[00:00:15] Speaker A: Welcome to episode 33 of biorassi's few and far between podcast. I'm your host, Chris O'Brien. So where does AI. Fit into drug development? We hear a lot of talk about AI in the life sciences these days, but are more practical applications here around the corner or still off in the distance? My guests today learned to embrace technology solutions at a young age, moving from early animations of photosynthesis in high school to exploring beyond the boundaries of computational biology and protein modeling. Dan Mandel, CEO of grow biosciences, continues to apply his creativity and imagination to protein design, expanding on the structure of genomically recoded organisms. Or grows from George church's lab at Harvard medical school and fine tuning the amino acid alphabet into protein based therapies to treat incurable diseases. In addition to AI. We'll also discuss some of the questions that today's biotech entrepreneurs need to ask themselves before taking the plunge into their year zero in the industry. I hope you enjoyed this episode. I think this is really one of the most exciting stories I've come across in biotech, and Dan and I had a lot of fun discussing his personal philosophies of porc Nolos dos or why not both? And the concept of building the seatbelts before you build the car. Okay, let's start the podcast.
Dan, welcome to few and far between. It's a pleasure to have you on the podcast today.
[00:01:45] Speaker B: Thanks, Chris. It's great to be here with you.
[00:01:47] Speaker A: We were chatting just before we got started. Producer Adam said, guys, don't use all the good stuff before we start recording. So, you know, I think our dynamic here is going to be just fine, which is really great. I thought I would start us off, if you don't mind, with asking a little bit about your backstory. Every superhero has a backstory. You have an interesting one. You didn't start out aiming for biotech, if I remember correctly, as a child, what did you think you were going to be doing?
[00:02:11] Speaker B: Yeah, I guess my interest in superheroes was, how do you make it look like they're flying? I really wanted to get into special effects as a kid and grow up and work for George Lucas, and I actually spent most of my science projects as a young boy and as a teenager making movies. So I started off doing stop motion animation where you kind of like take a frame of a lego one at a time and move it along. And then by the time I was in high school biology class, I was trying to animate photosynthesis, and instead of doing the requested assignment, I would try to imagine what it looked like to fly through a cell and expose other students to my vision of what that might be. So I really loved that, and it got me into computation and thinking about the inner workings of life and the forces that really drive it. But I also found that I wasn't a terribly talented artist, and I watch those movies if you and I hang out sometime. They're good for a laugh. They're not going to land you a job at Industrial Light and Magic.
[00:03:02] Speaker A: Well, I love the concept, and probably if you were doing it today, I don't know, maybe you'd become a YouTube TikTok phenom and we wouldn't have the benefit of you in, you know, fortunately, that didn't happen.
[00:03:13] Speaker B: Well, I love bad movies as one of my passions, and so I think we could start up a bad science video channel where people look at what people did, try to animate it in the 1990s, and have a lot of fun with it.
[00:03:23] Speaker A: I'm all in for bad movies and for the bad science animation. That sounds like a YouTube channel already. Look at all the stuff we're creating. Okay, so you start out thinking, when you get to college, are you still thinking you're going to be doing computer graphics, or have you shifted by then?
[00:03:38] Speaker B: I was studying AI and machine learning when I was in college, and it was before it was being applied very much to biology. And a lot of it was the early days. It was not at all cool. Not only did we not have Siri, we were sort of just being able to do voice recognition in a mediocre way. That was kind of where we were at. But there was some very interesting work getting done in modeling things like transcription networks, trying to understand how genes interact with each other. The field of systems biology, thinking about a cell as a collection of systems was just coming together, and so I found that fascinating. But quite frankly, I hadn't really studied biology, chemistry, or physics in college. I studied computer science, philosophy, psychology and linguistics, and a lot of AI in there. So it wasn't really until I went to Scotland, actually, where I did my master's degree, that I started modeling biology using these tools. And I tried to do a project.
[00:04:28] Speaker A: When are we in time now, Dan? When is this?
[00:04:30] Speaker B: This is 2002, 2003. So this is a full 20 years ago. Wow. Okay.
[00:04:34] Speaker A: Just fascinating to me that you were studying AI during what I think was the AI winter, right when things were a lot of people thought that landscape was really bleak, and it sounds like you were also ahead, in a way in thinking about applying computer science and biology in general. Did you feel like you were a voice in the wilderness then, or tell us a little about that.
