From PhD Research to Multi-Billion-Dollar Biotech Startup: Chris Gibson on Recursion’s Origin

Hey, everybody.

Yeah. Jenna joined us at Recursion as an intern-ish, like right around when we were getting the S-1 ready to go public. So it was like at the most tumultuous time probably ever at the company, and she jumped right in and has been an amazing part of the company and now the community, where she's one of the leaders at BioHive.

Sounds good.

Yeah, so I think everybody probably knows that there's a lot of diseases in this world that we don't have treatments for, and we all, I think, probably would agree that we want to fix that. Unfortunately, biology is really, really complex, and so despite all of the efforts of hundreds of thousands of scientists around the world, many here at the university, we still have this huge gap, this huge area of unmet need. And Recursion was founded to try and leverage sophisticated computational tools — like at the time we called it machine learning, and now everybody is talking about AI. It's kind of a big thing right now — leveraging these computational tools to look at the incredible complexity of biology to try and find new medicines that otherwise might be hard for us to discover. So that's the thing we're doing. We've got multiple programs in clinical trials now at the company, and partnerships with big companies like Roche to try and find treatments for the hardest neuroscience diseases, and partnerships with companies like Sanofi to try and find treatments for cancer and for inflammatory diseases. So it's a long way from where we started back in a laboratory right up the hill 13 years ago.

So I joined the lab of a guy named Dean Li, and I think if you've been at the university for a long time, you may remember Dean. Physician-scientist, he had this incredible lab, and what I loved about Dean's lab is that there were all kinds of different scientists there. There were molecular biologists, geneticists, there were physician-scientists, there were engineers. I was a bioengineer, and he brought all these different people together, and I think having all those different perspectives was really, really special. I didn't really specialize in Dean's lab. I was kind of like the fixer. So I ended up being on basically everybody's project in some small way, because they'd say, for example, 'Hey, I'm really interested in studying how shear stress, so like the flow of liquid, affects white blood cells attaching to the inside of a blood vessel. Can you help me build a machine to test that?' And so I was sort of the one that would put that together and help them on the project, and I'd go from project to project. And at one point somebody in the lab was studying a disease, and they were using a microscope to look at images of these human cells, and they were looking at all of these actin stress fibers. So if you've ever looked under a microscope, actin is this — it fluoresces. It's tagged with a green fluorescence, and there's all these — it's like the scaffolding of the cell. It's the 2x4s that hold up the structure of the cell. And they were like, 'There's way more of these in the disease state and way less of these in the non-disease state. Can you help me quantify this?' And as an engineer, I thought, 'Man, I'm too lazy to sit here and count these things because there's thousands of images.' And so I started writing software to help us count the number of these actin stress fibers that existed in one state and to count the number that existed in the healthy state. And that was, I think, where the insight came. A couple of things were obvious then. And again, this is really common now, but at the time not a lot of people were sort of digitizing images of biology. A couple of things were obvious. One, I'm not a very good coder. Two — and like, you couldn't vibe code then. It would be so easy to do this now. So I wasn't a good coder. It was actually a slightly harder problem. Like microscopes, especially academic microscopes, you know, there's a little more light on one side of the slide, a little bit less light on the other. Like, there's some real structural issues. And then the third thing was that, man, there's a lot of information in these images. And as we started quantifying things, what we were realizing is that it wasn't just actin stress fibers. It wasn't just this really obvious phenotype that we all saw by eye, but actually the computer was picking up hundreds of other features that, as you know, a pathologist hadn't even noticed in these slides. And I think that was the aha moment for us, that, man, there is richness in this imaging data. You've all heard that a picture's worth a thousand words, and it turned out that's true not just for all of us in the pictures we take in our lives, but also true at the level of the microscope. And I think that was kind of the founding idea for Recursion, and the project grew out of somebody else's project in the lab that I got lucky enough to get to work on.

