Building a Bicycle: Obvious Ventures’ Nan Li on Investing and Platform Biology

This is the AI Health podcast, where we explore the ways in which AI will transform healthcare, biotech, and medicine through conversations with entrepreneurs, investors, and scientists. Hey, I'm your co-host, Pranav Rajpurkar. And I'm Adriel Seporta. And you're listening to the AI Health podcast. Today, we'll be interviewing Nan Lee. Nan is a managing director at Obvious Ventures, where he leads investments in computational biology, AI-driven platforms, and intelligent robotics. Specifically in the computational biology space, he's led investments in companies like Recursion and LabGenius. All these companies are building platforms that aim to use biological data in more sophisticated ways in order to accelerate new discovery. And the use case that we're really going to focus on today is the discovery of new drugs. I thought before we begin our conversation talking about how these tech platforms could help accelerate drug discovery, we could talk about how drug discovery has traditionally happened. And maybe we can focus on a specific drug and walk through the science behind it a little bit. That's a great thought. And so in honor of the Super Bowl, and in being the only country alongside New Zealand, in which drug makers are allowed to advertise prescription drugs directly to consumers, why don't we hone in on the story of a drug that maybe our listeners have seen commercials for, which is Humira. Let's do it. What is Humira? It's an injection under the skin. And originally, it was designed to treat rheumatoid arthritis. It's now been approved to treat a number of different conditions due to inflammation, including Crohn's disease and ulcerative colitis. Cool. And how big is the drug? Humira is the best-selling drug in the world. It brought in $20.7 billion in sales in 2021 alone. Wow. No wonder all the ads. So how was it discovered? So the story starts in 1984-85 when tuber necrosis factor alpha, or TNF alpha, was identified as a key protein that leads to inflammation. And with that discovery, it became clear that inhibiting this protein's action could prevent inflammation. So would TNF alpha then be considered the therapeutic target? Exactly. A drug target is a molecule in the body that is associated with a particular disease process. The drug aims to change the behavior of this target, which is usually a protein. And then in this case, it's trying to inhibit this protein, TNF alpha, by binding to it. Okay. So we now have the target. Great. But now we need some kind of experimental setup in which we can identify molecules that bind to this TNF alpha. Exactly. So we have a few questions here. One, how do we find these candidate molecules, which in our case will also be proteins? And two, once we found these candidate proteins, how can we create a drug with it? To achieve one, the solution in this case is screening. So researchers will screen through millions or even billions of proteins to find good candidates. And in our case for two, the key is to have the gene that encodes for this protein. Once we have that, we can use it to produce and isolate these proteins and then make medicines out of them. Okay. So in the case of Humira, what was the specific process that allowed researchers to do this? So in the case of Humira, the key discovery was phage display in 1985. This is a method that allows researchers to find proteins that bind to this target in a really efficient way and then link it with this associated genotype. Okay. So let's try to understand how phage display works by maybe starting with what a phage is. Phage is short for bacteriophage. It's a virus that infects bacteria. In phage display, the phage expresses some gene as a protein on its coat. So basically, we're making the phage do the work of showing us what proteins different genes make. And once we have that protein displayed on the outside of the phage, we can see how it behaves. Got it. So now we have a lot of phages that are displaying different proteins. Now what do we do? So first, you take your target protein. And again, in our case, that's TNF alpha. And then we put it in a microplate. So basically a plate with a lot of wells. Then you get these phages to display different proteins, which could be in the millions or billions. Put the phages in the microplate and wait for the phages to bind to TNF alpha. And any that don't will be washed away. So I think I get the basic idea. You create a bunch of viruses that display millions if not billions of different proteins. And after this microplate process, you're left with proteins that bind to the TNF alpha and could be potential drug candidates. And through the phages, we also have the genes for these proteins so we can produce more of them. Exactly. And this is how Humira was discovered. In 2002, it received FDA clearance to treat moderate to severe rheumatoid arthritis. And then since then, it's received clearance for a number of other diseases as well. Cool. And so how common is this phage display approach for other drugs? So this process has created eight of the currently top 20 pharmaceutical products worldwide by sales. So it's huge. And definitely it's an important technology for understanding modern drugs. What other methods for developing drugs are out there? So this is an example of target-based discovery. We have a specific target that we've identified and then we want to find candidates that affect that target. The other main approach is phenotype screening. So in phenotype screening, you don't need to have a specific biological mechanism that you're targeting. Instead, you can just look directly at the cells to see if they're responding well to the candidate drug or not. And what are the pros and cons of either approach? Well, one of the main drawbacks of target-based screening is that disease