[00:04:52] Speaker B: Yeah, you're right. The field of AI was kind of in the midst of its so called winter. There had been a number of high profile failures and some what seemed to be fundamental limits to what you can do with AI that were then later disproved. But it was a time when people were kind of who were doing it, doing it because they were fascinated by the underlying abilities, kind of the ability to reason in uncertainty. And that's really what modeling biological data is. It's taking a lot of noise and pulling out signal in a way that's predictable, quite frankly. And so when I started doing it, to your point, my project was pretty misguided. And I don't think there was really anyone around me who could have pointed that out. I was trying to take protein sequences and then run them through a machine learning tool and predict whether or not they're going to interact. And I didn't know what an amino acid was. I didn't really know what a protein was. I knew nothing of protein folding. These were simply encodings of chemical properties like hydrophobicity or isoelectric points or charge. And I ran them through machine learning tools, and I got answers out, but nobody was going to test them. And they were very detached from any fundamental biophysics. Like I said, there wasn't anybody to tell me that that wasn't a great idea. And Odly enough, 20 years later, we're now doing that, right? You're seeing Generative AI taking these sequences, so that's all very much possible. But at the time, I really needed to take a step back. And the good news was I was super fascinated by these things I knew nothing about. And so I wanted to know, what is an amino acid? How does a protein fold? What are the fundamental driving forces that underlie biology? And so I went off and did a PhD at UCSF, which is one of the best places to study biophysics. And I got into this field that we call computational protein design. And in that area, we take computer algorithms and we try to understand not only how do proteins fold, but what happens if you change their underlying sequence. Can you actually make them adopt new structures? And can you make entirely new functions that we haven't seen yet in nature? And so that was a ton of fun because I was in one of the first labs where we had both the computational group where we could design these sequences and an experimental lab where we could test them.
[00:06:54] Speaker A: That must have felt cutting edge at that point. Did you feel like, okay, now I found my people?
[00:06:57] Speaker B: Like my tribe, I definitely found my people. I would say cutting edge is hard to claim because it still took years to go from any kind of computational result to something that you had proven out in the laboratory. So I don't know what the analogy would be, but it's like if you were to swing that cutting edge, it would take a very long time to make one slice through some butter. But we were doing it right. It was the forefront of the field, and we were just starting to be able to bring tools together that let us rationally engineer proteins with totally new capabilities, including those that hadn't yet been explored by nature.
[00:07:29] Speaker A: So when you're going through that program? Are you sensing, okay, I've landed in a place where there's real interest. Now I feel confident that either research money or professorship in my future or something like that, or how are you thinking about life back? Or are you just like, hey, I like what I'm doing?
[00:07:44] Speaker B: So towards the end of my PhD. I was probably 50 50 on if I felt like trying to go through a full academic career and open up a lab as a professor versus doing something entrepreneurial. And I did actually have a pretty strong entrepreneurial bug already. In fact, at UCSF, they had a really cool class called idea to IPO, where over the course of the class, every week, you would generate another component of a business plan. So it might be a financial model one week and the commercial plan the next and intellectual property the next. And by the end, you've done all the work to create a company that's an imaginary company based around a product that a bunch of people got together and thought might be a good idea from a class. But you did present it in front of a billion dollars of venture capital at the end. And every so often, somebody would actually fund a company out of this class. And you start to learn. How do you speak VC? What is a product? How do you actually identify a valuable opportunity? How do you work with a team to do that? A team of scientists who have very different capabilities, most of which are pretty raw. None of us coming out of the academy are, for the most part, are business people. So that kind of opened my eyes to it and to some of the challenges. But it also made me realize that the everyday life of an entrepreneur might be something I'm more naturally inclined to than the everyday life of an academician.
[00:08:53] Speaker A: I think that's a wonderful thing for bioprograms to offer. I think you're right. It's a big gap to cover for a lot of people who've grown up in academia and who might be interested in starting something, creating a company. Man, there's just a bunch of things you need to understand that are outside of the kind of core science skill set, aren't there? So that's a smart thing for them to offer, and I wish more institutions did things like that. Okay, so from there, do you go to harvard or what comes next?
[00:09:19] Speaker B: I did, yeah. And I wanted to go to a large technology development lab, where, again, I could kind of straddle the line between pursuing an academic career and perhaps being able to start up a company of my own. And that really led me to george church's lab at harvard med, where I felt this incredible sense of creativity and imagination, obviously very well resourced and really an attitude that there's very few limits on what you can and should do, which you have to be a little careful about, but nevertheless felt very much at home. And I had come from a very small lab for my PhD. So I felt very ready to go to an environment where I had basically no supervision and almost like a blank check to go after whatever crazy idea me and my collaborators wanted to.
[00:10:00] Speaker A: You didn't end up working for George Lucas, but you ended up working for another famous George.
[00:10:04] Speaker B: Another famous George, another powerful beard.
Yeah, they're both luminaries in their own spears, for sure.
[00:10:11] Speaker A: Okay, and then so walk us from there to the creation of, like, what's the story?
