How Flagship's Valo is reshaping AI drug discovery with human data

0:05Hello, and welcome to Beyond Biotech, the weekly podcast from La Biotech. I'm Dylan Kassane and this is episode 217 for the podcast. AI has been one of the biggest buzzwords in biotech for years now, but for all the technological progress, actual R &D productivity gains have been hard to find, and the cost of developing a new drug just keeps climbing. My guest today believes that's because AI alone was never going to be the answer. Valo, the biotech founded by flagship pioneering, is betting on a different approach, combining AI, predictive chemistry and, crucially, real-world human data from the very start of drug discovery, using access to more than 17 million longitudinal patient records, some spanning 20 years.

0:56That approach has already attracted major partners, including a Parkinson's collaboration with EMD Sirono worth up to$3 billion and a cardiometabolic disease partnership with Novo. Joining me to talk about it is VALO CEO Brian Alexander, a former practicing radiation oncologist whose years treating patients with aggressive brain tumors helped shape his mission to reimagine how we discover medicines. Brian, welcome to Beyond Biotech. Thanks. Thanks for having me. You know, the last couple of years, AI has just been this buzzword for us now, but R &D productivity hasn't really moved that much. why do you think that the gains that we've been promised haven't shown up yet?

1:42That's a great question. I think that there's this intuition that all productivity will go up with AI or new technologies. And I think in places it really has. But if you look at R &D productivity from kind of end to end, the real bottlenecks are in clinical development. And that's just taken longer and longer over the years. And I think there's for two major reasons. One is for most diseases, we have better comparator arms. So there's a better standard of care than there was maybe 20 or 30 years ago. So you're looking for smaller relative signals, which means bigger trials, longer trials. And then a lot of the endpoints that we use for clinical trials are more attached to the way FDA calls a patient feels, functions or survives.

2:28So, you know, maybe 30 years ago, you were looking at lipid lowering as the, you know, the provable endpoint for a statin against nothing. And now you're looking for major adverse cardiac events against kind of maximal statin use. So AI can do some things in, you know, make chemistry cycles faster or, you know, increase the ability for you to submit a regulatory documentation to the FDA in a couple months. But, you know, if you're talking about the endpoints that are over years and larger trials, it's probably getting washed out in those types of things. So I think AI is having productivity benefits, but it's getting washed out by the other kind of trends in clinical development.

3:07So it's not by itself some sort of magic bullet then. What's missing when companies start to apply AI to drug discovery, you know, and it's just not there? Yeah, I think it's kind of like any technology, new technology that you're trying to apply to a set of problems. I think you need to really deeply understand the problems you're trying to solve and then deeply understand the technology and its capabilities. And you don't often get a lot of people that kind of know both of those things. And so, you know, you can make up for that by having great teams that work together and really listen to each other and understand the problems you're trying to solve and the technologies that you can bring to bear.

3:44And I think we're starting to see that. I think, you know, you any technology that gets adopted makes the way we do things more efficient. so you know we don't really change the kind of way we do things but it just works faster or better and then i think you know when when the technology's been around for a while and you kind of understand how uh it works a bit and and one thing about ai it feels like it changes every couple months and it's like has a new capability and and and then you have to think about how to adopt it but then you know we're really thinking about how to do things like quite differently now that we have tools that we didn't have before let's talk about you a little bit because before biotech you were a practicing oncologist and radiation oncologist treating patients with aggressive brain tumors how does that shape how you're approaching now drug discovery as a as a person in the world of biotech yeah i mean 100 shapes the way i look at the world all together um you know i started off in academic medicine and i think it doesn't take long in a clinic seeing patients with metastatic brain tumors and glioblastoma and things like that uh for you to really think that academic medicine is an oxymoron.

4:51It's a very applied thing. You know, it didn't feel, it doesn't feel like you just want to sit there and, and, you know, you got to do research, but if research took years and years and years, you didn't know how it was going to be applied, the patients that were coming and seeing you the next day weren't going to benefit. And you'd often hear patients saying, you know, I know this won't help me, but if I can help anybody else, you know, I would love to do it. And I think now that we're in an era with all this information, all this availability to learn in new ways, there's the opportunity to actually help patients in real time during their lifetimes.

5:27And so, you know, I think I was, we're seeing all these patients and the first, the first kind of thing that really bothered me was that we didn't really, there was often not clinical trials available. And that was, you know, I was at, past general brigham dana farber where you know there's as many trials available anywhere in the world and sometimes there's weren't trials available for a patient with an incurable brain tumor and that just seemed just like a failure of the system in in two ways one you didn't you weren't the the patient wasn't able to try something new and you described the kind of standard of care quote unquote to the the patient their families and it was just like you felt so guilty.