and the human body are pretty complex. So even though a protein can inhibit a target in phage display, for example, there are a lot of factors that can prevent it from behaving as helpfully in the human body. And you need to have hypothesized a target in the first place, which certainly isn't easy. And so the main advantage of phenotypic screening is that you just don't need to start with a specific hypothesis about disease mechanism. But this also has its challenges. So you often have to work backwards to figure out how certain candidates are working. And is one newer than the other here? Phenotypic screening was the norm before some of these technologies like phage display really tipped a lot of R&D agendas towards target-based screening the last few decades. But there's definitely a lot of excitement now for future avenues for phenotypic screening. New biological technologies as well as data sources, tech infrastructure, and machine learning methods have really ramped up the scale and sophistication of phenotypic screening. And these approaches can be used to find new targets too. So there's definitely synergy rather than a clear cut distinction. These are the types of companies that Non specializes in. So I think it's a great segue to turn the questions over to him. Non, thank you so much for being on the show. We really appreciate your joining us. Of course. This is many months in the making, so we're really glad we found the time. So I want to start very high level and talk about tech and life science investing in general. Because you do both, but often investors will sort of focus on one area or the other. And so maybe you can talk a little bit about how they're different, how they're similar, how you think about investing in the space jointly. Yeah, the intersection between those two worlds is an emerging category. It's one that started as, I think, not so much of an overlap. And increasingly, there's this really great greenfield space that's being carved out at the intersection of tech and bio or life science. There are predecessors on both sides. So there's a really mature and thriving life science industry of founders and investors and pharma companies and incumbents. And then on the tech side, there's also a really healthy ecosystem of sources of tech IP and software and processor business models, and also a pretty healthy investment landscape. That's where I come from. So just a quick personal note, dual background in computer science and math for me. And I've been a tech investor for over a decade at this point, primarily focused on frontier tech. And how I came across this new intersection is really from the tech side. Back in 2012, I was looking into the exploding field of genomics and bioinformatics with the success of Illumina. It was very clear cut, at least on the genomic side, that we were about to be drowning in cheaper and higher throughput genomic information. And this might be an interesting treasure trove to apply data science principles. So that's where I got started of just investigating genomics and investigating the increasing trail of bioinformatics information that would be out there and what researcher scientists would do with it. I think that was directionally correct and maybe the founding intersection between these two fields. But fast forward eight, nine years, the intersection is much greater. It's much bigger than genomics alone. To put it simply, there's a combination of the biospace and the tech space where a lot of the modern toolkits to investigate biology and assay biology, sort of run experiments, are getting more robust, higher throughput, cheaper, the sort of following the footsteps of genomics, generating higher resolution and multifactorial data sets from biology. And then that pairs really well to tools that are coming out of the tech sector that interrogate data and make sense of data. Can you talk a little bit more about going from experimentation to data generation? Are there parallels in typical technology companies that have a similar type of business model or a similar focus? I think an analogy that you could make is really the way that internet services companies are run when they are first stood up, which is there is an exercise around data instrumentation, which is if you're building a consumer app, you have to think about the user journey and getting into product management mode. You think about what are the interesting data points that I would want to collect that would tell me about how these users behave and what they want. And there is a real academic science around how to retool internet companies so that they give off a stream of information from all of the users engaging with any platform. You talk to anyone that works at Google or Twitter or Facebook, and they're very familiar with that discipline. The analogy holds when you look at biology, which is to say when we're interrogating biological systems, we have to do that design thinking from the early days to think about how do we instrument biology? What experiments would we want to run? Not only that, but when you lock into the type of experiments or the lens through which you want to interrogate biology, you then want to start thinking about, well, what kind of data does that generate? What are the characteristics of that data? What is the throughput? What are the interrogation cycles? So experimentation cycles? What are the costs? Thinking about biological experimentation from that lens of what's the data trail that it creates is a new discipline. It's a different way of thinking about how to run an asset, for example. But if you have that intention and architecture from day one, the way that a lot of these modern hybrid tech bio, platform bio companies do, then of course, when you have that in mind, they tend to run their experiments in a totally different way than a traditional