We chat about longevity with Alex Zhavoronkov, CEO and founder of the artificial intelligence (AI) specialist Insilico Medicine.

Born in Latvia, Alex founded Insilico in 2014, with the company raising a huge $110M Series E round this year.

Alex gives us a tour of Insilico’s sprawling pipeline, a blunt rundown of the competitive landscape in AI drug development, and explains how Insilico is navigating the huge challenges faced by longevity-focused startups.

⭐️ ABOUT THE SPEAKER

Alex founded Insilico in 2014 after serving in senior roles at ATI Technologies, NeuroGNeuroinformatics, and the Biogerontology Research Foundation.

With an academic background in biomedicine and computer technology, he has published more than 200 peer-reviewed articles, and has bachelor’s degrees from Queen’s University, Canada, a master’s degree from Johns Hopkins University, US, and a PhD from Moscow State University, Russia.

🔗 LINKS MENTIONED

• Insilico Medicine’s website: https://insilico.com/
• Insilico Medicine’s phase 2a results: https://insilico.com/blog/1112
• Insilico Medicine’s Series E round: https://www.prnewswire.com/news-releases/insilico-medicine-secures-110-million-series-e-financing-to-advance-ai-driven-drug-discovery-innovation-302401040.html
• Thomas Clozel, Owkin | What does AI Biotech really mean | E10: https://flot.bio/episode/thomas-clozel-ai-biotech/

Transcript

Intro [00:00:00]

Alex Zhavoronkov: Over the past 12 years, we’ve seen dozens of very promising companies that collapsed, even though they raised considerable amount of money, they tried to go into the clinic. But so far everything collapsed. So nothing worked clinically. So we help do research in sustainability in energy, batteries, agriculture.

So we’re very broad and we also realized. That we have to develop our own drugs in order to stay afloat and to to grow. And now we have over 30 programs that are internal. Some are running you know, all the way to phase two. And it looks like we’re making small steps, but it will require many, many, many, many small steps to go a long way.

Philip Hemme: You have new to a new episode. I am your host, Philip, and on this show I’m entering the best Europeans in biotech to help Duke grow. AI drug development could change biopharma, and one of the most advanced AI drug development company in the world is in SCO medicine. So I call up with Alex while he was in Shanghai.

I didn’t know him personally, but I followed him online for a while and I heard. Many good things about him for my network. We talked about growing up in Lavia. We also talked about being the first to show efficacy phase two data for an AI generated drug and why we should invest trillions more into into aging research.

So he’s my conversation with Alex. If you’re enjoying it, please hit the like and follow button. Alright. Welcome to show. 

Alex Zhavoronkov: Happy to be with you for today. 

[00:01:43] Longevity: work, romances and bromances

Philip Hemme: So let’s start more personal. I think you’re European. You were born in, in Ladia, but then you started in, in North America, you worked in Hong Kong, in East Asia.

And now you, I think you are how the company’s bus in, headquartered in, in Boston, Cambridge. So can you just like, walk me a bit through, like, or tell me more about your, what, your past. 

Alex Zhavoronkov: Sure. So born and raised in Latvia, in Riga. So one of my citizenships is European. And I immigrated to Canada at an early age, so I’m also Canadian and got my first two degrees at Queens Univers in Canada.

Worked in GPU computing originally. So I thought that computer science will save us from aging and death early in the, in the process. Made some money when I was in the GPU business and decided to go into biotech. So with my credit work, Hopkins was a professor here and there, and then started in Sica at the dawn of the deep learning revolution 2014.

So now 12 years with the company and still doing the same thing using computer science for. Longevity research. 

Philip Hemme: Yeah. And so you were in longevity even before, before sco basically like yeah. 

Alex Zhavoronkov: Yeah. I’m very interested in aging research and now it’s my 22nd year in aging research. 

Philip Hemme: That’s cool.

That’s cool. We’ll talk more about the anti-aging as well. Hmm. On the, on the more personal note as well. I, I saw that you’re quite friend with Thomas Lau who was on the show actually for I think episodes nine or 10 or something. Can you tell me a bit more that your relationship with him and how, like, how this friendship came, came about?

Alex Zhavoronkov: Sure. So Tamaya is a close friend for many years. I followed his company since the very beginning. I really liked the fact that they started publishing in peer review journals and some of those papers appeared in very high peer review journals. So it’s kind of French coding ninjas, so to speak. So we first met, I think around 2016 or 2017, so around that time at actually at the wedding of a common friend and then.

