Catalyzing Drug Discovery with AI and Scaled Chemical Libraries with Devon Cayer

Welcome to the Axial Podcast. Axial is an early-stage investment firm based in San Francisco. We partner with great founders and inventors investing in early-stage life science companies, often when they are no more than an idea. Axial is fanatical about helping the rare inventor who is compelled to build their own enduring business. Perfect. Okay, we're recording now, Devin. Um, cool. You just said some really insightful stuff last like 5 minutes. We got to start recording now, but great to have you on the podcast. For context, Devon Keier is the co-founder and CEO of 1859, a leading small molecule drug discovery company enabled by AI. But I'll let Devon give the background. He's the best salesman for the company. But good to have you on.

Yeah, thanks a lot, Josh. Really appreciate it. We've been going back and forth over the past few years. Really inspired by how you keep the community kind of cohesive and always plugged in. I think the Axial Substack is fantastic. So really kind of excited to share our story and hopefully it helps.

Yeah. Oh, I think maybe we can start off, we can talk about like how you got started, maybe how you got started in science, you know, or wherever you want to begin. It could be at Scripps, it could be earlier if you want.

Yeah, yeah, yeah. So let's see, going all the way back. So I was always really interested in science. You know, early days in high school reading Discover magazines and always kind of reading the latest and greatest and seeing how the fields are moving. I got really interested in genomics. I remember like when they, you know, they did the Human Genome Project and I was pretty fascinated by the ability to, you know, sequence the human genome and for the first time kind of understand what makes us tick. So I was really interested in biology. You know, watching movies like Jurassic Park and just kind of thinking outside the box of what's possible. I think, you know, that you're only really limited by your creativity and your imagination when it comes to science. And so, yeah, my scientific journey, I guess, started off at UCLA, was part of Frieser-Stoddart lab at UCLA doing supermolecular chemistry, and I took this organic chemistry course with this guy, Stuart Cantrell, who's the editor for Nature Chemistry, and just got hooked on the ability to kind of design molecules, design matter from scratch, following just kind of like these rules of organic chemistry. And it was a little bit more art than science. Of course, there's science thrown in with orbitals and electrons and pushing things around, but ultimately with chemistry you really just have to visualize nature and really understand kind of this multifactorial rules dimensional universe to make things happen. I was really attracted by that, and so when I joined Fraser Stoddart's lab at UCLA, he was focusing on rotaxanes and catenanes and molecular machines and making these complicated, huge 3-dimensional molecular structures and using them for molecular electronics and, like, these tunable LED types of, types of components. So just kind of jumped in, I had no idea what I was doing, I'd only taken a few chemistry courses, and so it was not that useful, but learned a tremendous amount. Ultimately got to the point where, you know, really understand what was, what was going on. But with chemistry, it's, you're designing things from first principles, And when you start to look at enzymes, they're doing chemical transformations that, you know, a chemist can only really dream about. And so I got really kind of fascinated with biology. And so me and some colleagues at UCLA wanted to kind of test the waters in synthetic biology, and so we tried, tried our luck at building these genetic circuits in Arabidopsis thaliana for controlling the production of different types of proteins or chemicals. And so started to transition away from chemistry and get a little bit more into biology. Um, learned cloning and, um, genetic engineering, protein expression, um, kind of just a completely different skill set than what I've been focused on. And so I went to Scripps at a, um, in La Jolla to get my PhD in chemical biology, kind of focusing on bridging both disciplines and thinking about biology more at first principles than kind of biology as modules. So I joined Guy Rezagadir's lab at Scripps. They colloquially called us the Wizards and Warlocks Lab because we were all just doing crazy types of, you know, just a lot of crazy science, just like slapping things together and, and, uh, you know, really kind of think about, you know, how do you make self-replicating systems, or how do you use DNA as a logic element, or how do you do molecular computation with DNA as an information readout. And I was just drawn to just crazy ideas, um, things that really hadn't been ever built before. And so my thesis work, um, it was a, you know, Herculean effort. It was just basically me kind of trying to get this initial concept off the ground took a long time. So the benefit of doing something complicated is you get to fail a lot. But in the process of failing, you create this understanding of science and nature that otherwise you wouldn't have if everything worked perfectly. So it's really kind of the negative data that's really, really valuable for getting good at science. And so that was a good experience. Ultimately, my thesis work culminated in being able to barcode enzymes and barcode antibodies with a really efficient manner and using those constructs for what we call plug-and-play therapeutics. Basically having the ability to recruit different types of constructs to cell surfaces, where I could use those constructs for creating some sort of therapeutic modality to treat cancerous cells. Um, and so after that, you know, I, I had gone to Europe and was really attracted to maybe kind of doing a stint in Zurich or Italy or London like that. So I was going through the find yourself phase. And, uh, I was walking on the beach up in Encinitas, um, here in San Diego, and was like trying to think of what do I do, what do I do, do I go to Oxford and do this postdoc to figure out a way to sequence proteins, um, or do I focus on kind of this entrepreneurial idea that I had been kicking around, um, and I was walking on the beach and I was like, well, if you could only figure out if you could do like high-throughput barcoding of antibodies So kind of taking a step back, my thesis work was around if I knew how to target cancer, I could create this plug-and-play therapeutic platform. But if I didn't know the biomarker, I don't have anything. So I was like, well, how do I figure out how to quickly identify with a small amount of sample the target that's specific, a biomarker that's specific, specifically expressed in a cancerous cell in a really rapid way? And I was like, well, I know how to barcode antibodies. And I was like, okay, well, if I can barcode antibodies well and I can do that in a high-throughput fashion, what if I had a library of antibody constructs that went after all these different biomarkers that potentially exist on the cell surface? Okay, that'd be cool. Well, then how do I deconvolute it? Okay, oh, I just use the DNA barcode as a sequence tag. So that was kind of the initial idea for this initial company that me and a couple friends started. Up in SF is barcoding antibodies with specific sequences, applying those to some protein sample or cell sample of interest, and then identifying at the single-cell level, um, what are the biomarkers that are specifically expressed. Um, so yeah, walking on the beach—