life science company that doesn't have that broader macro goal or view in mind when they start to run their experiments. Am I right that you focus a lot on platform bio in general and your investments? Definitely. Okay, for platform bio, who are the users for the most part? Do you categorize them in your mind as their researchers, their big pharma? I think the constituents in platform bio or this next wave of biotech companies are very similar. Some things stay the same and there's quite a bit of innovation that we can dive into. But the things that stay the same are the goals, which is to advance medicines and get treatment to patients. And then related to those goals, the constituents that are around the table in terms of scientists, researchers, large pharma partners that could be useful in bringing assets forward in development, clinical partnerships to run clinical trials, regulators. So the ecosystem of players that are involved are very similar. I think what's different is the way that these companies build, which is that implicit in platform bio is an intentionality of what that platform is and architecting the platform. And I think that's altogether a very different way of building a company versus a traditional single asset life science where it's really around commercializing a biological discovery that happened to have occurred. So a lot of traditional life science is really around corralling around these chance events that happen in academic labs or even in research labs of pharma companies where once in a while a really interesting biological discovery occurs or a really interesting molecular discovery on the asset side occurs. And then there's this race of how do we commercialize this? How do we surround this with the right team to go do something with it? Traditional biological discovery is a series of chance lightning in a bottle moments of the Eureka of science is hard to predict. And I think the way that the industry is set up and the way that companies are built mirror that when you shift into a lot of the modern tools that are available and what platform bio companies do is they're trying to industrialize discovery. They're trying to build these platforms that are architected from day one to increase their chance of discovery. They're building the engine first before they get to the discovery with the goal and association that that engine will yield many discoveries, ideally over a pretty broad swath of biology. And that looks a lot more like the tech platform playbook where there's a J curve. There is an initial phase of platform architecture, engineering, and investment. And then the fruits of that labor are back loaded and extracted for a really long time. That looks like the story of Airbnb or Pinterest or Twitter, right? No, that's great that you brought up like Airbnb and Twitter because I was thinking to myself, okay, is this basically SaaS? But no, you're actually really thinking of true platform players in the tech space or like Airbnb and Twitter. Exactly. Yeah, you see the seismic shift where there's still a really healthy ecosystem of chance discovery and biology that's going on and we can expect that to continue and even become more productive with a lot of these tools. But the new opportunity here is for an interdisciplinary team to piece together the right elements of the in-house biological wet lab side and then the in-house data science and compute side and build these highly focused discovery machines that are interrogating biology, building patterns out of really complex data sets that they are generating. And hopefully that translates into more discovery at a more even cadence than what the industry typically sees. We've been talking about platform bio, abuse-to-term assets to our listeners who may not be familiar with these terms. It'd be useful for us to dive into maybe one of the platform bio companies that you've invested in, sort of how you think about what the platform is and what the assets they have are. Sure. I've been really fortunate to be part of this ecosystem from the really early days back in 2012. And I work with a bunch of companies that take this philosophy and approach into different therapeutic areas and different drug types. So everything from small molecule to gene therapy to antibody therapy, there's a whole ecosystem of companies being built up and I am a part of quite a few of them. I was fortunate to be an early investor in Recursion Pharma, now traded publicly as RxRx. I was involved in that company and on the board from Series A onwards to the IPO. And that's a really interesting platform bio story. You know, a company was founded in 2015 with the high level viewpoint that accelerating cell phenotyping with computer vision would yield step function results. So what they do is combine two fairly well-established ideas into a new company and the intersection is altogether a different product, so to speak. So investigating cells in terms of what they look like is a really widely accepted practice in pharma. It's called phenotyping. So the idea here is that a cell is like a tiny person. They have the same genetics that the rest of the organism has and they go through their own life and metabolism. They live and they have biological function, they eat things, they divide. So they're sort of tiny versions of you and you're made up of billions of these cells. So the idea here is that cells, when you look at them under a microscope, they can look healthy or they can look sick, just like you can look sick when you have the flu and then when you get better, your aura changes and your friends will say, oh, you look like you're getting better. The same thing happens to a cell. You know, sick cells look sick and healthy cells do not look sick. So this general idea is to investigate cells that are sick and see if they're changing through the