Yeah. So we didn’t know each other before, but then it turned out that we’re in the same space. So we’re just friendly. We don’t compete. They might be competitive in some, you know, small areas where we, which we also touch, but it’s not it’s not direct competitor very much. Like his company.

I like the fact that he is you know, socially active, but at the same time. Very, very business savvy. Mm-hmm. 

Philip Hemme: Yeah. That’s good. No, he is a great, he’s a great guy. Yeah. I think I met him around the same time, 2016 when kin was still very blurry of like, we’ll do something with ai, but we still don’t really know what we’re doing.

And then and then it took shape and it’s pretty impressive how much he, how fast it grew and, and where they’re right now. Hmm. So if people interested, you can check out the episode with, with Thomas. It’s, it’s online as well. And the, the last thing I, personal, which I mean, I, I follow you on, on X and I followed you some.

I I read a lot of the articles you write on s so actually it’s great to connect because I, I’m seeing lot of things online with always good to connect. And one thing that I found quite I. Like quite fun is like you’re organizing this conference in, in Denmark, and I think, and I saw you like, propose to your wife, like on stage.

I will show the picture to the audience. Like how did you have this idea? Like what, how did it came about? 

Alex Zhavoronkov: Well my fiance I met her at an aging conference as well. And she’s very much into longevity research and when we first met I never thought about, you know a potentially romantic relationship.

So even though she is kind of significantly younger than than I am, but she was already quite deeply engaged in longevity research and when we connected, that was our common topic. So we actually are very committed to. Making people live longer, healthier, and more productive lives. And she’s been committed to that kind of goal for a very long time before she met me.

So it was only reasonable to you know, propose at an aging research conference. However, I did not expect to do it on stage. It kind of seemed along serendipitously. She didn’t know, so I didn’t know what she’s gonna say beyond, she was like, oh, what are you proposing, by the way? Right? Like I originally said, said like, would you be my longevity research partner?

And then I kind of explained that she, she said yes. Now we’re a happy couple. We just celebrated the anniversary of our engagement at A RDD at this conference in Copen again. I help organize. I originally founded it, but now it’s run by Mornin and Daniella Bakula at the University of Copenhagen.

That’s the largest event on aging research in the world, every end of August. And the entire industry convenes in Copenhagen for five days to present and discuss their latest longevity research. So. We really liked that event, and hopefully we’ll do it again and again until we solve aging.

Philip Hemme: That sounds good. Yeah. 

[00:07:56] Survival in the graveyard of longevity drugs

Philip Hemme: I’ll I, we’ll talk about the, the conference and anti-aging. I think it’s, yeah. I, I saw so online and yeah. And I heard also about the conference actually maybe to, to, to switch gears. I mean, what you mentioned was, was inco, and one thing I’m curious is like, obviously I’m, I’m looking at it from a bit of an observer.

I’m, I’m, I’ve been watching the anti-aging field since I think at least, you know, 20 15, 20 16. When kind of started when there’s the first big waves of investment, which are looking at the AI field for like, quite in details as well. I’ve seen a bit of ups and downs as well. And then I’ve also seen all the Hong Kong investment where there’s kind of massive amount of cash coming in.

Also, some, like some correction as well. So you are, you are kind of in these three spaces and kind of adding up the three I’m learning how, like, how, like how much of it, like do you, I don’t know how. How does it fit together for you? Like how much are you like using these trends versus how much, like, let’s say, not real value, but value are you building?

Alex Zhavoronkov: So, sure. One thing to know about aging research and drug discovery is that, and development is that so far over the past maybe 70, 80 years, that people have really tried to use credible science to advance human productive life lifespans, using aging research and extend healthy, productive lifestyle lifespans.

So everything has failed in a very dramatic way. I’m talking about specifically going after aging. Of course, there are beautiful therapeutics and immuno oncology eBIC diseases. So, or maybe GLP ones are the first longevity therapeutics that we’re gonna see. But, so far, except for those, you know, really focused therapeutic drugs, everything has failed.

So it’s the, it’s basically a huge graveyard. Even during the time that I co-organized the conference. Over the past 12 years, we’ve seen dozens of very promising companies that collapsed even though they raised considerable amount of money. They tried to go into the clinic, but so far everything collapsed.

So nothing worked clinically, and there are many reasons why things collapse. Some of those reasons are actually non-scientific, so companies are running outta money or the founder is go and chase some new shiny things and just leave it be to something else or. You get into the situation where you sell to a big pharmaceutical company and you have to, you know, close shop on aging research.

