How Recursion Is Using AI to Rethink Drug Discovery

Ed: Hello everyone, we've got Ben Taylor on the show today, Chief Financial Officer and President of Recursion UK, bringing more than two decades of healthcare, finance, and biopharma transaction experience, including senior roles at Accenture, Goldman Sachs, and Barclays. At Recursion, he helps lead the company's financial strategy and UK presence as it advances its technology-enabled drug discovery platform. How are you doing today, Ben? Ben Taylor: Great.

Thanks for having me on. Ed: Where are you calling from? Are you based in the UK or is 'cause I Ben Taylor: I am, yeah, yeah. Well, I technically live in London, although I'm on the road so much, sometimes it feels like this is just where I change the clothes in my suitcase.

So Ed: But it's but the company's based out of ⁓ US. Is that right? Or were founded there? Ben Taylor: So ⁓ we've got a good mix actually.

So ⁓ the Legacy Recursion Headquarters was in Salt Lake City, the Legacy Acentia was in ⁓ the UK. ⁓ and so we have ⁓ large operations in both, and then we also have ⁓ some smaller offices around, including New York, which is where our CEO is. Ed: Okay, okay. And just for everyone who doesn't know the company, I thought we yeah, just give a brief overview of of what Recursion is.

So if you could do that, that'd great. Ben Taylor: Yeah, of course. I'll keep it really tight. Basically, we use AI and lab automation to ⁓ improve the quality and efficiency of making novel medicines and thereby improving the probability of success of really getting that that idea into a final drug that can be used with patients.

Ed: And I think I'm right in saying you're coming across from Acentia ⁓ in a relatively recent merger. What's what's changed at Recursion or how has that influenced ⁓ you know, the company at large? Ben Taylor: Yeah, it was ⁓ really transformational all around, both for us and I think for the industry as well. So Recursion and Acentia were both ⁓ two of the leaders in the AI and drug discovery and development space.

And ⁓ they had started around the same time, almost ⁓ 14 years ago now, ⁓ but really focused on two different areas. So Recursion had focused on the biology, Acentia had focused on the chemistry, and how do you take AI and lab automation to get better solutions than we're traditionally? Possible. And so both companies had developed their own internal pipelines, partnerships, ⁓ as well as a fairly large platform.

And by bringing them together, what we created was the ability not only to find novel targets, but also then create the drugs that ⁓ could be used to hit those targets. And that's a really important part. So not only the idea, but also the translation into a real product. And since then we've actually been able to take the ⁓ expertise.

Both companies had a little different expertise in in some of the AI and the capabilities. So we've been able to bring that together and we built out ⁓ a clinical technology platform as well to really round it out. And we applied that to all of our clinical trials also. Ed: Mm-hmm.

Yeah, I've been excited to have this interview because ⁓ obviously ⁓ everyone knows AI is advancing at an exponential pace and it's really impacting certain areas, but I think we're yet to see what is possible in healthcare yet because you know there's there's a long cycle that it has to go through. It's not immediate like getting access to a chat GPT or whatever. but yeah, so AI is advancing really quickly. If that continues, well I'm assuming yeah, we can assume that it will continue.

⁓ what would the drug discovery process look like in five to ten years based on what you're seeing today? and where does Recursion fit there? Yeah. Ben Taylor: Yeah.

I mean I think it'll be wildly different in a lot of ways because if you think about it, we talk about the industrialization of the sector. Like right now, almost all aspects of healthcare are very artisan. You've got a, you know, a very manual process or a creative process that people are doing to put it together. And all of the different steps are very segregated.

And so what AI is able to help ⁓ bring together is really how do all of those different parts function more effectively? Efficiently, be able to evaluate data in a much better way, and then bring them together so they actually build on top of each other. ⁓ And so that's part of why we've seen such differentiated results. And I know we'll talk some more about that, but just being able to actually drug targets that haven't been drugged before or go into ⁓ new areas of biology.