introduction of different drugs. And this is called phenotyping and it's widely established in the industry. What recursion brings to the table is phenotyping happens in the industry with actual scientists looking through a microscope and annotating tags around what the cell looks like, describing its level of sickness, measuring things and sort of subjectively just coming up with what they think about the cell. And they're swapping that out, that manual process out with deep learning driven computer vision. And that's a concept borrowed from peer industries like autonomous vehicles and robotics. And computer vision driven phenotyping is much faster than a person looking through a microscope annotating one cell at a time. So we're talking multiple orders of magnitude faster, maybe a thousand X or 10,000 X more productive. But much more important than throughput is that the resolution of the annotation and how accurate and descriptive the annotations are. When you look through things from the lens of deep learning, aka math versus having a person try to describe what a healthy cell looks like and what features there are, it's incomparable. Computer vision is way more accurate and nuanced in describing an image than a person would be. And this applies to landscapes, it applies to cats, it applies to cells. So what recursion does is they built a platform for auto annotation of hundreds of thousands and millions of cell models that have been exposed to different combinations of diseases and potential solutions, different drugs, to have the computer vision system annotate levels of disease rescue. So there's sort of a disease score and then a disease rescue score. And that serves as the foundation for discovering drug disease hits that may not be apparent through the study of disease or through the knowledge of that particular molecule. But the proof is in the pudding. They can see the evidence through the cell and the cells behavior. And at this point, recursion is a publicly traded company. They went public last year and they have four clinical programs that are all assets they discovered this way through computer vision. And those are now being dosed in patients in live clinical trials. The most advanced one is in phase two. So that's an example of a platform, which is the productization of this idea. And then the assets that the platform can generate and the asset pipeline of recursion is well over 50 at this point. So they found 50 drugs that rescue diseases based on the platform's output. And they are meticulously pushing those drugs through preclinical and human clinical trials to get them to market to get them to patients. So generally, when I think about platform companies outside of the bio space, I think of companies that benefit from having some sort of a first mover advantage. So, you know, because they were sort of first to market, they can get enough adoption and they can sort of take off from there. Is there a similar situation here with recursion though? Is there that same inflection point that you would look for and say like an Uber or I don't know an Airbnb? Yeah, I think that essentially here you're replacing network effects in terms of that sort of sustained competitive advantage or a head start. You're essentially taking out network effects because there's not really a marketplace dynamic going on in these platforms on the bio side, but then you're swapping in data network effects or data modes. So what happens with these platforms is that once you anchor on experimental data that you deem to be important and predictive of treatment development and biological discovery, there is essentially an underlying hypothesis in any of these biological platform companies, right? They believe that some type of experiment is important or they have to generate that data, their data sets in a certain way, that becomes the foundation for their platform. And then as they run the platform, they build this running head start in terms of amassing that data in the right way. And in biological data and this intersection of biology and technology or compute, you know, intentionality of data generation is really important. So this is probably a truism across all of machine learning, but more data is not better, especially biological data because it's complex and sometimes not interoperable. Just measuring data by gigabytes or petabytes is not the right measurement. So I think a lot of these platform bio companies start with a novel contrarian hypothesis of what data is important, what experiments they should stand up to generate this data. If they make investments in lab automation and throughput, they're doing it with a particular intention in mind. So when they run that platform for a year, two years, three years, recursion's case, eight years, running at eight years now, that's a huge head start. It's not even a head start measured by number of patients sampled or gigabytes of information or, you know, a genomic archive, but it's even more nuanced and tailored than that, which is, you know, recursion believes that images from cells should be stained a certain way, collected a certain way. They've run these experiments at high scale. It's a tremendous advantage that they have over either an upstart, trying to compete in their space, or even a pharma company that wants to throw hundreds of millions of dollars in this direction. It's really an insurmountable lead that a platform company has. And I want to ask just on that note, how do you quantify competitive advantage? Just, you know, if you have to think about how much more of an advantage recursion has in 2022, as opposed to in 2019, do you think of a particular sort of metric along which to assess that? I think it would be number of compound screened and the number of diseases that are modeled. So that's a simple way of looking at footprint. So what's the sort