So there are many reasons why they didn’t make it. So from the very beginning when we started at Silica, we decided to develop a very sustainable business model. So, so that you know, when funding is skiers. We could piggyback on something that is revenue generating and decided to develop AI solutions for the entire industry for target discovery, for small molecule chemistry, for biologic design for clinical trials, outcomes, predictions, clinical trial analysis, and even all things science.

So we help. Do research in sustainability in energy batteries on agriculture. So we’re very broad and we also realize that we have to develop our own drugs in order to stay afloat and to, to grow and to validate that AI really works. So, so far. Since 2019. So 2019 we got our first big check.

So previously we were kind of bootstrapping and working on minimal funding. So 2019 we got our first big check and decided to go after on our own software and decided to go after our own pipeline as well. Because in order to show that the software works, if you have not developed, at least, you know, few, developmental candidates, your software doesn’t work. It’s usually like that. That doesn’t matter what, what you, what you’re saying to investors and to yourself if you haven’t demonstrated that you have what, you know, at least a dozen developmental candidates you didn’t really push the needle there, so in any way.

And we decided to actually build a model where we test our ai. Get it to a certain level. And also, sorry, it’s hot in here on so to take it to a certain level and get to the point where we can license this product out. So to the pharmaceutical companies. And so far we’ve licensed a bunch of those molecules.

So now. Even if you know, things don’t go well dramatically, you know, if all of us died due to the pandemic or something like that within the company other companies have unlicensed the drugs and on they’re very likely to develop them further. So, and since we have for out licensed several, just looking at the overall probabilities, there is a good scan that, you know, it’s going to sustain to some extent, even though biotech is biotech, you never know. So and now we have over 30 programs that are internal. Some are running you know, all the way to phase two and we constantly validate the hypotheses that we generate using aging research and then try to validate them experimentally.

And it looks like we’re making small steps, but it will require many, many, many, many small steps go a long way. Otherwise, if you make, you know, a big investment right away in something that is an ultra risky, the probability of collapse is like close to a hundred percent. 

Philip Hemme: Yeah. Hmm, that makes sense. 

[00:14:57] Insilico Medicine comes of age

Philip Hemme: The, I think, I mean, I like the, the thing about like starting with AI but then demonstrating value and at the end of the day, I mean, we end up like.

In the biotech or drug development world. And what matters the most is, is, is the clinical data and, and having especially efficacy data, but even, even just phase one data is, is what matters at the end. And I’ve seen quite a lot of AI companies who like. Had the AI part and try to license out the AI part, which kind of works, but also you just don’t capture that much value and it’s very hard to prove.

I think to, to your point, which I think is also a trend of lot of AI companies going much further into owning their own pipeline and their own assets which I think makes, makes sense. And, but you, you realize that from the beginning or you realized it like a bit later, right. 

Alex Zhavoronkov: 2019. So when we started, we were all about ai.

Ai, deep learning. Deep learning. You know, look at how cool this algorithm is. Look at how we are outperforming here and here and here. Look at can generalize, look at, can generate stuff. 2016 was our first big paper showing that we can generate a molecule using desired properties. Desired properties.

Using generative design. And there we used the generative auto quarters. And then 2018 we developed the generative sensorial reinforcement learning platform and started designing using that one. But the so I didn’t have the personal brand or the backing of an top tier VCs from the very beginning.

To start to raise a lot of money right away. So some people are, they manage to just on their personal brand and their investors, or very often investors incubate those companies without you know, doing a lot of technical due diligence because they trust that they 

Philip Hemme: trusted team in the, like the team.

Alex Zhavoronkov: Yeah. So we were not like that. And, we had to prove to the others, but also to ourself that we want to you know, change the industry, but at the same time, we can really solve practical problems within the industry. And our first five years of existence, we’re focused on developing the algorithms and then deploying them within big pharmaceutical companies to try to see where they work.

And we took projects from. You know, disease modeling from target discovery. There is, by the way, no money in that. I mean, unless you figure out how to sell it for a lot of money just on the promise of delivering something. Those partnerships unless like when they’re of too big, I don’t know a single successful one.

Like everybody partnered with an AI company on target discovery. Like, show me the targets in phase two, like or phase one and we have one with force and pharma in phase one, low target. And we also did a lot of work in chemistry, in even operational management, process engineering. So we took all kinds of projects to understand how pharma works.