[00:10:17] Speaker B: Yeah, so when I got to the church lab, they were just finishing these new organisms that they were engineering, which they call Genomically recoded organisms, or Grows. That's actually the acronym of our company name. And I was thinking about, with George, actually a very different area than we're in now, which safety. So one of the properties that happens as you modify these organisms in the way that we were is that they become resistant to viruses. And that's a very powerful tool for biotechnology, because a lot of production organisms are susceptible to viruses. And if you get an infection, it can shut down a whole factory and can actually leave patients suffering, as was the case with a very famous event at Genzyme, which shut down productivity for a rare disease patient. And so you can make virus resistant production organisms. You can prevent these kinds of catastrophes from happening in the future. On the flip side, some people might wonder what happens if one of these virus resistant organisms escapes from the laboratory? Could it actually take over an environmental niche occupied by what we'd call wild type microbes and have an economic or an ecological excuse me problem? And it's never happened in the past, really. These engineered organisms tend to be quite sick in their own right as they're diverting all their energy to making something for humanity, typically. But on the same time, it's a fair question, and it's something that we wanted to get in front of. So we built the seatbelts before the car, so to speak, wherein we did a project to figure out ways to biocontain these organisms. And the strategy that we used actually drew from all of my past experience in computational protein design and combined it with these incredible new organisms. So the idea was, can we use design to create organisms where a lot of the key proteins that are essential for life will only fold if you put a synthetic amino acid into them? Right. And so these organisms, after several years of engineering, reached a point where if you don't feed them this synthetic amino acid, these essential proteins misfold, and the organism dies. And to the extent that we could test them, which is up to about a trillion organisms in the laboratory, we couldn't find a single one that could escape this what we called synthetic oxytrophy, this biocontainment mechanism. And so this really made us feel confident for the first time that we could engineer proteins using these new amino acids in a way that's both predictable and valuable. And where that led us with Grobio was to think about, well, now that we can do that, what are some of the really pressing challenges in the clinic facing patients that can't be addressed with the standard 20 amino acids? And how do we bring this powerful new amino acid alphabet to bear to address these unmet needs?
[00:12:49] Speaker A: That's fascinating. I pulled you forward maybe a little bit too quick. Will you explain protein design? And you just referenced the kind of core 20 amino acids. Will you sort of walk through how that works and how you guys change to the way to think about this?
[00:13:02] Speaker B: Yeah, absolutely. So protein design as a field is really asking the question of, given a protein structure, what are the amino acid sequences that would fold up into that structure? And what we've been able to do as a field is move from structures that we know to novel structures with novel functions. And as computation has gotten more powerful and our algorithms have gotten more advanced, we can now dream up entirely new protein shapes and then design sequences that will fold into those shapes. That's really what we mean by protein design. And so, up until these first new recoded organisms were engineered, we pretty much only had the 20 amino acids that life has used for the last three and a half billion years to work with. And as you can see, there's an incredible amount of diversity around us. Life has made wonderful things using these 20 amino acids, and some horrible things, too, right? But it's all built from the same 20 amino acids. And it's extraordinary, right at the same time, as engineers, we don't have three and a half billion years to wait for the next important new breakthrough. And so, as we do what's called forward engineering, where we have a design goal and we want to fit something to that goal, we need to expand that toolkit, and we need to go beyond what nature has already made available to us so that we can access powerful new modalities to treat patients. And that's really what we're doing at Growbio. So we're taking these design tools, which started off using only 20 amino acids and now adding totally new building blocks so we can dream up entirely new ways of making therapies. But how do you make them? You can put anything you want into a computer. That's where these grows come in, these genomically recoded organisms that have an expanded genetic code and allow us to put those new amino acids into proteins wherever we want. And that's the real power of the company, is we have this organism that lets us put in as many of these amino acids that are novel as we want. And we have these computational tools that let us design new therapies using these new building blocks.
[00:14:52] Speaker A: So it does feel like a science fiction story. How many new amino acids have you guys designed?
[00:14:57] Speaker B: So there are many that have been developed by the field already. When I got to George's lab, there were hundreds out there. The problem is you can't put them into a protein because there's no organism that lets you do that until the advent of these recoded organisms with any kind of efficiency. And so we already had hundreds to work with, and now we've gone beyond that. At growbio, we've dreamed up new amino acids that haven't been used before either. One of the key challenges is you can come up with a new amino acid, and you might even have an organism like ours that lets you install it, in theory. But you also need special, what we call translational machinery in that organism that lets you install those amino acids. And in a synthetic biology level, what that means is an enzyme and a tRNA, and both of those two components have to be engineered. And what's kind of cool is you can use a lot of the same tools that we use to engineer new proteins that are therapies and apply them to the process of engineering those cellular components as well. So you can kind of use the same tools you would use to make the car, to make tools you would use to make the car. Right. And so we can repurpose those and dramatically accelerate our ability to expand that amino acid alphabet. So we really use these tools both for the products and to enhance the organism.
[00:16:07] Speaker A: Dan, do you see the constraint? Is the constraint that you're addressing more about the grows themselves, the tools you're creating to insert novel amino acids or something else? Or is it kind of all of the above? Is it an ecosystem of capability?