of industrial footprint that recursion has that overlays in the pharma space? So, you know, how many drugs have they sampled and collected data on, and how many different disease models have they built? So I think that gives you a sense of their coverage area. You know, recursion started in rare disease, and they have moved since then to inflammation, oncology. They have a fibrosis program that's publicly shared with buyer. So there's an expansion of their biological mode. And then on the dataset side, it's really around, you know, how many experiments have they run in this sort of intentional way. And the focus and intentionality of the data generation is itself a huge head start. I think that's something that doesn't get enough attention, which is that when there is conversation around AI for drug discovery or platform bio, this new space, a lot of the conversation tends to hone in on how do you deploy machine learning into biology. So it's all about how to retrofit computer vision or data science or ML approaches or tactics onto the field of biology. I think a lot of the companies that I work with that have been building these platforms from scratch, they weigh equally the challenge of adopting biology to fit compute. So it's really asking the other question, which is, if you are trying to generate experimental data as the end all be all goal of running biological experiments, how would you design those experiments differently? You know, what falls to lower priority and what becomes higher priority? If the goal isn't to appease a single scientist that has a hypothesis to test out, but how do you industrialize biology so that there's this ever growing data archive generated by experiments? How would those experiments be run? And that allows companies to rethink everything from the design of the assay itself or the systems design of building robotics or lab automation systems to try to get high throughput. It's really effective when you start from scratch and design both sides. So it's not just computational biology, but it's also, you know, new biology, new wet lab design for the sake of compute. It's equally important. Is there anything different in the business model of platform bio companies or it sort of sounds like recursion is just smarter pharma, like they're doing what big pharma wishes they can do. Is that an oversimplification of it? I think that if you look at companies like recursion, they do have multiple assets. They have a partnership model live with two very large pharma companies. They have an internal pipeline. So they do look like a full stack pharma company. Right. They're not a services company and they're not neither are they trying to sell their asset through M&A. So they do look like a next gen pharma company and that's definitely the ambition for Chris and the team. I think in terms of the business models, there are so many business models that companies have now that are available to them in terms of building on your own assets, co-developing it with an external sponsor that might have more resources, signing up for upfront payment driven partnerships that are more R&D and research oriented. So there are a lot of different models that are available, but the most important difference between business models of new bio companies and call it legacy bio companies is really this multi asset point of view. I think once you shift your mentality away from this idea that biological discovery is a chance of that it cannot be predicted and it definitely cannot be repeated, then everything that you do, your behavior is driven by that core belief. If you believe that biological discovery is hard, you wouldn't try to have two of them. It's too hard to space out or to predict. So then you end up with, once there's a discovery, you want to push it to clinic, get some clinical proof points and sell that thing because that's your golden goose. There's not going to be a second one and that explains a lot of the capitalization, company building, fundraising approach that a lot of single asset companies have, which is to raise large rounds, give up a huge amount of ownership in their companies. So typically companies are raising $30 million, giving up 50 to 70% of the company to try to get the asset to phase two. That's the goal. And then hopefully you can sell that to a pharma company because they would know what to do afterwards. The key difference with platform companies is that they invest a tremendous amount of resources in building the platform. And they don't come out with a particular asset in mind. Some companies I work with don't even have a therapeutic area in mind yet. They're just building this platform. And there's a little bit of a steeper J curve. And the belief here is that if you build this right, then many biological discoveries will come afterwards. And there will be many monetization options for that long tail, for that pipeline. The race is the same, right? The destination is the same. You want to advance assets into clinic and then eventually into patients. That's unchanging. That's a goal for anyone in the biotech field is to help patients. But I think the difference is that at the starting line, a single asset company focused on development is just going to run towards developmental milestones. So they get going and they're called it, you know, running in the race. And that a platform biocompany is sitting at the starting line, building the platform and engineering, call it a bicycle, a system for propulsion that takes some time to build. So when you look at that at different snapshots, you have to have the right mindset to understand these companies. Because if you look at it slightly after the starting gun, you'll be left wondering why are platform biocompany sitting at the starting line? What are they doing with all that capital and resources? What the hell is going