Founder Stories: Chris Gibson, Co-Founder and CEO of Recursion Pharmaceuticals

All right, good morning everyone. My name is Andrew Nelson. On behalf of the Phil and Penny Knight Campus for Accelerating Scientific Impact, it's my great pleasure to welcome you to this latest installment of our Entrepreneurship Speaker Series. As I hope you know, the Knight Campus has a stated mission of science advancing society, and we're well aware that one of the ways in which that happens is through startup activity. And it's in that vein that we have started the speaker series to highlight people who have excelled both in science and in thinking about the commercial impact of their science. And honestly, I can't think of someone who better exemplifies that than this morning's guest, Chris Gibson. To introduce Chris, Chris, I actually want to introduce our own Ashley Walker. Ashley is an assistant professor of human physiology here at U of O. She's one of our night campus affiliates, and quite frankly, she's the reason that we were able to attract Chris to join us for this morning's talk. So without further ado, please join me in welcoming Ashley.

Think back to what you were doing 8 years ago. I can tell you what I was doing. I was a postdoctoral fellow at University of Utah. In our lab was collaborating with the lab of Dr. Dean Lee, and in that lab was an MD-PhD student, Chris Gibson. Chris was working on a project to identify and repurpose drugs to treat rare vascular diseases. So Chris is very— pretty good friends with my mentors, and so I would get updates about what Chris was doing in his life. One time I hear, "Chris has gone to Stanford for a summer program at the business school." That seemed a little odd to me as an MD-PhD student. Then the next update is, "Chris isn't going to go back and finish his MD. When he's done with his PhD, he's going to start a biotech company." Now, I'm a pretty risk-adverse person, so that seemed really wild to me. And so while in these last 8 years I followed the pretty traditional academic track, Chris took the big risk and co-founded Recursion. His company has grown to now be publicly traded, worth many billions of dollars, and employing more than 300 people in less than 8 years. So if you want the traditional introduction details about Chris, he actually is from Oregon. He grew up in Lake Oswego. He is back home in our Oregon rain. He went to Rice University for his undergraduate degrees in bioengineering and business. He went to University of Utah for his PhD in bioengineering, and now he is the CEO of Recursion. What those details don't tell you about is Chris's heart and his values. Recursion started as a company seeking to cure rare diseases, and as part of this, they made connecting to the rare disease community in Utah a priority. And I saw this when I would go to their holiday parties and they would invite members of that community. To their holiday parties along with their employees and their friends. They— he's also worked to promote biotech in Utah and support and mentor those setting out to found biotech companies, particularly those from diverse backgrounds. And one last story to kind of show these priorities and values is a story from when I went to the grand opening party for his company. So this took place in a conference room. It was the conference room that they were renting as the office space for everybody in their company at the time. And so they had food laid out on the conference room table, but Chris's computer is also set up on the conference room table that you're eating your food at. And you couldn't help but see that he had his computer screen open. Maybe I'm a snoopy person and I looked at it. But he had his Google Calendar up, and he had every last second of his day scheduled. There was no time in that day really for sleeping, which was a little worrisome. But what he did have time on his schedule for was spending time with his family every evening. And so while he has been really accomplished in creating Recursion over these last 8 years, I just wanted to talk about the fact that he made priorities and kept to his values along the way. So enough sappy stories about me. I'm going to turn it over to Chris to tell you about his journey.

