Chris Gibson, PhD, Recursion’s co-founder and CEO, discusses exclusively with GEN Edge his company’s ambitious plans to develop 100 pipeline candidates in roughly a decade, the savings in money and time from its AI and ML approach, the proposed combination with Exscientia, and workforce growth plans following a reduction last year. [Read Part I here]
GEN Edge: Would positive readings for those up-to-10 candidates be the solid proof that AI is indeed more hope than hype?
Gibson: Of course. Positive clinical data is what all of us are here to do. Regardless of the tools we use, all of us want to get medicines to patients. Not all of our programs will be successful, but if some of them are, that’ll be really good. Ultimately, I don’t think we can really prove to the world that our approach works as well as we believe it may until we have enough readouts to actually start to calculate some statistics.
We’ve always said we wanted to get 10 readouts before we sort of made a judgment of our baseline. And I think we’ll know that in the next 18 months. My hope is that we’re as good as the industry average, if not better from a probability of success standpoint, because we’ve already demonstrated that we’re faster and that we spend less money and that we can scale this approach. The industry bar though is as much as 10% or less. If we can do all those things and meet the industry average probability of success at a baseline, I’ll be really, really thrilled.
Then from there, if we can build on top of that and show in subsequent generations of molecules continuous improvement in the probability of success while going after areas of high unmet need, first-in-disease in many classes, or going after really hard areas of chemistry—as I’ve said before, I’m here to change the way, at least take a swing at changing the way that medicines are discovered. And we’re willing to fail at trying to do that and running this experiment. So we’re excited for the next few years.
GEN Edge: Is that 10% industry success rate not a low bar?
Gibson: In terms of the benefit we bring to patients, yeah, I think everybody wants it to be better than 10%. But to call it a low bar would mean to say that the hundreds of thousands of people across our industry who’ve developed or worked their lives to develop medicines, if that’s the best we can do, maybe the way I would describe it as biology is a high bar as opposed to what we’re doing is a low bar.
The reason that we have a 10% success rate through the clinic on average in our industry is not because people aren’t good at science. It’s because biology is really, really, really complex. And that’s the whole reason why we founded Recursion: To leverage sophisticated computational tools to disentangle these thousands of parameters of biology that no human can hold in their head. Again, I think many people are waiting to see how we do, but our belief is that this approach is sound and that it will just take us time to be able to kind of get the flywheel moving and really start to show people.
What I always tell the team is that if we could fail 80% of the time in the clinic from start to finish, not in any one clinical trial, but from getting into the clinic, getting to market, we could be twice as good as the industry average. And it suggests that over time, as our tools get better and better, there’s not a little bit of improvement to go. We’re going to keep at this until 80% or 90% of our medicines make it all the way through the clinic.
And the question is, will that be 15 years from now or 50 years from now? My hope is it’s closer to 15. But I think these new technology tools in the hands of companies like ours, other companies like us, large pharma companies, are going to dramatically shift that probability of success in the next couple of decades.
GEN Edge: Recursion has quantified the savings of its AI-based approach by comparing its cost vs. the industry’s of drug development to an IND—under $10 million for Recursion vs. about $25 million for industry. Recursion has also cited a 10-month timeframe for validating leads, vs. 30 months for industry. How are those achieved?
Gibson: The industry numbers we get from a paper that is referenced in our corporate deck. It’s one of multiple different publications that people use to quantify the industry averages. I think it’s a reasonably well accepted one.
How do we do it? I think it’s from the compounding efficiencies of putting together all of these tools. It’s from using LLMs to discover high unmet need. It’s from building maps of biology that help us hone where we go. When we want to start a project at Recursion, we don’t have to build a new team. We don’t have to go build new assays. We don’t have to go build new experiments. We can just look at a map where we’ve already looked at millions of molecules and knockouts of every gene in the genome. And we now have data from Tempus [AI] and chemistry predictions all in this giant set of software.
Our team can essentially do about the first year or two of a traditional drug discovery process almost entirely in silico now, using the data and the tools that we’ve built. We still go run the experiment to validate that the predictions we’re making are real. But we’re not spending time and effort to work on a disease until we have some reasonably high confidence that there’s something worth working on.
We can do this for hundreds of programs a year if we so chose. We have that many programs we could be working on right now. We just, in this capital markets environment, are being a little bit cautious about the ways that we spend money.
However, we don’t have enough downstream clinical development know-how and expertise yet to handle all the programs that we might be able to drive forward. So I think that’s how you see these early statistics where we have enough scale to show cost and time to various stages like validated lead and IND. Then, as we start to read out clinical programs, you’ll see us extend those statistics as we get a sufficient N [number of candidates] into the clinic, first at Phase I, then at Phase II and beyond. But you’ll have to give us a couple more years!
