Founding Y-Trap and Protein Engineering with Rishi Bedi

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. Hey Rishi, thanks for doing this podcast today with us. Really appreciate it. You might be the out of— we've done like 20 of these so far. You might be the person I've known the most well personally-wise than all the other guests. So, you know, be a little nervous. We might know each other a little too well here. Glad to hear that.

Glad to be here. Yeah, appreciate it. Excited to chat.

Cool. How about you do a brief introduction on you and Ytrap? Sure.

Um, so my name is Rishi Bhati. I'm the CEO of Y-TRAP. Um, Y-TRAP is a biotech startup. We're focused on developing multi-specific fusion proteins for cancer immunotherapy. So you might be familiar with immune checkpoint inhibitors like anti-PD-1, anti-CTLA-4, which have been transformative for the treatment of many challenging solid tumors, diseases like late-stage non-small cell lung cancer and melanoma, you know, that with previous standard of care chemotherapy were very, very tough diseases, very limited survival rates. They're still tough diseases, but immune checkpoint inhibitors have been transformative for some patients and really meaningfully extended long-term survival. The flip side, though, is that it remains the case that most patients in most cancer types don't respond. Indications like colorectal cancer, prostate cancer, ovarian cancer, huge disease burden, but have proven essentially completely resistant to immunotherapy despite a lot of clinical development effort and dollars from, you know, top companies in the space going after it. So it's not for lack of trying that many of these indications remain very challenging to treat with immunotherapy. So Y-TRAP is focused on addressing the reasons that we believe immune checkpoint inhibitors don't work in these tumor types. We've constructed over many years an understanding of why simply activating and expanding T cells, which is essentially what immune checkpoint inhibitors or ICIs do, is insufficient, or in some cases may even be counterproductive. And coupled with that, we've developed an engineering platform of modular protein components to build our internal portfolio of multifunctional molecules that we hope are capable of counteracting these myriad mechanisms. So that's sort of the, the Y-TRAP story in a nutshell is what we believe is a deep understanding of the reasons that the tumor immune microenvironment is resistant to immune checkpoint inhibition and the sort of accompanying engineering platform to build a pipeline of multi-specific molecules that can address those multiple mechanisms.

Okay, so on that biology side, you have, you know, essentially checkpoint inhibitors take the brakes off the immune system to attack cancer cells and hopefully clear that out. And then X number of patients respond, maybe 30% or something, or some heuristic, like some, some range like that, um, depending on the indication. Um, maybe we can dig in deeper, like what's your belief and why most a lot of patients don't respond to like a Keytruda or other checkpoint inhibitors. Is it because of the microenvironment? Is it because of the target engagement or maybe a combination of all these things?

Sure. I think the brakes analogy that you give for immune checkpoint inhibitors or rather releasing the brakes being what immune checkpoint inhibitors does, I feel like that's quite an instructive analogy. I think you can think about our thesis as follows. If immune checkpoint inhibitors are releasing the brakes, we're helping steer the car. Oh. Right? So, an immune checkpoint inhibitor is agnostic to the phenotype of the immune cell, the T cell in the tumor. So if your tumor is chock-full of good T cells, T cells that if activated would evoke potent anti-tumor immunity, immune checkpoint inhibition is going to do a great job because you're releasing the brakes on a car that's fundamentally steering in the right direction towards anti-tumor immunity. And we think those are the cases where you get good responses. But in some tumor types, for a variety of the reasons that, that we can dig into, the immune microenvironment in the tumor is not full of the desirable anti-tumoral T cells steering in the right direction. In some cases, they're steering off in a suppressive direction where activating them suppresses anti-tumor immunity. It's a little bit counterintuitive. And in other cases, their activation might in fact be counterproductive as they might do things that are tumor-promoting through promoting angiogenesis, promoting metastases, even directly promoting tumor cell proliferation. So we believe that in order to make immune checkpoint inhibitors and immunotherapy in general work in a larger diversity of patients and cancer types than has been the case so far, you need to do your immune cell activation and expansion through ICIs, but you need to couple that with blocking the factors that are steering the immune response in the wrong direction. So tumor cells have many mechanisms of skewing the phenotype of T cells, myeloid cells, other stromal cells in the tumor into a phenotype that is immunosuppressive and tumor-promoting. So simply releasing the brakes in those cases won't help. You have to release the brakes and block the key signaling mediators that are skewing the phenotypes of these of these tumor-polarized cells.

And so you have like this, you know, you have a solid tumor and it's like a little ball of cancer cells and immune cells and a bunch of other cells. And when you take a Keytruda PD-1 inhibitor, well, I kind of make, hopefully make immune cells more likely to clear out those cancer cells, but there's other immune cells and maybe they're becoming suppressive and there's other protein factors are causing that. Maybe it's released by cancer or some other, other cell. Could you describe what actually a Y-trap is, uh, and like why it's able to not only release the brakes but then like, um, what's the— like steer certain immune cells? And how does— also, how do you identify protein factors as you build up these libraries of proteins? Like Is it— I need to know which factor in which cancer type is important.