on? And then it's not until they build the platform and start running it towards different disease areas and start generating their biological insights and outputs that you start to see this takeoff event. And that's exactly, I think, what all biocompany are aspiring to is to have that type of machine that can propel them quickly after it's built up. But the business model difference is that machine takes time and effort and capital to build. So you have to have the right leaders and the right investors who think about things the same way. Because if you're an inpatient investor, you might say, you got to get going. Where's the asset? Where's the clinical milestones? Why are we not engaged with the CRO? And I think with a lot of platform biocompany, they delay that further than a normal company would. It's just a different way of building a company informed by slightly different underlying primitives. That's the key is the baseline assumption of how biological discovery works is different for these new companies. Have you seen any response by Big Pharma itself in terms of investment into this next gen way of discovering these drugs either through building out their own teams or investing in some of these startups? Yeah, this is a very timely topic circa the beginning of 2022, because we're seeing an explosion of interest and activity from the broader pharma industry that, as it always does, is sort of delayed from the activity from the startup ecosystem. So if you look at companies like a Recursion or a Ginkgo or an Accentia, these companies have been around for a few years, putting these models into practice and building out these platforms. But I think recently, there are plenty of pharma companies that are investing efforts into building up in-house data science teams and machine learning teams. If you just look around the executive roster, a lot of pharma companies are adding ahead of machine learning or ahead of data science. A lot of those roles are less than three, four years old overall, but you're starting to see these new sort of hybrid CIO roles coming into the fold. So there's in-house activity to go and get up to speed on this discipline and start baking that into the internal R&D process. So that's happening on the pharma side. And then recently, what's been happening only as of the last few years, really, is a healthy amount of upfront partnership that looks very different than traditional pharma partnerships. Pharma typically acquires assets. When they do engage with startups, it's about buying molecular IP from startups so that they can bring drugs in-house and develop them downstream. With these new platform companies that are building out in different areas, pharma is starting to engage with these companies in R&D partnerships, where they say, if I could have a say in your R&D direction, if you could do some work in an area that I'm interested in, I'm willing to give you a bunch of money up front to help fund the expansion of your platform. And I'm willing to give you really interesting downstream milestones and even co-share royalties on programs that we mutually build. So I think you're going to start to see more of these partnerships that offload discovery to platform companies upfront and then pharma companies get to hone in on where they're best, which is the running of clinical trials and downstream development and all the way to sales and marketing and the entire commercialization engine that it takes to actually get a drug to your pharmacy. That's something that biotechs can't touch. Pharma companies are so far ahead in their ability to go do that. So there's a really nice symbiosis. And you start seeing that Incetro has a deal, Recursion has two deals, one with buyer in fibrosis and one with Genentech Roche. Accentia just announced a deal last week with Sanofi. It goes on and on. So I feel like the really interesting inflection point right now is this idea of the bicycle of biological discovery was laughed at five years ago. There was almost no industry engagement from anyone that knew what they were talking about. Plenty of engagement from naive and wide-eyed VCs. But now you're starting to see almost a deal a week. There are so many interesting partnerships being forged. So I'm definitely seeing a lot more pharma buy-in and a lot more industry transition following in the path of these trailblazing pioneering companies. And it's very rewarding and very cool to see that play out. And just sort of looking into the future, do you imagine the way this will play out as you have all these different biotech companies coming with their own platforms that strike up exclusive deals with these pharmas? Or do you think there is a world in which they're able to say we're doing this disease area with this pharma company, this disease area with this pharma company in order to diversify their risk there? Yeah, that's what you typically see is that the carbots are disease specific. When you look across the top pharma companies, they do all cover most of the disease areas, but they all have a slightly different order of operations in terms of where they think that they're the least covered and where they want external help. So I think so far, if you see all the terms of all the deals, they're typically centered around certain indications or even subindications. And it's also for a limited amount of time in a limited number of assets. So it's typically X economics for up to 10 assets in fibrosis over the next three years. And then there is opportunity for the platform biotech to find a different deal after that time period to re-up and double down with the same partner. So there's a lot of flexibility there. And I think very few companies are open to signing multi-category exclusives. It's not good business practice to do that. And I think the pharma side, they've been understanding of that. And