Good afternoon, everyone, and welcome to Morgan Stanley Global Healthcare Conference. I'm Sean Lammon, Head of U.S. Midcap Biotech Equity Research here at the firm. Before we begin, for important disclosures, please see Morgan Stanley Research Disclosure website at www.morganstanley.com/researchdisclosures. And if you have any questions, please reach out to your Morgan Stanley sales representative. For this session, we have Recursion with Co-founder and CEO Chris Gibson, CFO not here, Chief R&D Officer and Chief Commercial Officer Najat Khan. Welcome and thank you both for your time today. Very, very kind of you for having us.
Yeah.
Welcome.
Thank you.
So we're doing this for all our companies. We've got, what, three macro questions just to set the tone. But with China's rising biotech innovation, how do you think about Recursion's competitive position here and will this influence your R&D and business development strategy?
Yeah, so look, we're excited to see new medicines coming from wherever they come from, and China certainly has had a meteoric rise over the last few years. At Recursion, our focus is on going places that are first in class or best in class in undruggable areas of biology that I think are maybe not the forte of the Chinese biotech market right now. We're really focused on going places we don't think other people can. We think as that market continues to rise, it's gonna create a differentiated opportunity for us in this kind of going where others can't, focus that we have here at Recursion.
Wonderful. A last question. As an AI, taking over a biotech company, can you describe the key ways your platform is leveraging AI and thinking about AI's future disruption potential?
Do you wanna take this one?
Sure. Happy to, great to be here. You know, for our platform, I think it's been really important to build an integrated end-to-end platform that touches on the two, three highest drivers of the high failure rate that we see. So the fact that Chris alluded to, you know, Recursion really started in the biology space. How do you use multimodal data, generate that data, which doesn't exist today, from phenomics all the way to real-world data to identify novel biology, novel insights, etc.? And I think that's really important for the China conversation in terms of being able to differentiate ourselves by having unique biology. That's where first-in-class products really come from. The second is the integration with Exscientia.
I think that was really key to also build out the second module, which is focused on chemistry, right, and really ensuring that we can drug the undruggable as well as design molecules using generative AI and active learning, both your hit ID, your starting point is different from what others would do, and then rapid lead optimization, using active learning and other approaches, and then very recently, we have also built out the clinical development AI component. That's the third part of our module. We call it ClinTech. That's leveraging both the multimodal and patient data to stratify patients, incredibly important for early development, so you start the strategy with enrichment in the patients that will respond, and then using other approaches with AI and real-world data to rapidly enroll and execute programs.
So we really focus on the areas that we feel are the highest drivers of the failure rate from biology, chemistry, to clinical trial execution.
And across all of those layers, we're combining wet lab and dry lab. I think that's another really important differentiator. If you were to visit our offices, you'd see a room full of robots doing hundreds of thousands or millions of experiments every week across all of these different areas. Or in our partnership with Tempus, we're getting real-world data from patients, and then we're combining that with some of the most sophisticated dry labs in the world. In fact, Recursion today still owns and operates the fastest wholly-owned supercomputer in all of biopharma, where we're training hundreds of different AI models, exploring lots of different frontier foundation models across all these different data layers.
I don't think there's many companies that are able to bring the sophistication both in the wet lab and the dry lab, and most importantly, iterating across those in order to, you know, effect change in all these different areas that Najat mentioned.
Wonderful. Thank you. And last question on the macro before we get to Recursion specific, but you know, what's been the most impactful on Recursion from the regulatory side of anything? Is it tariffs, MFN, but some way off, or FDA regulatory?
Yeah, I'm happy to take that one. I think you're right. MFN so far, a little bit off, way out, ways off, but we are keeping track on that, right? In terms of whenever we are looking at with our, with our automated workflows to actually start a program, you have to think about not just the unmet needs, the chemistry, biology, etc., but also the potential commercialization, pricing, impact, and so forth. So we're keeping that's more of a long-term. We're keeping that. We're watching that closely. On the regulatory front, I think there are areas that are tailwinds for us. So very recently, some of the announcements made and the guidance shared on reducing the reliance in preclinical work and translational work on animal models to leverage more predictive models.
