Recursion Pharmaceuticals - Q1 2024

Transcript

Chris Gibson (CEO)

Hi, everybody. I'm Chris Gibson, co-founder and CEO of Recursion, and I'm so delighted you've joined us today for our Q1 2024 earnings call, learnings call, as we call it. And I wanna run through some of the exciting updates from the company over the last quarter as we move forward to achieve our mission of decoding biology to radically improve lives. And so with that, of course, some disclaimers, and Recursion, I think, is uniquely positioned to hit the tech bio escape velocity. And what I mean by that is, I think we have pretty uniquely put together the pipeline, the platform, and the people that are giving us the opportunity over the coming quarters and years to really start to demonstrate a shift in the pace, the scale, and hopefully, also the probability of success of drug discovery and development.

And we're just delighted to be in this position, to be helping to lead tech bio, because we feel so confident about what the future looks like. We feel that the future of tech bio is the inevitable future of biotech, and we are delighted to be leading that, and so thankful that all of you are supporting us. So with that, I'm gonna go ahead and dive in to a little bit of what we're working on, and I'm gonna start with the pipeline. And the pipeline, in particular, I think is very exciting because we have an opportunity to start to demonstrate catalysts on a roughly quarterly cadence. These catalysts are gonna start in Q3 with our first phase II readout, and then again, we're gonna start to be able to demonstrate phase II POC readouts on a quarterly cadence.

We'll kick off with REC-994, the Sycamore trial, and cerebral cavernous malformation. This is a first in disease opportunity, where we're really leading out. As the first institutional sponsored program that's gone through the FDA, we are nearly complete, not only with the study, but enrolling almost all of the patients into a long-term extension. And what we'll be looking for in the context of this trial is the... You know, looking across all the evidence from all of these exploratory endpoints, not only at safety and tolerability, where we believe we've got a really strong opportunity, but certainly at a variety of different potential efficacy readouts that we could work with the FDA to move forward with, to try and get this medicine to patients as quickly as we can.

We'll follow that up with another program, REC-2282, in neurofibromatosis type 2, where we've given guidance that we think we'll be reading out the preliminary safety and efficacy in Q4. We've got then REC-4881, with a preliminary safety and efficacy readout in the first half of 2025. We've got another REC-4881 program, again, with another preliminary safety and efficacy readout in the first half of 2025. And then, a number of other programs coming behind that, either initiating phase II studies or moving into IND submission in the near term. So really excited about this pipeline. I think it gives Recursion a unique opportunity to start to demonstrate, in a really robust way, shots on goal.

Of course, we know all of these programs won't be successful, but we believe some may be, and ultimately, what's exciting to us is not only the opportunity to bring medicines to patients, in many cases, first in disease, but the opportunity to begin to learn, the opportunity to begin to take the data from these trials, feed it back into the platform, and whether it's a success or a failure, to be able to improve the platform, because we really are on a multi-decade journey to build what we think will be the Techbio company that defines the future. It's not only our pipeline that we're excited about, it's our platform, and we're delighted to share with today's earnings, a lot like the deal we did with Tempus back in December, a deal here with Helix.

Helix is a fantastic company. We've been getting to know them for a while. They've partnered with healthcare systems all across the country to bring really significant scale of exome genomics and longitudinal health data into a robust software environment, where, like we did with the Tempus data, we've now signed a collaboration with Helix that's gonna give us access to hundreds of thousands of de-identified records, along with Omics data, that we can put together with all of the rest of the data we've generated at Recursion and start to continue to train these causal AI models to help understand the gene networks that are underlying now, not only oncology diseases, but also non-oncology diseases.

And what's fantastic about this collaboration is that, as you know, we've got large partnerships in the context of Roche, Genentech, and neuroscience, and maybe other partnerships coming down the line in non-oncology areas, and this collaboration will allow us to be generating even more robust hypotheses using patient data to help drive our really, really exciting platform internally. So very excited to announce this collaboration with the Helix team, multi-year collaboration here, and we are gonna be kicking that off imminently. Second, we shared this on social a couple weeks ago, but just wanna emphasize, you know, Recursion pioneered the use of phenomics to try and understand biology, but we now also have started to scale multiple other layers of omics: transcriptomics, working on proteomics. We have in vivo omics. We have the patient data, like I just talked about.