[00:16:23] Speaker B: It's a great question, and it really is all of the above. So even when we left Harvard Med with this really cool organism, it's an academic organism. And so we had a lot of work to do to industrialize that organism to show that we can grow it at scale in a cost effective way in what are called fermenters. You have to be able to grow these to very high density at a low cost. Otherwise you can't compete with other approaches. And so a lot of the early years of the company were built around that. And we now have six years of industrial age excuse me, industrialization behind us, which are a key part of the intellectual property of the company. But as you just alluded to, that's also true for the components in those cells. Right? So the enzymes and tRNAs that I just alluded to and the products, right. So you can put an amino acid in all kinds of places in a protein as a product. That's a very large combinatoric space that you really can't screen, even with a high throughput screening setup. You need to use computation to reduce that size to something manageable. Right. And so that's another key area that we focus on, are design tools that let us take that astronomically large space of possible, even with a single protein where you could put these new Nsaas, as we call them, nonstandard amino acids and reduce that to a size that could be screened in a laboratory, even in a high throughput setting.
[00:17:35] Speaker A: And what are you trying to get down to? What kind of number is screenable?
[00:17:39] Speaker B: Depends on the apparatus and on the context of the screen. But typically it's not uncommon to screen anywhere from thousands to billions of variants. Yeah, once you get beyond that, you start to get into regimes that are very difficult for any biological system to manage. So we'll often start with a combinatorial space which is larger than the number of stars in the sky. Fantastic. Right. And that is, we're talking so many orders of magnitude beyond what we can screen in a laboratory that it wouldn't fit on a page. Right. And so this is where the computation comes in, because we can eliminate almost all of that space, because almost all of it's going to be unusable. And the better our tools get, the more we can tune them to the right purpose, the more likely we are to enrich that final space for the right answers that then will be pulled out of that screen in the laboratory. And we've done that now.
[00:18:25] Speaker A: That's very cool. So it really is this kind of comprehensive, I guess you could say it's a system that you're building. You're building the new amino acids, you're building the grows, you're building the tools to insert them, and then you're building tools to reduce the size of the combinatorial set that you're going to have to test. And you need all of that in order to hopefully get to a therapy at some point.
[00:18:45] Speaker B: That's right. It's a full stack of technologies, all of which are essential and all of which form core parts of the company. That's right.
[00:18:51] Speaker A: So it's incredibly exciting. It must also be I'm picturing the conversation you would have had with Funders early days, because it's transformational technology. If it works, which of course, in any new thing, that's always the question, what is the reaction when you say to people, hey, when you said, in the early days, we're going to build this system? Talk a little bit about that, if you would.
[00:19:09] Speaker B: Yeah, that was sort of the key question is how do you figure out what to do? Because I'll raise my hand and say, as a scientist coming out of an academic laboratory, I didn't really understand what it means for something to be a product and how important it is for somebody to tell you that they need to have that product. I think that it's very easy for those of us who build technologies to get positive feedback and turn that into the illusion of a need to have product. And so I think a lot of the early days of the company were focused on exactly that question. How do we identify something, that problem we must solve, and furthermore, a problem that only we can solve. And that's where the value is. And if you talk to an investor, that's the same question they want to ask, where is there a key unmet need that has a lot of value that only you can address? And that's what we've done. And so the early days of the company were sort of around posing several of those areas, building out the technology and showing that we can do that. And now we have focused onto some very specific areas where we've made, I think, incredible progress to the point that we now have animal data in multiple disease indications showing that these therapies work using totally new modalities.
[00:20:16] Speaker A: That's incredible. When you were having those early stage conversations, you also talked about the sort of safety, there's a safety story here, right, to the idea that these can't escape into the wild. How important was that compelling was that when you're having conversations about funding, how important is that, do you think, to the story of the company? In other words, you could almost say there's cure disease story here and there's an enhanced safety in these therapies that are being developed. I don't know. How do you think about those two things?
[00:20:41] Speaker B: Yeah, it's a great question and I would say it's overwhelmingly the potential to help patients is what drives the value. I would say the biosafety aspects of the organism itself are kind of icing on the cake in the eyes of an investor. It's again, because there hasn't been a widespread release of an engineered organism that's run amok in the wild. People don't look at it as a problem that is in need of an expensive solution. Right. But the good news is, with us you get that for free, right? And so it's like the seatbelts come in the car, right? You're not paying extra, not an option.
So sorry, but you have to drive safely. But yeah, it's definitely overwhelmingly what are the unmet needs of patients that we can address that drives the value of the company in the eyes of investors and I think in our own eyes as well.
[00:21:24] Speaker A: Okay, I'm going to switch to some kind of business nerd questions. So you had this intellectual property and these tools that you had built. You must have been weighing, should we build a company? Should we partner, should we outlicense? How did you think about all that stuff?
[00:21:38] Speaker B: Yeah, you're getting to our favorite saying in the company, which is porcano los dos, which you have to be a little careful. Right? Obviously we always need to maintain focus and I think we've succeeded in that. But because the organism itself can be utilized in so many different ways, we've always been open both to partnering and to internal development. And that's actually a big part of our BD strategy. And so we do have programs that are enslaved for internal development. We're not out trying to sell them. And we also have programs that we do talk about externally for partnering, where we push them to a certain point, and we would look for a partner to carry them from there. And to your point, there's also opportunities to partner on the organism as we've made so many advances with the organism. And then we have so many, as you said, so much important IP around it, we can do that. Right? And so if somebody else has a great idea for using this organism to produce a product that we don't intend to make, great, right. We can work together on that too. So we very much built our BD strategy around carving out those pathways, that is to say, products that are going to have the highest value for the company, but in a manner, that is to say, for a cost at a timeline, which is fundable with VC dollars. Right. That's realistic versus those that we think really should be partnered off to another company and triage your risk that way.