Yeah, so to start with, you know, your first question, the majority of our pipeline is— the majority of our pipeline consists of multifunctional proteins that we call Y-Traps, also the name of the company. And so Y-trap simply alludes to the shape of the molecule. So antibodies shaped like a Y, our constructs are fusion proteins where we take a normal antibody, a normal IgG-style antibody, and fuse an additional trap domain onto the molecule. So that additional trap domain is generally a naturally occurring receptor extracellular domain. So it continues to perform its natural function, that is bind its cognate ligand. The antibody continues to do its job, bind the antibody target. So by fusing these two moieties together, you now have a bifunctional molecule where the antibody is doing its job, the receptor ECD, or trap as we call it, is doing its job sequestering a particular ligand. So that's kind of the Y-trap concept. We've since evolved from that and built many different multi-specific formats and sort of expanded our tool chain beyond the original sort of Y-trap structure that others have also developed. We're hardly the first to fuse an antibody and a receptor.

One thing on the protein side, like traps are kind of, like you said, they're like since the '80s at least, like Regeneron got started with trap technology. What was the key breakthrough? I remember you, uh, uh, the Nature paper published on this work in like 2014. What was like the key breakthrough? Was it using these traps in a certain like cancer type, or was it the fact that you're using a natural receptor, or what was the kind of the key technical breakthrough that kind of enabled all this to happen?

Right, so as you alluded to, folks have been constructing recombinant proteins comprising natural traps for a long time. People have been engineering those traps as well. You know, Regeneron has a marketed product, a Flibersept, which is an engineered chimera of VEGF receptor components. So that technology and the idea of using receptor ECDs like this has been, you know, in the works for a long time. I don't think YTRAP sort of precipitated any sort of technical breakthrough. For us, it's really been about target selection, and not just target selection, but which particular targets make sense to go together in a particular bifunctional. You know, some folks develop bispecifics simply because there are two targets and you have to hit both of them. So why not make one molecule instead of two for expediency and cost-saving reasons. For us, that's certainly part of the benefit, but a huge part of the story is blocking particular cytokines and particular ligands in the immediate microenvironment of particular cells. So if we have an antibody that binds a particular cell surface receptor, for example, our receptor ECD the trap is now decorating that targeted cell in a very specific and very complete way. And we've found this to be very important since a lot of signaling in immune cells in the tumor microenvironment is very paracrine or even autocrine, meaning the same cell is producing the ligand that it is then responding to. So in order to counteract, you know, really tight signaling loops like that, it's really important to make the cell completely impervious to any response to that ligand. And we found that the Y-trap construct—

we're both sneezing here. We're both sneezing. I made you sneeze. I sneezed first. We'll edit this out, don't worry.

No worries. So we found that, you know, the Y-trap construct is really effective at allowing us to decorate target cells with ligand blockade. And we found in many cases that to be impressively more effective than even combination administration of the two agents. So it's not just bringing two targets together into one molecule, there's also a lot of selection in, which moieties you fuse together and how you get that therapeutic effect of localized ligand blockade.

Yeah, it's kind of a situation where two is maybe not better than one, or maybe one construct has some sort of advantages in a cancer— a solid tumor type. Uh, that's kind of complex, a very complicated conversation. Maybe we'll touch on in the end too. It is one observation I've had as an investor is like Some companies, especially like in cancer, there's always somebody, a KOL, that can identify what targets to pursue or like what the key drivers are. And it's based on intuition. It's based on just a vast repository of knowledge. At least for Y-TRAP, who do you refer to to get that knowledge, to like to get a sanity check on your targets, to make sure like, hey, if we're gonna go after PD-1 plus TGF-beta, Is that going to be the correct approach in this cancer? As a founder, who do you refer to, to ensure that you're making the right decision? This is the first step for drug development, and it could lead to massive ramifications down the line.

Yeah, sure. So this might tie into talking about the origin story of Ytraps. This might sort of naturally segue into that. I guess my first comment on your question is that at least Y-TRAP thesis is to not outsource that work. We feel our fundamental contribution is in the identification of these targets and the construction of these specific combinations of targets in these multi-specific molecules. So for us, that's very much our differentiating secret sauce is the biology and the specific therapeutic approach that we've chosen to pursue. So Y-Trap, the company, started with the work of my father, actually. He's, or he was for a long time, an oncologist at Johns Hopkins. And, you know, over many years had built up in his group a deeper understanding of how tumors are able to reshape the tumor immune microenvironment in this way that I described to be suppressive with respect to anti-tumor immunity and hospitable to tumor growth. So, you know, our scientific story and fundamental beliefs about tumor biology and immunology really stem from insights that he had in his academic lab. And that was definitely beneficial for the company, you know, for those ideas to be able to incubate and, you know, mature over many years, not on a startup timescale, but sort of on an academic career timescale. So that was, you know, certainly helpful and allowed Y-Trap to kind of, when the company side got started, to hit the ground running in a sense with some already, you know, relatively strong beliefs about the right way to address the tumor microenvironment. The other piece that I would say is that, you know, you mentioned the idea of PD-1 plus something. And I think that's where a lot of the field has gone. That's, you know, the thesis for a lot of companies is, you know, have your new one target. Combine it with anti-PD-1, and you're off to the races, partner with Keytruda, and see where that takes you clinically. I think our hard-won lesson has been, and perhaps the field's hard-won lesson in the last several years, is that it's not that easy. And despite the phenomenal success of PD-1 as a single agent, bringing in a second thing and sort of throwing it on with PD-1 might make some theoretical sense and you might be able to convince yourself that you've found the critical piece that PD-1 was missing. If you just bring your one agent in with PD-1, it's gonna be great. I think the clinical results of the last few years have unfortunately spoken for themselves and suggested that that's probably not going to be the case. There's not going to be a second magic bullet that sort of slides right in with a PD-1 and just solves the problem. So, we've come to the conclusion that you really need to hit multiple targets. And, you know, correcting the polarization in the tumor microenvironment is not a one-target job. And that's why we build multifunctionals. That's why we believe that our optimal clinical regimen will actually be a combination of two or more multifunctional molecules. So it's maybe a little bit of a tough pill to swallow if you really want to believe in your one target plus PD-1, but I think the right way to approach it is really what is the set of targets that are critical and how can you build the smallest number of molecules to address that multiple or that set of multiple targets. And that's where, you know, we've landed on thinking about that is rather than, you know, putting all of our eggs in a single target basket, making sure that we are hitting the multiple, sometimes redundant or synergistic pathways that are really working together to create this problematic tumor microenvironment.