But I really wanna emphasize here our work in transcriptomics. We've now sequenced our 1-millionth transcriptome at Recursion, and we are leveraging our unique take on doing transcriptomes in order to work across a whole genome transcriptomics map, the first of potentially many in that space, and also starting to layer in data from transcriptomics across chemical perturbations as well. We think this will be a really exciting orthogonal data layer to what we're building with phenomics, what we're gonna build with proteomics, what we're building with in vivo omics in the animal organism step, and what we've got with genome-scale, population-scale data, not only with Tempus, but now with Helix.

And so not only are we training our AI models within each of these data layers, but increasingly starting to train AI models between and across these different data layers, these multimodal, omics, data sets that we're generating, I think, pretty uniquely here at Recursion. So really exciting, and congratulations to the team for the incredible scale that they've put together with our transcriptomics technology. Finally, another exciting opportunity on the platform side, we've really been leaning into active learning over the past quarter, and I think you're gonna see us continue to increase the cadence of this work, this research work, coming out of the team at Valence Labs and being done here at Recursion, which is allowing us to move away from having to do every experiment, towards making predictions about what experiment would give us the highest amount of learning.

What you can see here is one of our early pilots looking across just about 50 or 100 genes, where by using this iterative approach of active learning, we can get about 80% of the value, 80% of the information you see on that top line, only doing about 40% of the experiments. We think this is gonna be a transformational tool for us to deploy with our unique scaled wet lab environment, because what this is gonna allow us to do is start to explore this incredible sort of high-dimensional space of biological perturbations, chemical perturbations, time-based perturbations across multiple different layers of omics. It starts to become trillions and trillions and trillions, in fact, trillions times trillions times trillions of different possible combinations one could explore, far too much to ever brute force.

Using these kinds of tools and technologies we're building out of Recursion, I think we're gonna be uniquely positioned to do the best next experiment, to use our resources in the most efficient way to broadly map drug discovery, to broadly map biology, to broadly map chemistry, and ultimately, bring medicines to patients more quickly. And then I also wanna talk a little bit about some exciting news around BioHive-2. Again, we announced a little bit of this on social. You know, we partnered with NVIDIA last year. They brought an equity investment in, and since then, we've been working so closely with that team. Announced in the fall that we were gonna build a BioHive-2, our next supercomputer. And you can see here that we actually now have built out this supercomputer. We benchmarked the supercomputer at 23.32 petaFLOPS.

We did that once all the materials were in place. We built it out and benchmarked it in three weeks, thanks to an incredible partnership with the NVIDIA team. If we were to take that benchmark and compare it to the previous top 500 list, it would put us right around a 30th position there. We'll see as the new top 500 list comes out, where we end up, probably in the top 50. This means Recursion now owns and operates the two of the fastest supercomputers in all of biopharma. We'll be combining BioHive-1 with BioHive-2. We'll re-benchmark that later, and see if we can continue to move up the path.

But I think in the space of tech bio, where we know scaling laws apply, where we know both data and compute are gonna be critical, Recursion really uniquely positioned with one of the most compelling data sets, one of the fastest-growing data sets, one of the most purpose-built for machine learning data sets, alongside one of the most robust sets of on-premise compute in the industry. And putting those two things together, we think gives us a really, really robust moat. So kudos to the team for all of their hard work getting this done. And now I wanna finish by talking just a little bit about the people, because ultimately, without the people, we can't drive this mission forward. We had a couple of really big announcements this last quarter.

The first was Najat Khan, I think one of the icons of tech bio, moving from J&J to join us here at Recursion. She's already joined our board of directors, and in the next sort of six-10 weeks, she'll be moving over as our Chief R&D Officer and Chief Commercial Officer here at Recursion. She brings, I think, incredible vision from hit and target discovery, all the way through to using digital tools for commercialization, marketing, and distribution problems, and that visionary arc across every step of drug discovery and development, her experience doing pipeline strategy at J&J, and I think J&J, really one of the leaders in thinking about digital tools in the biopharma space. We're just delighted to have Najat join us here at Recursion and can't wait for the next stage with her.