one thing, one quick question is, what was that spark that got you interested in business? Was it in childhood? Did you try to make money on the side, or was it in college you have a side gig, or did you ever sell surfboards on the side, or were you ever interested in money, or did it just kind of just happen, or were you more interested in the prom of like, oh, that came up during grad school research?

I think so. In grad school, I think I was always kind of entrepreneurial. My brain naturally gravitates towards problems. Cool. And how do you, how do you solve those problems in an elegant way that scales? Less kind of bespoke brute force problem solving, but more kind of how do I create an ecosystem of solutions around this problem? And there is a— yeah. And so that's kind of always— I've always had like entrepreneurial ideas like that. That I've pushed on a couple of things kind of in the background in grad school and college and high school. But I think kind of given the time constraints in college and in grad school, never really just had that push to see any of those through. But I would say like, yeah, always kind of pretty entrepreneurial by nature.

And so this was like 8 years ago, right? Or maybe 10 years ago or something. I don't wanna date you too much, but like, but just like, it's like 8 years ago. And do you have any role models, you know, in terms of like your Scripps? I think Scripps is very entrepreneurial in general. I think Scripps has the most drug approvals of any institution in the world. So you have that exposure, but did you pick up any like mentors or role models on the translational entrepreneurial side?

Yeah. I had a lot of friends that were part of like the UCSD Entrepreneurs Challenge that were kind of my early initial mentors and some friends whose family members were CEOs of other types of companies. And kind of what happens is like, I think I'm naturally just pretty open and transparent. And so I'll talk to someone and one conversation leads to a networking kind of connection, and then I'm like talking to some other person and then some other person and some other person, and all of a sudden like you kind of find yourself in a room with a bunch of people that have, that are really entrepreneurial, and then you're just having a conversation about how you guys like to do things different and, you know, what are the lessons and learnings that people have had in starting companies.

