Hi, folks. Ready to get started. Here with Dr. Michael Secora, a scientist, investor, and operator in the space, and here to talk a little bit about Recursion. So without further ado, get into a couple of questions. First question is, you know, Recursion's seen as one of the leading tech, TechB io companies who's helped define the space. What is TechB io, and what problem is Recursion looking to solve?
Well, thank you very much for having me here. Thanks for having Recursion be a part of this great conference. So to kick off with what is TechBio. T echB io is looking at life sciences by starting with a technology industry-first approach. And what I mean by that, it's how we generate, how we integrate, how we utilize data. I think some of these approaches are perhaps new within life science, and maybe a little bit what we take for granted on the tech side, because so much of the data we have is indeed inherently connected. And that data, importantly, is important for it to be relatable, meaning any observation or any experiment can relate to any other, and in so doing, you start to understand increasing connectivity.
That it can be scaled, meaning with greater volume, we're able to drive a higher signal-to-noise. And this is very important when you're thinking about complex systems like biology, chemistry. And lastly, that the data itself is orthogonal, meaning different kind of complementary data can be added together, and when so, you're able to confirm, refine, or refute a given insight and drive that to some kind of program. And I think that, you know, these ideas, I think, have been within the tech space for a while, but more recently have been taken into the healthcare space. If we talk a little bit about the fundamental problem that I think Recursion looks to solve, is in drug discovery and development, there...
It takes approximately $2 billion and approximately 10 years to bring a new drug to market, and the likelihood of success is less than 10%. So you see a system that is characterized a lot by inefficiency. But yet we find ourselves at this unique time, where we have tools to our advantage. We have tools with high-performance compute. We have tools like AI and ML. We have the ability to systematically and with automation control how you are knocking out a given gene and understand how a gene can relate to a gene, a gene to a compound, compound to a compound, and build these massive datasets.
By combining all these tools together, integrating all these tools together in an operating system, you know, we have been able to find novel insights across biology, across chemistry, across clinical data, and with those insights, turn those into programs. And from there, with those programs, start to try to drive value, either through our pipeline, our partnerships, our data. And we believe that, you know, if we continue to advance this operating system, we not only can have value to this industry, but can have a dramatic impact, which I would believe has true scientific and societal consequence, on how do you find the right drug for the right patient at the right time across a person's lifespan?
Within the development, you talked about the $2 billion, 10 years, and the probability of success. Where do you have the most impact on? What, which of those metrics are you solving for the most? Is it the $2 billion cost? Is it the timeline?
Well, that's a great question. I think, you know, it kinda gets into some of the, you know, tangible benefits that we're already seeing with utilizing such an operating system. And so for some years now, we've actually been making public different statistics for what it takes in terms of time and cost to bring one of our potential therapeutic candidates through a certain stage of development. And compared to industry averages, we find that we are half the time and half the cost compared to the industry average. So we're already starting to see major changes in efficiency metrics. But beyond that, it's also how you start to kind of quantify novelty. Where are you finding true scientific arbitrage? Meaning, where are you finding targets, biological targets, that how they operate are not known within the corpus of scientific literature?
How are you finding novel chemical scaffolds that can interact with those targets and start to understand certain causality within all of this rich patient-centric data? So whether it be on the efficiency metrics or whether it be how we start to kinda quantify novelty, I think we're seeing dramatic and tangible impact across all these different measures.
So maybe you can take the efficiency and reinvest into some of the novelty, if you're a customer. I'd be interested on the AI and machine learning side. You know, we see AI being embedded in all sorts of industries across the board, and, you know, how should we think about this technology being utilized in healthcare, and what are some of the tangible benefits you're seeing of AI at Recursion?
Mm-hmm. Well, I think, you know, as we talked a little bit about the efficiencies already, we see a system that is characterized by significant inefficiency. So I think that the inclusion of some of these tools of our time, AI and ML, make a ton of sense. Secondly, I think if you think about the world today, by different sources, folks will put forward that the amount of data worldwide is doubling almost every two years or so. And so we're awash in different kinds of data, and I think that's true as we try to build out additional datasets within life science.