And so as we go forward five and 10 years, I don't think of this as being like ⁓ the industrial revolution. Where it was ⁓ craftsmen and now you're suddenly on an assembly line. This is really more like the creation of CPUs. You know, we're going from people putting together vacuum tubes to now how things are microfabricated, right?

Like if you think about the level of complexity behind what we do, ⁓ biology is effectively ⁓ this infinite space that we're looking in, all of the possible combinations. Medicinal chemistry is 10 to the 60th. Like these are products. Problems that we can't really comprehend if you're trying to write something down on paper or even with traditional computing.

And so this is where AI is really opening up completely new ways of being able to tackle problems that have been plaguing us for decades. Ed: Then it's it's fascinating to think about what the potential is because there was so much we don't know still about so many things in this area. ⁓ and it feels like there's been a long time where we haven't moved forward dramatically, and this could be a step change in how things happen. Ben Taylor: Yeah.

Well, I'll give you an interesting ⁓ fact. So just to layer in how little we know, if you take the entire pharmaceutical and biotech industry for all of the time that they've operated, all of the hundreds of billions of dollars that have been invested in and all the good people that have been working on it, and you look at all of the drugs that are currently approved or in clinical trials, you only cover about 10, 12% of the entire genome. And so the rest of ⁓ that biology is basically unexplored from a ⁓ therapeutic perspective.

And realistically, if we can't explore it from a therapeutic perspective, we don't actually understand the underlying biology that well. And so this is where we need new tools. We need new ways of both creating data and then analyzing that data to be able to get outside of the very small corners that we've been working in so far. Ed: And then if we talk about things that you can show today that sort of you know tells us that we're on the right path for how we should be looking at things, what's the clearest evidence that Recursion is is actually developing these improvements for humanity?

Ben Taylor: Yeah, no, we we call them green shoots. ⁓ and we've seen a lot of green shoots coming through. Obviously, the goal is to ⁓ get approved medicines over the line and really start treating more patients with them. ⁓ as you said, there's a long cycle time in getting there, but there's a lot of ways that you can actually tell that you're making a difference earlier.

For example, we don't ⁓ you mentioned chat bots earlier. We're not ⁓ doing sort of quick consumer things, these are actual products that come out. So when we make A drug, it's a chemistry that you can take into a lab and you can test its properties. And so, more than a double ⁓ dozen times, we have successfully reached a development candidate milestone, which is basically you're able to take a chemistry, take it into a lab, test it, compare it to everything else that's been done, and see from that laboratory testing it's in a better ⁓ position than what has traditionally been done.

In addition, and most recently, most importantly, we had a drug that reached proof of concept. So that's taken that step beyond not only ⁓ looking and saying, hey, this looks better from a biology or chemistry perspective, but actually putting it into a patient population who has no approved therapy and seeing real impact ⁓ for those patient lives. And that's our 4881 program. That we just put out proof of concept data at the end of last year.

And so that was a biological connection that was made through AI and the platform. Now we see other drugs advancing in. One of the reasons we're really excited is we're on the cusp of a lot of upcoming ⁓ data points around ⁓ the clinical programs. In addition, we've had large partnerships.

So, you know, sometimes internally you can really get excited about your. own stuff, but to get a partner to actually pay you money for it is a statement that they are equally as excited. And so we've brought in over 500 million from our partnerships and we've hit ⁓ not only the upfront payments, but now just in the last ⁓ about two years, seven different milestones with Sinofi and Roche, two of the biggest pharma companies in the world. Five with Sinofi that were ⁓ novel chemistry ⁓ milestones showing that we were able to achieve something that they hadn't been able to internally and the industry hadn't been able to.

And then ⁓ on the Roche side, we were creating novel biology and neuroscience and they paid us two milestones of 30 million each for being able to create these maps of neurobiology. Ed: Mm-hmm. And when people hear AI drug discovery, ⁓ I assume some people might think it's just, you know, asking Chat GPT to to help solve this particular problem, but I'm you know, I we we can all probably safely assume that that's not the case. ⁓ can you tell us more about how is AI specifically used in these different processes that you have to go through?