And then 2000 17, 2016, actually some investors realized that we are like one of the best teams in the world. And even pharma. And they were, they started suggesting like, look, if you believe in your algorithms and your ability, why don’t you go and develop a drug? 

Philip Hemme: Mm-hmm. 

Alex Zhavoronkov: And we’re saying, no, no, no, no, we’re not going to do that because you know, we are D eight guides, right?

We’re not the drug only 

Philip Hemme: company up. 

Alex Zhavoronkov: We, we don’t have that expertise. We’re there for you. We want to support you. Let’s do a small biotech. Together. Right. And we tried to do a few of those geo joint ventures NewCos with kind of smaller teams and it didn’t work. So 2019 we got our first serious funding, so $37 million.

Nowadays it’s. People smile when they hear this kind of number because you know, five years into your existence, that’s your first round, 37 million you are like a tiny fry. But for us it was a huge amount at that time. And that allowed us to do two things. So launch proper software. So we thought, well, people like our services, like our algorithms.

But it’s actually quite difficult to partner because every time you need to go through the due diligence process you know, CDAs, NDAs then you do a contract and then you implement it, and then you get, you know, maybe a hundred thousand dollars or a few hundred, you’re like, okay, well how can I make it less like frictionless or at least with less friction?

So we decided to launch software. Some software is very cheap and accessible to academics. Some software is priced to perfection for the pharmaceutical companies because we know that the development of the software costs enormous amounts of money, and we also need to support them with additional people, engineers.

Deploying their platform security costs a lot. And of course all kinds of privacy compliance. So we actually have to price it competitively. Yep. And at that time we also said, well, let’s try at least like couple programs. 

Philip Hemme: Yep. 

Alex Zhavoronkov: And in 2022 on actually 2021, we nominated our first two developmental candidates.

So a developmental candidate is basically one stage before clinical trials. So you do one formal step of existing testing and you can apply for IND and go into phase one. And in 2022, we nominated nine. So it was quite, quite a big jump. Usually even within big pharmaceutical 9, 9, 9. Nine developmental candidates.

Okay. So okay. Yeah. In 2022, I think we had one IND. That means that we actually started phase one. And so yeah, so it was then we nominated 22 developmental candidates. So basically starting 2021. Then reached clinical trials, clinical stage. So it’s a pretty you know, pretty good validation for the software. But the fun part is that things that are currently in phase two are the demonstration of what our AI could do in, let’s say, 2020 

Philip Hemme: mm.

Alex Zhavoronkov: Okay. Because. The actual computation costs you, I don’t know, three weeks, four weeks nowadays at that time, maybe a little bit longer. But the rest of the time you spend experimenting, you need to develop the experiment. A validation demonstrated it works. 

Philip Hemme: Okay. I like that. I like that. I think we’ll go deep into, into the area also.

[00:23:12] Landmark efficacy in AI drug program

Philip Hemme: I’m just curious, but on the, to stay on the pipeline, I think. One, one very interesting thing and what I, what I looked at, I looked at in cco and when I looked at industry, I’m always, okay, cool, ai, you go fast, you nominate a lot of candidates, but I, I really wanted to see efficacy data and what I liked with you guys is that you had a phase of phase two A with some efficacy data.

I think it was November, 2024. Then you actually met your primary endpoints. I think it’s one of the first AI generated drugs with actually efficacy data. What is the first time I’ve kind of seen, can you, yeah. I mean, first is it true and then can you just expand a bit on the results? Yeah. 

Alex Zhavoronkov: Well, I don’t know another one myself, but no, I’m seen this people re would love to would love to learn if there is another one.

Who knows, right? Maybe somewhere in countries that I don’t operate in. You operating in all, in all the continent. Yeah. Well, you never know, right? So some companies in Japan and China, they might not advertise at all, right? So maybe if somebody is quietly doing something in a basement but but so far I have not heard of another example where the AI generated drug even reached phase two. Because most of the drugs that were on phase two now, many of them got deprioritized by the companies. They were designed using traditional approaches, and some of them were actually unlicensed.

So it’s not that difficult to get into phase two and try it out if you in-license a drug. Many companies have done that or repurpose a common reagent or repurpose a a drug that somebody else developed. So when you’re not doing it with your own chemistry, it’s relatively simple but then people are likely to not sell you the best drug, right?