Yeah, even back then, too. It was like, why would I sit here and manually count thousands of these little lines in these images? Like, this is a super dumb way to spend time. And so, yeah, I started coding, and after three days, I was like, I am not going to make this happen. And Google did exist then, and a Google search helped us find software written by a woman named Ann Carpenter from the Broad Institute. She's actually now just moving, I think, to Indiana University. So this famous scientist at the Broad — she'd already written this software. It already existed. You can go download it today. It's open source. It's called CellProfiler. And I was like, oh, this is awesome. So I just used her software to actually do all the things that I wasn't able to code the system to do. And now Ann's software and derivatives of Ann's software are part of like every major biotech and pharma company in the world. They're being used across so many different academic labs because this core premise is obvious, right? There's probably clinicians in the room — this idea that images contain a lot of information is really obvious to all of us, right? Osler talked about the way that physicians would examine patients, and just from looking at a patient, in many cases, you could diagnose the disease that they have, right? Somebody with heart failure, they have clubbing of the fingernails. There's all these clues in images to what's happening. Phenotype is an incredibly powerful thing. It's not everything. There's lots of other data you need, but it's incredibly powerful. And what was happening in sort of 2011, 2012, there was this thing called ImageNet. So Google sort of — there's this huge data set that got released. I think Google released it, ImageNet, and it was tens of thousands of images with labels. And we were just on this precipice of this massive revolution in machine learning and AI sort of helping us digitize two forms of data. One was text, and we've all seen how that is turning out — like, it's still turning out now, but I'm not going to say it's going well. It's super interesting and a lot of stuff is happening. And the other was images. And so these two things were happening in parallel, and we got so lucky to have this problem at the moment that all these tools were starting to come to bear. People were using them to like — you know, who's done the little CAPTCHA thing where it's like click on everything with a bus in it or everything with a fire hydrant in it, right? Like, it's not perfect yet because you can still tell a machine apart from a human using images. But it's come a long way, and our ability to — if you've ever just taken a picture of a mole and uploaded it to ChatGPT or Claude and been like, 'Should I go see my dermatologist?' It's come a long way. There's a lot of information in images, and we got lucky enough to be right at the moment where that was something you could do. Like, five years earlier, we just wouldn't have been able to do this.

My dream was to become a cardiothoracic surgeon. So I was doing an MD-PhD here, and Dean's lab was a great one for the research side of that. There were cardiothoracic surgeons in our lab, and I used to carry a pager so that whenever there was a heart transplant at the university, I would get called in to be like an intern assistant with some of the teams that were taking the heart from the patient who was getting the new heart, and we would take the old diseased heart out and cut it up into all these little pieces for research. So I used to actually carry a pager, which is like — I mean, that really dates me. But the idea was, like, I'm going to become a surgeon, but I need to learn how to do research because I want to be kind of on the research side of things. And then this idea hit. And I remember being really torn because it felt like — I think all of us feel this — like you're on this wheel that you're running on, and like, you know, I'd already done undergrad and I was doing an MD-PhD, which feels sort of endless. And then I was going to go do an internship and a residency and a fellowship. And I was like, man, I'm going to be like 150 years old by the time I finish school. And I was like, can I really take the chance to go try and start this company? Because if it doesn't work, I'm going to have to get back on that wheel and start running again, and I'm going to be 156 years old before I finish school. And what I ultimately got to — and it was with a lot of counsel from friends and family — was that the value that I wanted to measure my life by was impact. Like, that's the thing that I want to look back on, whether it's impact as a good father or impact as a good member of a community or impact in terms of new medicines. Impact is the thing that I want to measure my life by. And so I did this super nerdy — and I'll be honest, I actually did this like a net present value of impact calculation, where I was like, okay, so if I'm going to become a surgeon — and it was very back of the envelope. I don't remember the exact numbers, but it was probably like, I'll maybe see 10,000 patients in my life, and I'll probably help like some reasonably large percentage of those patients. So that's good. That's a lot of people helped. If I go start this company — I do remember that I put down a 1% probability of success. But if this company could be successful in the way I truly envision it might be, what if we get to affect like 100 million or 500 million people? And the numbers were really, really clear. Right? 1% probability of success across something that has so much impact felt to me like the better decision to make. And so I ended up going. And the beautiful thing was my wife was, I think, a resident or a fellow still at that time. So we had like a backup doctor just in case it didn't work out for me. So that was good — like some security there. And so I had a lot of privilege to take that risk. But in the end, it wasn't a huge risk. And what I learned is even if Recursion had failed in the early days, even if Recursion fails in the future, like there's so much learning that happens in the process of trying to build something that you're going to be just fine. So if you're thinking — I mean, if you're thinking of starting a company that's a really bad idea, don't go take that risk. But if you're thinking of starting a company and you can't stop thinking about it and you're like staying up at night and you're — as Dean used to say, if you think about it in the shower, that was his litmus test, right? If this is the thing you think about when you're getting ready every day, then it's actually maybe not as risky as it sounds. Like, what's the opportunity cost, right? Maybe you'd go for a year and it doesn't work out and you learn a ton and you come back to what you were doing before. And I think a lot of people, when they think about risk, they think about the risk of doing a thing, but they don't think about the risk of the other choice. Right? Like, what's the risk of how I would feel if I never took that bet? Right? What's the risk if I stayed on that track, became a doctor, had this really, really successful career, but looked back and always wondered what would have happened if I would have done that? Right? Like, those are real risks, too. And so I'm sounding super nerdy and analytical, like building tables here to make career decisions, but that's how I thought about it.