6:10It was like, you know, the look in their face like, that's it? Like, that's what I can expect. And then we weren't systematically learning from any patient journeys either. So, you know, if it wasn't in a clinical trial, you weren't learning and the patient wasn't, what wasn't exposed to anything new. So there's like this lose-lose in the system where the system wasn't learning, the patients were getting any exposure to novel therapy. So I really wanted to work to kind of change that. And that's really informed most of my career since then. you know you were also working in policy too you're a white house fellow across two different administrations did that experience have some sort of impact on you especially when it comes to thinking about innovation and how to do that at scale oh absolutely um that was a completely perspective changing experience from uh you know being in both administrations like you said you know i was kind of like on one side of the island and just to see like the real people on the other side and really trying to solve problems, seeing leadership at its highest levels and how you would create the mission, vision and conditions and culture so that people could perform to their best and really attach to the mission and vision directly.

7:22So it's definitely impacted the way I lead. I got to work for a lot of great leaders, um rich and seki the secretary of veterans affairs secretary of the army really set this kind of like amazing culture uh and tone where you know you just wanted to do what was right for the veteran and he would always back you up so it gave you a lot of autonomy but you the mission vision and understanding of what we were all trying to accomplish collectively was very clear um and so yeah that really impacted me um leadership and the ability to kind of like take a step back of my own, you know, you always want to do stuff, but being able to kind of step back and make systems where people can connect directly to the mission vision and their own kind of talents and effort can manifest in the world.

8:06And I think that's how you scale in organizations. And then, yeah, the scale that the government has, you know, within the VA healthcare system, there is the ability to do kind of like local experiments and different kinds of ways of delivering care. And if that worked in what was called Veteran Integrated Service Networks or visions. If it worked, you could expand it kind of more globally. So yeah, just the actual systems in place, the incentive structures that like either promote like good incentives and behavior and innovation or tamp those things down. And you really see what kind of government is good at and what it's not good at.

8:40So it really started making me think about not just government, but the entire ecosystem, whether it was government or academics or philanthropy or our industry, how it all can work together in ways to really impact patients in a positive way. Well, you would go on, you'd leave clinical practice, you worked in industry for a little while with Roche and Genentech, and now you've taken up this role at Velo. Tell me a little bit about what Velo does and what your role is there. Sure. Yeah, I came into industry first through Foundation Medicine and Foundation Medicine became part of Roche and Genentech.

9:18And then I stayed at Roche and Genentech for a year and saw kind of pharma at its, you know, what it could do at scale and things that it was focused on. And then Valo, the opportunity for me was kind of back to when I was in academics, the things that I really like to focus on are ways that you learn with information in the in kind of the current day when we have like instead of paper charts, which would happen when I first started my residency. Now you have like access to millions and millions of patient records over decades as a starting point. How can you learn from those things? And so for me, Valo was the opportunity to do that.

9:56So really, Valo has kind of two sides of our platform. One we call human causal biology, leverages the idea that, you know, most drugs when they go into clinical trials fail and often they fail in phase two or phase three when really this is the first time you're doing a human causal biology experiment. You know your causal means like is the intervention doing the thing that you want it to do and it's not because of the way the population was selected or anything like that and and so the kind of gold standard for that is doing a randomized trial you hold everything kind of constant outside of the you vary the the intervention And then, you know, there's variation on both sides and you can kind of control that statistically.

10:39And you kind of try to isolate the signal from all of the random variation or non-random variation that happens outside of the therapeutic. Unfortunately, like everything upstream from that has been like, well, let's do causal biology. We can do that in the lab. We can do experiments and understand those things. Or like human, we can make observations in human data. But when I started my medical training residency, anything that you got from observational data was considered hypothesis generating. And we shouldn't take too much from it. And let's wait for the trial to come out. And, you know, that's you know, you do have to have a lot of caveats from observational data.

11:18But now that we have so much of it and other kind of techniques like Mendelian randomization, because we know people's genetics and things like that. How can you actually have a better starting point on the human cause of biology? So the first kind of evidence that we're getting isn't in the clinical trial, but we can move it up front. And if we're looking at different targets or trying to discover new targets in a different disease, we think that the amount of observations that you can have can really give you a better starting point. For two ways, a lot of diseases are classified by the limited observations that we've been able to make.