[00:22:52] Speaker A: Hi, this is Chris O'Brien, host of Few and Far between. We'll be right back with this episode in a moment. I personally want to thank you for listening to our podcast, now in our third season, it continues to be an amazing opportunity to speak with some of the top thought leaders in the clinical trials industry. If you're enjoying this episode, please leave us a review on Apple podcasts. It really helps people discover the podcast. And don't forget to subscribe to Few and Far Between so that you never miss an episode. One last request. Know someone with a great story you'd like to hear me interview. Reach out to us at
[email protected]. Thank you. And now back to the podcast.
Ten years from now, what do you think this looks like? Are these tools are you licensing these tools to other people? Are you guys you're obviously licensing some of therapy, some of some of the drugs that you're creating or solutions you're creating. What do you think the world looks like? You know when you kind of when you're relaxing over a drink on the weekend and you're fantasizing about the far future? I don't know if ten years of the far future, but whatever time horizon makes sense.
[00:23:54] Speaker B: Yeah, the far future is getting shorter and shorter these days. But I think it's a great question. And ten years from now, I would expect us to have multiple products on the market that utilize some of our first chemistries that we've developed with these new amino acids that are fulfilling unmet needs. And then where we're moving to. In the medium term is new organisms that allow us to combine multiple Nsaas into the same protein. So right now, a lot of the recoded organisms that are developed are mostly efficient with a single nonstandard amino acid at a time. So we can put as many of those Nsaas of a single type as we want into a protein. But if we want to have multiple functionalities that are novel at the same time, that's a lot more challenging. And so we're now finishing an organism which is going to allow us to put multiple Nsaas into the same protein at the same time and combine many of these novel functionalities. And so that's going to be coming very shortly, actually. So that brings us kind of to the medium term. And I do think that those organisms have a lot of partnering potential and I think we've certainly seen that because that's just not possible right now through any other modality. So those are some of the things I see coming now. On the other hand, we have totally new chemistries coming down the pike as well, right? And so today we already have chemistries that let us modulate the immune system and make proteins more stable and to do conjugation for other purposes that are very important. But there's lots of other chemistries we can work with, too that could, for example, make it easier to access the inside of a cell, right for delivery. Or to make a protein that can switch on and off in the presence of different wavelengths of light or chemical signals or those that have total resistance to proteases. So they can be very stable and not be degraded, even in very harsh environment. These are all chemistries that we can work with today that could be made into very viable products in the medium to long term. So those are all things that will happen, I believe, in the next ten years, as well as new production organisms that are also recoded. So right now, everything is bacteria. And bacteria can make many different proteins.
Most of the proteins that are sold today are made in bacteria, but therapeutics tend to hew a little bit more towards human cells or mammalian cells because a lot of them are antibodies and those are tough to make in bacteria. So as we move towards recoded yeast and recoded mammalian cells, that's also going to open the door to a broader set of modalities that can be accessed and can be enhanced with nonstandard amino acid chemistries.
[00:26:15] Speaker A: Fantastic. We talked a little bit about applications in rare, ultra rare, and orphan specifically. And do you think that this technology broadly moves us closer to personalized medicine and N of one solutions and stuff like that? How do you see the role for grow in rare conditions in general and then maybe where you think it goes?