Great, that makes a ton of sense. That's fairly unique. And it was fairly unique a couple years ago when the White Shark Vet started, given IO boom was still kind of in full war, even in 2018-ish, '17-ish. Maybe we can talk about the White Shark story and your story. And how you got to build a company. I think when I first met you, you were director of machine learning at System One, now called Herophilus. Great name, Saul. I think Saul probably listened to this, so great name. And you were also building Y-Trap, so you had like two jobs, very stressful, a lot of work. I was like, wow, how is Rishi doing all this? It was amazing. Maybe we could start off, like, how did you get involved in science? Were you always entrepreneurial? What kind of formative experiences that maybe set you up to lead and build YTrap?

Sure. Yeah, absolutely. Happy to talk about that. Happy to talk about that. Talk a little bit about Herophilus as well. Really respect what they're building.

And you're also a distributed bio, and Jacob Glanville is awesome. So that's a whole nother experience too, where it's like you got exposure to venture-backed and you also got exposure to— would you call it distributed bootstrapped? Or I don't know, distributed is more like— is it a whole different approach to building? You got the full flavors of building biotech companies.

Yeah, I think bootstrap is probably fair to describe distributed bio. We can talk more about it. I would say beyond just bootstrapped, I think that doesn't quite do it justice. It's really an incredibly long game played by their original founder, Jake Landau. And I think now he's really seeing that long game play out on the end, and it's actually quite remarkable to see the whole picture. But maybe we can come back to that.

I'll talk to Jake up some other time. I'll do that.

Jake, yeah. Right. So, okay. So maybe we start at Stanford. I studied computer science and biology, focused on machine learning methods, AI methods. Just one quick question.

You go to Stanford. Did you want to do biology? Is it your dad's biology? Or you're like, screw biology, I'm going to do social networking. You know, I'm going to do AI, or doing CS and bio, was it part of your master plan?

It's funny, it was part of my master plan, or it's what I sort of had envisioned, but I didn't really know what it meant. And I certainly didn't appreciate when entering Stanford the extent to which computational tools and AI machine learning tools specifically would be just critically important in various aspects of biology research and drug development. So I think I had a very hazy idea that it was something that was exciting to me, and I liked the story, you know, at least as I sort of had it hazily put together in my mind. But I didn't really know what it meant or sort of how it would be reduced to practice. Um, so I think I had a couple of, you know, formative experiences at Stanford that really showed me what it actually meant. And, you know, kind of conveniently for me, um, what it actually meant was exciting, and I liked it. So the this sort of vague idea that I had of what I wanted to do translated, I think, quite well into what it ended up being, or at least I liked what it ended up being. So at Stanford, one, I was super lucky to meet Jake Linville, who we talked about briefly. He was a graduate student while I was an undergrad. I walked into a graduate clinical immunology seminar. I think I was pretty easily identified by the program administrator, Maureen, as an unmoored undergrad, but with his heart in the right place. So she spoke with me for about 45 seconds, and I think, and then immediately concluded, you need to meet Jake, and rushed me over to meet him. And we worked together for a number of years while I was at Stanford. I worked with his, first company, Distributed Bio, worked on a bunch of projects with him, and he was incredibly generous in his time. And, you know, now I look back on that, and that was certainly a formative, you know, research experience for me, and definitely learned a ton about protein engineering and, you know, the space that we're in.

So you were Jake Glanville's, like, assistant? Like, he's in Mark Davis's lab, and then Jacob got like 20 papers in grad school. Jeez, that must have been a lot of work, Rishi.

Uh, yeah, I was not involved with the vast, vast, vast majority of what he did. We sort of had our, you know, um, side quests.

Uh, one thing in college— okay, right now I'm looking at you. This will only be a recording, not video. When did you start growing your beard? Because when I was in college, my— I started growing a beard around like junior year. I was a late bloomer. Did you enter college with a beard? Or did that kind of grow over time?

So I went to a pretty buttoned-up high school. The old school.

Pretty preppy, Rishi.