Awesome. Thank you, Ashley. Dr. Walker. So I didn't realize you were going to say all those things. You pretty much gave the whole talk, so I don't need to dive in. But yes, I'm back home. It's fantastic to be back. I was in Portland last night for a couple hours before I drove down this morning, and it brought back a lot of memories. Super creepy, I actually drove to my old high school on the way down here this morning, so there's there's a piece of me that feels like it wants to be back in Oregon one day. And huge thanks to you for the invite. We spent a lot of time in lab meetings together over several years with very different mentors, but mentors who were collaborating and learned a lot and worked on lots of projects together, and it was fantastic, and I'm glad that you found a great spot here. So I looked at the history of the University of, the Knight Campus, of this speaking series, and I was very intimidated. You had some amazing names here. And as Ashley just shared, I'm less than 8 years out of grad school. So rather than talk about all the amazing things, I want to just tell you a story and tell you about some of the things I learned along the way. I'll weave in some science, but I want to talk about a little bit of what it's like to actually create a company as well. And hopefully the learnings will resonate with some of you as maybe you set off or have set off to build something of your own. So this was me much thinner, which tells you something about what it's like to build a company, 8 years ago with my 1-year-old at my dissertation. And this was me doing an awkward fist pump 6 months ago and a day when we went public with our bell-ringing ceremony for the NASDAQ from our headquarters in Salt Lake City. And I'm gonna just tell you the story of how I transited from this place to that, and along the way, what I've learned. So the seed of this story actually begins a little bit before grad school, and it begins when my wife, who Ashley mentioned, is coming back to her home, coming back to do her residency at the University of Utah. The sad thing about that for me was that I was all the way down here in Texas working in a tissue engineering lab doing an MD-PhD. And so I did a lot of this Google search for cheap flights for the first year while my wife was an intern and I was an MD-PhD student in Texas. And ultimately after a year of that, I decided that that probably wasn't the right priority set for me. And so I decided that I had to move to the University of Utah. And it turns out you're not really supposed to transfer during an MD-PhD program. Oh yeah, this is, this is where I was. And so embrace serendipity is the first learning. There's a lot of serendipity through this story, and I think a lot of us try to plan our lives in this incredibly specific way. And I think sometimes if you try to hold yourself to that path,, you miss out on some huge opportunities. And the best example of this for me was I had everything planned out in the lab I was in in Texas, working with Rena Bizios in tissue engineering, and all of a sudden I was gonna transfer, start my PhD over, which was not part of the plans, 'cause I was 3 years in at the time, to be with my wife and to be with my family. And so I flew up one weekend, and I met this guy. This is Dean Lee, who, some foreshadowing, would become my co-founder, and he's now also the president of Merck Research Labs, so the head scientist of Merck. And I walked into his office, and I'd been warned by some people that he's pretty intense, and you can, I see some laughing, pretty intense guy. And we had a 2-hour conversation, and it was incredible, and at the end of that conversation, what Dean basically said was, he was the director of the MD-PhD program, at the time, I should say that. He basically said, you can transfer to the University of Utah MD-PhD program if you join my lab. And I said, but I'm a bioengineer and you're doing molecular biology and genetics. And he said, exactly. And so I went home, I kind of thought about it, I talked to my wife. I was like, I really don't like when these molecular biologists go up to the whiteboard and they write like a protein and then arrow to another protein, and then there's like maybe a feedback loop if we're getting complicated. And like genetics, like I hated that class. I really don't, I mean, I love you, wife, honey, but I don't know if I can do this. But I talked to some people, I talked to some mentors, and everybody said the same thing about Dean, which is that you'll learn a ton. And so I embraced the serendipity, I embraced the moment, and I transferred. And the funny thing about transferring medical schools that I didn't know at the time is you're not allowed in the US to be enrolled in two medical schools. So I had to resign from medical school in Texas with a promise that I would be accepted to medical school again in Utah as I drove up over the weekend. But there was three days where I was not part of medical school and was just going off a promise that they would accept me. And luckily, Dean was true to his word. They accepted me. And the next week I started in his laboratory and it was an amazing, amazing journey. So as Ashley mentioned, Dean studied vascular biology. And in particular, he was trying to understand the endothelium, which is this layer here, which keeps our blood in and other things out. And he really wanted to understand what makes it leaky and what makes it not leaky. He always told us that every disease at the end of the day is related to the vasculature. That's all you have to understand about medicine, Chris. If you understand the vasculature, you understand everything. And it's funny, like the immunologist will tell you the same thing. If you understand immunology, you know everything about medicine. But Dean was very, very focused on vascular stability. And so when I joined his lab, they'd been working for 7 or 8 years to study a disease, a genetic disease called cerebral vascular disease. Vascular malformations or cerebral cavernous malformation that is caused by mutation in one of three genes, CCM1, 2, or 3, that leads to these leaky kind of bulbous vessels in the capillary bed of the brain. And they had done some of the foundational work in this space, including publishing some really big papers. And when I got to his lab, this was the state of the art. This is what we knew. Here's one of those pictures I don't like, by the way. So lots of arrows between proteins. And in particular, Dean had really focused on this bottom part and really, really focused in on the activation of RhoA as a driver of this disease. And so right as I was kind of getting up to speed in the lab, we were building out the animal models to study this disease better. So this was, I think, one of the first papers that I got to to help with, and we built animal models of CCM1, CCM2, and CCM3. And what was really interesting and amazing is that unlike animal models usually, the disease we were able to create in the mice very closely mimicked what you see in humans. So I think this is actually human CCM2 pathology samples, and these are mouse pathology samples from the CCM2 mouse model. And it's very unusual to see basically just a mini version of the human disease. That's pretty rare. Usually mice end up manifesting the disease like in a slightly different place or a slightly different way. So we had all these amazing tools and we wanted to make an impact for patients quickly. So rather than try and start discovering new medicines from scratch, we asked what new medicines might actually intersect or what known medicines might intersect this Rho pathway and allow us to mitigate this disease for patients quickly. And we actually looked at simvastatin, a very, very safe medicine that through HMG-CoA reductase is able to pretty significantly inhibit the activation of Rho. So we did the experiment. We took the animals we'd built. We dosed them with statins. And 5 months later, we read out this. And I remember being in lab. Aubrey was the one that read this out. He was up at the board. We kind of blinded ourselves. And this was the big unveiling in lab meeting. And in mice fed a standard diet, they had a normalized number of lesions to 1 here. In mice fed a statin diet, they had more lesions, trending towards more lesions. This was a massive disappointment. We had published all kinds of papers talking about this hypothesis. We believed deeply that rho activation was the driver of this disease. We knew in these animals from studies we'd done that I'm not showing here, that we were inhibiting rho activation with statin. And we made the animals worse.

Failed.

Oop. Big stamp. One of the things I admire most about Dean from my time training with him, and I don't think it's always true, was that when we saw this failure, he was happy that we had gotten to an answer that we could trust. We had an answer that we believed. And it wasn't the answer we hoped for, but there was positive reinforcement around an answer that we could trust. And that's what he always pushed on in the lab, whether you had positive or negative data, could you, could you get to an answer you could trust? And if you were the grad student that got to an answer you could trust, even if it wasn't one you hoped for, you were rewarded. That was part of the culture of that lab. And I think it's an important element that all of us have to create in academics and in industry. So learning 2, embrace failure. That was a rough day. I felt like I was gonna have to start a new PhD project for the third time 'cause of the one I had started before I transferred. But here's Dean to tell you a little bit about the way he thinks about risk. And I thought it was best if he did it in his own words. We've had some issues with the video, so I'm gonna try the trick of press Escape. Boom. So let's listen to Dean talk a little bit about the science of risk. Also, his brain goes really fast, so he usually doesn't add the last word to every sentence. So just like, imagine.