I think that plays into very much the area where, back to what just Chris said, the dry lab, wet lab, not just creating the models, but the moat really comes with the data. How much data on failure or successes? Some successes are available, but failure when it comes to ADMET and other experiments done preclinically exist in the real world today or exist in publication. It's limited. So we need to create that data in order to then build better models that can predict. That's an area of huge focus. We have our automation studio both in, of course, Salt Lake City, but then also in Oxford. The other part I'll also say, there's been guidance from the FDA for the last four, five years in terms of the use of surrogate endpoints, especially in rare diseases, in the use of alternative endpoints, accelerated approval, etc.
I think that's very important with how we focus on clinical development, rapid go, no-go POC, because unlike a one or two-asset company, we have the volume given the platform to have more shots on goal. So we prefer to be good stewards of capital and make quick go, no-go decisions, for instance, what we did with the portfolio prioritization a few months ago.
Wonderful.
Maybe one thing I'd just add to that, you know, as we look out over the next decade, it's pretty clear where healthcare is going, and it's going in a place where there's gonna be increasing pressure on every company in the space. We're gonna need better medicines, and we're gonna need to get there more efficiently. And I think as we look out at the investments we've made over the last decade and the investments we're making going forward, we are investing extraordinarily deeply right now to build a platform that's gonna enable us to repeatedly discover and develop medicines, which we believe over time will lead to a higher probability of success. And we're already demonstrating decreasing time, decreasing cost to advance those medicines, and we're doing it at scale for a company of our size and scope.
I think increasingly in that pressured environment going forward, we're gonna need medicines, and so you're gonna need companies like Recursion and others who can do it in a more efficient way. So, you know, very much, regardless of whether it's MFN or tariffs or what kinds of pressures come, we know pressures are coming, and I think this is the kind of company that's built for that kind of future.
If I can just add on one point, it's also important to be prepared, and so the investment being made right now in order to develop the data, the models, the algorithms, which we are using in our programs that go into the packages that we submit or the pre, pre-specification discussions that we have with the FDA is how you actually change guidance to reality in programs and actions and do that at scale.
Wonderful. Thank you. To get a more specific on Recursion, I've got some questions on the platform. How has Recursion OS 2.0 evolved since the merger with Exscientia, and what differentiates the platform from other AI-driven drug discovery platforms?
Yeah. So look, we're really excited to bring the two companies together. I think Recursion and Exscientia really represented the biology and chemistry-centric approaches to what the future of this industry looks like. And by bringing those two together, we now have the multiplicity of those technologies. It's been a really exciting first nine months, I guess, at this point. Lots of cultures that we're bringing together, lots of actual tools that we're bringing together. And I think it's safe to say at this point, the teams are fully aligned and driving forward, and we're leveraging the best tools from Exscientia and from Recursion across every single program internally today. And we're starting to leverage all of those tools in our partnerships with Roche, Sanofi, and others. And so I would say it's going really well.
What's differentiated, it's back to what Najat talked about before with these three areas that we're focused on from, sort of understanding the biology, designing the right molecule to attack that biology, and then leveraging tools in the clinic to really develop our medicines efficiently. And at every one of those steps, again, real data, real sophisticated compute, and iterating across those two.
Sure. Can you walk us through how Boltz-2 is being integrated into workflows and why you've chosen to open-source it?
I mean, you know, Boltz-2 is a really exciting collaboration. It's just a highlight of one of the many models that we build internally. What Boltz-2, for those watching, is allowing us to do is if you think about the top of the funnel in terms of where ideas and hypotheses come from novel biological insights, our funnel is very much, much broader because we are looking at these whole genome, large maps of biology and generating hypotheses. Instead of using something like Boltz-2, which helps you do much more accurate FEP-based analysis much later on in the design process, we have integrated in our first module of our workflow. That is when you're figuring out your biology and also getting a triage understanding of are there molecules that can bind, and what would that look like?
The fact that you can move up the triage process so much earlier in discovery means the whole notion of going from a V, which is it takes a long time to really attrit and figure out which molecules would have the largest, highest probability of success, you turn that into a T, which means a much broader start, and then you have much higher probability and consistent probability, and you do it better, faster, so that's how, and every single program that we have in our discovery portfolio is leveraging Boltz-2 or Boltz-2-esque iterations of models that we have in our portfolio. The other thing I'll just say, there's new models coming up all of the time.