These tools of our time, AI and ML, are not a luxury to be used by a few, but rather, I think are a necessity to be able to drive these kinds of efficiencies and find these novelties across all of this kind of connected data. Lastly, I would frame, just, you know, as a scientist by background, how much of biology and chemistry do we actually know? I would argue actually very little. And so from that perspective, it seems absolutely necessary to kind of use these tools to systematically, and I think augmented with robotics and automation, to conduct experiments and collect data, to generate data.
Collect data so that one is indeed starting to understand more and more of these great, rich, connected relationships for which then they can apply those tools, for which we can drive insight, for which we can then drive down those inefficiencies. And I think to kind of get a little bit to those tangible benefits, I think I would just kind of highlight again what I, you know, said a little bit previously around those statistics that we've already showed. I would see those as leading indicators of us having an impact on the system overall. And I'd also would just kind of highlight the way that we are able to, you know, find novel targets and novel chemistry that is not known or well known in the corpus of literature.
I think as a scientist by background, that's very exciting to have that kind of impact in just the body of knowledge that we are aware of.
That's very cool. And what are some major-- what are the challenges that you encounter when you're building an AI and trying to uncover, you know, those, those novel concepts? What do you, what do you see as challenges in meshing that world of taking AI and ML and the tech together with the world of the science world and the bio world, which might be a little bit of a different... You know, they might be speaking a little bit different language than the folks on the tech side?
Yeah, absolutely. Well, you know, I'd previously been an investor both on the tech side and on the bio side. And it's great to be a part of this conference on the other side of, you know, of the microphone, so to speak. But it's those... We can acknowledge that these are just different worlds. They're different ways of, you know, how you think about value, the different kinds of metrics that matter, and even different kind of personalities that live in these worlds. And I'd say the first challenge has been on the cultural side, the people side, and I mean that mostly internal to the company, where you want to build out a culture that has equal parts of tech folk and equal parts of bio folk.
And those are two different worlds, as we, you know, I think can, can all agree upon. But you want these groups. And I think at Recursion, our, our demographics have been relatively equally proportioned, about 35%-40% of each group is at the company. But you want them to be able to speak, speak to each other in a common language. You want them to be able to trust each other. You want them to be able to work collaboratively with each other on these very profound problems, and I think give unique perspectives around different problems that each group may not have been thinking about in isolation. And I think that's number one. The second is, when we think about drug discovery and development as a system, right?
I think I would really encourage folks. There was a really interesting paper put out about a year ago from the FDA about the use of AI and ML across the full stack of drug discovery and development, from novel target identification to next-generation manufacturing. You see that it is a full system. And so you can think about each one of these pieces as some kind of point solution, but it's really in the integration of these point solutions within one encompassing operating system that I think one can have true systemic change, bring about true systemic impact.
And so the challenge there is not only formulating sometimes the best point solution, but how all those point solutions can be connected together in service to that system, and that system in service to tackling, I think, some of these fundamental questions, which really take a holistic perspective to have impact.
I know you mentioned you have both at Recursion. Do you see your customers doing... You know, when they're organizing their teams, are they, are they going through the same process of bringing tech folks and the bio folks together to work on their problems?
Well, I think that's a great point. I think just to highlight some of the human element, I think we've seen some dramatic changes in how people, both on the pharma side, have been thinking about tech, and how the tech folk have been thinking about life science. That certainly wasn't that way a few years ago, 5 years ago, 10 years ago. It's more of a recent change, and I think it's wonderful. I mean, if I were to comment about a few different personalities, I would look at Aviv Regev, who was brought in to lead early research at Roche Genentech. Now, Aviv Regev, she has a background as a computational biologist. So someone who's not a traditionalist, but a computational biologist leading early research, that is really fascinating.