Cool. Okay. That's really something. Just like, it's just kind of being more problem-oriented. And then if you kind of—

people will—

you'll gravitate towards people who are aligned. Well, let's shift back to the thread. Is it like— so I think the company, your first company called MYI, I remember.

Yeah.

So focused. I think you guys are the first class of IndieBio.

I remember.

Yeah.

Class. Second class. We're the second class. Bio. No, yeah, and he went to AnyBio. We also kind of got funded a little bit by the Illumina Accelerator. We got a grant. And I think, you know, coming straight out of grad school as a scientist, you're very technology-focused, technology-first. And, you know, our initial pitch decks were terrible. It was, look at this really cool technology and these 100 things that we could potentially do. And the VCs and the investors are like, don't care, I like your technology, but really what's the one problem that you're in, that one product that you have? And then the technology solves it, great, but why is it better than everyone else's? Then how are you going to make, making it money? And so kind of going through the accelerators, that reworked, um, you know, how I think about things. And it's much more kind of product-focused versus technology-focused. And it was kind of the boot camp of how to start a company and just the, you know, a lot of the kind of mentorship that we got through— what's his name—

Ryan, Ryan and Ron.

Yeah, yeah, Ryan and Ron and Arvin. They were fantastic, super honest, super helpful, and we learned a ton in that first foray with starting a company. Ultimately, we figured out kind of where the technology best fit, but unfortunately, just, it ran out of money. And after that experience, you know, after we kind of licked our wounds a little bit, we're like, okay, well, let's see how other people start companies. So I was recruited to this company called Omnium down here in San Diego. And, you know, really just wanted to learn, you know, what works, what doesn't. Take a big dose of humility once, you know, your kind of first startup experience doesn't necessarily work out. And, you know, I think naturally just my personality is, I'm pretty straightforward. I would say I have a strong personality but kind of minimal ego and good at kind of cutting through noise. So it kind of naturally just kind of emerged as— I don't want to say a leader, but someone that could herd everyone together and solve complicated problems. And so at Omnium, worked with a broad swath of kind of interdisciplinary scientists and Learned how to, you know, put the pieces together to essentially build a sequencing system from scratch. It was good experience, you know, met a lot of great colleagues that I still interface with today, some of which are kind of part of the original kind of team at 1859. And was recruited to Singular Genomics, another sequencing company here in San Diego. Was really kind of impressed by how Eli and Drew managed to make a lot of progress in a short amount of time and in a field that necessarily wasn't their core expertise. Drew was kind of a good indirect mentor, I think, for how to start a company. And I think that was kind of the experience that gave me the confidence that, you know, starting a company is complicated, but with just enough experience, you kind of see kind of the minimal viable specs required to successfully launch a startup. And I think that was kind of the big push that I had. I remember I was like talking to some friends at this little local brewery here in, up on the Mesa. And, you know, we were kind of all kind of ambitious and we're like, well, could we do this thing? And ultimately it came down to, you know, what skill set do we not have yet? And the conclusion that was that, you know, we thought we could build it. We had kind of a successful track record of getting really complicated things to work. We knew how to kind of manage teams and communicate and work well together. And that was kind of the last kind of missing piece that we needed to kind of get back into the entrepreneurial journey.

Cool. So, okay, I think that kind of moves to 1859. So you found in 2018, I think before anything, the most important question when you start, why did you name it 1859? Was that like a name you always had in the back of your head? Or was it just reading a book? Oh, this sounds cool.