[00:26:33] Speaker B: Yeah, I do think that's the case. And I can probably bring to bear an example of some of our current work. And so rare diseases, they tend to focus on cases where we have hundreds of thousands to a million or so people in the US. Have rare diseases and so they tend to be sort of underserved and not as well characterized or because it's a smaller patient population, it's difficult to monetize because you're not going to have quite as many people and you have to make them very expensive. Therapies that said, because of that, there's a lot of effort to help treat these people and to find ways to both to monetize and to do the necessary technological development to get there. And so there's lots of ways to incentivize that both through regulatory pathways and through funding mechanisms. And we're now foreseeing that right, with fast track approvals of the FDA and with, I think, some much more egalitarian funding mechanisms, as you know. The other thing, though, is the technology. And I think some of the things that we've been developing using Nsaas can really help. And so one of the areas that we focus on is autoimmunity. And in an autoimmune disease, a patient is reacting to a cell or a protein or multiple cells and proteins in their body in a way that they've decided is foreign. So your body mounts an immune attack against itself and that's something that is typically treated by hitting the entire immune system with a suppressive agent. So basically we say, well, their immune system is overreacting, so let's just turn the whole thing down. And that has two critical flaws to it. One is you're not really modifying the disease in any way. So typically you'll alleviate symptoms, but the patients are still sick. And secondly, you're not leaving the immune system at full strength anymore. And so even though it's got this small flaw that it's attacking some of the proteins, you've now knocked the whole thing down so that it's not able to fight off other diseases. And that means you're more prone to infection and sometimes even to cancer. And those are really big problems. And so what we want to do is treat the underlying cause of that disease specifically. And this gets to the root of your question, right, because number one, many times the disease is driven by a very small number of what are called antigens, such as the protein you're reacting to. And number two, we want to make sure that we can address all the different patients who have this disease who may have a different genetic background from each other. And so the way that we've been pursuing that is by using nonstandard amino acids. So what we do is we use these special Nsaas that have what are called glycans on them. And these glycans are the sugar molecules that decorate most of the cells and proteins in your body and are really one of the key signatures by which your body distinguishes what is self from what is foreign. So, for example, there are pathogenic bacteria that will steal your glycans and put them on and pretend to be cells from your body to evade your immune system. And there are cancer cells that will decorate themselves with these protective glycans in order to hide from your immune surveillance. And there are cancer therapies that try to chop off those glycans like a lawn mower in order to access those cancer cells and reactivate the immune system. So this is a very powerful language, these glycans, but it's very difficult to engineer and very messy. And so what we've done is we've turned this glycan language into the language of amino acids, where we can say we'll take a glycosylated amino acid, which confers tolerance to a protein, and now we can put that onto any protein anywhere we want, as many times as we want. And so we can take an auto antigen that an autoimmune patient is reacting to and make a version of that antigen that now has a tolerating signature on it. So we put that back into the patient, it reeducates the immune system to treat that protein as a self protein, and now they stop reacting to the auto antigen. And so this wasn't possible before because there was no way to engineer that auto antigen to have these glycans on the surface. Well, by turning those onto nonstandard amino acids or into nonstandard amino acids, we can now do that in a very defined way. Can really take any protein that causes an immune response and now make it silent or toleragenic. So we're doing that for autoimmune. We're doing that for enzyme replacement therapies that become ineffective because you develop antibodies against them. Same thing. We can put these glycans on them and make them tolerating. And it can be applied for gene therapy as well. In gene therapy, you're reacting to proteins on the delivery capsid, typically, which is a viral protein. You can take that viral protein and create a toleragenic version to turn off that immune response. And there are other modalities, too, right, that you could use this for. So right now, we have two programs underway, one in autoimmune and one in antidrug antibodies for enzyme replacement therapies. And we're also now pushing into gene therapy as well. And we have animal data already for two programs there showing that, number one, in autoimmune, we can profoundly improve disease progression in an autoimmune model, in an animal model of autoimmune disease. And number two, we can take an enzyme replacement therapy, which is a marketed product and extraordinarily effective for a few weeks or maybe a few months, but then becomes inefficacious because of that neutralizing antibody response. We can basically eliminate that in this animal model using these progly glycans. And so that opens up a whole battery of new modalities and applications that we're now building out internally and also working on partnering.
[00:31:35] Speaker A: Do you think tolerazation is scale it for us? A little bit in comparison with direct development of therapies is tolerazation as big bigger? How do you think about that?
[00:31:45] Speaker B: Oh, it's huge, right? Because you've basically got all of autoimmunity, and then you have every therapeutic that's ever been beset by issues of immunogenicity. Right. And so these are what we'd call Holy Grail problems. Now, coming to your other question around personalized medicine, you kind of get that for free with this approach. So to go one level deeper, when you have an autoimmune response, the auto antigen you'reacting to is taken up by an immune cell, and then it's broken into pieces, and each of those pieces are presented to immune cells, and each of them has a chance to become a reactive piece or an epitope, as it's called. And every patient could be reacting to different pieces from that antigen. And that relies on what's called your HLA variation, which is something that we can measure in every person. And you and I, Chris, very likely have different HLA variants. Therefore, if we had an autoimmune disease, we'd be reacting to different pieces of the antigen. So if I want to make a therapy, do I treat your version of it or do I treat my version of it? So one of the really cool things about our approach is we tolerate the patient to all the portions of the antigen, and we let their immune cells break it up into the same pieces that they're reacting to because all we're doing is putting those glycans on. And when you put the glycans on, you switch those presenting cells to a tolerogenic state. When they break up the protein into pieces and present it, they present the same pieces. So now for my autoimmune disease and your autoimmune disease, it's the same therapy, but I'll tolerate to the pieces that I'm reacting to, and you'll tolerate to the pieces that you'reacting to, and the same thing is true for everybody else. So it's one composition, but it becomes personalized medicine inside the patient. And so that makes it so we don't have to break up the patient population. We don't have to have multiple different therapies, but we can reach, hopefully, all the patients.
[00:33:27] Speaker A: It's incredibly exciting. Okay, I'm going to switch gears for a second. Everyone on Twitter has become an AI expert in the last nine months. You've been doing AI stuff for a couple of decades.