Ben Taylor: Well, yeah, and this is one of the the reasons why ⁓ healthcare is such a big problem, but ⁓ it's actually ideally suited for AI, but ⁓ is a multi parameter problem that you have to really come at. So if you think about it, right now there's about a ninety-five percent failure rate ⁓ in the ⁓ drug discovery and development industry, which is absurdly high. Those problems aren't all coming from one thing. It can be chemistry, it can be biology, it can be patient selection or clinical trial design.

And so we actually have used AI in different ways across all of those different areas where something could fail. And what we do is we try and look at each program and say, why could this fail? And then do we have a technology or can we create one that'll help us better predict how to keep it from failing? Now, what that means is we use all sorts of different AI.

Sometimes it's big data. ⁓ problems like a lot of our biology problems are big data problems. That's why we've got over 40 petabytes of data that we've created internally in our own labs. And then we we bring in another 25 petabytes that's not even including all of the public ⁓ publicly available data.

So those are big data problems where you're really looking at and trying to figure out correlations or understand ⁓ you know mechanisms in a better way. But sometimes we're doing sparse data problems. And so a lot of chemistry is actually how can you learn fast? Because the data is not going to be in existence, and the number of potential solutions are effectively infinite.

So you really have to go back to principles of generative AI and information theory and active learning and try and say, how am I going to design the right set of ⁓ predictive models paired with experimentation to rapidly learn? my way into the right answer. And so if we're doing that sort of work, we may use a generative AI to create, you know, hundreds of thousands or millions of different virtual molecules, test them in literally thousands of different systems, modeling systems.

And each one of those has their own way of functioning. It could be ⁓ all sorts of different ⁓ AI, or it could be a classical system as well. And then you're scoring, you're selecting, you're making a few and producing it, and then feeding that data back into the model again. And so it really comes down to figuring out what is the right tool that you need to use to answer a particular problem, because there's not one problem that we're trying to solve.

Yeah, I think it's actually also interesting how important data quality is here. and this is this is one thing where we just put out something in Nature Biotech ⁓ that showed our ⁓ there's something called transcriptomics, and you're ⁓ people are including ourselves are are trying to predict biology better using this ⁓ transcriptomics data. And what we showed is we were able to much better predict it. Ed: Yeah.

Ben Taylor: ⁓ even using far smaller data sets because the data quality that we had is so high because everything that we've done since inception has been created for machine learning, created for integration into our system. And so that takes a lot of the noise out of the system. Ed: Yeah, yeah. That's a really inter interesting ⁓ I think Elon Musk said recently he was talking about ⁓ ML modeling and like eighty percent of it was basically before you actually do any of the synthesis and analysis 'cause if there's rubbish coming in it just affects it so badly on the outcome.

Yeah. Ben Taylor: Yeah, it's really true. And that's actually the same point on the agents. So we've absolutely loved all of the development of agents.

And we have ⁓ partnerships with a lot of the major tech companies integrating their sort of agent-based systems or different technologies into our system. ⁓ because what that allows us to do is basically ⁓ workflow optimization. But you have to have the underlying modeling systems, the foundation below that, because otherwise your agents are just going to be using systems that don't work. And you actually lose fidelity by adding an agent to a system that doesn't work versus, you know, each one of those had to be independently validated and then you can integrate it into a bigger system.

Ed: sense and from what you you've experienced today, do you find the most benefit from AI? Is there a clear area that's that's that's stood out so far? Ben Taylor: Well, I think what we're really excited about, I think ⁓ you know Chemistry was one where you could get a lot of early hits. ⁓ And obviously, we've been a leader in that for a long time.