Everyone is so focused on being the best at what they do, but you have to match that with the concept that really in this day and age you have to see association and links that other people don't see, or even if they see, they don't act on. I need to understand the things that I don't understand. I need a different type of scientist. I need to understand how do I make something that has legs to stand on. You bet on young people, you give them the resources, you take the bet. Some of them are going to fail. We as an institution have to continue to be willing to take that risk. Without taking that risk, we will not change medicine. We won't change patients' lives. We will treat the patient the same way as we did 10 years ago. If I go the other route, which is give it to the most established and only the established, that was the old Asian system and the old German system.

I'm not sure if I'm supposed to say it, we're kicking their butts.

I mean, if I could help matchmake really smart people who have taken a chance to come together and build something that they can't even believe they could build. You tell me who has a good job. You have to interact, you have to collaborate. All we're trying to do is crystallize that innovation, that entrepreneurial spirit that is embedded embedded in the university system. And that requires the same spirit but a different sort of emphasis institutionally. If there's a place that can do it, it's us.

All right, I should have stopped it before the advertisement for the University of Utah. No offense, Matt. I know that that's a little dangerous. I don't know when we're playing this year in football, but the point of that was to understand a little bit about how Dean thinks about science. And that's what I learned working with him was how to embrace these failures and how to look for opportunities at the intersection, at the interface. And what happened about a week and a half later really started Recursion. I was sitting at a seminar in a room a little bit less awesome than this one, but about the same size. You probably remember that. And Stephen McKnight was giving a talk from UT Southwestern about what I call the animal screen that probably sacrificed like a dozen postdoc lifetimes, where they screened, I think, a couple thousand molecules in mice and then sliced up the brains of all these mice and looked for brains— or yeah, looked for brains where you could see more of whatever this thing's staining for, which is like pro-neurogenic neurons or iPSCs or something. I don't remember the study. What I do remember is that this looks like a pretty basic phenotype, right? You have normal brain on the left, brain with a pro-neurogenic compound on the right. And after that talk, I went up to Dean's office and I was joined by Kirk Thomas, who worked in the lab. He actually was the postdoc that worked with Mario Capecchi and was the first author on the paper that won Mario the Nobel Prize, and with Dean. And the three of us were talking. And as we talked about Stephen McKnight's work here, and in some ways kind of like were amazed that you could find these compounds just by slicing up brains and looking at these black dots, something struck us. This was a picture of human endothelial cells. These are healthy ones on the left, and on the right are human endothelial cells where we used siRNA to knock down CCM2, which is one of the genes that when lost causes the disease. They look really different. You don't have to be a pathologist to see the difference here. And as a bioengineer talking to these two great scientists, we decided we were going to try to do an unbiased screen. Rather than start with a hypothesis that Rho was driving the disease, We were just going to add compounds to these cells and ask if any of the compounds rescued the cells so that they looked like the healthy ones. And as a bioengineer, I didn't want to do this in a manual way. I wanted to automate it. So I started coding. And I pretty quickly quit coding because I'm not a very good coder. And I started Googling and found software written by this woman named Anne Carpenter, who's at the the Broad Institute, she had built a piece of software called CellProfiler. And it allowed a biologist like me to basically use a graphical user interface, which I was much more comfortable with, to take examples of cells or worms or whatever you're looking at and drag and drop them into positive and negative controls and then train machine learning classifiers to understand the difference between them. And so this is a screenshot from some of the early work. Where we were generating these huge images using, at the time, a state-of-the-art high-throughput microscope. I will foreshadow again and tell you that Recursion now generates more data in 15 minutes, every 15 minutes, than I generated in my entire PhD, which kind of makes me sad about all the time I spent pipetting. But we basically trained on healthy cells and cells in which we had knocked down CCM2, a set of parameters that help describe the differences. And we did not input these. We let the machine learning classifier pick these differences. And then we did a screen. We took a whole bunch of drugs, known drugs that were on the market that we thought could quickly get to patients. And we added them to the sick cells and asked the machine learning classifier if any of them had rescued. And I also remember this lab meeting very well, because I had spent $200,000 of Dean's money on this project, and these were some of the best results. So from visual analysis, we actually found simvastatin again. So we reconfirmed all of our human biases. From this automated analysis, we found some stuff that didn't look very good, rosmarinic acid. I don't know what that is, but that does not look like healthy stuff. Cells. In fact, of the 40 things that we found, very few of them actually looked like they were having any benefit. One of the ones, actually, this molecule was suggested by your mentor to be added to our library. It was not yet an approved compound. Foreshadowing again, Recursion is now kicking off a Phase 2 trial to study this molecule, Tempol, in patients with this disease. Maybe it looks better. I don't know. It doesn't really look that much better to me. But what we decided was to persist, to embrace the serendipity, to embrace the failure, and do the next experiment. And the next experiment was to ask if you took another functional measure of this disease, which is simply looking at the barrier function of these endothelial cells in a dish using electrical resistance. We knew there was a phenotype there in this disease that if you knocked out this gene or knocked down this gene, These cells got very leaky, and we asked if any of these molecules rescued. And of the 39 molecules that people identified using their eyes in a blinded manner, one of those, simvastatin again, actually made it through that screen. Interestingly, 8 out of the 40 molecules that the automated imaging system picked out actually rescued this cellular monolayer phenotype. That was very compelling. This, I think, is the moment when I showed a version of this data to Dean, where the early kernels of Recursion started in my mind and his. And so we ended up going through an acute animal model. 7 compounds made it. In the acute animal model, 2 of the compounds made it. And then we put them back into that MRI model that we had failed on about a year before. And what was incredible and amazing is that if you look in the top right, when we actually did the same measure that I showed you before on failing, both of these compounds ended up rescuing this disease. This is the first dose. There's been lots of optimization since, but this was a pretty compelling moment where instead of trying to understand the biology, we let the biology tell us where to go. And the biology told us to go in a place we totally didn't expect. Vitamin D, which we published some other papers on later, and we think has a role in stabilizing the endothelium. And this molecule that your mentor had suggested we include, that's now called Rec-994, which was a preclinical model that had never been in clinical trials and seemed to have this really nice effect in this disease. So in that moment, we came up with this idea that you could use images of human cells, not just to do individual screens, but perhaps to build a map of biology. And there's lots of other ways that you could do this. You could look at things like genomics, and we do that now. You can look at proteomics. We're doing that. You could do more specific things like ELISA. But at the end of the day, what we loved about Phenomics was that it represented this kind of holistic representation of cellular state. It wasn't narrow. It was broad. And by just staining cells for 6,000 nucleus, nucleoli, Golgi, ER, mitochondria, actin, and membrane, we believed that we could train computer vision systems, machine learning systems to actually recognize tens of thousands of differences in these images. Just think about it, the size, the shape, the spatial distribution of every one of these organelles in every one of these cells, the textures. I mean, there's so much that you can measure. Now it's not very interpretable, but it's super, super data rich. And it