Oh no, it was, uh, it was my friend's idea. So, um, we were down in La Jolla at this place called The Living Room, and we were trying to come up with a name for the company. And from the first company, MyEye, the, the way we kind of came to names before was what makes a good logo. And so MYI, MyEye, was you could have the antibody, and we're like, oh, and I could be information, like it'd be really cool. It's probably not the greatest way to go about getting a name because everyone's like, I remember our first company, they were like, M-Y-E, MyE, like it was like this like Japanese or Chinese, like, you know, 3-letter name that everyone was confused. But we were at the living room in La Jolla, we're trying to come up with a name and I was always kind of drawn to evolution. And this concept that 1859 is very, very akin to kind of the evolutionary process. And so me and my friend were trying to come up with a name, and I was like, well, what about Galápagos? And he was like, no, there's Galápagos Biotech, that's a terrible name. And I was like, yeah, that's true. Um, then I was like, what about Darwin? And he's like, it's, it's okay, but it's like, you know, you don't want to be named after a guy. It's not like, you know, you don't want to wear a jersey. Um, yeah. Basically showcase your biotech name. And then he was like, well, you know, what did Darwin publish? And I was like, Origin of Species. And he's like, when was it published? And I was like, no idea. And he's like, 1859. He's like, that's it. And there was this recent company called CD47, I think, that was another number-based name. And, you know, typically with like biotechs, it's, you know, Aflasia Therapeutics or da-da-da Therapeutics, and we're like, yeah, we don't want to do that. So, um, 1859 was more kind of like, it's either, it's either going to be a whiskey company or a biotech, but hopefully, uh, we could get a big enough of a name that it turns into a biotech company. Um, so that's how we, uh, we honed in on it.

Okay, so you're at Singular, and how did you make the jump from having a job to a startup? Was it something that you were prepping and you touched base with Andrew and other co-founders, or did you just get out of Singular and then do 1859? I think the latter. I think actually just verbalizing it, you left Singular and start the company because no conflict. And so once you left Singular, I answered the right question already. When you started 1859 in 2018, How did you get the team together? How did you get the co-founders? How did you think about the problems you wanna work on? And how did you find the technology?

Yeah, so there was a seminar at Scripps that GSK was presenting at. And Richard Lerner was hosting a conference on DNA encoded libraries. And I was sitting in the audience and I was like, They were showing kind of the workflow for DNA-encoded libraries and how it's useful for discovery. And I was familiar with like David Liu's work and a few others in the field, but I always thought of DNA-encoded libraries as kind of like tinker toys, not necessarily valuable, kind of like an interesting academic exercise. But seeing what GSK was showcasing, I was like, okay, this field is sufficiently mature. This can be pretty interesting. The problem was that DNA-encoded libraries was really relegated to binding-based assays, not necessarily probing for function. I couldn't do cell-based assays. I wouldn't understand the degree at which a molecule had potency because it was more of like a binary readout. So I'm sitting in the audience and I'm like, well, this is cool, but if I had to rebuild it, how would I do it? And if I could kind of miniaturize an assay into something that could fit into the picoscale, and I could, I could park a molecule in that picoscale feature, and I could visualize what was going on and then deconvolute with sequencing because I'm at such a miniaturized format, I could have an ultra-high-throughput screening platform. And it was kind of a big jump. But kind of just based on my expertise with sequencing and the first company and my thesis work, and it all kind of just came together. And I'd reached out to Andrew McConnell, Brian Pagel out of Scripps Florida. They were doing this in a microfluidic format. And Andrew moved out to essentially pursue a new venture, but me and him became friends. For the, for like first year and a half that he was there. And it was just like a meeting of the minds, like we're super practical. I'm, you know, pretty relentlessly objective. I don't care, you know, I'm a scientist by nature, I don't kind of follow kind of dogmatic perspectives, I follow kind of data-driven perspectives. And, and Andrew was really the same. So we could have like a conversation and completely disagree and then hone in through the data and be like, okay, yeah, that makes sense. Um, and so we could work really, really well together.

Um, how did you two meet? How did you all three meet? Is it like through friends, or did you email them, or—

yeah, um, it was an email, um, and then, uh, he was kind of on his way out from Florida, um, And we were working at J Labs at the same time. I was at Singular and he was at this other company. And we would just kind of have lunch together and play volleyball and, you know, go to breweries and he would kind of get me up to speed with where he was and I would get him up to speed where I was. And then sometime in 2019, we decided to make the leap and kind of pursue this concept. Together with one of our other co-founders, Ramesh Ramji. He was part of the advanced research team out of Illumina. And then two other co-founders, Brian Pagel and Sarah Cortez. And, you know, we spent— it was like February to June, May really. It was just nonstop pitching. So I had lunch meetings with someone new every single day in the San Diego area, pitching this concept for the company, putting pitch decks together, like one new completely revised pitch deck like every day or two. Um, and just like a lot of energy.