Similarly, you know, if you take a look at JP Morgan at the healthcare conference, which we also spoke at, back in January, we did an event with NVIDIA. We had an event where Jensen Huang, who's been a great supporter of Recursion. We've done some great work together in the construction of. You know, we recently just turned on BioHive- 2, which is our most advanced supercomputer. It is the most powerful supercomputer in all of pharma. It's also a top 30 or so supercomputer across any industry. And so you see Recursion not just commanding massive automated wet laboratories to build out these datasets, but also control these computational resources to extract insight.
You see Jensen actually talking with our Chairperson, Martin Chavez, around his own ideas, and you see a tech visionary really understanding the depth of life science. And I think what's been wonderful is just seeing these worlds continuing to mesh, and I think that's only gonna continue to accelerate, and I think you even see other large tech companies thinking this way too.
It's fascinating to see that convergence and hear about it. And maybe one question, kind of in with the tech framing, is: When do you think AI-enabled drug discovery will have its ChatGPT-like moment?
Yep.
Is that coming?
Yeah, well, this is a bit of a question I do get from time to time, and I think it's a great question. When I think a little bit of this, I think... You know, I think about it from a scientific perspective, a technology perspective, even as a trading perspective. And there's this kind of phenomenon that incremental changes often go unrealized or often go unnoticed, and I think that's particularly true when you're in a system that is evolving rapidly. And I think that is something that characterizes AI enablement within healthcare today. And so to kind of think about the ChatGPT moment question in that kind of framing... I mean, ChatGPT, you've kind of looked through its background. GPT-1 came out in 2018. GPT-2 came out in 2019.
GPT-3 came out mid-2020. It was then out there, the technology that was embedded within ChatGPT was out there for about 2.5 years. And it was only when that, that piece of technology was wrapped into some kind of web-based application, did it gain this pervasive notoriety, where that value became so self-apparent, self, self, self, you know, so, so apparent to everybody. But before that, there were many, many folks that it would seem, you know, saw the value in these transformer models. And I think that there is an analogy of this happening with an AI-enabled drug discovery today, where if you look at, at, at some of the proof points, you know, last week or 2 weeks ago, you see AlphaFold 3 being released. That's incredible, right?
That kind of protein-folding problem continuing to be refined. You see, you know, drug companies advancing novel biological targets, novel chemistry, novel insights across, you know, clinical data. And then you see a lot of life science companies being founded really on the these kind of core philosophies of TechBio. And one example of that, recently that came out of stealth was Xaira. Significant financial backing, significant, you know, personalities associated with it. I mean, Scott Gottlieb from the FDA, Jennifer Doudna, who's a Nobel Prize winner. And it's and you see this kind of commitment, where, you know, as far as I can tell, it's as if the life science companies of today are being founded as TechBio companies around this, these, these core beliefs of digital nativeness.
It reminds me a little bit of, you know, in the early 2000s, you know, the early 2010s, when a number of SaaS companies were being founded, if they needed to be founded in a way to be cloud native. And it's as if some of that history is, I think, playing out again in how we think about life science companies. So I think that a lot of these ideas, I believe that these kind of moments are happening right now, and I'm confident that it's not a matter of if these ideas become increasingly entrenched in how we think about life science and pharma, but really, really just a matter of when do all of these people kind of come around to it.
I think that, you know, I'm pretty also confident that, you know, large pharma companies are gonna look more like Recursion 10 years from now than the other way around.
It's really interesting to hear about the TechBio convergence and how that might be a cloud-native type moment for the industry. Would love to shift gears a little bit to narrow how that impacts the business and spend a couple minutes talking about the business, and maybe with starting with what capabilities has the company developed to tackle these problems for TechB io, and then, you know, how do you drive value in terms of the business model?
Sure, sure. Well, it, when I, when I think about a lot of companies within technology, there's often this kind of recurring arc, and it's something like the following: You make a system for which one then is producing this kind of connected data, for which one is then producing these models, for which one is then producing these different kinds of insights and predictions, for which one is then producing improvements, and with that, one is producing a product. And I think that's how we, we look at whether it be the social network data, whether that be preference data around you know what we wanna watch on Netflix, whether that be how we, how we have you know driverless cars connecting telemetric data. All of that kind of follows that arc of improving that final product.