Then we also announced Michael Bronstein, the DeepMind Professor of Artificial Intelligence at Oxford, has joined us as a Recursion Valence advisor, and we're so excited to join him in London in a few weeks as we formally open our London office, where we think there's just such an exciting talent arbitrage, so much great talent in the computational biology space exists in London, and we aim to consolidate a lot of that great talent as we advance this exciting mission forward for Recursion. Before I move to questions, I just want to end with some high-level guidance.

I think, as I shared before, Recursion really uniquely leading tech bio, and we believe that over the coming quarters, and coming years, I think Recursion has the opportunity, with a wide variety of catalysts, to really start to demonstrate the potential, of our philosophy for, for drug discovery. On the pipeline side, we've got multiple phase II trial starts in 2024, multiple phase II readouts over the next 18 months on roughly a quarterly cadence, and our hope is that if our pipeline continues to operate, we'll be able to meet or exceed that cadence in the future. We've got multiple INDs that we think are, are gonna happen in the near term.

On the partnership side, we continue to move forward with our colleagues at Roche Genentech on a really pioneering, visionary collaboration, and so excited to be working with both the gRED and pRED teams. We've got the potential for near-term program and map options on top of the program option we already announced in oncology last fall. With our ongoing collaboration with Bayer, we see significant opportunity in the near term in the space of undruggable oncology targets for some program options. We've got a fantastic collaboration with Tempus, and we've already identified some pretty exciting targets in the context of non-small cell lung cancer, and we're starting to integrate larger chunks of their data for broader, kind of pan-cancer causal AI models that are giving us really exciting hypotheses in oncology.

And then, as we've shared, you know, we see the opportunity for additional transformational partnerships in the context of non-oncology areas, areas perhaps like cardiovascular and metabolism. And then finally, as you saw, as we shared our Phenom-Beta foundation model for image-based drug discovery with NVIDIA on BioNeMo, we continue to do business experiments with our data, using it as a value driver, and have the potential to put some of our data, some of our tools, into a variety of different marketplaces, a variety of different partnerships, and we hope to be able to share more there soon.

Then finally, you know, our platform, we're moving the Recursion OS from automated discovery, which is where I would argue we are today, increasingly workflows and processes that are being driven in an automated way, towards autonomous discovery. Our new platform that we unveiled at JPMorgan is a stepping stone in this direction towards autonomous discovery, where AI agents will be leveraging the tools that we've built at Recursion, both wet lab and dry lab, to automatically hypothesize about biology, automatically look for high areas of unmet need, and automatically prioritize experimentation to give us the fastest route to impact for patients. We feel like that's where the industry is going to move, and we wanna make sure we're leading that. Then finally, I just wanna share some of the things that I feel like are on our roadmap in the near term.

Active learning, I've talked about. Proteomics, I've talked about. But we're also doing a lot of exploration around the potential for organoids and spheroids to help us increase both translation and kind of predictive ADME tox at scale at Recursion. Those are kind of bottlenecks that we're working on now. And then, of course, on the automated synthesis side, we continue to think that the time that it takes to synthesize small molecules, eight-12 weeks, is... We've been bringing that down through our collaboration with Enamine, but we see a real opportunity to continue to accelerate that through things like automated synthesis and automated microsynthesis. And finally, we're ending Q1 with nearly $300 million in cash at the end of Q1, and I think that gives us robust runway.

And, you know, Recursion really feels poised with some of these milestones that I've just mentioned in the near term, these potential milestones giving us pretty significant runway extension as we begin to hit those. So we feel like we're in a really fantastic space. We're honored to be helping to lead the tech bio evolution as we move biotech into tech bio, and we're so thankful for all of your questions and support. And with that, I'm gonna go ahead and move over to start answering some questions. So let's dive right in. Looks like we've got a question from Dante Noah, Eric on the, Eric Joseph on the team at JPM, Gil Blum, et cetera.