turns out, you can generate one of these images for about 2 to 4 orders of magnitude less money than any other high-dimensional biological dataset, gene expression, proteomics. At Recursion today, we generate about 12 billion images like this every week, and we do it for around 35 cents per image. So it's much, much less expensive than something like genomics. Recursion started growing. Learning 3, stand on the shoulders of giants, but not for too long. So everything built up to that point had been standing on the shoulders of giants, standing on their shoulders to understand the causative genomics of the disease, learning from people like Kirk and Dean. But ultimately, there was a lot of risk in this idea, and we just had to go do it. And so Dean sent me to— man, I'm just feeling really old every time I see these pictures. Dean sent me down to Stanford, as you alluded to, and I got to spend 4 weeks building the idea for a biotech company that would use images at scale to build a map of biology and then would use that map to navigate between diseases and health without any a priori hypotheses. That was the idea for Recursion. We built this amazing, amazing pitch, and as I shared with some of you earlier, We sat down for the final project. We got to pitch 3 real Silicon Valley VCs. All 3 of them fell asleep. I mean, one of the guys was head down on the desk. It was super embarrassing. We were last. Maybe they were just tired after lots of pitches, but we were learning. We were learning how to tell the story of what we were building. And ultimately, what I also learned when I was in Silicon Valley was a lot about the way the technology world felt and the way that they think about the world. And this is Moore's Law. You all might have heard of Aroon's Law, which is Moore's Law backward, which is essentially what happens in the pharma industry, where things get less efficient over time. We were sitting there learning about how to build companies, and there were people talking about automated vehicles, and SpaceX was landing rockets back from space. A lot was happening in that world that was pretty cool. And then here's the drug discovery world, where there's hundreds of thousands of scientists who are incredibly smart, dedicating their careers. And over time, you see the number of new drugs approved per year in gray stays about the same, and you see the cost per new drug approved going up. 40-year trend of this. Why is that? Well, it's my belief that it's because biology is massively complex. And in the face of all that complexity, we build mental models that are whiteboardable and understandable that we can publish in journals that are two-dimensional in many ways and have lots of arrows between proteins, which is like my archenemy. And as humans, we have tons of biases. So many biases. This is like the abbreviated list of human bias. I think most of those biases help us survive as humans, but some of them are really damaging in research, things like confirmation bias. And so ultimately, my belief is that the reason we have 90% of drugs that go into clinical trials failing is because we picked the wrong target in the very beginning of that project. The industry is actually pretty good at finding a molecule that hits the target they want to go after. It just turns out we're not good at understanding the actual relevance of that target in the entire system of biology. And so you can see the failure rates. They're extraordinary. And you can see that this Phase 1 to Phase 2, where you go— first you look at safety, a lot of drugs fail because we didn't understand all the ways they were going to do bad things to us. And then you look at Phase 2 to Phase 3, a lot more drugs— oh, sorry, this is success rate. So still a third of drugs are failing because we didn't understand the ways that they would hurt us. And then two-thirds of them fail after that because we didn't understand perhaps that that target wasn't actually useful in the way we thought it was. And all the while, patients are waiting. And this is a picture from where we had our Christmas parties and our holiday parties, where we had all kinds of patients that not only did we get to know them and their families, they were invited came, as you mentioned, to a lot of our, a lot of our events. And one who always came was Ricky, who you can see here, and we actually have a medicine going into Phase 2 trials for a disease she has now that she inspired us to work on. So Recursion keeps growing. Why did we feel like it was the right time to finish my PhD, drop out of med school, and start this company? It was because we were operating at this convergence. We were seeing not just one technology, but many coming together. And the first of those were tools to control biology. RNAi was challenging, but there were TALENs and zinc fingers, and CRISPR was on the horizon. And we felt like there was going to be the opportunity for the first time ever to actually control biology. Just in the last 2 years, as Recursion has transitioned to CRISPR, we've now knocked out every gene in the human genome, in multiple human cell types using multiple different CRISPR guides. That would have been completely crazy to think about just 7 or 8 years ago. Automation— people don't think about this a lot, but from the auto industry and other places, there's been this trickle-down where we now have robots that can operate, as I mentioned before, at dissertation scale every 15 minutes. Today, we do more than 1.7 million experiments a week at Recursion, 50 weeks a year, thanks to robotics. And what's best is that robots don't have that special wrist flick that someone in your lab has that always makes the data look better than when I did it. They just operate the same way every time. And of course, on the world of technology, declining cost of storage. At Recursion today, we've generated and store 10 petabytes of data. And to give you a sense of that scale, every feature-length film in human history put together in every language in high definition— they make me specify 1080p, that's the lawyers now as you grow a company— that's about 3.5 or 4 petabytes. So we have about 3 times all of the movies in human history put together. That's a lot of images of biology., and now it doesn't cost a lot to store them. There's been a move over the last 10 years from expert systems to modern AI, which we'll talk about in a minute. And of course, there's been incredibly continuous scale in the ability to compute thanks to Moore's Law. And at the intersection of these innovations, we saw the opportunity to begin something. The fourth lesson is that humble beginnings breed excellence and discipline. Some of our competitors are doing an amazing job, but many of them started with very large Series A rounds. We started in a different way. Here's the conference room that you were mentioning. We rented this from the university for the first few years, and I told a story earlier of the first day we were in the conference room. Some folks walked in and said, hey, I think we have the room. And we said, I don't think you do. Don't think you do. And they said, well, how long are you going to be here? And we said, If things go well, probably a couple years. Very confused looks, and apparently somebody in an admin position at the U got yelled at because this was everybody's favorite conference room with the great view. But we were scrappy. We were hungry at the beginning. This is actually our office warming party where we invited all our friends to come and to bring us things that could help us on our journey, like a coffee maker and Cup O' Noodles. And I think this is like paper. But that scrappiness in the early days defined the discipline and rigor with which we did science. We didn't have $100 million to spend. We actually bought all of our lab equipment at a used lab equipment store in San Diego, loaded it in this U-Haul, and then drove home through the night. Actually, Blake and I, my co-founder Blake and I drove back with all this equipment. And then we got everybody out of the office, moved it into the lab, and realized a lot of it didn't work, which was a good early learning. Used lab equipment doesn't always work. But ultimately, that scrappiness in the beginning, I think, taught us a ton that would become really useful later on in the company. So in my second and final video before I finish the talk, I'm going to see if you can get a little glimpse into what the vision of the company was in 2016. And if it doesn't work, then I'll just keep going. There we go. So this is a video made in 2016, so about 5 years ago, that tells you how we were thinking at the time. There are thousands of drugs that have been studied by probably tens of thousands of people. Many of them ended up not being useful in the way they were originally intended. That actually represents decades of work, and in some cases hundreds of millions of dollars that have gone into each one of those drugs. But it doesn't mean that all the work they did was in vain. Rather than just sort of say, "Well, that didn't work," it might actually be easier to take all of that work that's represented in those drugs sitting in freezers and just find a new way to use it. In some sense, we're standing on the shoulders of giants.