And so by taking that kind of thinking and applying it in a life science context, let's kind of walk through that again. You know, I think it's important for one to construct the system. So what is the system here? Recursion Operating System that runs almost like a high-performance manufacturing facility, where we're running experiments nearly 24 hours a day to generate connected data across a number of different kinds of biological contexts. With that data, we are extracting insights, constructing models. With those models, we're making predictions, predicting predictions around novel biological targets and how different kind of compounds could be interacting with those targets, and how they could have a causal effect with, for a person afflicted by some kind of disease.
We then take those predictions, and we're going through a period of improvements. We're going through how we're validating it, confirming that insight. We're optimizing that compound, and then from there, we arrive at our product. What is our product? It's a, it's an insight that we have confidence around, for which we want to then take, as a program into our internal pipeline, into our partnerships, or perhaps think about how to make those insights maybe available through one of our data strategies. And so that's a little bit of how, you know, I think about the Recursion approaches, the Recursion approach kind of en masse, as this kind of connected system to drive, you know, those insights into, you know, the, the value proposition across our three-pronged business strategy.
Which, you know, the business strategy itself is the internal pipeline. We focus on rare disease and precision oncology. There is the partnership component, which focuses on more complex therapeutic areas, like neuroscience, for example. And then there's also our data strategy, which focuses on the potential licensing of data or some of the tools that we have already been able to develop.
Would love to double-click on that partnership element of the strategy. You know, how are you driving value from the Recursion platform with partnerships in particular?
Yeah, absolutely. Well, I think we've been fortunate to develop great partnerships, both on the large pharma life science side, as well as on the technology side. And so happy to kind of unpack a little bit of what is gained for us across both of those groups. So on the large pharma side, you know, we're able to apply our platform to often more complex therapeutic areas like neuroscience. And in so doing, we are able to, as we have crossed some kind of milestone, some revenue payment, there's some kind of program option that could be paid to the company, or there's some kind of research initiative for which we then are also receiving an option payment. So there's, you know, revenue, and there's also credentialization validation for the platform as we conduct this, you know, this research together.
On the technology side, it's often what we gain there is access to different kinds of capabilities, like high-performance compute, like as we talked about with BioHive- 2, our supercomputer. Or it's access to certain kind of datasets that, again, can be related back to all the other data that we've integrated, and in so doing, have a more complete understanding of biology and chemistry and how that can translate into patients.
You mentioned NVIDIA. Who are your other major partners today, and what might we expect to see from... in terms of additional partnerships moving forward?
Sure. So to unpack the different partnerships we have, first, on the life science side, we have a partnership with Roche Genentech, in the area of neuroscience and one GI oncology indication, for which, they actually last fall had already optioned their first program and continue to do great work together on some of our mapping efforts within neuroscience. There's also a partnership with Bayer in the space of undruggable oncology, so we're going after, a number of different targets in oncology that it's been deemed or thought to be undruggable. And I think we're bringing a new approach to go after some, I think, very hard scientific problems. I think it's very exciting work happening there, too. On the technology side, there is, first, our partnership with NVIDIA in high-performance compute.
NVIDIA first helped us construct our first supercomputer, BioHive-1, about 3, a little over 3 years ago. And then just recently, we went live with BioHive-2, our next-generation supercomputer, which is a great resource for us as we construct massive foundation models and I think are, you know, mining the corpus of our over 50 PB of data, which represents one of the largest such datasets on Earth. And then other partnerships that we have are partnerships with Tempus and Helix around multimodal, patient-centric data. These, how we then, how we think about generating causal AI models. So whatever we're finding at an in cellular level, could that indeed translate into patients as well? And lastly, we have a partnership with Enamine in the space of chemoinformatics and chemical synthesis.