Similar question here, and they're asking: "What does success look like for each of your upcoming CCM, NF2, FAP, AXIN1/APC phase II trial readouts, and what might your next steps be?" I think just given the time today, I won't go through each of these programs individually, but what I will share is for each of these programs, we are likely to be first in disease or near first in disease. When you have an opportunity like that, I think we really have to work with our colleagues at the agency, work with key opinion leaders, work with patient advocates to look at the sum of the evidence. Of course, we'll be looking for a therapeutic window, but we'll be looking for early signals of efficacy across a variety of different readouts.

In the context of CCM, for example, we could imagine looking for objective improvement in things like hemosiderin deposition around lesions, as well as subjective improvements in things like patient-reported outcomes or other kind of neurologist reported outcome tools that we have in the secondary endpoints. And I think as we look at the sum of the evidence for each of these, we'll work very closely with key opinion leaders and patients and the FDA, and what we wanna see is moving the biology. If we see that we're moving in the biology in a way that is gonna be meaningful for patients, potentially, that's gonna be the signal we wanna see to drive forward and have discussions with the agency.

The next steps will be to aggressively pursue whatever it is we can to move these medicines to patients. In some contexts, like NF2, it might be moving to start our phase III trial in consultation with the agency. In other contexts, we might even have discussions with the agency about the potential for accelerated approval. But we're gonna really need to see what the data looks like, and we'll be looking forward to reporting that in Q3, in Q4, in the first half of 2025, with more programs coming in the future... All right, moving on to another question here.

Steve Decker from KeyBanc and Vikram Purohit from Morgan Stanley are asking, "How should we think about the significance of your phase II readout for REC-994 in terms of validating your platform and the potential for other programs in your pipeline?" This is a great question, and one we get asked pretty frequently. So Steve, Vikram, what I can say is, you know, I think we've already got a lot of leading indicators of the power of this platform. As I've shared before, you can go back and look at the paper that we published, a preprint in April of 2020, where now we are nine for 10 at predicting the outcome of FDA-approved drugs in the context of SARS-CoV-2 virus, and we made all of those predictions well ahead of time.

But as we all know, there's a lot that goes into every clinical trial, a lot of resources and trial design that can really influence the outcome, and the probability of success of the average phase II is somewhere between 25%-35%. And so we know that there will, we hope, be both successes, and we know there may be some failures. Ultimately, I think there's an uncorrelated opportunity with each of the programs we have moving forward, and by that I mean technically uncorrelated. Because even though we're using the same platform to identify each potential opportunity, we are validating those opportunities against the same gold standards, standard animal models, PDX oncology models, et cetera, that anyone else in the industry would use.

And so while we certainly think a positive trial gives people a lot of optimism around what we're building, if any of these trials are negative, then we certainly could imagine that the technical risk is relatively uncorrelated between each of these, and that's why we're pushing so hard to advance a whole pipeline of programs, with readouts coming on a quarterly cadence. What's more, we've got programs that are now moving forward with pharma partners, we've got the potential for additional programs to move forward, and we've got the potential for driving value through our data and through our computational and software opportunities. So I think Recursion, unlike a traditional biopharma company, really doesn't have the same bimodal risk that many other companies in this space do, who will typically advance one or two drugs to this big sort of phase II readout.

Great, thanks for that question. Next, I'm gonna go to Mary F. and Gil: "What success have you had to date with using the Tempus data? What other population genomics data might you look to access, and how could such data complement what you were able to learn from Tempus?" Great question. So we've already leveraged the Tempus data. We signed that collaboration in either late November or early December last year, had data coming in within weeks. The team worked over the holidays, and we had deployed some of our early AI models onto Tempus data in the first, really by the start of JPMorgan in the first couple of weeks of January. We've continued to refine that work, and as I shared earlier, we've already identified an exciting, novel opportunity in the context of non-small cell lung cancer.