11:50And they weren't classified to reveal anything about their underlying biology. And so you have things that are very well defined, like sickle cell anemia, because you were able to, when the disease was defined, you look under the microscope and see the sickled cells, and that distinguished it from every other kind of anemia. And so when Linus Pauling was describing the biochemistry of the disease or other researchers of the Mendelian inheritance pattern of it, it was because they had isolated that particular anemia from the rest of the anemias. And if you had tried to find all those causes for anemia, you would have a lot of noise there.

12:28And so, you know, we think a lot of the diseases have these kind of large syndromic classifications because of the inability to really observe things over time. And I think you can have a better starting point now in classified diseases in ways that reveal the underlying causal biology. I saw this in cancer when I was in the clinic and at Foundation Medicine when things were, lung cancer because it was in the lung and then pathologists could look under the microscope and it was non-small cell or small cell lung cancer and then you know by the time I was practicing it became like EGFR mutant non-small cell lung cancer because there were failed trials of EGFR inhibitors and you went back and looked at the responders and you sequenced them and found EGFR mutations and this was kind of like talked about as reverse translation we took like something that failed and went back to the the science and found the responders and I think we had the opportunity to have more forward precision now with all the information that we have available rather than reverse translation.

13:22And, you know, I saw how that changed the lives and practice of medicine by being that precise on the specific cause of biology of an EGFR mutant or an ALC fusion positive lung cancer versus just all comers. And so we'd like to bring that kind of approach to other diseases. And then the other side of the platform is a predictive chemistry platform where now that you can synthesize any drug that you could possibly imagine, any chemical structure you could possibly imagine, the problem is more of a search and optimization function, where it's like, okay, we're in this whole, you know, hundreds of trillions of molecules that you could make, where do you want to make those things?

13:59And so you, you know, we have proprietary ways we embed or represent chemical structures and learn over time. So we do that on the patient side with patient embeddings, we do on the chemistry side with chemistry embeddings, and we let the kind of AI ML techniques help us learn and encode that knowledge over time as we have more programs go through the process. You know, I want to dive into that science a little bit more in those two parts there, the human causal biology and also the predictive chemistry. On the human side, you're starting with human data instead of animal models and animal data.

14:34Why does that make such a big difference? I think it makes a big difference for maybe the kind of two reasons. One is to think that the diseases as they are currently classified maybe have multiple underlying causal structures. And so if you're trying to have a hypothesis in the lab about how you want to impact disease, we have a big partnership with EMD-Sorono on Parkinson's disease. And that disease was defined by James Parkinson in 1817, noticing patterns in six of his patients and writing essay on the shaking palsy. And that was enough. Those were enough observations to really define a set of patients that had a final kind of common pathway where you lose dopamine secreting neurons in the substantial nigra of the brain.

15:22And if you replace the dopamine, then you can get some improvements in the symptoms. But like, why did those neurons, why were they lost in the first place? Maybe there's multiple pathways to get to that. And so now being able to observe patients 10 years before they have the diagnosis and try to come up with some clues about what might be some of the causes. And not only like for one patient subgroup, but maybe there's two or three different subgroups of ways that. So I think better disease classifications with with the more amount, more observations that we have as a starting point gives you more kind of like a better chance to have a reveal the underlying causal structure of those subgroups.

16:01And then, you know, connecting those things to knowledge graphs and like what's been published about connecting the phenotype to biochemistry. And then things like statistical genetics, network mendelian randomization, or a bunch of causal inference techniques that we have that, you know, start prioritizing these targets in ways that you wouldn't have otherwise. You know, it's interesting to hear you talk about someone writing a paper based on just six patient records and drawing some conclusions. as I understand it you've got access now to more than 20 million longitudinal patient records and some of those for 20 years what can you learn from a record like that in 20 year patient histories that you're not going to be able to get another way or a regular way yeah I mean we're still learning that so uh you know I think if you just think of the scope and scale of the amount of patients that we were able to observe previously and the amount of observations that we'd make per patient, you know, you could, you could see the way that they kind of like, like their, their symptoms, the way that they, they kind of behaved.

17:04But now, you know, we have the, not only the 20 million patients over 20, 30 years, but often things like genomics that go along with it. So there's like lots of data points per patient. And this is where you need things like, like deep learning or a foundation model or some kind of other techniques to start embedding those patients. And what that means is like, how do I represent a patient that has like 5 ,000 data points over 20 years as a point in some abstract space so i can see how close that patient is to another patient and and so like coming up with these noticing these patterns amongst patients those are things that kind of ai and ml are are good at doing and it gives like a um it gives kind of tools to to to to people to make those those observations so you you can flag up these patient groups, you can flag up these targets.