How do you think the current fervor stacks up? Do you think AI is underrated, overrated, or kind of appropriately rated right now? And talk a little bit, if you would, about where you think it's going and the application of these large language models and other kind of emerging tech here. I'd just love to get the benefit of your years here.
[00:33:54] Speaker B: It's porcanolos dose again, right? It's both underrated and overrated. And you really can't overstate the profundity of the improvements that we've seen in the protein modeling world in the last three years. It's just been a rocket ship taking off in our ability to model, for example, the structure of an unknown protein to a point that it's to say it in a somewhat cynical way, almost usable. And the reason why I say that is, to be truly usable, you really have to get the structure right to atomic level accuracy for many applications, but for drug development, especially. And so we're getting close, and we can do that for some proteins, particularly those that look kind of like other proteins that we really understand. And our ability to go further off into left field has room to go. We're on our way now, insofar as it applies to our world of nonstandard amino acids, one of the things that really drives the current form of AI is being trained on enormous amounts of data of preexisting examples. That's how these generative AI approaches work. And for things that involve non standard amino acids, there isn't so much out there. Right. And so the field has some catching up to do. Now, for protein design, it kind of sits between those two. Right. Because we're able to model the structure of existing proteins really well. But our ability to design new structures and new functions, again, kind of depends on how similar it is to stuff we've seen before. But we're definitely getting there. I think that the reality will catch up with the hype probably faster than we might think. But our ability to use nonstandard amino acids really depends on our ability to combine a fundamental understanding of physics with these generative AI approaches. And that's something that's happening now, too. So we're excited by it. We're bringing these tools into the company. We know that we're beginning to see the vanguard of what you can do with them. And both the previously inaccessible accomplishments, as well as the crash and burns. Right. It's all happening at the same time.
[00:35:44] Speaker A: Yeah. Right.
[00:35:45] Speaker B: And you got to find the edges of the picture frame. And I think we're finding it. We're crashing into it. But, yes, we absolutely are using those tools. And we'll be using those tools in the near future to rapidly engineer proteins with totally novel capabilities that use novel chemistries.
[00:35:59] Speaker A: Do you expect another step function improvement in the models over the next couple of years? We've heard folks say we're going to reach a point of diminishing returns where the training costs are too high to really drive a huge improvement. But right now, I know there are a number of models that are headed towards release, and kind of the pro AI camp says, no, we're just getting going. Where do you fall on that spectrum?
[00:36:23] Speaker B: Yeah, I think it kind of depends how you frame the question, because I think what we really need right now is tuning. Right. And so a lot of the devils and the details here, and as I mentioned before, what you really need is atomic level accuracy and so let's say right now, you're sitting at, just for argument's sake, we've gone from ten angstroms of accuracy to two angstroms, right. Where the length of a bond of an atom is like, around one angstrom. And so if we can go one angstrom further, we'll be atomic level accuracy. Right. So to get that last bit is really hard. Right. And so the step function might just be a lot of tuning and some more data that lets us tweak things enough that we stop making mistakes on the last mile there. And once we get to that point, even though it's only squeezing out a little bit more accuracy, the models become dramatically more usable. And I do think that that is achievable.
[00:37:11] Speaker A: A currently locked door becomes unlocked.
[00:37:13] Speaker B: Yeah. Okay.
[00:37:13] Speaker A: That makes sense.
[00:37:14] Speaker B: That's right.
[00:37:14] Speaker A: That makes a lot of sense. Okay. In this last section, I'd love to talk a little bit about careers and advice you might have for bright young person who's considering this intersection of computer science and biology and kind of where they should go, what they should think about. Do you have anything to share there about you? I don't know, Dan. You were either really smart about this and saw it coming a long time ago, or you got a little bit lucky. But you sit right at an intersection of two of the most exciting trends, I think, that are happening right now the biotech revolution and the use of advanced computation. So we riff a little bit on that.
[00:37:48] Speaker B: Yeah, I definitely did not see this coming 20 years ago. And as I said, I didn't even know what an amino acid was when I was modeling them. Right. But I think what I did always do was follow my nose. And I felt like this intense curiosity about how biological systems work. Bringing it back to the beginning of our conversation, I mentioned how I used to make these animations in biology class in high school, and that was how I got into excited around life sciences. I would be the first to admit I was really bad at doing experiments. Like, I didn't think I could do it. It wasn't really until I got to George's lab and he has this philosophy that nobody has bad hands, that I kind of discovered that if you really care, anybody has good hands. And I got pretty good at doing experiments. Right. This fascination I had with biology, I kind of viewed computation as a backdoor to get into the field.
[00:38:31] Speaker A: Wow.