When you say that rare diseases or orphan diseases are truly rare, as a group they're not rare at all. We're talking about 10% of our population. The question is, how do you actually begin to view that population population as a population that you can effectively develop medicine for.

I mean, you just walk into Primary Children's every day and you meet hundreds of families that have been just diagnosed with some rare genetic disease and they're just told there's no options. You know, like, we can, we can give you a wheelchair, but can we actually stop it from getting to that point? Where the child might die, and there's nothing. And so this is the type of thing where we can do something, and it's not a small thing. It's that we feel that we can take these drugs, repurpose them, and get them to the patients quickly.

I think we bring this fresh approach, we bring this mindset of exploiting everything that's been coming together and all the convergence of different technologies over the last decade and especially the last several years. In drug discovery and in understanding these diseases. It's a statistical and computational question to try to figure out what uniquely defines that disease in a mathematical way, in a statistical way. Instead of the traditional drug discovery approach where we spend 10 years working on one disease, we have a team who's going to spend 10 years working on thousands of diseases. You can imagine if we model 1,000 diseases in 10 cell types and look at the effect of 5,000 drugs on those. And for every one of those interactions, we're looking at 10,000 or 100,000 cells. And for every one of those cells, we're looking at 1,000 or 2,000 different parameters. That's sort of the level of, of what we're doing. And it's not like what we're going to be doing, that's the kind of stuff we're doing right now. If you combine the best elements of technology with the best elements of science, the ability to answer really complex questions exponentially increases. And that's, I think, a pretty exciting place to play. Awesome. So I mentioned that team that was going to study 1,000 diseases over the next 10 years, and that was like all 8 of us in the video strutting into our closet laboratory. And just a year or two after that video, we were operating at some scale. We had modeled more than 1,000 human diseases in more than 10 different cell types. And we were starting to build these high-dimensional representations of biology that allowed us to essentially ask, could you recognize all the elements of some disease as it manifested in an image or some disease model and find a molecule that would make all of those things go down back to normal? Normal, could you also measure all the other things that would change? Because every medicine has side effects, and we could actually start to measure all of those changes. And we were looking for drugs that had this high-dimensional therapeutic window, and we found a bunch, and we started advancing them forward. And it was really hard work, and working hard matters, especially in, in company building. And fast forward to where we are today. This is a figure we made for the IPO 6 months ago. What we've built is what we call our operating system. And I want to just walk you through the scale of what we've built now. So we have a team of more than 100 software engineers and data scientists who have built tools that allow biologists to operate at genome scale. So you can actually go into our own software today and you can design an experiment that will, for example, look at the effect of every gene in the genome in a cell type. And we've built tools that make sure that those experiments are designed with rules that enable all the data that comes out to be relatable to all of our other data. And that's actually really, really important. So there's specific controls and other elements that we add to make sure that data aggregates over time. We do all the work in our highly automated laboratory, and we turn the lights off to make it look awesome. With all the LEDs. I don't know why. And we generate now not just images like I showed you in the top right, but all kinds of high-dimensional data sets. These are beating cardiomyocytes that we actually ended up scaling into 1536-well plates so that we could look at the effect of molecules on the beating of these. This is in 384-well plates. We're doing now nearly 3,000 exomes a week and scaling that hopefully tenfold in the next few— in the next few months, and then doing proteomics at scale as well. All of these go into our supercomputer. So we now have a computer, this is one of the 100 fastest supercomputers in the world, BioHive 1. It's 40 DGX-100 systems, and it's allowing our machine learning researchers to actually train algorithms across huge swaths of this data. And instead of waiting for those things to train for days, literally, they can actually train in a few hours. So we can iterate and build better models quickly. And all of those data allow us to serve insights about biology to our scientists. We generate hypotheses that we test from only our own data. The fastest way to get a program killed at Recursion is to go read a Nature Science or Cell paper and then to come in and say, I read this paper, I have an idea. That will kill your program. The only way we start programs at Recursion is if we have a unique insight from the data, then we drive. That's the place that we think we can uniquely build. And what this is allowing us to do is to start to scale drug discovery. So on the left, these are the industry standard success rates of a molecule from the start to the clinic. We're only getting to the clinic now, so we don't have strong statistics. And what you see here is that we kill programs early. This is that tech mentality of failing fast. We set a really high bar, we try to eliminate confirmation bias, and