17:55Once you validate that target, you bring in the predictive chemistry side. How fast can you actually move from identifying the target to optimizing a molecule that might do something there? Yeah, it can happen relatively quickly, but there's a lot of variability depending on whether the target was kind of known before. And often, if it's a novel target, you know, the way that we do kind of do chemistry, which is a little bit like agnostic to the structure and we look we look um for the results of a biochemical assay so in this approach that we may find novel chemistry that means that we have to find if there's a new target or we we have to find an assay that kind of tells us whether the target's being engaged and so sometimes that that's a that's a process um if it's a really hard target um that people haven't found things before that can like require a lot more optimization so it it can happen a lot faster and i think like a lot of things get reported kind of publicly about how fast something is and if it's kind of like similar uh to another uh chemistry that's already out there and it's on a known target with a lot of known kind of like readouts then those things happen relatively fast but if you're looking for like novel chemistry against novel targets um you know that still could be a process i mean it's much faster than trying to do it like without the the availability of all these models that we've really encoded a lot of knowledge in but it is it's like talented medicinal chemists with tools that encoded knowledge from like programs over decades.

19:23That's really like the confluence of things that can help us either like find novel things or things faster. Let me pick up on that term optimization because you talk about things being optimized for safety, optimized for efficacy. What does that mean at a really early stage and how can you even predict something like safety before you actually get anywhere near the clinic? Yeah. Well, predicting safety in the clinic is challenging. I mean, that's going to require a lot more data because there's a lot less information of things that have gone into the clinic and have safety profiles in the clinic.

19:55But there's a lot of things known about things like the way drugs are absorbed and metabolized and those types of things. And we and others have a lot of models that relate chemical structures to those types of outcomes. So if you look at developing a drug to a target, often you talk about, oh, can we hit the target and how potent is it? And potency is one of 30 things that you need to have a drug work in a patient. And so the more of these things in this what we call a target product profile that you can model, the more you can make educated guesses about where in the chemical space you should be and then optimize those things quickly.

20:38So the things that have happened over and over again preclinically have a lot more data, have a lot better models. As the things kind of move into patients, we'll have better and better models over time to be able to predict those kind of things. But the data becomes more sparse when you get into the clinic. So if we can talk about this system that you have, this pipeline that you have, where is AI adding the most value? And I guess the follow-up to that is, do you still need a human scientist somewhere in there? Is their judgment still relevant or as relevant? as it was before? Absolutely. I think the way that we use AI and ML is really unlocking human ingenuity.

21:18It's, I think, in the kind of human-caused biology space, it's the ability to notice patterns, but it's really like connecting those patterns to biology can happen through connection with knowledge graphs, but it's still people kind of looking at those types of things and saying whether it makes sense or not. I think, you know, I think the the the systems will get better and better and better. But I think that it just makes it more powerful for to unlock the really talented medicinal chemists or biologists or epidemiologists that are using these tools to make inferences and predictions and go in a direction.

21:53And it's like, you know, it's it's everyone's still trying to figure out that interface. You know, like when you get to a point where you're like, oh, should I trust my intuition or should I trust this model? and then there's places where the model like works much better and I think if you know the chemists kind of come along and there's places where like you know it's still it can't you can't tune it as as precisely as you need to be to a certain parameter so you know you still need that still need the human let's talk a little bit about the the positioning of the business now you've positioned yourself as this external R &D engine for pharma companies rather than a standalone drug developer wanting to push things through to commercialization why build the business that way um i think from kind of maybe first principles uh back to maybe economics 101 and what i was saying before about the ecosystem i i think like early r &d and discovery is quite different than clinical development and especially late stage and even earlier stage which is kind of more of an optimization function i have a long background in kind of clinical development and I understand kind of like how that system works and what good looks like.

23:02And so I don't think that, you know, it's good to assume you have expertise in everything. And, you know, back to the kind of ecosystem comments I made before, I think the more that we can specialize in the things we're really good at and work with others who are good at those other things, then the whole system works better. You know, patients get the drugs they need faster and everybody kind of like wins in that world. you know it for a biotech that we we feel like we're good at that human cause of biology and predictive chemistry once things go into clinical development we have we have less expertise there the systems work less well and it's really expensive so if like we're going to raise money and put money into a certain thing and create value for the world we feel like that value is better created earlier the caveat to that is that like when you're working on novel biology often you need to get to clinical proofs of concept before people think like, yeah, though that target makes sense, you know, like they can kind of go along with your rationale.