Super nerd.

Yeah, so we heard from Mayor Mendenhall, Troy, just a minute ago about all the incredible resources that exist in the city and in the state today, in the world today. I think there were a lot fewer back then. And in particular, you know, there was not a lot of receptivity to this idea. When I actually told Dean that I was going to go do this, he was like, 'No, don't do that.' And then he was like, 'I have to talk to your parents.' And I was like, 'What? You have to talk to my parents?' What Dean didn't know is that my dad had left Intel when he was on this rise with 14 other employees to start a company. So it was like in my blood, and I did not tell Dean this. I just let Dean figure it out as he was worriedly telling my parents about how his son was going to go take this huge risk, and my dad was like, 'Yeah, that sounds awesome.' So I was like, 'I got Dean.' And as soon as Dean knew that we all knew what risk we were taking, he was 100% in. And the first thing we did — well, the first thing we did is we actually talked to the U, and we're like, 'Hey, we want to go build a company around this.' And I think that was a really hard thing for the university. You had myself, a grad student, and you had Dean, who was a pretty high-profile, you know, vice dean of research and all of that at the time. He was right after that the interim CEO of ARUP, right? This prolific physician-scientist. But I was going to be the CEO, and Dean was going to be like, you know, the Bob Langer, like the kind of part-time CSO. I think that was a hard thing for the U. There had not been a lot of student-run startups out of the university at the time. And so we spent a lot of time with the university getting them comfortable with this idea of a student running the company. So that took a while. And then the very first thing we did when we actually started the company is we sat down. I had been so blessed in Dean's lab to have the opportunity to work really closely with him on writing grants. And when I left his lab, he had five parallel R01s funded at the lab, and I had gotten to help him write — I had gotten to be the primary writer of two of those grants with him and a guy named Kirk Thomas, who was Mario Capecchi's first author on the homologous recombination paper, as like the people who were editing and giving feedback and all of us working together. And I learned so much about writing grants. The very first thing I did when we started Recursion on day one — literally on day one when we sat in our conference room — was started writing an SBIR grant. And we used it — we did a direct-to-phase-two SBIR grant. So we asked for $1.5 million from the government, from the NIH. And my dissertation was the equivalent of our phase-one data. It was like our whatever — I can't remember that section, like the background material. And I was so lucky to have been trained by Dean. And we got really lucky. Funding was good at the time. We got an almost perfect score on that grant. And the big unlock for us was three months — three and four months after we started, we submitted that grant. A couple months later we got our score, and we knew we were going to be funded. It took another year to get the money. But we were able to go take this letter from this group of like 28 really smart scientists that the government trusts to dole out money to all of these investors and be like, 'I know you don't understand what we're doing, but all of these people do, and they're going to give us money, so you should, too.' And that helped us raise another million dollars from private investors. And we basically spent three years with just a handful of people, like four people, working on trying to prove out to ourselves that this idea of using images as a new form of data to be digitized to understand biology was worth doing. And it took us until 2016 to get confident. And as soon as we were confident, we went out and signed like three small pilots with pharma and raised a Series A venture round. And that's when things went from like this really stable but hard work phase with a handful of people to this crazy growth trajectory phase, which is a very different phase of building a company.