We have a program that's now moving forward, using the Tempus data, and I think we're really uniquely positioned to take the Tempus data alongside the proprietary data we've generated at Recursion, to bring those together to identify targets that really you wouldn't be able to identify without these complementary data sets. And so you'll see us continue to move programs forward that way, and we hope, as those programs hit kind of the preclinical stage, we'll be able to share more about them. But as we just announced today in our collaboration with Helix, we're also now looking at sort of non-oncology scaled, population scale, genomics, transcriptomics data. And we think that's a really fascinating opportunity, not only for the same play.

Can we take that data, looking across large non-oncology diseases, maybe in neuroscience, maybe in cardiovascular metabolism, combine it with our internal data to identify sort of this, this combination of forward and reverse genetics that can move the company forward, probably in, in some of our partnerships, either existing or, or future partnerships. But also, I would say I think there's some opportunity in oncology and non-oncology space to actually use both the Tempus and the Helix data, along with our underlying data, to get a sense of how these genetic, these gene networks really work. Knowing how they're perturbed in the context of oncology settings and how they're perturbed in the context of non-oncology settings, I think will give us a really robust, a field from which to work, robust substrate.

You know, it takes a while to take a discovery program and get it into the preclinical space, but rest assured, Mary and Gil, as we get those programs into IND-enabling studies, we look forward to being able to share quite a bit more. All right, next, we're gonna go to Laura, who asks about our London office. "What are we looking for in the London office? Why are we opening a London office, and what are our international growth plans beyond that?" Great question, Laura. You know, we like to operate in cities, in communities, where we feel like there's an arbitrage, where there's great talent, and maybe fewer companies that are leveraging that talent. We're based in Salt Lake City. We've got fantastic teams, really focused in software engineering in Toronto, and other related areas, digital chemistry as well.

We've got a great team in AI and AI research in Montreal. We've got a fantastic team in San Jose, Milpitas, with our in vivo facility, and London felt like an opportunity for us to accelerate our computational biology talent. We think the U.K. has done a really tremendous job of training folks at the intersection of data science and computation with biology and chemistry, really probably ahead of the universities in the U.S. in terms of that integrated training. And we had nearly 300 applicants in just the first couple of days when we announced our London office for just a couple of dozen positions that we posted, and these were extraordinarily talented folks. So we feel like that bet is already paying off with fantastic talent in London. As far as other international plans, you know, I think-...

That office is probably gonna be a fantastic step for us internationally. We're still only a team of 530 or 540 folks, so I don't think you'll see us do a lot of additional international growth in the, in the near term. But certainly, as the company begins to, you know, move into development, begins to scale our development ambitions, maybe even thinks about commercialization in the intermediate to long term, we'll have an opportunity to grow in places like Asia, Western Europe, and beyond. So I think those are more intermediate to long-term plans. Thanks, Laura. All right, next up, we've got a question from Lucille M, who asks, "What do you think about Xaira being founded, and how much are they a competitor for Recursion?" Great question.

So for those who don't know, there's an exciting new tech bio company with a great cast of characters, that got announced a couple of weeks ago. They've got significant funding, really, you know, Marc Tessier-Lavigne, and others who are leading that organization. There's a lot of disease that needs to be treated, needs to be, cured, so we welcome everybody to the space. And our belief is that the biopharma industry in a decade is gonna look a lot more like scaled versions of companies like Recursion, than it does today. And so we welcome companies like Zira to the space. We look forward to potentially collaborating with those companies, competing with those companies. What I can say is that we believe in tech bio, the primary bottleneck will be data.

We're seeing that where data exists, companies are making extraordinarily rapid progress with computational tools like machine learning and AI. Where data is sparse, it's much, much more difficult. So what we think Zira will have to do is generate and aggregate high-quality data sets to make, to make progress there. The reality is that cells take time to grow, organoids take time to grow. So, you know, we know that they've got an incredible team, and we look forward to seeing how they start to work in that space, continue to work to build the right data sets. Certainly from our perspective, the more the merrier.