24:03But, you know, we always say like if we had a platform in the human cause biology side, increase the probability of success from drugs going to clinic from 10 percent to 20 percent, that would be like world changing. But for any one program, you wouldn't know, you know, and it was still more likely to fail than not. And so when we have conviction of things, we may need to push those into longer into clinical development. So we're trying to make a system where, you know, we have these, the two side platform, novel biology, novel chemistry. We can use the novel chemistry to known biology. And so we have a few programs of our own where we're not finding a novel target, but we're finding novel chemistry to known like high value targets.

24:41and then kind of a good mix of working with partners like Novo Nordisk or EMD Sirono to do target identification all the way through drug development, hand things off when things go into the clinic and then have kind of a fractional ownership of our portfolio as a kind of a model on that side. And then there may be areas where, you know, we just focus on our own portfolio into kind of clinical proof of concepts if needed, if they're novel targets or challenging targets. And And we've been thinking about that in the INI space where we have cardiovascular metabolism. We're working with Novo and Parkinson's with EMD-Sorono.

25:19We may do other partnerships with some of those kind of bigger things and take on risks together, but then focus our own portfolio in other areas. I want to ask about those partnerships there. emd serono that's with uh up front some milestones that's up to three billion dollars there for the company and that's focused like you said on parkinson's what are you actually bringing to that partnership what can you do that they couldn't do by themselves um i mean what we kind of what i started with which was starting with large amounts of human data and able to take that data we had already built a lot of like neuro neurodegenerative disease what we call phenomaps or clustering of patients with neurodegenerative diseases, not only kind of into the like similar categories that are that are already labeled, but like within things like Parkinson's disease, noticing subgroups of those subgroups of patients in Parkinson's disease, which have different disease courses, maybe different comorbidities, which we think are just like clustering of diseases down to the cell type.

26:26And so we have like that and we had done some work with Michael J. Fox Foundation on and kind of a preclinical consortium on trying to understand targets better. And so we had a lot of work done there. And I think with this kind of approach where we bring a lot of real world data to the human start with human causal biology is quite different than what most people do. And so I think that's at least an orthogonal data set to what people are kind of using to prioritize targets already. and then once we kind of get to those points then we kind of goes into our chemistry platform and we start working we'll start working on the targets that we kind of collectively decide are the ones that are the highest value and move those forward you mentioned this deal this collaboration with novo as well there you're not working on on cns you're not working in parkinson's but it's cardiometabolic it's obesity cardiovascular disease you know you're working in some very different fields there is the approach always the same starting with the data and working from there Yeah, that's I mean, it's a lot of different fields.

27:25And that's another way reason we kind of like focus on some certain areas with partners and other areas on our own, because I think you do need a, you need expertise in the kind of biology of these these areas. But but for us, because we start with the disease and like the approach is relatively disease agnostic, where you kind of like look at an ideal world, we wouldn't even start with the disease because like we're already putting labels on patients that we you know, our argument is that like those are the wrong labels. And so, you know, if we were completely agnostic and we were kind of moving more towards kind of starting the starting point being more of a clinical genomic foundation model where you're like very patient centric because it's not like you're not labeling them with a disease to say like, OK, here's patients with this disease and let's find some subgroups.

28:14You know, we already see in like Parkinson's is a lot of overlap with with like Alzheimer's subgroups and things like that. And so we did have to start with like patients diagnosed with a neurodegenerative disease because we're as a starting point. But we'd like to be even more agnostic than that. So in any of these partnerships, it's kind of like a therapeutic area specific starting point where you kind of organize the data, notice patterns amongst the data, subgroups connected to biochemical pathways through like our proprietary knowledge graphs, layer on causal inference techniques that we have and use that to prioritize targets for a specific therapeutic area or suggest that the therapeutic area should be defined somewhat differently.

28:58So it can be applied to pretty much anything. And so that we kind of organize it that way. You know, you've got EMD Serrano, you've got Novo. But then you also mentioned that you did some work with the Michael J. Fox Foundations. It was a targets for therapies initiative. That's a different sort of partnership. Again, it's a different sort of collaboration. Tell me a little bit about what you did there. Yeah, I don't know. It's not a completely different kind of partnership. It's like people in the ecosystem that care about a specific disease that want to come together and apply like our approach to, you know, the problems.

