have a great therapeutic, you know who it should be given to and who it shouldn't be given to.
00:37:30
Speaker
Then you have to build that into a pretty complicated clinical protocol. That's the design of the clinical trial. You have to build enough flexibility in that protocol that you can see new science and respond to it in real time, which is hard to do in a highly regulated clinical trial you know that has been set up.
00:37:55
Speaker
And then most importantly, you have to find the patients out there that fit all of the criteria for the inclusion exclusion for the protocol. And that becomes a problem, like again, an information theory problem of which patients are showing the diagnostic characteristics that would make them a good patient for the trial, as well as the different disease characteristics, you know, molecular basis of the disease that would make you want to include them or exclude them from the trial um to show the statistical significance and get drug approval.
00:38:33
Speaker
So I
Opportunities for Young Professionals
00:38:34
Speaker
see the big area next in the acceleration of approvals being in clinical development. Yeah, I think what I hope for is like a ah future where you have a certain issue going on with you and then scientists can eventually understand your specifically why does that happen for you? Not as like a simple pathway, but what is the exact interaction going on? But it's a difficult problem. You have the data acquisition side. You need to understand the basic science and biology of how these things work, which are still getting there for a lot of different things.
00:39:04
Speaker
um treatment areas. It's also, as you guys are working on, a data analysis side too. That's a huge piece of it. But I'm hoping AI and just humans are kind of working almost side by side in some way can get us to that point. But it's really exciting. I think examples like SapientBio are really good demonstrations of what are the key steps that need to be taken over the next 10, 20 years to really get there. So I'm super excited to see where you guys go and really I appreciate you coming on today, Dr. Yusaka, for um really sharing what's going on in the industry, but also what you guys are doing day to day. So thank you so much and looking forward to seeing what you guys do in the future.
00:39:38
Speaker
Yeah, that's great, Nikhil. I mean, thanks very much for having me on. And to those people who are considering careers right now and you know looking at the disruption of AI in a lot of traditional careers in banking, um you know across the board, AI is is changing career opportunities for young people.
00:39:59
Speaker
And I think in pharmaceuticals and in science and in disease, like There are so many opportunities opening up um and the fundamental economics of shortening the drug development cycle, having more therapies on market and higher patient benefits that leads to more investment in the space. And
Conclusion and Further Content
00:40:22
Speaker
so i I'm extremely bullish on opportunities for young people working in therapeutics discovery and development.
00:40:33
Speaker
Thanks for listening to The Healthcare Theory. Every Tuesday, expect a new episode on the platform of your choice. You can find us on Spotify, Apple Music, YouTube, any streaming platform you can imagine.
00:40:45
Speaker
We'll also be posting more short-form educational content on Instagram and TikTok. And if you really want to learn more about what's gone wrong with healthcare care and how you can help, check out our blog at thehealthcaretheory.org.
00:40:58
Speaker
Repeat, thehealthcaretheory.org. Again, i appreciate you tuning in and I hope to see you again soon.