We look forward to leading the space, and we're so glad to see so many super competent companies joining, joining us and others as we move towards what we see as an inevitable future. I think I'm gonna do two more here. We've got a question from Hamida Al Gazwi, who says, "My daughter has Batten disease, CLN6, an ultra-rare genetic disease. Is Recursion willing to help labs who are interested in helping these kids, since we all know that pharmaceutical companies would not work for 35 patients, and how to establish that kind of relationship?" Thanks, Hamida, for the question. My heart goes out to you, your daughter, your family, everybody else with Batten disease, and everybody else with a rare disease.

I think Recursion believes that by building maps of biology, by decoding biology, there will be a path forward to working across many of these diseases. We have a track record of working with patient groups. You can reach out to us via [email protected] and get connected to our patient advocacy team. There are scenarios where we have used our maps to work directly with patient advocates to try and advance programs forward.

And ultimately, we do believe that companies like Recursion and others, as tech bio comes into the space, even if we don't have a clear hypothesis today around CLN6, and I don't know, I don't have the map pulled up right now, but even if we don't have a clear hypothesis around CLN6 today or other areas of Batten disease, that these kinds of approaches, these scaled approaches, are gonna be really exciting in the medium to long term. And I know that's no consolation to you and your daughter today, but my hope is that in five to 10 years, it's not gonna be hard to see a biopharma company working across diseases that have 35 patients or maybe even less. Thank you so much for your question, and again, reach out to [email protected].

All right, and finally, we've got a final question here. There's a question about my beard, which I'm going to not answer. And I will move on to a question from... But thank you to Alec at Bank of America for the question about my beard. I'll go to Amir Shaheen, who asks, "What's on the wish list, the next big pieces of the puzzle, that we need to get put in place in the wider community over a five-year horizon and a 15-year horizon in order for us to progress as fast as possible in using state-of-the-art computation for drug discovery?" Amir, I think it really comes down to the data sets. We believe that ultimately, to fully understand biology, people are gonna need to build out really deep, broad data sets.

You're not gonna need to build out hundreds of these, you're gonna need to build out 12 or 24 technologies. Maybe it's phenomics, proteomics, metabolomics, lipidomics, transcriptomics, invivomics at some scale, alongside predictive ADME data sets, tox data sets, alongside automated synthesis, and on the large molecule side, moving in the direction of other modalities, RNAi therapies, antibody therapies, other kind, you know, gene therapies. I think there will be a dozen or two dozen scaled technologies, and the company who can bring together the highest number of those over the next five-10 years in a disciplined and robust way, is gonna start to be able to pull out compounding efficiencies.

So that even if you only make each step of drug discovery and development 5%, 20% better than it was before, as you start layering these technologies together, as I think Recursion is really doing, at least my belief, more and better than any other tech bio startup in the space, you're gonna start pulling together these compounding efficiencies, and that's gonna create this flywheel of momentum and opportunity. And of course, we've got programs that are gonna be reading out from our first-generation platform over the coming quarters. We're excited then for our second generation of molecules to start, reading out after that, and we hope a third, a fourth, a fifth generation. And at each stage, we'll be able to demonstrate higher scale, lower cost, more rapid translation of these programs, and ultimately, the biggest lever will be probability of success.

As you all know, 90% of drugs fail in the clinic today, from start to getting to the market. If we can get, as an industry, to 80% failure, and then 70% failure, and then 60% failure, we're gonna dramatically improve the access to medicines and dramatically reduce the price of medicines over the coming decades. We wanna make sure that we are doing an experiment to ask and answer whether the kinds of tools that we're building can help lead out with that kind of vision.

So, watch for us to continue to build the vertical with small molecules, and then as we make a lot of progress in that space, we start to demonstrate successes in that space, you can see Recursion thinking about moving into complementary modalities as well, so we can go after a broader range of diseases, both with our internal pipeline and with our biopharma partners. All right. Thanks, everybody. It's been fantastic to connect with you all for these 35 minutes here at our Q1 2024 earnings. Please follow us on social. Please post questions at our future earnings calls, and please engage with us at conferences and in other ways. We're so excited to be having this conversation with all of you and to be leading the tech bio field as we move biotech into tech bio.

Thanks, everybody, and have a fantastic evening. Bye-bye.