I see mostly indication focus partnerships, if not even more granular than that. I want to go back to talking about sort of the long timeline for these new bio companies, because you mentioned that it requires patients both on the side of the operators, but also on the side of the investors. And I'm sort of curious, like, who has a stomach for this kind of investing? Is it a different character than you would normally see in life science investing? And also, how do you manage fund life cycles based on this? Because I imagine sort of the typical fund cycle is usually shorter than maybe what, you know, you would see a success in. Yeah, you know, I think that I think these companies are going to be around a really long time. They're going to continue adding to their pipelines. I think this is just the beginning for both this industry overall, but even the leading companies are just getting started. I really feel that to your point, this will go on for a really long period of time. As an investor, I think that, you know, we're typically investing on the venture side with 10, 12 year fund life cycles where we're able to, you know, hold equities and nurture a company, even hold on after an IPO. That's typically the timeframe. And in that window of time, it's about enough time to actually get an asset through end to end. If you look at the timeline of drug development, you're typically looking at 10 year end to end between the biological discovery and you see it on a Super Bowl ad. That's typically the timeframe. It happens to align. But more importantly than that, I think these companies inflect value and prove out their hypotheses much sooner. You know, thank goodness. You know, we're not sitting in a black box for 10 years waiting for the drug to come out, but really along the way, there are so many proof points, if you know what to look for. And if you, you know, are a student of these disease areas, you know, whether it's positive controls where there's a rediscovery of a commercial drug, which happens all the time. And we like to use this proof, which is these discovery engines, if they can build a capability of finding a drug that was discovered the gold fashion way, but through happenstance and almost at a monthly cadence. Oh, interesting. That's a very powerful proof point and investors anchor to that. There's a lot more industry engagement now than there was five years ago. So I think partnership revenue, co-discovery deals, there's a lot of commercial progress that can happen. And even if you look at recursion, this is a company that's been around for six years and they have four clinical programs that are being pushed into patients and a backlog of 50 plus preclinical programs that are headed towards IND. That's a proof point within itself, just in terms of the efficiency of that discovery. If you look at how much capital they've had to run with and how many programs they've been able to produce, it's an early indication and sign that this platform is more efficient. The sort of dollars to discovery ratio is way off compared to standard, what you would see in the industry. So there are these clues along the way that investors are anchoring to. And it's a way for us to, I think, validate or invalidate different investment ideas over time. But it's also a way for early investors to exit companies and later growth oriented investors to come in, you know, Accenture went public in the last year, Recursion went public in the last year, Incetro raised a big crossover around, you know. So these are all opportunities for investors with different timelines and internal horizons to self-adjust and say, well, I'm on board for the long-term ride. So I'm going to buy in now and other investors are selling. So there's a lot more liquidity there. And I think the ecosystem is developing quite well where investors can get involved either very early stage and get rewarded for that at the IPO or they can even be a fidelity management or a BlackRock and they buy in at the IPO. And they're thinking about, boy, this platform is going to be amazing 10 years down the road. And that's my mandate. So you see this lineage come together that spans well over 10 years. And that's what these companies need is an ecosystem of investors at different stages that are all vision aligned. And I think that's just starting to happen led by early stage investors buying into this idea, you know, five, 10 years ago. But now there's plenty of companies that are spearheading this in public markets and being held by mutual funds. And that's a good thing. I think it's a good thing for the whole category. And it's a good thing for founders and new upstarts that are encouraged to dive into the space and feel like they're not taking a tremendous amount of capital risk. I want to ask about what are areas or technologies that you have strong conviction will be important over the next decade? That's a meaty question. The first thing I'd say about that is the amount of really interesting development coming out of platform bio is really exciting. And it's sometimes reduced to look what AlphaFold is doing or look what DeepMind is doing on the compute side or, you know, look at a certain type of sequencing. But it's actually a whole family of technologies that happen to be maturing at the same time. So split between biology and technology. So on the bio side, you know, shifting from genomics, which is one way to interrogate biology into multiomics or looking at biology at different levels of abstraction, you know, what's encoded in the genome, how much of that encoding is actually translated into RNA that might turn into protein at the transcriptome level? How much of that actually gets made into protein and and how is that protein interacting with proteomics and protein motion? There's a company that just raised a mega around 500 million called ICON