How do we start to construct diverse chemical libraries to understand what could be the right structure to drug a given target with? And so those are a number of the partnerships that we have in place to kinda answer your second piece there around what other partnerships we would potentially consider. I think they would potentially be of a similar sort of what we have already kind of formed. I think on the large pharma side, there could be the potential for other partnerships in large therapeutic areas. Areas like cardiovascular metabolism, inflammation, immunology, infectious disease, I think all could be very interesting areas for us to apply our platform.
And then on the technology side, I think, you know, being able to access other forms of data that could indeed relate back to some of the data that we already have, to have a more fuller operation of the operating system, as well as other capabilities, and some of those capabilities could be around different kind of therapeutic modalities, like large molecules, genetic medicines. And I would also even put this out there as a potential capability at some point in the future: I do think that quantum computing is going to have a tremendous role to play in how we think about compound optimization and all the different kind of chemical permutations that could be possible. That technology, I think, a lot more to be developed there, but one day, I think it will have an impact in life science and in tech.
Maybe more near term, how do you see tech companies entering the industry and driving innovation in the next coming 1, 2, 3 years?
Yeah, well, it's been really incredible over the last couple years to watch large tech companies increasingly wade into this space. And just to unpack that a little bit, I mean, I think, you know, NVIDIA has been really, you know, forthcoming with respect to how they've even thinking about applying, you know, computational tools, launching, BioNeMo, which is their, sort of bio marketplace for a number of different kind of, tech tools on their own platform. And then there's other companies like Oracle, who acquired Cerner, which seems to me to try to become that kind of system of record across electronic health records. Very fascinating. You look at Microsoft with OpenAI, them launching BioGPT. You look at Amazon acquiring One Medical, acquiring PillPack, being involved with a phase I trial.
Of course, Google, as we talked a little bit before, with DeepMind and Isomorphic, advancing, AlphaFold 3. So it would seem to me that large tech is increasingly wading into the future of healthcare, and it also, seems to me that the future of healthcare is very much predicated upon those that can generate large, relatable datasets that are fit for the purpose of machine learning, and AI. I think that's very much a space that Recursion has been operating within in the last 10 or so years, and certainly a space that, you know, we are happy to continue to operate in.
I'd also like to touch on your upcoming clinical data readouts. You have many of them, and can you touch on them and how, you know, the opportunities that they create and how they validate your platform?
Sure. Yeah, so, so in addition to the partnerships that we talked about, and in addition to the supercomputer and the data that we generate and the different kind of technology tools that we have, we have a large pipeline of assets, large pipeline of programs. And to your point, we actually have a number of clinical readouts that are coming. We have 7 readouts that are going to be reading out over the span of about 18 months. And whether you're an AI-enabled drug discovery company or a traditional company, that's an extraordinary number of clinical readouts that are coming at approximately better than 1 a quarter.
Just to kind of unpack a little bit of what that timeline looks like, next quarter, we're going to have a phase II readout in cerebral cavernous malformation. That is a large, rare disease, no approved therapy. We'll be reading out a phase II. It's characterized by these lesions in the brain. We could potentially be the first in-disease therapeutic. Q4, we have another phase II readout in neurofibromatosis type 2. This is another form of rare brain tumors. Here again, rare disease, no approved therapy. We again could be a first-in-disease therapeutic. We go to first half of next year, we've got phase II readouts both in familial adenomatous polyposis, as well as APC mutant cancers. Again, both of those large and significant disease areas.
This year, we also have another phase II that'll be initiating in Clostridium difficile infection. That's a very large population as well. Also, watch on the near term for IND submissions, that's the phase right before you kick off a clinical trial, in RBM39. That's a target that we found to be associated with HR-proficient cancer, as well as a novel target we have identified called Target Epsilon in fibrosis. Lots of readouts. If you think about. When I look at this pipeline kind of en masse, what I see here is a pipeline that is broad and deep and mature, and a product of an operating system that's been able to classically work across therapeutic area.