It's also an offensive play for us to commoditize, quite frankly, some models that should be, when we think about some of our competitors and the models that are used, but also the rapid iteration of how we can integrate open-source models that we might be interested in, fine-tune, tweak them into our platform. The modularity approach of how the platform has been built will include us to the engineering and the tech teams that have done that at Recursion.
Great. Thank you. And how are you leveraging causal AI and multi-omic data to improve patient stratification and clinical trial design?
Yeah. So what you asked in terms of Boltz-2 was that first module. I'll go to the third module, which is ClinTech. This is where we're leveraging all of the preclinical work that we do combined with clinical genomic data, as I like to call it, genetics, transcriptomics, patient data. An example of that is the CDK7 program that we just initiated the combination for. Now, with CDK7, which is a master regulator, you can go in many different indications.
What we did is we looked at our preclinical data, but also leveraging our partnership with Tempus and internal causal AI models that the team, this fantastic team called Frontier Lab, which is our next-gen AI research team, developed proprietary models that basically showed that when you have a higher overexpression of CDK7, you have lower OS survival rate, and the model and the analysis, the Kaplan-Meier curves, which is just statistically significant. Why is this important? It helps us narrow in on the disease area and the patients, patients that we select, but it was also consistent with what we had seen in preclinical data, CDX/PDX models, as well as some of the data we have from the clinical outcome so far, which is our one PR was in the ovarian cancer patient population. I mean, I think this is something Chris has mentioned consistently.
It was very important to not just take data sets by indication, but be pan-cancer because then it allows you to look at it from a much more pathway and mechanistic approach. CDK7 can have impact in breast cancer, ovarian cancer, you know, non-small cell lung cancer potential, right? And same thing with our other programs like RBM39, which is focused on transcription stress and DDR defects in tumors. You can go in so many different directions. How do you prioritize? That kind of pan-cancer causal AI analysis augmented with what we can do internally, but with our more, you know, traditionally what you would do preclinical work and clinical work is really helping us increase the accumulation of evidence on where we go.
And it's not just in oncology. We've got a partnership with Helix in the non-oncology space. And in general, we believe in, we're new school in many ways, but old school in others. Like, we believe in the power of forward genetics and the power of reverse genetics. And we believe even more deeply that when you combine those two, you actually get real synergy. And I think the kind of data sets that we've generated at Recursion, the kind of data sets we partnered to bring in from groups like Tempus and Helix are enabling us pretty uniquely to combine those two things, and we think that's gonna give us a lot of opportunity in the future.
Wonderful. And what are some of the key milestones or success criteria for REC-617, in PROC, second line PROC?
Yeah. I mean, 617, this is the CDK7 program that I was just mentioning. We started the combination. I mean, combination is gonna be critical for us to see, you know, whether response, of course, is one, but also do we keep a cleaner safety profile, which has really plagued this class of medicines for a long time, right? Therapeutic index and being able to optimize that. It's not ever gonna be an easy endeavor given the importance of CDK7 broadly. So that's gonna be the selectivity. And that's where the design element comes in and how the molecule was designed. And also, this is why patient stratification becomes really important. How can you improve or enhance the signal and not the noise? But I'll say that there's also other programs.
For instance, our FAP program, which is additional data coming up in the next, by the end of this year, which is one of our first, I would say, proof of concept of the platform program. So there, FAP, this is a MEK1, allosteric MEK1/2 inhibitor, that is focused on the disease called FAP, which is driven by a loss of APC mutation. The idea that you could actually use an, a MEK1/2 inhibitor in order to actually reverse, the disease actually came from a phenomics platform. So that first module again through unbiased, we did not know MEK1/2 would work. It's never been tested before, but really looking at how do you take disease and reverse to healthy and we screen thousands of compounds, and MEK1 inhibitor was one of the ones that was top of the list.