Not a religious, but yeah, definitely a secular preppy school. And one of the sort of strong commandments was no beard. And I think I would not characterize my college experience in any meaningful way as rebellious, but I guess growing a beard as soon as I got to college as my one small act of— my small act of—

Okay, she's from LinkedIn research, valedictorian, right? Okay, another personal— what was your speech about? What was your valedictorian speech at Gilman School? Was it just like, you know what, these rules stink? Was it very anti-authoritarian or was it more like buttoned down? I did a valedictorian speech. I didn't know that thing was going on. And my speech was really boring. Yeah, I had a lot to say, but I just like, you know what, I'm not gonna say anything. I'm just gonna just, you know, do 5 minutes and walk off.

This was interesting, you know, I think I was well coached by the faculty member whose job it was to, you know, sort of coach the person giving the speech. And I think he accurately pointed out to me that in many ways the purpose of the graduation ceremony is really for the parents, not the graduates.

Oh, okay.

And I think, disappointing as that might have been to my grandiose ideas for what I wanted to talk about, I think that was good advice. And I think my speech generally reflected that, which is really, I think the point is a demonstration of gratitude to the many people that have brought the class to that point, parents, teachers, the sort of broader community. So I think that's probably sound. I will say that, Our school also had a tradition where we had to give senior speeches. And that was really an opportunity to talk about whatever you wanted in front of the whole school. And that was fun. I talked about curiosity.

Cool. Rishi, you're just building the tapestry of your story, that you're creative, a little subversive, but you know how to follow the rules. So you're a Stanforder, you work at Jacob, you're doing all these projects. Was there some sort of like, what was your first exposure to startups? You're at Stanford and this is like, yeah, sure. This is when like Facebook's huge and everybody wants to do social networking and everyone wants to make an iPhone app. How did you think about startups? How did you buck the trend almost? Because most people are, when I was in college, everybody wanted to do an iPhone app. Or some sort of like Uber for X. Very few people were like, oh, biology, life sciences is interesting. Now it seems kind of too popular right now, but at the time in college, what was your first exposure to being an entrepreneur?

Yeah, I would say I kind of got into startups through sort of a back door. I have huge respect for the very entrepreneurial people who sort of go to college maybe even only for a year or two and then figure out what it is they want to build and just go fucking build it. I think that's That's impressive. And, you know, very cool for people that have that kind of clarity. For me, I didn't feel like I had that. I felt like I was interested in, you know, I think that in areas that people, you know, now classify as like deep tech, right? I think I was interested in sort of digging deep on, you know, challenging technical problems. And I didn't feel like I had the experience, you know, after college to, you know, go start a company. It seemed like an appealing future thing to do, but it was certainly not part of any sort of master plan. And I think as most people interested in biology figure out pretty early is if you want to do something exciting in the space and you want to really have a deep enough understanding of the science and the engineering in order to do something useful, you probably have to get a PhD. And, you know, totally open to that. It seemed like a very reasonable path. It seemed cool to spend a few years, you know, doing self-directed research. So I was sort of on the PhD train. I had finished my master's and I was about to start the PhD at Stanford. I had a really good time working, you know, as an undergrad and as a master's student with Ron Dror. Just a great person and a great mentor to me. Learned a ton about structural biology and protein biophysics working with him. Also learned a ton about machine learning working with him and one of his graduate students, Rafael Haunschin, who has also started an exciting company, Atomic AI, that maybe you'd be interested in chatting with him about. Anyway, all of this is to say that I was sort of on the academic train and pretty excited about the PC life, but also wanted a little bit of a break. I'd been at Stanford for 5 years. I was about to sign up to be at Stanford for 5 more years. So I thought maybe a little interlude would be beneficial. So I was gonna take 6 months to work at System One, now known as Herophilus. This neuroscience drug discovery company in Dogpatch in San Francisco. And when Sean and Saul, the founders, pitched me what Herophilus was, it was this scrappy startup growing brain organoids to do better data-driven, ML-first drug discovery in neuropsychiatric disorders. One, it just sounded like mad science to me because I had very little background in stem cells or neuroscience. Just seemed like crazy futuristic future of drug development, was stoked to check it out. So I joined for 6 months, but then 6 months became a year and then a year became 2 years. And I sort of fell off the going back to grad school train. I think they were very progressive as a company in not being very credentialist. I think a lot of places are very fixed on what people with a PhD can do and then what people without a PhD can do. And I think they both came from interesting academic backgrounds themselves, so sort of had appreciation for the fact that it didn't have to be that way. Saul founded two companies before going back to grad school for his PhD and then is now a professor at UCSF.

Saul's a crazier story. Saul's like two venture-backed startups, one backed by Sequoia, and then this final— and this is like the dot-com bubble, all this stuff. And then he goes back to grad school as like, what, a 40-year-old, a 35-year-old or something, and then becomes a professor at UCSF, then starts another company. It's like, it's— I'll do one of Saul one of these days. Saul's story is nuts. One quick question is like, you know, you're on this academic track, joined System One, what was your dad or family— do they still want you to get a PhD? Or what's the— like, what was that pressure like and how did you deal with that?