And it's exciting to see that continuing to evolve. The novel biology is the real long-term game changer. Because for the point that I was talking about earlier, you know, almost 90% of biology is effectively unexplored. ⁓ And so being able to ⁓ create and analyze data in a new way.

is how you get to new understanding of disease, new understanding of patients. ⁓ And that's going to be a massive long-term unlock. We're just, we're just, you know, putting our toe in the water of what we could be doing on the biology side. Ed: ⁓ and how do you decide, how does the process work for deciding what drugs you're gonna put some resource towards like ⁓ researching and is that impacted by the partners heavily and what money's available for certain areas, or do you find that AI gives you more advantage with certain types of drugs with certain characteristics and that's what you focus on?

Ben Taylor: Yeah, that's a really great question. I think there's two elements that go into it. One is what do you have good data for? ⁓ so ⁓ You can't take one set of data and hope that it's going to solve all of your problems.

And so I think there's one level of you start producing data. I mean, early on when we did it, it was very inductive. We didn't know what data was important. And so you start to sort of, you know, create a lot of data and then see what sticks.

But over time, what happens is you actually start to learn the language of the data. And then you can get much more ⁓ intelligent and targeted. with what data you create and what questions you ask. And so that's been the exciting part for us recently is being able to say, hey, I want to go after this disease area or this patient population.

And here's the data set that I need to really achieve that. And I don't need to recreate all of the data in this area. I can just do a sampling because now I understand the language and I'm able to extrapolate out some of it. So I think that's been that's been a really ⁓ exciting thing.

to ⁓ see change over the last few years. Ed: Mm-hmm. And is there a ⁓ if you had unlimited funds, w is there a certain area you'd you would sort of attack in terms of a t type of disease or something like this? Maybe it's you know, you know it's really h low chance of getting somewhere and that's why people don't attack it today.

Is that a Ben Taylor: Yeah. Well it's it's ⁓ it's really interesting. ⁓ what we've tried to do it was going back to our original investor mandate. A lot of the people who gave us the funds to really drive this from the beginning said, I you know, they were big ⁓ long-term innovation investors.

So they wanted to see transformational change in the industry. So that's always been a part of our mandate, and that's still what we're focused on. But they had another aspect because we were going towards biotech, they basically said, I want to invest in biotech, but I hate the binary risk models, right? Like biotech is filled with binary risk.

Use that technology, use that efficiency. to create a better business model as well. And so that's really what we've tried to do. And if you look at our own internal pipeline and our partner pipeline, we don't have a single area that all of our ⁓ pipeline is focused on.

There's not a single indication or type of science or you know even type of drug ⁓ that goes through it. And on the partner side, you know, we've been able to diversify into two really large partnerships with Sonofi and Roche. We also have some other smaller ones, but those are the the the real flagships. And ⁓ that gives us more diversification as well.

So with Roche, we're ta tackling one of the biggest problems in in medicine and healthcare, which is neuroscience, right? There have been so few new targets and new drugs in neuroscience. And you look at things like Alzheimer's and Parkinson's and all of these big diseases that we really have not changed treatment of in decades. And we don't have good understandings of the underlying disease as well as how we could therapeutically treat it.

So going after something massive like that with Roche, who is a world leader in the the Area, well, then you flip over to our own internal pipeline, focusing in on areas like oncology and rare diseases, because those are things that we internally ⁓ could tackle on our own and and really ⁓ move forward. And so you see that that diversification of business model as well as ⁓ indication. And just to ⁓ answer your earlier question as well, like when we're deciding ⁓ the targets, part of it has to do with our expertise.

So a Rochar. Sinofi, we're going to work with them very closely to agree to a target. There's going to be not only the synthetic validation of something, but also the experimental validation of something. And this is something I don't think the rest of the industry has grasped quite as much as it it ⁓ it will need to.

Is there's no existence of one without the other. you really have to be marrying them together because otherwise you just don't know how good your model is until you can validate it experimentally. And so you always want to have that paired. ⁓ What we what our partners really bring is that expertise in neuroscience and immunology and oncology and all of those areas that we work on with our partners, versus, you know, we internally again are just focused on those more targeted areas.