We saw consistent polyp burden reduction in mouse models, and we're also, we just shared some very, very early cut of the data, and we see a polyp burden reduction from 30%-80% so far in the patients, side effects on target, rash, etc. So there's a lot more work to do there. We have more data coming up later this year, but I just point to that because that uses multiple components of our platform as well. And again, FAP is a disease where there is no approved drug that's available, and the off-label use of celecoxib is about 20%-30% polyp burden reduction. So we're looking for those trends, holding. So much more there, but just wanted to point out as one of the important programs to watch out for.
Sure. Thank you. And I'll still get some more pipeline questions. So you're currently targeting MALT1 inhibitor REC-3565 in B-cell malignancies. What makes MALT1 a high-probability program, and how are you tackling liver toxicity concerns in the space?
Yeah. I'm happy to take that. So MALT1, unlike RBM39, which I just mentioned, is a first-in-class program, the novel target, CDK7, really, really hard to drug. And then also, FAP MEK1/2 inhibitor, again, these are all first-in-class. For MALT1, there have been other MALT1s in the clinic. So a little bit more of a validated target in the clinic. So J&J, Schrödinger have assets in the space, ORRs in the 20s. So from your first question, there is some inkling that there's been monotherapy activity. However, one of the challenges with some of these assets, or just in general in the space, is UGT1A1 inhibition. And there's been improvements on that front, of course, with every iteration of the compound. We see very, very, minor to non-UGT1A1 inhibition. And why is that important? It's actually important not just in monotherapy.
It's really important in combination when you go and combine with BTK inhibitors, and given all of the prior treatments that these patients have, this is in B-cell malignancies, late-line patients, and so forth, being able to have a compound that you can combine with a better TI is one of the areas of differentiation that we are focused on, and also, in some cases, with some of that, some other assets, we've seen that those that have UGT1A1 polymorphisms have had more AEs, so that's another patient population that has unmet needs, so we're gonna look through all of that. The program right now is in monotherapy dose escalation, some ways to go in terms of getting some data, and we'll keep an eye on what that TI looks like.
Wonderful. Still on the pipeline, what insights from Phenomaps led to the development of RBM39 degrader REC-1245 as a CDK12 analog with better selectivity?
I love the questions. You've now gone through every single program. So RBM39, this is a degrader compound. So just to say from a Phenomaps perspective, and I wish I could show you, these are incredibly large Phenomaps. I mean, they are whole genome CRISPR knockouts, trillions of different relationships that come through. However, what we did see is CDK12, again, really important for DDR modulation, but it's been hard to drug because of the homology and the similarities with CDK13. So, we did see strong gene-gene association from our Phenomaps with RBM39. Now, I wanna pause on that for a second because sometimes people think that some of the targets that we are identifying is just from a one interaction. But the beauty of the Phenomaps, and I wish we could show it here, is the fact that it turns into a much more systems-level biology.
You start to see the connections between different pathways and so forth. And this was an orthogonal way in terms of being able to modulate and have the effect that you'd want to with CDK12, which is important for DDR modulation without the challenge of CDK13. So that's what we saw in our Phenomaps, that it does not interact or inhibit CDK13. The next step was to actually design the compound. And the compound was designed, it's a molecular glue, from target ID to IND enabling studies, only 18 months, which in industry is four to five years. So you see the speed of what you can do with the second module, which is the design module. And now it's in monotherapy dose escalation, and we should have some early data on safety PKPD, first half of next year.
I do wanna say that once before we got into the clinic, we actually went back to those Phenomaps, expanded the aperture to see what other pathways are actually perturbed in some way. And that's what helps us understand that it's really the high replication stress and DDR defects that are important, and that helps you then stratify the right patients that we go after. So there's a lot of work in terms of once you've created these maps, you can actually go back and query it just like, you know, you'd query like a search tool. And you have all of these relationships. So this is why it's important to create these data sets. And then actually that is a huge competitive moat for us because you can use that for target ID, for patient selection.
Now that you combine that with some of our real-world and Tempus, Helix, other data sets, you can do the forward and reverse genetics, validation and query that Chris was mentioning earlier.