[00:38:31] Speaker B: And it wasn't really until I did start combining them. And I was lucky also to have a Pi in grad school, tanya Kartemi, who had both labs and encouraged me to do both, that I could put them together and kind of gain the confidence that I could understand both sides of the coin. Now, as far as career choices coming out of the academy, it's one of those things that there's kind of two parts to it. So one is really asking yourself about what you want your everyday life to look like and the other is completely unknowable, right? Which is I'll just say nobody really knows until they try. This job is very strange, this sort of transitioning from the academy to being CEO of a biotechnology company. And none of us are good at it when we start. Anybody who says otherwise, I would question, I'm the first to admit it, right? We know nothing about how to do this. And it's all about being teachable, seeking out criticism, assuming a growth mindset, and surrounding yourself with people way smarter than you and absorbing as much as you can because you must be shaped by this process. You are not cut out for the job when you start. And so that's something you have to be like wanting to seek out and be excited to have that part of your life. And that's not for everybody, right? And I'll leave it at that. But so there are some things you can ask yourself around. What do you want your everyday life to look like? And so I found myself wondering, do I want to pitch investors or write grants? Right? How much do I like teaching? What kind of collaborative models do I like? How important is longevity to me? And on the one hand, these different features bifurcate somewhat between academia and biotech. On the other hand, they're blending more now than they ever did. And a lot of the questions I asked myself are irrelevant now. For example, I loved mentoring as an academic and I was kind of like worried I wouldn't be able to do that. You totally still do that now. And I have multiple students I mentor, I have other companies I mentor. I myself gained tremendous insight from people who mentored me. And not just like people have been doing it for a long time, which I've also been lucky to have. But companies that came before me, both out of George's Labs and others who are two, three years ahead of me, say, gave me tons of invaluable guidance throughout this process, right? That's all there now. I have written multiple grants when I was a startup CEO, founder, right? That didn't go away either. I still had to do that. Right, but you do change how you present the story, right? And so pitching an investor is very different than getting a paper published and having to actually find as going back to the early part of our call again, a need to have problem that you can solve, right? Because in the academy, if you write a cool paper and it's going to get into a nice journal and that's your currency, it can lead you to presume that a cool technology is therefore commoditizable. And that's just not necessarily the case, right? And so wanting to embark on that journey to truly understand how the customer discovery process works? How does a product development process work and do you care? And are you interested in intellectual property regulatory, as I said before, market analyses, all these things that you really only think about in the buy tech world, whether or not you like it or not, are not exposed to in the academy so much. Right. Do you have a passion for those things or are you at least excited about learning them and some people do and some people don't. I get really excited about IP. I love the kind of logic puzzle of it. I turned out that I really like exploring market opportunities and building strategy around that. Right. And I really love all the kinds of different professionals I get to interact with. I like interacting with really smart lawyers and really smart business development people and actually investors too. And so if you like those kinds of things, the day to day lifestyle may be what you're cut out for, but at the end of the day, you're never indexing for success. You're indexing for the drudgery. Right. You're not spending your days like building a groundbreaking product in biotech, nor are you spending them publishing high profile papers or celebrating your tenure. You're grinding through something, and it's a question of which grind you feel is most naturally suitable for you.
[00:42:05] Speaker A: That's great. You said a bunch of things there. I think you should give a I don't know if you give a talk on this, but you should give a talk on this for academics who are considering making the move into running a company. And in my job, I love that I get to work with lots of really talented professionals in disciplines that I am not a master of. I think that's part of what makes it great. But if you want to spend all your time in your core discipline, it's.
[00:42:25] Speaker B: Not going to be a very happy place for you. Right?
[00:42:27] Speaker A: So I liked your point about you need to accept that you will be changed and change as a result of the process and developing that. And do you talk about this? It feels like something it feels like I heard a mini Ted Talk in there already. Advice for startup CEOs in Biotech.
[00:42:40] Speaker B: I've done some panels for sure, and it does come up. And I've gone back to both universities that I've worked at and given talks or panels on this kind of thing. And like I said, I benefited so much from my friends and colleagues who came before me. We all try to then pass on what we can and of tell you what it's like to run a biotechnology company ten years out, but I'm happy to tell you everything I know from year zero to year six. And that's the most important information you can get as someone who's kind of just been there. And you have to kind of index that along with someone who did it for 20 years and can look even further out. You need that whole short to long term perspective, really, to think about, do I have a company here? Do I have a technology here? And what is a product?
[00:43:19] Speaker A: You need to be in the detail and up in the clouds.
[00:43:22] Speaker B: Both absolutely.
[00:43:22] Speaker A: Well, listen, Dan, pleasure to have this conversation with you. This is one of the most exciting biotech stories that I've come across and really, really fun also to kind of hear how you're applying this technology and where it's headed. Thanks for talking with us today on Few and Far Between.
[00:43:37] Speaker B: Yeah, really fun speaking. We did, Chris. Thanks for having me on.
[00:43:48] Speaker A: Thank you for listening to the latest episode of Few and Far Between. Our podcast is now available on Apple podcasts and other major streaming services. Please take a moment and leave us a user review and rating today. It really helps people discover the podcast, and we read all the comments. Those comments help us to make Few and Far Between better and better. Also, be sure to subscribe to Few and Far Between so that you don't miss a single episode. Got an idea for a future episode? Email us at
[email protected] or contact us on our
[email protected]. I'm your host, Chris O'Brien. See you next time.