Ed: Mm-hmm. And do those partnerships unlock a lot of unique data sets as well that you're able to use? Yeah. Ben Taylor: Yeah, they do, absolutely.

And ⁓ the nice thing is we're actually doing all of the work. So when we agree to a program with Sonofi or we're doing ⁓ work with Roche, it's actually not a back and forth between our labs and coders and their groups. We actually agree to what the goals of the project would be, and then we do all of the work ourselves internally. We can use most all of the data for the improvement of our models in different areas like that.

⁓ It's just areas where you know we we have ⁓ like we're working on a particular target with Sanofi or we're working on neuroscience, as I said, with ⁓ Roche, where we limit it and that data ⁓ then we keep inside of the partnership. Ed: People ⁓ will probably they're probably asking the question, okay, you've got all these other biotech companies out there who might be saying they're using AI to approach things. Fundament and you've talked about the business model being different at Recursion, but could I suppose could you just bring up that comparison and how Recursion is is different to these other biotech companies that might talk about using AI, but I assume not in the same way.

Ben Taylor: Yeah. Yeah, well, and look, everyone should be using AI. It's just a better tool. It would be like saying, you know, back in the 1980s, ⁓ you know, we're we're gonna use a computer, right?

Like it's just gonna be a better tool than whatever you were using before then. ⁓ and so everyone should be using AI. How you use AI, the depth that you're able to use it, how you're able to problem solve all of it. That ⁓ is a very wide variance.

And so what most of like if you look at the large pharma surveys or what they're really using it for, most of it is workflows or early research, right? Like sort of ⁓ ideation. there some of them are starting to work on more fundamental systems behind drug discovery and development, but really, you ⁓ I think that'll be a progression. There's just a lot of infrastructure.

and it's hard to go in when you've got, you know, a ⁓ 2000-person ⁓ Clonops group to say, Hey, I want you all to work in this different way. Versus when we come at a problem, because we're AI native, we just say What is the AI way that we could try and tackle and solve this problem? ⁓ And it's not always there, but a lot of times it is. And so I think that makes a big difference.

⁓ The other part is for the smaller companies, ⁓ a lot of times you'll see ⁓ more of a single point solution where they're focused on a particular area, ⁓ which honestly is great. We want to see more people doing it. Our competitors are not other people using AI, it's the traditional way. Way of doing drug discovery and development, which is where almost all of the resources in the industry go right now, right?

Like we want we want more and more people to adopt this. So I think when we think about our differentiation, what we have is obviously the data that we talked about before. Second, I'd say is really understanding how the models are validated and then being able to integrate those models together. So another ⁓ recent ⁓ piece that we put out was on predictive biology, and it was basically ⁓ The test was okay, you've got this, you're going to train your models.

Try and predict how these cells will react, these cells that are not in your trading data and not associated with it, to you know, different biological perturbations. And so we created a system that utilized a mixture, it was basically a multimodal ⁓ AI driven system to predict that. And we were able to do it about 75% of the time. Now that is a very targeted thing.

This is this is a sector that is very ⁓ hypey or can be. So we wanna we wanna keep it under wraps, but that's really exciting because once you can start to predict the underpinnings of biology, then you can start to actually take away some of that experimental time and cost, but also explore new novel biology. So yeah, those are those are sort of the areas that we focus on. Ed: Yeah.

Yeah, it's it's interesting to hear you talk about that because I've definitely seen not in biotech, this it seems to be a general thing, there's a there's two types of companies that are ⁓ appearing in the AI age. ⁓ the first one, traditional more more traditional companies that are using AI as much as they can in processes to improve things slightly in comparison to other companies that have said, right, let's rewrite how the company works because things have changed like things have changed.

The old model is is not the same anymore. And you know, it needs to be built with AI from the beginning up, fundamental principles. And then you're actually able to unlock, you know, a huge change in in the the way the company operates. ⁓ and I suppose ⁓ Ben Taylor: Yeah.