I think it's also important to acknowledge that, you know, a lot of these diseases that I highlighted have no approved therapy, large populations in need of treatment, and are ones that, you know, many of these ones that I've cited have the potential for peak sales opportunities well in excess of a billion dollars.
That's fascinating and great to hear about readouts, and we hit on partnerships. But I think when I think about the three prongs of the business strategy you mentioned earlier, the last one is data. Can we double-click on the, you know, your business's data strategy, and what are some of the proprietary tools that you could make available in the future?
Yeah, absolutely. So as I talked a little bit about the data strategy before, it is one where we are potentially licensing access to some of our data or some of our technology tools, and let's unpack some of the tools that we've been able to advance thus far. There is Phenom. Phenom is our large foundation model in biology that's built off of all of these high-resolution microscope images that we have been able to collect. This gives rise to how one might want to think about drug discovery. Also gives rise to how one thinks about diagnostics. Also, there is LOWE. LOWE is our large language model-orchestrated workflow engine.
So how you connect different workflows across an organization simply by having kind of no code, no-code ability to kind of connect aspects of data, to aspects of different kind of computation that you want to maybe carry out in chemistry, to how you might order compounds, to how you get those compounds tested on a platform. So these are some of the tools that we've developed for ourselves. Also, you know, happy to make those available to partners. And I think, you know, I think that there's great opportunity for what, you know, what these tools could mean for additional partners going forward.
That's great. And maybe I, I'd love to shift into learning a little bit more about the vision of the business, and starting off in the near term, you know, what, you know, what's most exciting to you, on the horizon in the near term?
Yeah, yeah. Well, I think, you know, 2023 was an incredible year, and I think that there's a lot of incredible things that could happen, both for the space of TechB io as well as for Recursion. I think it's been great to see, I think, these ideas increasingly be adopted by life science companies, and I think that will continue in this space, as you know, new life science companies are founded and mature. And then for Recursion, I think that there's a number of milestones in the near term that we can watch for. I think we talked about the clinical pipeline, where we have 7 readouts over the span of approximately 18 months.
There's also the potential for options for different kind of programs, also the potential for options related to these research initiatives, particularly with Roche Genentech in the space of neuroscience. And there's also the potential for additional partnerships, maybe, in the large pharma side or on the technology side, similar to some of the other partnerships that we've announced in the past.
What about in the long term, in 5, 10 years? What do you hope Recursion looks like in that time span, both from a capability perspective and also a business model perspective?
Yeah. Well, I think, you know, five or 10 years from now, there's a few things that, you know, come to my mind. I think, you know, firstly, I would love for us to see... I would love for us to continue to build out the Recursion OS, the Recursion operating system, and in so doing, really provide a full-stack solution that goes from that novel target identification through all the different stages to go to next-generation manufacturing and distribution. And I think even there, start to really take learnings, lessons from the technology space about, you know, how do you maybe go direct to patient, instead of, you know, as we would go direct to consumer, as a way of thinking about that.
Or even, I think there's been different ideas put forward about how to even think about drug pricing from a subscription model perspective. I think these are all really fascinating ideas that could be explored more and more. I think also, watching Recursion not just go into the clinic and through the clinic, but into the market and have a potential, you know, bolus of programs that are being brought to the market will be outstanding and having that impact on patients. And lastly, as we continue to, you know, build out the operating system, have additional modalities, perhaps across antibodies, across genetic medicines, and find what is the right treatment for a given person, at when when they're being afflicted.
I would also, beyond the operating system, I think it's also, you know, important that as we generate increasing amounts of data, it's my belief that all disease becomes rare disease, all medicine becomes precision medicine. And that's, for me, it's with the advent of increasing data that has relatability. It's about finding the right kind of patient cohorts that can respond to something. And that means, I believe, driving drug discovery increasingly to be that of a search problem, where you get to that kind of important call of trying to find the right drug for the right patient at the right time, and it becomes that kind of optimization problem at that time and place.
Well, this has been fascinating. I look forward to watching all of that. Thank you very much for the time today.
Thank you.