I think I've interrogated the pipeline enough for the minute, but, you know, maybe to have a series of questions on partnerships. But before I do, I'm just sort of wondering how, you know, AI-driven biotechnology company business models evolve. And, you know, as you generate data to help train models and you've got a proprietary pipeline, which is contributing to the training, but then you've also got part of the business, which is, you know, an outsourced service to Big Pharma, to essentially get molecules into the clinic faster, which is also generating data. But do you foresee at some point that the, you know, prioritization of proprietary pipeline, you know, diminishes somewhat? And then the greater proportion of the business is really providing services to Big Pharma and Big Biotech is the first part of the question.
And then I guess, well, yeah, let's answer that part first.
Yeah. So look, never say never, and we're always open to the data. But I think from the early days, our belief has always been that the most upside, both in terms of economics for the company and for shareholders, but also in terms of impact for patients, is composition of matter and medicines that meaningfully improve the lives of patients, and with that as our North Star from the leadership, through the company and the board, and I would say frankly our largest investors, we have believed that that's how we're gonna create the most impact in this world. That's how we're gonna build a really iconic company in this space. And so I don't see that changing. There are certainly mechanisms by which we could imagine sharing some of our data.
In fact, we've open-sourced some of the largest biological data sets on Earth in order to help the field move forward. Some of our RxRx.ai, you can go to the website, you can download like massive scales of data to help train algorithms if it's interesting to you. But at the end of the day, those represent like 1% of the data we've generated or have access to in-house because ultimately we think in the long term, making medicines is what matters, and I think it's very unlikely we're gonna stop doing that. Now, in our partnerships, we are partnered with Roche and Sanofi and others. In every one of those partnerships, we are part of discovering and developing the medicines themselves.
We participate in the upside of those medicines if they are to move forward. In addition, we get to learn from our partners so that we can build a company that isn't just, you know, working well on the discovery or translational side, but one day is able to do really well in the clinic developing programs in really, really hard therapeutic areas like neuroscience, where we're partnered with Genentech and Roche for a decade. These are great learning opportunities for us to build the kind of company we wanna become, which is one that's gonna be hopefully working across many therapeutic areas, many modalities one day. Small molecules, precision oncology, rare disease for our internal pipeline, partner with the best in a small number of other big areas.
We see that as a way to subsidize both the cost and the learning of developing that platform along the way.
That's fantastic. And when you do have partnered programs and you are generating data for a third party, who owns the data?
Oof. Boy, there's hundreds of pages in these contracts.
And I'll just say, I think it was interesting. Really probably the Roche Genentech collaboration was probably one of the pioneering ones in terms of an AI collaboration where we're anticipating these questions and not just who owns the data, but who owns the algorithms. And in particular with Roche and Genentech, we're co-building AI models in some cases off of data that will be owned by Recursion or data that will be owned by Genentech, Roche. And in some cases, they have options to actually buy the data from us independent of the programs that move forward. So it's actually a pretty complex, new area of law in this space, pretty exciting, and it just depends on the partner. Some people value that data. Some people, I think, haven't learned that they should value it yet.
Great. Thank you. And I guess more specifically on partnerships, but thank you for that. You've noted that you expect $100 million partnerships by year end 2026. What assumptions underlie your guidance for $100 million milestone inflows by 2026, and how are they probability-weighted?
Yeah, so at the end of the day, we looked across all of the partnerships we've already signed, all of the programs or sort of map-based initiatives where we have line of sight, and we probability-weighted those, and I think took a nice middle of the road, and that's how we got to that $100 million plus by the end of 2026. That does not anticipate new partnerships. It does not anticipate new programs that could be initiated as part of those partnerships, expansions of those partnerships, so I think there's certainly upside in terms of the revenue that we could see coming in not only through 2026, but also into 2027, but we feel very comfortable with that guidance through 2026.
And certainly in addition to that, there are, you know, cost-cutting measures that we've put in place, since we brought the companies together from 2024 to 2026, about a 35% reduction in the cost basis. And I think still able to execute across most of what we wanted to, despite that decrease in cost. And that's because we're getting more efficient all the time. We're building new tools that are enabling us to do more with less.
Great. Thank you. I guess your partnerships with Roche, Bayer, Merck KGaA, you know, I guess we don't know a lot of detail around that, but how could you think about, or how could investors think about whether or not they complement your proprietary programs we do know a bit?