we actually try to create a T where we can explore all of biology and chemistry really broadly, narrow in on the right medicine, and then take it all the way to the clinic. We want all that failure early. This is a really foreign idea, I think, to a lot of people in the drug discovery and development industry, but it is a great way, I think, to start to spend less. So we spend about 80% less to take a drug from start to clinic, and we go about twice as fast. And these are leading indicators, and we're doing it at scale. This is our pipeline. We draw our pipeline in a different way, because like we're not going to be constrained by the status quo. So we draw our pipeline in what we call this lava chart, and you can see we've got 4 programs that are in clinical trials now, and 44 more more programs coming behind those across a wide variety of areas. And we're not just doing repurposing like you see here. We're also building our own molecules now. In fact, when we looked across a lot of rare genetic diseases, there were the vast majority of them, there was no drug that was already discovered that we found that could have a really strong effect. So we're building new molecules from scratch. So the last, second to last learning is to operate at the interface. This is the makeup of our team today. Over 350 employees, 40% of them are biologists, chemists, and drug developers, but about 35% are data scientists, software engineers, and automation engineers. That's what we think the biopharma company of the future looks like. But to operate within this context, you actually have to learn to speak different languages. We build project teams where a data scientist or a software engineer can actually kill a project just like a chemist or a biologist. And that's kind of a foreign idea in the traditional way that drug discovery has been done, where folks with these skills tend to be kind of siloed. We're trying to build teams that are built from all of these different folks, but it requires a new kind of scientist, one who can think at the interface of how all their colleagues interpret the data. And I'll give you one example. So that we have— so we finish on time, of what this is now allowing us to do. As I mentioned before, we've knocked out every gene in the human genome in multiple human cell types. We've also added thousands of proteins and soluble factors and studied nearly a million small molecules at multiple doses as well and profiled all of those in what you can think of as just like a high-dimensional space. And we can compare and look at the vectors in that high-dimensional space as kind of representations of biological state and ask how similar or different they are. And so if I just go into oncology here— hard to project the entire genome in PowerPoint, so we're going to narrow this down to just a couple dozen genes. And if you cluster these things by similarity, you see that we can recapitulate a lot of known biology. So for example, here we have the RAS gene family, and here we have the RAS family negative regulators like RAS1 and and NF1. And what you can see is not only do the expected things cluster together, but actually these negative regulators are highly opposite in this high-dimensional space, the things that they regulate. So in this high-dimensional space, you can think of their vectors pointing in opposite directions. That gives you tremendous power. This is a map of biology. And you can look at this map and find really cool stuff. So here's one from a few weeks ago, or a couple months ago. We looked at the PI3K, PI3 kinase gene family, and everything you'd expect to be there— if I were to list all these, it's all the genes you'd expect. But this one here, that's a gene the world hasn't studied very much. It seems to be encoding some protein. In fact, it looks like it's an enzyme that is hanging out, strongly clustering with the PI3 kinase family. I mean, if I was an academic, I would start working on that. That would be pretty cool. You could get a good paper out of that if it holds. We've now done some work and think this is real. This is a real novel member of this functional group. We don't know exactly how it interacts, but this is how we start drug programs at Recursion, is by finding these novel insights that we don't think the world knows. I won't take you through all these. So the final learning is collect and inspire great people and create an environment for their success. I think I told some of you this morning, some of the grad students, some of the faculty I met with, that the science has been hard, building a business has been hard for sure, but by far the hardest thing, much harder than those, was the people stuff. Science has truth and people have truths and trying to figure out How to navigate that, especially as a grad student turned entrepreneur, has been incredibly challenging, by far the hardest thing. And I think we've done a great job of creating an environment that people love. And these are just tiny subsets of our team, a slide I already had ready. But we've created a place where we are bound by values that are really, really strong and where we all share those values., but where people can also really be their authentic selves. And we've created a place where people can do super exciting science and take massive risks and be rewarded not for success or failure in the outcome, but in the way they operate in trying to take that risk. And I'm really proud of what we've built. And ultimately, if this all comes together in the end, we're still early, the early innings of building Recursion, my hope is that we can create an incredible impact on the world. And that's why I'm doing this, and I think why everybody on our team is doing it. So thank you so much for spending the time with me. I hope that story was helpful for some of you, um, and those learnings were helpful as well.