So I think 1859 is a really unique business model Yeah, because you're kind of a— you have a tool that makes a drug. You're definitely a drug company, right? But you have many elements of developing breakthrough tools to then enable drug development. And those early days, was it always a drug company, or did you think about tools and maybe going down the path of like instrument sales? And how did you kind of hone in on your current business model? A lot of daily iteration, maybe part of it.

Oh yeah, so as soon as we got funded for this company from FusionX, we flew out to Boston, started pitching to biotech after biotech after biotech at our biotech, pitching our platform, pitching our technology, really seeing what resonates, what works, what doesn't, really being product-focused. And, you know, the core concept is encoded libraries. We encode them with DNA or other types of molecules. And when we initially started, we were kind of binned into, oh, you're a Dell company.

Yeah.

And so we tried to copy the Dell company playbook of, you know, we tried to take some of HitGen's concepts and XChem's concepts and the business model. Because what we were selling was pretty unique. We have the ability to make hundreds of thousands, millions of molecules every few weeks. We make discoveries. How do we monetize those discoveries to make sure that we're maximizing value? And so we tried every business model variation you can think of and got a lot of feedback. What works, what doesn't, and ultimately kind of learn from the experiences of AbSci, AbCellera, Adimab, a lot of the antibody discovery companies, and started to adopt some of the things that they do. So essentially the core of the company for us, we have our quote unquote 1859 discovery engine, which consists of A target, disease target comes in from a biotech or a pharma company. They want us to design tailor-made molecular libraries that affect that disease target in a desired way. We build a couple hundred thousand molecular library in 2 weeks. We then screen it on our platform in a dose-response screen, so It's a minor detail, but it allows us to effectively see the SAR library-wide. So identifying the best molecules, the worst molecules, and then the gradient in between. We then leverage that dataset to train machine learning models to predict lead-like or the optimal set of molecules that a biotech or pharma company would be interested in. And so the tech is really just the means of generating millions of functionally annotated data points. And then leveraging that data to get to the high-value molecules that partners are interested in. And so effectively, because we're going through this design-build-screen-learn cycle, our core metric for our company is something we call cycle starts. So the number of effective cycles that we can initiate in a quarterly period, because every time we go through a full cycle, we generate data, which is, which is the associated value that a partner is going to purchase.

Um, and so how do you measure— so this is okay, the platform consists like 4 main parts, right? Your target and all the way to then like a SAR map and hits. How do you think about, just take a little step back, like building each part of the platform? You know, when you got the company started and you raised money, like did you try to build it all at once? Would they try to build first piece once really well and then add modules over time? And then also to overlay that, how did you think about like, you have like 3 different types of people. You have machine learning people, chemists, biologists, and probably a bunch of other types of people. How do you think about managing an interdisciplinary team as well?

So when we initially started, there was a lot of work that had to be done to essentially kind of scale up and commercialize the technology. A lot of kind of technical nuances that had to be solved. And so we focused initially on screening. So the first question was, can we build a screening system that kind of functionally acts in the way we want it to, to sort molecules that do the thing that we're looking for? Cool, we did that. And then the next question was, with controls, the next question was, okay, can we build a library that effectively fishes out those controls? And if it does, can it effectively predict and sort molecules that we otherwise didn't know work? Okay, that worked. And then the next question was, okay, we could do that, so we can build this, we can screen, and we can build libraries. The next question was, how how effective is this system at making true positive calls and not false positive calls so that we can have confidence in the data so that we can leverage that data as a valuable resource for machine learning. And so then we did that, and then the next question was, OK, the screen works, we can build libraries fast, we can be confident in the data, we can use machine learning models to predict high-value assets that are the types of molecules partners are interested in. Does that work? So that worked. So it was more kind of dictated by kind of the minimal viable need to demonstrate progress, and then dictated by the need of the partners where, say, they would come to us with, this is my problem, how does your workflow solve this? Or this is my problem, how does your workflow solve that? And you can either build the a vacuum for everybody, or you can build an adapter that takes in lots of different complex problems and put them all, pipe them all through the same workflow, which is, I think, is the hard part. And then that adapter then feeds back out value in the ways that they want to have it.