Yeah. I mean, I, I think, I'll just start with the Roche one. I mean, Chris mentioned it's focused on in the neuroscience space as well as GI oncology. So again, in that, we have created around five maps already, five of these Phenomaps. You know, one of them that we've talked about publicly and we just received a large milestone for this fall used, you know, we had to develop, I think it was one trillion iPSC-derived neuronal cells. This is incredibly hard. So many different relationships, and now the team is focused on really prioritizing the programs that come from these maps. These are in the neuroscience space. We don't have internal programs in the neuroscience space, so there's no overlap. We also have another program that's already optioned in GI oncology, and we're also developing maps there.
When you think about Sanofi that's focused in the I&I and oncology space, again, we ensure that there's no overlap in terms of the targets and so forth that we're partnering on there. Great momentum, four milestones in 18 months. These are very challenging targets, and this is where our design platform really becomes critical in terms of the lead series development candidate and so forth, milestones that are coming up. The last thing I'll also say, I mean, Chris mentioned this, we learn a lot, in terms of various disease areas, but then it also expands our aperture. You know, we're in oncology across when I think about portfolio, I think internal partners, oncology, rare diseases, neuroscience, and I&I.
And, you know, there's also additional maps that we have built that are not in that space, you know, in some of the other chronic disease areas as well. And so I think that plethora of understanding the breadth and depth in biology is incredibly important for us for our internal portfolio, but also potential new partnerships that we have. But we are, from a legal perspective, from an IP perspective, very clear on the areas that we work on, the guardrails and so forth.
Thank you. And how do you see the Tempus and Helix partnerships enhancing the real-world data capabilities and clinical trial efficiency?
I mean, I think, look, from the Tempus data, having that type of genomic, some transcriptomic, but really connecting that to patient outcomes is critical, right? If you're trying to ascertain, you know, which patients may respond versus not, if you're trying to understand what is the contemporaneous standard of care, trying to design a program and where that might be in a few years when you're actually, you know, either commercialized or you have a broader or deeper POC, that's critically important. And also novel targets in biology. So this is again the forward reverse genetics that we just talked about, where what we see from the phenomics, our transcriptomic, and marrying that to what we see from the genomic patient data as well as all the way to outcomes, patient outcomes. Same thing with Helix.
Helix is much more complementary, in terms of non-oncology areas, so I&I, cardio, and so forth. So that really gives us a huge breadth. And we recently did another partnership with a company called HealthVerity that's focused on bringing together claims, EHR lab results, right? And that helps us in recruitment and then also in some of the other areas that I mentioned. So this is an important point, which is what do you build and where do you partner? And I think we have a healthy balance of both because we don't want to build everything. We wanna be cost-effective and cost-efficient. We are building proprietary tools like in the pathway area, which doesn't really exist. It is a moat because we have our proprietary data, but we're partnering also in, with other companies in this ecosystem, to make it more robust what we have.
Thank you. Thank you. I next have a couple of financial questions. But you could talk about your current cash runway, the assumptions behind it, and how are you managing burn rate?
Absolutely. So as I mentioned before, we were able to cut costs by about 35% between 2024 to 2026. We finished last quarter with a little bit over $500 million in cash. We've given guidance that this year we're gonna spend sub 450 and next year around 390. And when you take all that together and then you take the $100 million of revenue that we talked about through 2026, plus the potential for significant revenue in 2027, and you look at our historical ATM usage, I think it's pretty clear. We've got multiple paths to achieving, you know, cash runway through 2027. And frankly, one of our most important and highest duties of the company is to make sure we're well capitalized. We're gonna continue executing not only what I just said, but potentially other financing options and/or other partnerships to extend that runway even more.
Yeah. Wonderful. That's clear. Thank you, Chris. A few minutes left. I've got a few strategic vision questions. You know, what does the concept of a virtual cell mean for Recursion, and how close are you to deploying it in production?
Great question. And there's a great debate going on right now about what a virtual cell actually is. As one of the groups that has been an instigator of this new term, we're happy to comment. Look, in traditional biopharma and biotech, data is leveraged as a validation tool for hypotheses that come out of people's exploration of the literature. I think Recursion, when we started the company over a decade ago, was one of the very first companies who said, "Can we generate massive scaled data as substrate to build algorithms that we can then iterate on to generate new hypotheses and advanced programs forward?" A virtual cell is just the cool marketing term for a transposition of this idea where instead of generating data to build an algorithm, your algorithm becomes good enough that it can be the beginning point.