I'll pause there. Chris, that was incredible. Thank you so much. I, I hope you're all as inspired as I am. Uh, we have about 10 minutes minutes left, which is hardly enough time, but I do want to invite questions for Chris. Just logistically, I'm going to play talk show host and run this microphone around to you, and that's because we have several people joining us over the livestream, and that way they'll be able to hear the questions as well. So to kick us off, Bob Goldberg, please.

Hey, Chris, that was really an amazing talk. You showed early on an example where you were going from cell culture to animal models. And so forth as you were weeding out the different targets. My question is, there's been advancement in 3D tissue models also in organoids, and I wondered whether you're finding some utility for those in your drug discovery platform. Yeah, absolutely. They're harder to work with at scale, and I showed you the beating cardiomyocytes as kind of an early work that we've done in that space, and we think that's a really important area for us to grow into. We're also trying to take a similar kind of ethos to even animal models. So we made our first acquisition last year. Last summer we bought a company called Vyom in the Bay Area that actually put cameras in the cages of mice, and then they use machine learning to instead of like do a rotorod test in neuroscience where you time how long it takes for the mouse to fall off the slightly increasing rod, they actually just video the mouse like 24/7, and you can extract from that the way it moves around the cage, the way it runs on the wheel, all kinds of signal that's actually a lot more robust. So yes, we're building at every one of those steps. I would say compared to our competitors, we're probably the least built out on organoids right now, but it's a big area of growth for us, as are co-culture models, iPSC-driven models, etc. But there's so many different places to build. It's an interesting one for us to consider.

Thanks.

Yeah?

I have two questions, one science, one sort of logistical. The science question is simvastatin and your fancy compound— your other discovery compounds all seem to be going down the same path. At some point they must have diverged. At some point simvastatin was a failure and the others were better.

How did you find that divergence? Yeah, so simvastatin worked through the cellular models, and then it stopped working in the final chronic mouse model. So that's the experiment we'd kind of already done.

But you had the courage to go ahead and take compounds that acted like simvastatin and go ahead and push them into the mouse model.

Perhaps.

We also, for example, vitamin D and the compound we ended up taking, REC-994, 4 actually operated via, we think, pretty distinct mechanisms. Vitamin D, we're not totally sure of. We published some papers around this. We could never get to final mechanism, but it was actually maybe integrating the membrane or something and creating some stability in the lipid membrane of cells. We still could never figure that out. REC-994 actually is a superoxide scavenger, and that was kind of like an interesting result. And actually, the lab that Ashley was in did a whole bunch of work and showed that in these cells, superoxide dismutase 2 was very, very low. And so you had a lot of superoxide. And we traced that back up to FOXO1 transcription factor being irregular. And that seemed to be a more important pathway than Rho. In fact, Rho may have been just a branch of the tree that was activated by the increase in superoxide. And perhaps the superoxide was the more important trunk of that pathophysiological tree.

So your talk is like the coolest talk ever. I can hardly stay in my seat. It's really far out. And one of the things that I'd be curious about is if you were talking to the Knight Center, if you were talking to educators and you were saying, these are the things that your students should be prepared to embrace, what would you say?

We talked about this with, with faculty this morning, and I can't remember who it was— somebody from Microsoft had been here a few years ago and I think said a lot of what I believe, which is you all have great smart students here who are going to be able to learn a lot of the things. The hard part is working with their peers, working with other people, working on project teams. If you can build smart, intelligent students, even if they don't know the specific algorithm or the specific assay that we need, we still love those students, right? We want fast learners, but we want fast learners who know how to work together. And so as you think about building as educators, my ask as somebody who hires a lot, and I will end with, I'll end with just our current open positions There's 111 for anybody who's looking. Can't forget to do that. Lots of open slots. We're hiring fast. What we look for and want are people who can work across these cross-functional teams. And not just work like, not just a software engineer who knows how to work nicely with other software engineers, but a software engineer who's going to sit in a room and work with a biologist and actually create an amazing interaction. So the more you can foster that kind of cross-functional interaction, I think the better for not only us, but probably most companies of the future. Excellent.

Other questions?

Over here.

Hi.

I want to echo, it's an amazing talk. I can't pretend I understood a fraction of the science. But the thing I was really inspired by is your embrace of failure and your embrace of how that guide you.