Makes a ton of sense. So then when you were building out each part of the platform to make it in one cohesive object thing, Did you feel like you need to build out the whole piece first before you talk to partners, or was it feasible for you to say, we can do screening, let's talk to X pharma partner and then get feedback? And then it seems like it was an iterative process with partners to build up the platform. Not so much like, well, we have this vision of platform, let's raise money and build it, then talk to customers. Let's talk to customers first and make sure we're on the right track. And that seems to be a big part of your success is looking at problems and always talking to customers so you're not going down the wrong direction. 100%.

And I think, you know, that's— there's the philosophy of lay out a plan and just execute on that plan, not blindly, but that's what we're doing. And then there's the other philosophy of thought of take a minimal viable step and iterate. And I'm much more on the kind of minimal viable step and then iterate to make sure that— because it's a long-term commitment, you want to make sure that your compass needle is pointing in the right direction. And that just requires rapid cycles of learning. You got to fail fast and you got to iterate quickly to really just get to product market fit as fast as possible. I think on the biotech side, that's a lot harder because the learning cycles are slower. You have to build a complicated technology, get it working, iterate. And, you know, you see that with, you know, with sequencing companies. It takes a long time to build a sequencer. TwistBio, it takes a long time to build a platform that can make sequences. It just takes a long time with biotech. That's a blessing and a curse because You know, it takes years to get to something that's viable, whereas software can take months to get to something that's viable. But the benefit of it taking a long time is that you create effectively an IP know-how moat where it would be otherwise difficult for competitors to catch up faster than it took for you to get there in the first place. Whereas software, I think that's a little bit different.

That makes sense. I think the lesson is you really have to always touch base with your partners as you build the platform, even if they're not— like, would you recommend talking to a potential partner even before, even, even like before the paperwork signed? Or would you wait to present data to a partner, um, and not let them— to impress them, right? Is it— how much is the wow factor important versus touching base with partners? And maybe that's depending on the company and the people, but maybe from your experience, what was the most effective way to get partners?

I think for us, you want to break your technology as many ways as you can, because— and you want your partners to break it. You don't want to break it. You want them to break it in many ways as possible so that when you rebuild it, it's an elegant solution that solves all problems versus some. Not in the context of being everything to everyone, but being— because for a platform company, you can't be everything to everyone, but you should strive to increase your accessible market as much as possible. With as generalizable and scalable solutions as possible. I think that's at least my philosophy.

Um, that makes sense. That's actually pretty, really insightful. Uh, but on the platform side also, I think you alluded to core metrics. I think it's a really important topic is when you build a biotech company platform, it's like, how do you measure your success? Because you definitely, you know, you talk about cycle starts, um, but like for 1859 Are there some core metrics and how did you hone in on how to determine if a platform is working or not?

Right. I think it's you hone in on your product and then you work backwards from that product. So how long does it take you to get to that valuable product? That's step one. And then how long does it take you to get to that next valuable product along in the same vein. And everything kind of, I think all the metrics effectively fall out of that because you want to be able to show scalability, not just, you know, if my product was, I got to a, we were a platform company and I got to E3 ligase molecular glue type of molecule. Cool. How long did it take? 10 years, 5 years, 20 years? What's your next product? Is it going to take 10 years, 5 years, or 20 years? Because that's a really long iteration cycle and you're probably not going to survive. So I think it's, you know, it's really kind of honing in with kind of what your core value proposition is, like what your— where your competitive advantage kind of vectors point. And that's your product. And then kind of working back from the market for that competitive advantage kind of focused product.