I think I'm very lucky to have very understanding parents who are supportive of whatever I do. So, you know, yeah, my parents are both physicians, but I don't think I ever felt pressure from them to go into medicine at all. I think in fact, you know, being inside something, you are intimately familiar with the trials and tribulations that it brings. So, you know, I think, I think, I think for them seeing me get into biology and into medicine, you know, of course they were pleased, but they also sort of knew what I was signing up for. So I think, you know, they were always very hands-off in terms of letting me make, make my own decisions there. So I was, you know, lucky to not have to deal with that.

What came first? Did Ytrap get started or did you join System One and then Ytrap got started?

Right. So this was all sort of happening around the same time. So technically Ytrap had been started earlier. My father had started the process of spinning the company out, licensing the technology, the initial technology from Johns Hopkins. So that ball was rolling and You know, for me, for me personally, it was sort of something to work on the side. You know, I was, I was helping where I could, you know, helping with, you know, the, the, the, the protein engineering side where I could, helping with, you know, other sort of, you know, we were all learning for the first time, how do you start a company and how do you build that from scratch? So it was fun for me as sort of a side hustle to sort of help out where I could and learn about that process since it was cool. And it was kind of exciting for me to have a front row seat to, you know, getting it off the ground.

Um, what was it? So like, uh, White Chops getting spun out, you're getting your job start System One. Was there a moment where like your dad's like, hey, can you help me out here? Was it like during Thanksgiving, you're like, man, I'm trying to figure out this licensing agreement, Rishi, any thoughts? Like, was there a moment where like, hey, hey son, join the family business? Like, what, like, you know, like, was there like a moment, a moment like that? Or was Or did you just kind of just, did you raise your hand, say, let me help you?

I would say it was decidedly not cinematic, right?

I think it was like Godfather or something like Michael.

No, I think it was a very natural and gradual progression where it started off as one-off favors or brainstorming chats and then gradually evolved being regularly spending an hour or two a day to regularly spending, you know, all my evenings to, you know, regularly working the equivalent of two full-time jobs. So I think, you know, it was sort of a very gradual escalation. It, you know, it never felt like there was a, you know, turning point. I guess, you know, one sort of formative transition point for the company was in 2019, we were able to execute a licensing agreement with a top 20 pharma company. And I think that was a huge moment for us as a company and also for me to sort of realize that I wanted to commit 100% to building Ytrap. So it all felt very natural and organic, but after that in 2019, then in 2020, I left System One and was working on Ytrap., full-time after that.

So two quick questions. First question, always amazes me. How did you like, okay, you're doing a full-time job at System One and then you, you know, Y-Tracks at night where you're helping out, helping out the company out. Um, how did you like get sleep? Did you get sleep? How did you like maintain your, because you did this like a year or two. Um, it's almost Elon Musk-esque or something, you know, like Any tricks there? I'm trying to get more productive, so maybe help me. How did you figure out how to do 2 jobs at once and do a really good job at it?

Yeah, I would say it probably wasn't the healthiest time of my life and I probably wouldn't do that again. Okay. I think after I left my day job and was just working on YCharts full-time, I suddenly realized that I can have hobbies again. You know? And so I would, I would say it was not a, it was, it was, it was super productive and fruitful and it felt like the right time in my life to like be super buckled down and, you know, just spending all my, all my time working. But I also would not recommend it as a strategy, or I also certainly wouldn't suggest any of my productivity hacks as being sustainable or beneficial. For anyone. So, you know, maybe I'll respectfully decline to give advice besides just saying that it's just, that makes sense.

Maybe the takeaway is only do stuff like that. Only work 2, 3 jobs if it's necessary, you know? Yeah.

And that's exactly right. It was that I felt like I had to, I really loved what I was doing at System One. I felt like I was learning a ton. I felt like what we were doing there was super important, but I also felt like, you know, the work that I was doing on the Y-Trap side was super important and had to be done. So it was absolutely, um, consequence of necessity.

It makes a ton of sense. And then another part too, around 2019 you do the licensing agreement. What made you want to go, I say, I'm going to join full, I'm going to do Y-Trap full-time? Did you feel confident? Like, was there like, you build the skill sets, that certain knowledge base, but like, hey, I kind of understand how biotech companies work, I think I can take my own shot? Or was it more like, oh, we did this licensing agreement, we need somebody to do this full-time? Was it internal or external in terms of the motivation to go full-time?

That's a good question. I'd say it was both. I think it was really those two factors coming together, right? It was that I felt more confidence that we could pull it off and that I could meaningfully drive the effort. But then from the external side, it was also, I mean, you can't You can't build a company where everyone's part-time. You know, that's just, it's just not a, it's just not a thing. You know, I think biotech is cool and it's kind of fun to talk to people who are not in biotech and sort of have them appreciate how lean biotechs can be. And I think that's an interesting distinction, you know, for most tech companies where early on it seems like, you know, headcount growth is viewed as a meaningful proxy of company growth. It's kind of cool that in biotech that's not the case. And I think that allows for a more diverse set of company formation and growth stories, which I think is neat. But at the same time, you do need some critical mass of folks who are waking up and going to bed every day thinking about just that company. So I think it was necessary for the company and that sort of paired with my personal belief that it was the right time for me, you know, as an individual to do it.