And that—

I don't know if that is why you went from an academic setting to industry, but in academics, like, that is— that can be stifling, right? Because there's so many external pressures, whether they be tenure or graduation rates or funding or all these things related to not just in the sciences but everywhere on campus. Any suggestions for how to get people to embrace that it's okay to fail? Because I encountered so many people that are just afraid of failure and therefore afraid of taking risks?

Yeah, it's a great question, and I wish I had a great answer for it. It feels like a lot of what I've heard already about what's being built here is addressing some of that, where, for example, in the pre-tenure process, there's credit for things like maybe innovation towards startups and that sort of stuff. That's, that's fantastic. I think probably some of the more meta culture has to change, not at any individual university, but in the way the NIH and others fund. What I learned from Dean, I mean, he was just ingrained with like fail fast all the time. He would say, when you publish your Nature paper, imagine what Figure 5 is. Do Figure 5 first. Don't worry about Figures 2, 3, and 4. Do Figure 5. If it doesn't work, go to a new project. And what he then did is when he found something that worked, we would then go write grants for— I mean, he had 5 R01s, I think, when I was there at once. They were basically the same idea that we had found by failing fast, then applied into 5 different institutes in 5 different therapeutic areas. And so I think that idea of failing fast feels foreign in academics, but actually, if you truly embrace it, it might give you a pretty unique path to advance things quickly.

But it's scary.

Thanks.

And I think in your slide where you had your conference room, I think you had an Oregon beer on one of the tables. Oh, I probably did. I think I saw a Blackbeard Porter.

And there was probably some red wine from up the valley as well.

Thank you. Other questions? Thanks.

Great talk. So really interested in personalizing prenatals using genetics? Is your vitamin D research commercialized? What's up with vitamin D and venous malformations? Yeah, great question. So vitamins turn out to be really hard to do trials on. And we had a lot of ethical debates around publishing that paper because our suspicion, which we understand anecdotally is true, is that once that paper came out, people understood how reasonably safe vitamin D was a lot of patients with CCM actually just take it, right? It's like $5 for like a 600-pound box at Costco. So that— we had a lot of debate around that. And ultimately we decided that the safety profile of vitamin D— I think there's one clinical case report of vitamin D overdose, and the person was eating like pounds of it— supported getting the research out there. But it makes it almost impossible to go do the trial because everybody's basically already taking the medicine. So we just now assume that lots of people have good vitamin D. Vitamin D is probably one of the only vitamins that's actually survived the test of time in some of these really big prospective studies of vitamins where most vitamins have kind of— if you get none, they're called vitamins for a reason, right? You do get really sick. But the addition of most vitamins is not actually really panning out in clinical trials. To be beneficial. Vitamin D is the only one that sometimes looks like it actually has a difference, and it's pretty broadly across health. So especially if I were back in Oregon, I would be— I'd be taking my vitamin D. Sunny Utah, it's probably not as big a deal. But, you know, it's hard to study. What we understand, most CCM patients are taking it, and then you get all these anecdotal reports of people being better, and it'd be really easy believe that. There's no way that we can let ourselves kind of fall into that trap. But we are hopeful that REC994 has an additional benefit in those patients.

So time for one more question.

And, uh, thanks, great, great talk. Um, REC994, I think it's called Temple, is what I've known for, for years. Um, I'm just kind of curious, you can buy that at Sigma pretty cheap. So where— not to make this like about money, but at the end of the day, if you're trying to— if the company's trying to solve diseases, that's incredible, it's fantastic. That's the ideal, what I can tell, and that's really clear. But you also have to make money and pay a lot of people. So how you do that on something like Simvastatin or vitamin D or Temple when they're I think most of the money in drugs is making new ones.

Yeah. So two answers to that question. The first one is Congress passed something called, I think, the Orphan Drug Act, which created an incentive for protecting things that you discover if they're discovered in the context of rare diseases. CCM is a rare disease. And so because we have orphan drug designation from the FDA, because we were the first to come up with this unique insight, we get the equivalent of 7 years of patent exclusivity after it gets to market. Now, if it's something like vitamin D you can buy at Costco, that's not going to help. Most patients, not all, but most patients aren't going to buy low purity Tempol from Sigma. We're building it with GMP manufacturing at like 99 point whatever, 999% purity with high QC, and we're going to charge very low amounts for what you'd expect for, for a genetic disease would be my guess. Ultimately, that gets to the second question, which is why do medicines cost so much and how do we lower the price, which is a big one to end on in the last 30 seconds. But the quick answer is 90% of drugs fail in clinical trials. That $2 billion is not how much it costs to take a drug from start to finish. That's how much it costs to take 10 drugs through the clinic and for only one of them to work. And so what I always tell the team is if 80% of our drugs failed in clinical trials, we would be twice as good as the industry, which is an amazing way to think about it. And so what we want to march towards as we build these maps of biology and as we go to organoids and animals and all of this at scale, can we march to 80% failure rate and then 70 and then 60 and maybe in 20 years we can be at 90% success rate? Success rate in the clinic. If we could do that, we could charge way less than everybody else for medicines and make way more money. And I think that would be a great win-win for the world, and that's what we're marching towards.

Thank you. Thanks, Chris. Thank you so much.

Incredible.

Wow.

For all of our guests, uh, both in person and online, I encourage you to go to our website accelerate.uoregon.edu. You can learn more about the night camp there as well as future upcoming events. And please join me in giving Chris Gibson one more round of applause.

Thank you.