29:34And this is like finding targets and prioritizing targets and how do we modulate them? So, you know, I think Michael J. Fox Foundation is a fantastic organization that has really been able to kind of like organize the ecosystem in the community around Parkinson's disease. And so we were privileged to be selected to work on this grant with them and on a specific target. But it's not just us and them. They're able to create this network of groups from academic researchers to pharma that want to work on the problem. And so we were contributing to it in that way. You've got a couple of big pharma partners that you're working on projects at the same time.

30:13But then you can always develop your own programs internally as well. how are you making that decision about what you're going to work on that you're self-funding and what you're going to work on for partners or is better with partners that's a great question um i think when we start with the partnerships we're really going by a therapeutic area so that's like that depends on both what we're interested in and what partners are interested in and so you know if we're working on an uh obesity or a drug with with nova nordisk i mean the the things that we work on are and progress through the pipeline in the shared governance structure between us and Nova Nordisk.

30:53We're not choosing things within that space to work on. We'll take this one, you take that one. They get to select anything that they want to work on that they think is interesting in the work that we're doing together. If they don't want to work on something and we do, then we can work on that and kind of move that forward. And then outside of those partnerships where we really start with the disease area and prioritize targets in different ways. We have worked on some high profile targets with the chemistry platform just to kind of balance those things out so we don't have all biology risk. We have some chemistry risk and not biology risk.

31:29But we have overall, we look through the portfolio, the way that the data is kind of accumulating. And I think this is another example of ways where you have to think about doing things quite differently with the tools that you have instead of like picking a target and like moving it all the way through to the end until it's like either successful or it's like really really dead that you know we we do things like hey we can do really efficient hit finding on a basket of targets and let's like use that approach and see what's the best you know what's coming out of that step that's the best and then prioritize to like move those things forward so it's um required us to have a a bit of a different and and pretty aggressive portfolio review and prioritization process that we go through.

32:17And we can't be too wedded to any one idea because there's, you know, time is short and we have to kind of keep putting resources behind the best opportunities. Let's look a little bit forward now. If this approach that you're talking about, working with partners as well as doing some things by yourself, if this works the way that you think it will, what should change when it comes to clinical success rates over this you know five to ten year horizon yeah we believe clinical success rates will go up i mean there's the there's kind of an often cited statistic that if you have genetic evidence associated with the target that is more than two times as likely to be successful in the clinic and and that's just kind of like an association broadly of a target and maybe a heterogeneously classified disease and what we're trying to do is industrialize that process to have like not only target associated with disease, but maybe a disease subtype, better classified, and the potential mechanistic pathway by which that connects and the strength of that evidence that kind of comes out of the human data.

33:18And once you have that, then experimentation is very focused on accepting or refuting those assumptions that come out of that first pass. So by the time things get into clinic, we would think that the overall probability of success would go up over time. But that's a portfolio measure. And I think the partners that we've worked with understand the reasoning behind it and understand and see that some of the targets that we pick recapitulated in the preclinical systems. I think it'll take, you know, lots of time. Again, if you went 10 % to 20%, huge signal, you'd still need like 50 programs in phase two and phase three to kind of like prove that you're different than that.

34:05So, you know, I think the kind of statistical proof of that will be a long time coming. But, you know, we're confident in the things that we've been working on from first principles and the early results that we've been seeing. and the kind of interest in the approach from partners that we think that it will increase the probability of success of things that are going into the clinic. So what then is the biggest technical or scientific roadblock challenge, risk there that's standing between you right now today and that outcome, even if it's a 10 % to 20 % jump? What's the biggest challenge there?

34:41I think time and being able to find the right network and system of partners and ourselves to actually do the clinical development for all of these things, because that, again, is the bottleneck. So, you know, if we are able to, you know, work with lots of partners, put lots of things in pipelines, that's the ideal. But everything, you know, takes time and there's governance processes and, you know, different kinds of bottlenecks, whether it's capital or regulatory or anything when things go into the clinic to really have those clinical proofs of concept start bearing out. Brian, where can people go to learn a little bit more about what you're doing and follow along with the progress there at Velo?

35:26Certainly our website, www.velohealth.com, and our LinkedIn page we keep pretty active. Perfect. Brian, it's been a pleasure to talk to you today. Thanks for joining us on Beyond Biotech. Yeah, I appreciate having me. It was a great conversation. Thank you. That's it for this episode. Don't forget to check out all the latest news and analysis at labiotech.eu. I hope wherever you are in the world, you have a great week ahead. Thanks for listening, and I hope you'll join us again next time for another Beyond Biotech.

36:14Thank you.