Absolutely. Ed: believe you're looking at peptides as one of your sort of areas ⁓ at the moment. And I was just gonna what's your feedback on that sort of area? Ben Taylor: So yeah, yeah.

Right now we are small molecule focused, but the same ⁓ peptides is a logical expansion ⁓ to what we do. So basically, most of the design systems that you would use are effectively the same. And obviously, peptides are a huge area of focus for the world right now with ⁓ GLP1s. We've also historically done some work in doing large molecules and biologics and different ⁓ aspects like that.

What we wanted to do was really ⁓ take the small molecule aspect as far as ⁓ we could and then start to expand out into the other areas. And so it's ⁓ most of our platform is actually ⁓ molecule agnostic. So if you look at the biology aspect or clinical aspect, they can use peptides, large molecules, small molecules, we can do discovery in all of those different areas. It's really just when you're designing the molecule itself, you need a more specialized focus in it.

And so we've done a lot of ⁓ what you might call like a hybrid small molecule, which would be like a degrader or something along those lines. And certainly peptides is something that we we talk about a lot internally and you may see from this in future. Ed: And if ⁓ which Recursion program today are you most excited about over the next twelve to eighteen months? Ben Taylor: Yeah, so most of most of the world is focused on 4881, ⁓ which is ⁓ because it's furthest along ⁓ in the clinic.

That's the one where we announced the ⁓ proof of concept ⁓ phase two data ⁓ late last year. really ⁓ Hard situation for a lot of the patients. It's a chronic disease that there's no current medical therapies, lifelong ⁓ surgeries for the patients. They basically have to be having either minor or major surgeries every year throughout their entire life.

And so we were able to show a significant reduction in something called polyp burden, which is basically like ⁓ if ⁓ a normal person goes in and gets a ⁓ colon exam, they may have one or two polyps in there and they sort of remove them. For these patients, they literally get hundreds or even thousands of polyps all throughout their ⁓ digestive tract. ⁓ and ⁓ all of them are cancerous. There are not any benign ones.

And so if if left to their own, they'll progress to colorectal cancer and and other cancers. And so Really exciting to be able to see a real clinical benefit coming out of the drug. Again, this was a biological connection that the platform was able to identify, and then we were able to move it forward into clinical trials. We're currently discussing with the FDA the pivotal study for that.

So that would be the last study before you take it and try and get it approved with the FDA and other regulators. And we'll give another update on that later on this year. Ed: How so that's had a positive phase two signal. How does that work in terms of going from that stage to an improved drug?

Ben Taylor: Yeah. Yeah, so ⁓ basically what we have to do is go to the FDA and talk about the trial design. So the last trial before you ⁓ submit it for approval, you just need to make sure that you align with the FDA on, hey, this is what we're gonna measure, this is the patient population, this is the statistical powering that we're gonna do for that trial. ⁓ and that's what we're currently in discussions with.

As soon as we're done on that, we can just start ⁓ moving ahead, dosing patients. And depending on how many patients we have to dose and what the endpoints are, that determines the time that we'd have to enroll patients. But basically, what you're looking to do is show a statistical differentiation between a control group and the patient group that's treated. Obviously, the bigger the effect and the more the patients that have that big effect, the faster you get to that statistical differentiation.

And so that's what we're in the midst of ⁓ talking about with the FDA. Ed: And so so the time frames for ⁓ getting a drug approved are are different based on how long some of these tests take to get the statistic statistically relevant results and things like this. Yeah. Ben Taylor: Yeah, yeah, exactly.

Well, and there's a lot of different steps in there. That's also why so large pharma has traditionally been focused more on the commercial aspects of ⁓ the drugs. They they do obviously do discovery as well, but if you look at where they invest most of their resources, it's in the commercial aspects. ⁓ whereas most of the biotech industry is focused on the sort of the earlier phases of development.