Your algorithm actually helps you explore millions of potential hypotheses, pick the few hypotheses that you're most excited about for any given disease or any given pathway, and then you still go to the wet lab, but the wet lab becomes a validation tool as opposed to a data initiation tool. What that means is we can start to decrease the scale of data that we have to generate for every question that we wanna ask and answer. We're already seeing this, right? Recursion really founded the idea of phenomics for the biotech industry, this idea of using cell morphology as a foundational data set. We're seeing that because we've now done hundreds of millions of phenomic experiments; we've built industry-leading foundation models on these data.
We can actually now start to do less phenomic experimentation because we have algorithms that allow us to predict what experiments are gonna be most enriched for us to run. And we've also built out transcriptomics, and soon you'll see the transposition of transcriptomics as a data validation tool as opposed to a data substrate tool. And you're gonna see this across the entire, sort of, value chain, as Najat mentioned, from target discovery all the way through to ClinTech. A virtual cell is simply the way we talk about this transposition where the algorithm becomes the initiation point and the wet lab becomes the validation point.
And what that means eventually, one day, taken to its logical conclusion, is this change of shape of the funnel of our industry from a V, as Najat mentioned earlier, to a T, where, again, you'll never fully get there, but theoretically, eventually, because biology is deterministic, eventually, if you can simulate everything, you can explore all possible medicines for any disease, for any patient, completely in silico, and then pick the molecule that will work for that patient or that disease and take it all the way to the clinic with no attrition. That's how you truly get to that T. And our vision is to build a company that can approach as quickly as possible that shape change for our industry. And ultimately, that's one where you're just eliminating waste, and you're improving the efficiency of what we deliver for patients. That's what a virtual cell really is.
Where are we in that, in that place? I mean, it depends where in the pipeline. I think we are leading the industry in pathway-level algorithms. I think we're leading the industry in some of the causal AI work that's happening and connecting those layers. I think we are at the frontier in protein folding and atomistic work, and we'll talk more about those in the coming quarters. If you put that all together, I think there's this race for a virtual cell. Being able to predict what would happen in biology if you added any molecule or perturbed any gene, what would be the outcomes? I think we're probably among the front runners, if not leading that race right now.
Sure. Great answer. Thank you. We started publishing last week what will be a periodical called Looking for MOM .
But maximally optimized molecule.
Oh, I like that.
Essentially, if you look at, you know, how drug discoveries evolve from, you know, the multi-year discovery of Taxol to something that's, you know, really truncated.
Essentially, you get to a point where you can't iterate anymore. It might not be the best clinical outcome, but it's the best clinical outcome that we can design as humans.
Kind of in that context of a long term, you reach the best molecule that you can possibly get, and iteration is futile.
Yeah.
That's right.
So, you will end up winner takes all, and then at some point, everything goes generic.
Yeah. Well, we better wait until you get those. And I'll just say, if you look historically across the space, two things have been really true. One is serendipity matters. So if you look at all the great stories in our industry, so many of those stories had some incredible serendipitous moment where, without a little bit of luck, we wouldn't have ended up down that path. And at the same time, if you look across that space, most molecules are optimized in a serial way. Okay, we've got this figured out. Now we have to optimize this. Now this. Now this. If you can start to build real virtual cells, you can actually start to move to this many-parameter optimization problem where no human is able to keep all that complexity in their head.
Can we take the hundred things we want our molecule to do and simultaneously optimize for all of those and effectively remove the serendipity that's been required for many of the successes of the past? That's what the engineered future of biopharma looks like in the future, and I think that's where we and many others now are headed. It's become sort of almost a faux pas not to be using these tools in your pipeline and, you know, your clinical development and discovery pipeline.
Awesome. Well, we're just out of time. That'd be a perfect place to stop proceedings. But thank you both for your time today, and thanks for coming to our conference. We greatly appreciate it.
Thanks for having us.
Thank you so much.
Yeah.
Thank you.