I'm in the family business, Rishi. Come in the family business. You know, it's like, you know, your son's going to be in the business too. It's like, come on, come join Y-Trap. You know, it's going to be like, it's going to be like the Tatas or, you know, like a family. It's going to be like Valencia. It's going to be like one of these family businesses. But could you talk more around most biotechs, have like two acts, at least initially, technology to drugs. And some companies spend a lot of time on the platform and then make drugs hopefully at the end. Some companies are only drug companies. Uh, how did— when you did start joining full-time, did Y-Trap— did anything substantially change in Y-Trap in terms of its like tactics and How did you make that transition from technology to then actually developing drug candidates?

Yeah, that's a million-dollar question, right? So I think initially, Y-Trap, we had so many ideas. Our sort of modular engineering platform let us try a lot of things. There were a lot of possible ways to construct these molecules, and we kind of had like a kid in the candy shop fun. It was like, there are all these things that we can do. We have all these ideas, and we want to try all these things. And that, I think, is a reasonable stage to go through as a company. But at some point, it became clear that especially as we went out and talked with more prospective investors, more prospective licensees in pharma, We needed to have a clearer vision and we needed to tighten up the story and we needed to have a strong belief, not just on what large pipeline of molecules we hope to someday develop, but what are the first 1, 2, 3 things that we're going to develop? And do we believe that those first 1, 2, 3 things are going to have a meaningful impact, not just generically in immuno-oncology, but in a specific tumor indication, given the current clinical landscape, given the current clinical trials being offered, given the current other players in that space. We just really needed to boil it down to what is our treatment regimen going to be in a specific indication or specific set of indications, and have we built the pipeline to support that specific treatment regimen? So I think certainly in the last couple of years, and especially so in the last year, I think the phrase that my father likes to use is, we just have to get real. We just have to be very— and that's not to say the work we were doing before wasn't real, but we had to get real in the sense of making sure that our lead assets in our pipeline were directly in support of our near-term clinical objectives. And viewing everything through that lens is just a huge forcing function in how we spend our time and how we spend our money and in what scientific bets we make.

So how did you figure that out? So outside looking in, if nobody knew Rishi personally, knew you personally, I think, oh, who's this kid? What does he know about making drugs?

Right?

How did you actually internally build up that repertoire of knowledge, of different skills to be able to even know what questions to ask to develop a drug? And who did you lean on and stuff like that? You know, and your father, probably other people in your network. And then how did you, when you went full-time on Y-Trap, you know, how did you determine how to allocate your time? Because, you know, before you did this one, Y-Trap, your schedule was packed and you just had to kind of— you had scarcity of time. Then when you went full-time on Y-Trap, probably felt like, man, I have all this time in the world, what should I spend it on? And how did you determine as a CEO or the founder of the company that, like, allocate it to, you know, identifying which drugs are actually going to— how did you set those new milestones? How did you spend your time to figure out if these assets you're building are compelling?

Yeah, um, A lot to unpack there. So your first question of how do you build up knowledge base to figure out how to do drug development, right? I think there are kind of two pieces to that. One piece was the biology piece and how did I and we build up our knowledge of oncology biology and that, for me, there was just a ton of learning and I was very lucky and excited to get to undertake that learning process with my father, who of course has decades of experience in the field. And so on the biology side, I felt like that's— we just simply spent a lot of time on it. I spent a lot of time getting up to speed and then we both spent a lot of time digging deep into the literature, digging deep into our own data and just thinking really hard. I think there's just no substitute to deeply engaging with the data, deeply engaging with the scientific literature, and spending a lot of time doing that.

Do you have like a secret tool? Do you have like a master spreadsheet of cancer targets? Do you have like a booklet? Is there some sort of secret like, like thing, uh, uh, tool you have to like kind of ingest all this knowledge and, you know, kind of keep track of it?

I, I would say no secrets, but, you know, I'm definitely a, a tool user, the two things I would recommend to folks who are, you know, trying to assimilate a large knowledge base of papers— I really like Zotero as, you know, a paper manager, and I really like Notion for taking notes on papers. And, you know, there's an aptly named integration called Notero for integrating the two, and I found that to be, you know, quite effective. Um, there's no sort of a magic bullet there. I'm not like using GPT-4 to like just synthesize all of the insights for me, though I have found, you know, when I have a particular question and it's sort of, um, something I could go and spend 10 minutes looking up, I would say a very substantial fraction of the time I can punch that question into, into, uh, ChatGPT and get, uh, you know, if not an answer, at least a good lead on how to get the answer. Yeah.

I bet portfolio companies use GPT and similar tools to like get good prompt— prompt it and get like a good response in terms of like a plausible explanation for something. Like, what's happening here? If these genes and this disease, what could be happening? And it's, oh, in this pathway it's causing this, this phenomenon. I think, oh, it sounds plausible, it's worth, worth the leap. But okay, so you built your knowledge base, and how did you like spend your time, and how did you know How did you know at 2019, like, hey, for Ytrap, I should spend more of my time developing products? I think most founders, they would say, oh, let me scale my platform, let me build a larger library, maybe focus on licensing more. How did you know instinctually, or maybe you got advice to say, hey, let's build a pipeline now, this is the time, versus just doing another licensing agreement?