that looks at in vivo protein motion in a cell, which is amazing. You can see proteins interacting with each other. So there's an ever deepening toolkit on the biology side to look at biology through different lenses instead of only honing in on genomics or only honing in on some sort of functional readout. There's a much richer set of instruments to look at biology as a system and to look and revalidate ideas at different layers in that system. That's very powerful and didn't exist 10 years ago. And then on the compute side and the tech side, I think you have just the increasing maturity of enterprise grade AI computer vision, these systems being more turnkey, less custom, easier for a new lab or a new company to deploy. That's just happening overall. You have the improvement of distributed compute and cloud infrastructure supporting that so that the incremental, you know, comp cycle to investigate something is getting cheaper. That's just happening. That's a freebie. That's just a tailwind overall on the internet. So biocomponies get to use that. And then you have lab automation and robotics moving even more of the wet lab work away from manual scientist-driven work that might be bottlenecked in throughput or quality. You're getting a lot more standardization of data generation. We're at this perfect storm where there's a bunch of technology on the wet lab side maturing in parallel, each of them going through Moore's law like exponential improvements in capability and cost. And then the same thing happening on the compute side. So we just sit at the intersection of those ever improving tools. And the magic is downstream of that. I really like that decomposition into the developments in bio and the developments in tech and the infrastructure movement that's supporting that. I want to ask you one final question. And that is if you could share a story with us about a deep tech company that you believe was particularly promising on day one or in its very early days. I'll touch on a rather new company just because it's easy to talk about the companies that are older and have successes. But I think this one gets to your previous question as well, which is, you know, what's an area that might be growing in prominence over the next 10 years? I'm a big believer in the next phase of biology being a return to structural biology. So if you think about the way that we look at biology and disease, there's the sort of raw inputs, which is the biological state of a person before the disease sets in. So that's, you know, genetic predisposition or all the multiomic work. So it's kind of like the inputs into the biological system. What proteins are present? How are they interacting? You can measure some of those things and try to correlate that to actually what happens. And then there's the output, which is you can measure different disease biomarkers or different expression of a disease, but that's after it already happened. The pieces in the middle, you'll realize there are very few tools to actually see what's happening in biology. It's either what happened before or post facto measuring what happens afterwards. That's mostly what people do. And I think that there are a bunch of tools that are being built right now that actually give us a lens to seeing biology happening live as if it were a movie and you could watch it on television. And I think that's very, very special. And one technology and a company that I recently got involved with is called Gandiva Therapeutics. It's up in Vancouver. And this company advances a technology called cryo-electron microscopy. It's a mouthful cryo-EM for short. That is a mouthful. Yeah, it took me many board meetings to get the pronunciation. But cryo-EM is a way of deep freezing a biological sample while it's alive. So the sample is suspended in solution and the activity is frozen mid-movie. And then you put it under an electron microscope to blast it with electrons. And what you can rebuild from that is a single atom resolution image of proteins binding to other proteins, a drug compound hitting a drug target. So these are things that are happening in real time, suspended in solution. So that's really different than crystallography or other types of imaging techniques. Cryo-EM won the Nobel Prize a few years back. And this company is productizing it, making a more turnkey, becoming the Illumina of cryo. And that's one of those things. When you look into it, I know that it will be successful. I know that the ability to actually see what happens, forget about, I think that this drug cures this disease or I think that this target expresses on the surface of a cell, what about looking at it and seeing how the binding happens? And I think that capability and that company is science fiction. It's just mind blowing. And I think the implications of that, to be able to validate and ground truth the predictions of AlphaFold, to be able to add to the toolbox of how we think about diseases and what they mean and what drugs are actually doing, a lot of those questions are answered by structure and structural interaction. And I think companies like Andiva or ICON, which I mentioned earlier, represent the future of this space. And it's one that I'm very optimistic about. We're optimistic too. We can't wait to follow it with you and follow your success over the years. Non, thank you so much for being on the show. Thank you for having me. Well, a big thank you to Non Lee for joining us today. And thank you for listening. We're your host, Prana Van Adriel. And until next time, stay safe and stay healthy. The AI Health podcast is produced and edited by Oishi Banerjee and Mark Robbins. Music by Ethan Aichi. If you like what you just heard, let a friend know. Subscribe to the show and give us a five-star review on Apple Podcasts. Follow us on Spotify or connect with us on Twitter at AI Health podcast.