GEN Edge: Why did Recursion agree to acquire Exscientia?
Gibson: Well, we’re calling it a business combination. That’s an important clarification.
GEN Edge: The difference from a merger being what? Does Exscientia continue as a division of Recursion?
Gibson: No, we are fully integrating the two businesses. There’s some interesting nuances of U.K. law here, that essentially mergers aren’t allowed under U.K. law with the scheme of arrangement that we’re using. So it’s not a merger, it’s a business combination.
There’s a number of things that I think are important to note here. One is, we think there’s a lot of great tools and technologies and talent in the ‘TechBio’ field. And we see ourselves as a leader and we are willing to make bold bets to help consolidate the field to, as I shared before, put all of these pieces in place across the entire discovery and development pipeline. We have built a lot of great tools at Recursion, but I do not think we suffer as many do from the same degree of not built here that is often in our industry. If we see another team or tool or technology that we think is doing something really well that we’ve not yet built or complements what we’ve built, we are always very aggressive.
This is our fourth acquisition or business combination in just the last few years. So, I think you could imagine that we’ll continue to act as a consolidator when we see these opportunities.
Why Exscientia in particular? Well, we are great at what I would say is first-in-disease opportunities where biological understanding is the key differentiator. We think what Exscientia is great at in the ‘TechBio’ field is best in class, where precision chemistry is the differentiator. They’ve got a number of programs, both with partners and internally, that as we looked at them demonstrated this to us. And we thought putting those two things together gives us best in class biology understanding, and now best in class chemistry understanding all with ML and AI tools. What’s more, they’ve built an automated chemistry synthesis platform, which is part of our roadmap. They’ve already built it, demonstrated to us that it works. We think that’s compelling.
GEN Edge: Did Recursion go to Exscientia or did Exscientia come to you?
Gibson: All of this will be detailed in proxy filing, which will come out in the coming weeks or months. And I can’t say more until that time.
GEN Edge: What about Exscientia’s pipeline does Recursion consider compelling enough to combine?
Gibson: As you look at our differentiated pipelines, our first in class, their best-in-class pipeline, we see them as complimentary. There’s almost no overlap in the areas that we’re working on. It’s a nice kind-of bolt-on, from a pipeline perspective.
There’s an uncorrelated technical risk between the ML and AI that underlie those programs. So, it’s a nice opportunity to make the company even more investable to progressive investors who are looking to this kind of ‘TechBio’ space. If you put our partnerships together, it gives us compelling relationships across some of the largest pharma companies and will allow us to bring better tools to bear across each of our unique collaborations.
So, for example, we look forward to assisting in the Sanofi collaboration as soon as the deal is closed, just like we know how excited we are to use the Exscientia tools to assist in our Roche/Genentech collaboration or our Bayer collaboration as soon as the deal is closed. Putting all those things together and extending runway in the process, we see this as a very bold step that continues to cement Recursion as the clear leader in ‘TechBio’.
We’re here to run the experiment, to ask whether or not there’s a different way to do drug discovery that could improve the probability of success, improve the efficiency of discovery. And we will not stop taking steps that we think advance us on that mission, even if they feel a little bit bold for many.
GEN Edge: You spoke of extending [financial] runway. The combination with Exscientia is said to extend Recursion’s runway to 2027.
Gibson: Yeah, the beginning of 2027, without us taking into account any revenue from new collaborations or licensing molecules. We see that as a conservative estimate.
GEN Edge: What if any role will Andrew Hopkins play in the combined company, in light of his dismissal from Exscientia earlier this year?
Gibson: Andrew, I believe, is working on a number of other startups and is not involved in the going-forward business combination.
GEN Edge: You and Recursion recently confirmed that 10 to 15% of the staff was cut last year. Why did the company reduce its workforce?
Gibson: We are constantly looking at all the roles at recursion, at the performance of people on the team. We make decisions in the regular course of business to make sure we’ve got the very best team, and people are in the roles that are most important for our success. The company is larger today than it was at that time last year. So, I think maybe people are reading in a little bit too much into some focus we had around making sure we had the right roles and making sure that we had the best team in place at the time.
GEN Edge: What is Recursion’s current headcount? And how much is that expected to grow the rest of this year and next?
Gibson: We’re right around 560, something like that. I haven’t looked in the last couple of weeks. Right around 560–570. We expect moderate growth in the coming 12 to 18 months. We’re being pretty cautious. We also have a large number of potential new colleagues joining from Exscientia. So, we’ll be doing the important work of planning for the integration of all of those folks.