Yeah, great, great question. I think we got some clarity when speaking with other prospective pharma licensees. And this is something that I think is conventional wisdom. So I think probably many people would say this. I guess we just chose to go out and learn the lesson ourselves. The value inflection point for a clinical stage asset— sorry, the value of a clinical stage asset is just inordinately more than a preclinical asset. And for us, it just became very clear that the right path for us was not to spend a lot of time in protracted discussions and diligence with pharma companies, which we did do for a good portion of that time. And we, I think, just realized that the right path forward was for us to develop deep conviction in what we wanted to bet on. And go raise the money to bet on it. I think there's an alternate universe where pharma companies were more engaged on preclinical licensing, and maybe in that alternate universe, more companies can be built off of pharma licensing revenue, as opposed to raising money to take things through Phase 1 themselves. But my view is that the world as it is is not that, and the right thing to do for us, and I would suspect for a lot of biotech companies out there is to bring your assets into Phase 1 and then engage the licensing partners. Or at least, even if you've engaged them earlier, that's really the point at which the talks should get serious.

I think you're right. You want to do licensing once you have human data, then the numbers get a lot bigger. Licensing too early, there's a risk that you give up a really good asset too soon, and in retrospect it looks like a dumb move. So you speed up time till now and you build up Rytrap, get a lot of data, you have a lot more net— this is not the setting to make announcements, so let's not announce things, but you'll announce them down the line. Are there like lessons you've learned and Have you changed anyway since when you went full-time on YTRAP? Like, has your perspective on drug development changed, biotech, like, you know, versus when you started versus now in terms of what's important, how to make decisions, how to build teams, you know, kind of the reality of biotech versus the ideal of it?

There's a lot there. Okay, let's see.

Maybe any lessons you've made? That's too much of a— I've been reading this, I've been rereading this Clausewitz or whatever. There's a book called On War.

Clausewitz, of course. Yeah, I took a political science class in college where we had to read Clausewitz.

So I'm rereading it and I'm just getting influenced by this ideal versus reality. And a lot of my— the things you read like in any moment influence your thoughts and questions. So I'm kind of being influenced by Clausewitz to that question. It's like, what's the ideal of biotech and what's the reality of it? But maybe in general you can talk about any key lessons you've learned over the years building Y-Trap.

Yeah.

Sorry. Okay.

Just so like, there are like a whole bunch of things that I can talk about here. Do you want me to just like, I don't know, I think, I think nothing too spicy.

Let's do, uh, like nothing too spicy. But how about, uh, if like you had to— specific area that other people, you know, PhD students, people who are new to biotech and they want to start their own company, based on your experiences, what kind of advice would you give them?

Yeah, you know, I'm certainly early on this journey myself, so I sort of, I'm happy to share thoughts that I have, but, you know, I would say these should all be categorized as strong opinions, weakly held. You know, that's a framing that I like, is, you know, I'm certainly open to overturning these beliefs in the face of more data, and certainly encourage folks to to do the same. One framing that I found to be helpful when making a bet is what kind of risk you're taking on. And I think for us, you know, you can sort of, or for us, or really for anyone, I think you can categorize risk as ology risk, right? That's like, you know, does your antibody format work the way it is intended? Does it have the right, you know, developability, pharmacokinetic, tox properties, you know, to, to, to be an effective therapeutic? There's biological risk, which is, do your model systems appropriately recapitulate human disease biology? I don't think anyone advances a drug with no preclinical evidence, but the question is how well does that preclinical evidence translate into the human disease setting? And the third category is operational risk, which is, to put it bluntly, in biotech, not fucking it up, because there's sort of so many hoops you have to jump through. And I don't mean to criticize any of those hoops. They're obviously important steps that have to be taken in order to feel like a drug is suitable for use in humans. But the fact of the matter is that there are a number of operational hoops that have to be jumped through, and failure on any of those hoops can be catastrophic for a company. So I think it's helpful to frame the company and the choices you make in sort of this risk framework. And I think what we've found is we want to put all of our eggs in the bio-risk category, right? We want to feel very confident in our technology. We want to feel very confident that we're doing things operationally correctly. But the thing that we're never going to know until human trials is that translatability. If we're seeing really strong mirroring data, is that going to translate into clinical efficacy? Framing it this way is also helpful because it helps you focus on how to mitigate a biological risk. So we try to be very self-critical about our target selection and our proposed mechanisms. We try to frame questions as, What is most likely to make this approach not work? Right. And that's directly instructive for us because that's how we determine our pipeline, right? We sort of start with the seed of the strategy, a molecule 1, and we say, okay, what is most likely to stop molecule 1 from working? That directly admits theories on what molecule 2 should be and so on. So I found that, that framework to be helpful. And I think that bio risk is sort of the right place for us to put our eggs.

I mean, just, I think identifying what risk you're taking and then manage it. So whether it's biological risk, technical risk, or some other personal, some other risk that's, and just be able to manage it and just knowing what you're getting yourself into. So you're not like, you're alluding to drugs, You can't fool yourself because there's these really predefined steps. And if you can mess up on one, all the past work kind of isn't worth it. So you have to be really, really, really careful and measure a bunch of times. It makes a ton of sense. Maybe like— Yeah, sorry, go ahead. No, go on, sorry.