That makes a lot of sense. And so maybe we can talk about business model too, which is interrelated. You talk about Accelera and Atomap, these are wonderful businesses, right? Because they're like, they have technology platforms and partner up with large companies, generate a good amount of cash. And, you know, control their destiny. So 1859, how do you think about doing that for small molecules? And in terms of like, you know, antibodies have some advantages and small molecule has their own set of advantages. And it definitely seems that the partnering business model is more amenable to antibodies. I think for 1859, it seems like you're more focused on a fully integrated model, not just licensing a molecule, but a library. And a process. So how do you— maybe I'll start with traction so far, but maybe also think about like, you know, how did you develop your business model to make it tailor-made for small molecules and do something that maybe other people hadn't really done so well?

Yeah. So yeah, good artists copy, great artists steal. We stole all of our learnings from kind of the market. And so if our core metric is cycle starts, effectively that's the number of cycles we can execute per quarter. And then we think about, okay, how many annual cycle starts can we go through? And so that effectively becomes capacity for discovery. And so how we monetize, we have Discovery Access, which effectively is leveraging our in-house capacity in terms of a service-based offering where partners come to us with problems, disease targets of interest, and they need small molecules for. We apply our workflows, we build libraries, design libraries, screen those libraries, generate machine learning models, predict leads, and then we can also leverage those models to not just identify early leads, but also optimize molecules towards optimized leads or even candidates. So partners come to us where we essentially provide that tech stack that gives them the accelerator to high-value assets. Or if a partner wants to build out capacity at their site, we call that our technology access, where we effectively build out our screening infrastructure on site. Partner scientists can interface with our teams. We design libraries, we build libraries, we deliver libraries to that site. We generate, we do screens, we generate the data, we apply our machine learning models, we predict lead molecules. Partners can leverage those models to optimize the molecules that we discover for their targets, effectively keeping that in-house. So pharma, larger pharma companies and big biotechs are interested in that path. Kind of the mid biotechs and smaller stage biotechs are are more interested in the discovery access path, but it's not necessarily exclusive. We also have pharma partners that are, that just want to give us 20 targets. Here's 20 targets, go get us molecules. And so effectively our business model kind of falls into the, what Adimab calls funded discovery and technology license access. That's effectively how our business model falls out. And then the last kind of component that's emerging is what we call program access, where we might interface with academic institutes, or we pick our own targets. We apply our workflows to those targets, we generate molecules, we advance those molecules, optimize them further and further and further. This is more kind of a pipeline. And then we out-license those molecules to biotechs or pharma companies interested in assets around those particular targets. Or the future path could be potentially out-licensing those molecules to new cos and putting the associated experience and, and pharma veteran executive leadership team around those new cos to advance those programs even further. So effectively using our capacity in-house, we build capacity externally, or we leverage our capacity to pursue programs for out-licensing.

Concerned. Cool. And I think on this business model point, in biotech in particular, you know, like, come look at AbCellera. You know, they've done all this business model innovation, and what, what has it enabled for them? Right, it's allowed them to develop a lot of different programs with partners. It's allowed them to control their destiny, right, and not have to excessively sell equity. And there's a bunch of other advantages, you know. For 1859, you know, and it's just like, why do all this business model creativity? Or what's the long-term purpose in terms of like, what do you think, versus just the standard model of like, hey, let's make a drug, right? I think that's what most people in biotech do. They just say, make a drug and just do that one business model. Why take all this risk on the business model side and what kind of long-term value do you think that creates for the company?

Right, so I think we think of ourselves more of a catalyst. An accelerator for, for value creation. And I think the big issue is that if we focused exclusively on pipeline, 94% of the time we're going to fail. We would rather enable others that want to advance programs. Well, we're doing some internally more as case studies to kind of just stress test the platform and demonstrate that we cannot just find Hits, early leads, optimized leads, all the way to candidate, but kind of go end to end for a particular partner. So it's more kind of marketing material, if anything. But yeah, I think the big risk is it costs a lot of money to develop assets. And we would rather spread out that risk across the industry. And put them in the best chance of— give them the best chance of success to start with molecules that are much less likely to fail as, as those programs get further and further and further.