⁓ And often there is a transition between the two. I think ⁓ if we look at other companies that have grown from being you know smaller biotechs into real industry-leading companies, you know, your L Nylams, your Regenerons, ⁓ what you do see is a solid platform with a good partnership business, one or two ⁓ lead programs that sort of drive it through and then building it out beyond that. I think you'll always see us try and keep ⁓ at least one or two of our own drugs that we're moving forward and really keeping the full economics on.

⁓ but there's also a time period where you look at different programs and say, you know what, it's it's probably a good value transition for our resources, our skill set, and everything else to to partner that out. ⁓ and that's obviously. ⁓ the step up in valuation post ⁓ phase two is very high and post phase three is even higher. ⁓ but yeah, that's that's the general model.

Ed: Saying ⁓ I don't know if it's happened yet today, or it I I assume it will happen in the near future, is that this approach to ⁓ discovering new drugs for things ⁓ comes at a lower cost. So you can take some things that people may not have even explored before because the cost is too high for the benefit, ⁓ because in the end, you know, a lot of these things are businesses, they wouldn't have gone in. You know, that's mighty w we haven't had so many breakthroughs and things like hunting things, disease and things like this.

And it gives us more time to do that. Yeah. Ben Taylor: Yeah. No, you're you're absolutely right.

And if you step back and think about it from a ⁓ a risk equation perspective, so ⁓ we were talking earlier about like chatbots and a lot of the areas in tech where you've got maybe from inception to commercialization, you know, a two-year, a three-year time horizon versus ⁓ currently in the biotech industry, that's like 10 to 15 years. But let's ⁓ hopefully we are are helping to bring that down considerably. And I can talk a little bit about how we've been able to. But if you look at it from that risk perspective, in the tech industry, if you had 10 things that you wanted to change, you could maybe change seven or eight of them, right?

Because you're going to get fairly rapid feedback on did my changes have the effect that I wanted? But for Traditional biotech, if you look at it, what people did is they only changed one, maybe two things at a time. Because the investment horizon is so long, they said, I don't want to take risk on everything. I just I'm gonna focus on risk in this one new area.

And so what we're able to do, let's say we can compress that down. whatever is the right number. Let's say we we get it down in half. All of a sudden you can start to explore more areas because you can take more risk.

There's less of a investment horizon. You get more feedback on if you're being successful or not. And so that should have an accelerating ⁓ impact across the entire sector of now I can explore more science. Now I can take more risk on biology or chemistry or patient populations or whatever it is than I could before.

Ed: Mm. And is that that means you've got more of a diversified risk profile than maybe some traditional biotech companies, which is one of the risks that people get scared about if what you know, one of the main drugs doesn't get through. Yeah. Ben Taylor: Yeah.

Yeah, exactly. And I mean, it's funny because we always focus on we want to improve the probability of success, right? Like a ninety-five percent failure rate is just it's it's almost not even a functioning industry. And so how can we get that down as low as possible?

But because we're using technology, we also do it so much more efficiently. So for example, when we put out some of these numbers, like on the chemistry side, we've been able to develop the drugs with 90% fewer ⁓ chemistries that you have to experimentally make. That's a massive time saving. So we've been able to take what's normally about a four to five year process and bring it down to 17 months.

⁓ at the same time, it's also, you know, ⁓ a massive cost changer because each one of those molecules that you're producing is five to ten thousand per molecule. And so that is a huge efficiency change on the ⁓ the discovery. But you also see it in the clinical side. So by being able to understand our patient populations better, we use a lot of real-world data to understand where are these patients, what are their comorbidities, what could ⁓ what other drugs could we potentially used for combination therapy.

What we've been able to do is actually drive faster enrollment and expand our patient population. So the faster enrollment, where we've been able to see a 30 to 60 percent improvement in the enrollment ⁓ speed, that is directly affecting your ⁓ time and cost of the trial. ⁓ and 70 cents out of every dollar in pharma RD is spent on the clinical trial. And so those changes are just absolutely massive to the overall.

Roll system. Ed: And if people are looking at Recursion today, ⁓ what type of company is it? Is it a biotech pipeline company, a platform company? How do how do you perceive it?