I was just gonna say, you know, I think that's sort of a good way of thinking about things in the startup setting. I think there's a sort of parallel framing of the taxonomy of academic labs and taxonomy of how academic research can spin out technology. I think you could view most groups as being either target-centric or technology-centric. When I think of target-centric, I think of Jim Allison and CTLA-4 or Richard Flavelle and TGF-beta, where these folks are sort of pioneers in particular target biology and they've sort of built up a deep knowledge base and a deep set of contributions on how these individual pathways contribute to disease biology. Then there's technology-centric groups where, you know, are agnostic to specific disease mechanisms, maybe even agnostic to diseases of interest, but have brought forth some fundamental technology that allows revealing the mechanism of disease in a novel way. I think of Aviv Regev's group and single-cell RNA-seq as being an example of a group that has really bet on single-cell RNA-seq and has been able to exploit that technical knowledge across a variety of areas. The third group I think is less common, and I think it's what Y-Trap emerges from, and it's what my father's lab emerges from, which is a synthesis-centric academic lab where it's not about betting on deep discovery around a specific target. It's not dependent on being the first or among the first to bring a certain technology to fruition, but it's about synthesizing the vast literature out there and developing disease-oriented theses on how it should be addressed, and then constructing a translational research program to test very specific hypotheses in a very unbiased way. So I think that's, you know, it's not quite the same as thinking about startup risk, but It is, I think, sort of a parallel view of how academic labs might see themselves and see how they interface with spin-out startups.

That makes some sense. Yeah, I think also academic labs, I mean, some people are trying to, they get really narrow and deep in terms of a pathway or a set of genes. And then there's tool developers and sort of trying to identify, like trying to do, doing both is really hard because it requires different sources of talent where—

Right. I think that's right. But I guess maybe my punchline is that I would also encourage people to do neither, right? And sort of, you know, that sort of a third group where you're not making a bet on going super deep on a specific target. You're also not making a specific bet on, you know, deeply developing a particular technology. But you're instead making a bet on your ability to synthesize the insights of the field and construct treatment regimens based on that is I think something that also has a lot of value and is kind of where YTRAP and the academic research that it spun out from sort of sit.

That's particularly true for like areas where there's like decades of work and maybe you often see this phenomenon of old technologies being ignored or forgotten. And even mRNA, all this stuff, it's like, oh, it's just going back a few decades and then bring it to the modern era. Yeah, I think a lot of value in synthesizing. There's a lot of value in just knowing the history of a field and then be able to find an old idea and then use it. That's something I've observed as an investor. Sometimes the best platforms or ideas are just old and they might sound new or no one's ever talked about them, but they've been discussed and implemented for decades. And so any final— what's the last question? Any final thoughts? Like if you're entering college right now, maybe entering grad school, any advice in terms of like skills to build up, things like that.

Yeah, great. I think having literacy across domains is incredibly valuable. And I think we just sort of alluded to that a moment ago with the idea that having literacy, not just about the current forefront of the field, sort of the history of what has been tried before, and what hard-won lessons people have learned, in the field over decades is hugely valuable. So in that sense, having broad literacy is crucial, but also even in the present, having literacy across technical domains is a pretty valuable place to be. For me, the obvious example of this is the value of having literacy in both machine learning and computational tools, as well as in biology. And I think I've seen how helpful it is to have folks that are deep in ML and deep in biology regardless of where they started from. I think, you know, I've been lucky to have mentors that fit that bill and sort of, you know, set a precedent for the kind of, you know, dual, dual scientist that I want to be. And I think it's luckily becoming easier and easier for folks to do that. I think the barriers to entry to gaining computational literacy and data literacy are falling and have fallen a lot. So I think it is a very good idea for folks in the biospace to build up some of that computational ML literacy as early as they can. And I don't know if this is the audience of the podcast, but conversely, my experience has been as someone who was primarily trained in computer science, that there's a lot of fun in getting into the biology space and getting a deep, not just a superficial conversational understanding, but really developing a deep understanding of the bio side has been very, gratifying for me, and I would suspect that would, that would be the case for other computational folks who are interested in, you know, getting, getting deep into biotech.

Yeah, I think on the bio side, you got some great mentors. That's one thing you got lucky with. Definitely didn't get lucky with, but just like walking into your room and you get introduced to Jacob, you kind of walked your way into some really great biology mentors. And I think that's probably the key thing. If you can actually do physical work in a laboratory, you need to find the people who set the standard and Jacob is one of the best immunologists in the world, so you're extremely lucky to just walk into that room. That's probably one of your key moments of your career, going to that talk.

Absolutely. Yeah, I feel super lucky. Obviously the luckiest part for me is my father, right? I mean, that's why I had I've had many opportunities of serendipitous luck in meeting folks who I'm not related to being great mentors to me. But I've also been very lucky in that my original great mentor has sort of been around my life for a while.

You're lucky to be born, and you're also lucky that your parents didn't try to make you a doctor or force you to— allowed to be subversive too. Like the tapestry of Rishi, you just kind of You know, you know how to follow the rules, but you know how to write outside, color outside the edges once in a while and be creative. Well, thanks for doing this. This was awesome. We'll catch up soon. We'll record this 2 years from now and you'll have more announcements, but I really appreciate you taking time to do this.

Yeah, man, absolutely. This was great. I really appreciate it. Had a lot of fun. Happy to, happy to do it again.

Absolutely. Well, have a great day, man. We'll talk soon.

Yeah. One sec. Sorry.