I totally agree. I think, you know, I think business model innovation allows you to have a lot of leverage because you're going to use the resources of a bunch of other people, ecosystem. And I think you said this at the beginning, right? You like to build an ecosystem of solutions for a problem. And so maybe you do the same thing for 1859, right? You're building an ecosystem of solutions to make drugs. And so one quick side note, maybe, you know, talk about AI drug development as a side note. Machine learning is like the last part of your platform to optimize it. How have you seen machine learning affect various parts of your platform, especially recent advancements and How do you foresee that progressing over the next maybe like 5 years?

Yeah, so machine learning is going to be a critical component for us moving forward. And the reason for that is if I handed a chemist a million data points where those millions of data points all had SAR information around how a particular molecule affected some target of interest and the degree it affected that target of interest, they would be overwhelmed with information. They would be like, I have no idea what to do. But what machine learning is really good for is being able to tease out those very subtle nuances that make things work or not work. And we're starting to leverage all of the data that we're generating to create models that take into account all the positive and negative, negative data that we've generated. And we're creating a kind of real-time feedback loop for our medicinal chemists where, say, we give them a molecule that's validated, it works, it's an early lead, it's a good tractable hit, and then they start making these bespoke edits to the molecule. And normally the cycle for learning would be CRO, go make a molecule, and then we'll figure out if it works. But what we do internally is we actually leverage the models that we've generated to guide the chemists to make viable edits that optimize the model's predictive power to effectively— or optimize the, um, the model will essentially predict the best molecules to work. So effectively, the probability that something goes is going to work or have higher activity, the model's predictive metric goes up. If a chemist starts making deleterious edits or poor choices, the model's predictive score goes down. Effective, the metric says don't make that. And so we're using the model effectively as a real-time guide that medicinal chemists can effectively do the design work manually, and only when they get a high score it's kind of like Foldit for protein folding. Does a chemist actually place the order for the CRO to go and make the bespoke molecule? And then do we actually test it?

So it seems like, for the company, like, have you seen a correlation between number of cycle starts and then predictive, like, predictivity of models?

Yep, so if we do one cycle, we can predict early leads fairly well. Our Predictive power is pretty, pretty reasonable. As we generate more and more SAR data points, our predictive power goes up. We've been able to go through 2 cycles, start from, you know, 10 micromolar or so hits down to 3 nanomolar IC50s. So pretty substantial improvement after just 2 iterations. And the more data points that we generate, the better the predictive power model go, or improves. And effectively, we're going to start building out the scoreboard for activity prediction, potency prediction, selectivity prediction, liability prediction, ADMeTox, so that a chemist, when they're designing molecules, they'll have a real-time score to essentially guide them towards that needle in the haystack that they're trying to find in the first place.

Cool. And that actually might make small molecules like more feasible to license, more feasible to do something similar to AppCelerator, where more, more, you know, from antibody you can predict its activity. Small molecules are a little trickier, but if you can build more predictive models of very large libraries of small molecules, you, you can enable like really creative business models. But maybe to like wrap up the conversation, you know, talk about your story, talk about 1859, and we'll do this like 2 years from now, but any kind of lesson so far especially for that emerging person, the scientist who's working at Genentech or who's at Scripps right now, trying to figure out what to do with their career? Any kind of lessons you're willing to share?

I would say never stop being curious, never stop being kind of afraid to innovate, and kind of navigate towards the solution. Don't be kind of afraid of the problems that you're going to face.

Okay, that's really insightful. Well, I've learned a ton actually. I think you've, uh, you've given me like a bunch of cool ways to view business models, which I'm always a fanatic about. So yeah, this is awesome. Seriously, this is awesome. You know, for me personally as an investor, I love business model innovation because I'm trying to minimize dilution.

Yeah.

And then on the opposite end too, I also want a larger impact across many different programs and products. But, uh, okay, Devin, this is really awesome. I know you're super busy, so thanks for doing this. I know it's gonna be useful and I really appreciate it.

Awesome. No, thanks, thanks, Josh. Really, really appreciate it. It's been fun.

Awesome.