Ben Taylor: Yeah, well ⁓ we've held true to our mandate and we've maintained our diversified base. So we are a hybrid. Definitely because we are developing our own ⁓ pipeline, we're more biased towards the therapeutic side of that. I would say we're a therapeutics company with risk diversification in our partnership side.

So that 500 million that we've been able to bring in from our partnership business not only increases our MPV because we actually have phenomenal economics on those programs. ⁓ you know, for Sonofi alone, for example, we can do up to 15 programs, each one. Those programs has the potential for $343 million in milestones and low double-digit royalties. So they're really great economic deals.

but that also allows us to ⁓ fund development of our platform and bring in non-dilutive capital. ⁓ so Definitely therapeutics will be ⁓ the long-term future of the company, just how we will be traded and and ⁓ grow as a company, but the platform or the partnership side of it definitely balances us up. Both of them are driven by our platform, so ⁓ that's that's an ongoing piece. Ed: And today do you believe ⁓ milestone payments are more important to the business or eventual like product revenue?

Which one do you optimise for today? Ben Taylor: Yeah, ⁓ it's it's a great question. Well how we run our partnership business is actually trying to keep it neutral during the first couple years. So we're not ⁓ spending any money out of our own pocket.

So we basically get paid in advance for our direct costs in our partnerships. And usually within the first two to three years, we try and hit milestones that will start to drive profitability into it. By about the third year ⁓ of a given project, we hope to be past our operating obligations. So when we reach something called development candidate, for example, ⁓ all of a sudden that payment would be all profit to us.

And all of the milestones and royalties after that would also be all profit to us. And so we're just trying to keep it neutral until we start to get to that point and and move on. Long term it could be incredibly profitable. Ed: ⁓ well yeah, thanks so much for the conversation today.

I really I've really enjoyed it. And my l last question ⁓ before we wrap up was what how do you ⁓ create and maintain a moat for Recursion in the AI age, 'cause there'll be competitors coming out and what do you see as your USP to double down? Ben Taylor: Yeah, I definitely think ⁓ the data is a huge part of it and understanding how to use the data, ⁓ because there is definitely a period of time where you're just doing that inductive learning and you haven't figured out what data is important.

⁓ so that that is definitely a massive part. I also think the integration of the technologies, because each one of those components you have to independently validate. ⁓ and so we've just ⁓ we do applied development. Almost all of our budget actually goes into working on pipeline programs or partnerships.

And what that allows us to do is develop leading technologies, but do it against a project where you can see this is actually making a difference or it's not. ⁓ and so that's another really important part. we can build those and then we integrate them into a system. And of course, we've got our own supercomputer, which helps.

So we don't have to wait for you know the cloud to become available. It's also a lot more cost efficient ⁓ for us. And I mean, underlying it all is the people. It has taken no joke, ⁓ more than a decade to put together the right group of people because you really want them to be bilingual.

Like we want them to understand medicine and sciences as well as coding. And so we create these multidisciplinary, basically pods ⁓ that are working together. and they are they are incredibly effective. It's it's fun to watch because I I remember once a chemist coming up to me and saying, you know, I worked for more than seven years at, I won't name the pharma company.

And during that entire time period, I worked on two programs. I've been here less than two years and I've worked on seven programs because I can just be so much more productive with actually, you know, using the systems and doing it. And so what we what we are always focusing on are who are those strategists that also have the the deep understanding of their industry because those are going to be the people who can take AI and really maximize their use. They're not being replaced by it, they're being empowered by it.

Ed: Yeah. Yeah, thanks Ben. I mean it's been really great to talk to you today. ⁓ it feels like we're going into a golden age of drug discovery and healthcare.

It's really great for everyone. It's it's truly fascinating. So I'll be I'll be ⁓ paying close attention to how the company does and new new announcements. But thanks again for your time today.

⁓ and yeah, hope to see you again soon. Ben Taylor: I hope so. Yeah. Sounds great.

Thanks very much. Really appreciate it.