To a startup who wants to go directly to phase two. Or if you’re doing it with a repurposed compound you will not have the intellectual property protection, even though you might be able to see if your target hypothesis works. We have some of those validation experiments as well. One has completed on an IIT so equivalent of phase two in China for a LS, whether in rips compound with a partner.

We’re not allowed to talk about this data, but from what I understand, it’s very promising. Okay. And what we did we have the first kind of ai identified targets. So we used AI to identify new target that has never been in the clinical trials before. And then we generated from scratch the molecules for this target that inhibit this target.

And then we took it all the way to. Phase zero, then two phase ones, and then phase two A and actually two, phase two, a’s we have two parallel ones. And in China it has completed with safety and tolerability, but also with expected very promising efficacy trend. So we published the Nature Medicine paper on that, so you can actually read everything that’s disclosed.

The safety and deliverability profile and the efficacy data. And we also released the proteomics analysis in that paper. 

Philip Hemme: Okay. No, that’s fine. And I’ve seen the, in, it’s in, in the pulmonary fibrosis. Correct, yeah, correct. 

Alex Zhavoronkov: Like and pulmonary fibrosis, we decided to go after it because. I think it’s one of the most age-related diseases out there that can be used as a test bed for anti-aging interventions because it’s kind of like Alzheimer’s of the lung, quote unquote.

The average age of S onset is around 65 and most of the people with IPF, they expire reasonably quickly. But they resemble many signs of accelerated aging and fibrosis in the lung. So that’s why we actually decided to prioritize that indication. Yeah, so I wanted to see if our compound can make their lungs just in general, systemically younger.

’cause that’s what they have in common, actually. 

Philip Hemme: Yeah. 

[00:28:15] A sprawling pipeline for Insilico Medicine

Philip Hemme: Because that’s exactly what I was about to ask, of how much is it the antiaging disease and also in your, in your, in your pipeline. I mean, you have lot of oncology drugs which I don’t know if I would classify that really on, on typical antiaging.

I don’t know. What’s you some, 

Alex Zhavoronkov: how drugs there are anti-aging. I mean, the main hypothesis was that you would purpose them to cancer, get it approved, and then not. Look for alternative dosage and mechanisms and administration combinations to go after aging or probe some of the aging pathways.

So, and you also utilize them to understand aging a little bit better. And many of the drugs, I would say 75% of our targets. Have dual purpose. Okay. But some of the targets we decided select them just because they’re commercially viable. Yeah. Okay. 

Philip Hemme: That’s, that’s exactly what I was asking. The so 75, that’s, that’s interesting.

And then what I see also, I mean, you talked about our partnering, I mean. This, this IPF asset is, is fully owned. I think you, you didn’t partner it, which is also interesting that your, your lead candidate is not licensed. I think from a, from a biotech strategy, I think it makes sense from a value creation.

It just makes sense. If you have the financing to be able to do it is always what you want. In, in biotech, 

Alex Zhavoronkov: it’s not only about that. So during those trials, we actually do much more than a traditional IPF drug developer would do. We also collected and analyzed the proteomics data. And while we were doing trials, we also did a lot of preclinical work and then other experimental work demonstrating additional properties of this target.

Okay. So we’re actually doing a lot of work just in fundamental biology and chemistry, if we were to let it go. We would not be allowed to do that.

Philip Hemme: One thing. I’m also wondering, because your pipeline, I mean you have it behind you, but I will show it also on the screen. I have it, I have it on my, on my iPad. I think you basically have, if I read it well, you have basically the, I mean one phase two or two phase two in parallel, but for the same target drug and what we just discussed, and then I think you have one phase one completed, and then you have two other phase one or not even.

Five was a phase one, I think 3000. So basically one phase two, one phase one completed, and five phase one after that. 

How? 

Alex Zhavoronkov: Yeah. And we also have another one that is not on the pipeline chart. So during COVID we developed one C3 like protease inhibitor all the way to phase one complete. But now when it’s over, it be, people don’t want to go into antivirals.

Philip Hemme: Yeah, makes sense. But my, what I’m wondering is also like how, how do you prioritize the D, the different programs and how are you able to run? That many in parallel, especially resource wise. I mean, you have raised a lot, I think you’ve raised whether 400 million and the, I saw the series EI mean, you’re, you’re well financed, but at the same time, it’s also super expensive to run, you know, whatever, a lot of trials.

So how do you, how do you prioritize them? 

Alex Zhavoronkov: Yeah, so whenever we’re in need of capital, we would license some of the drugs. Or we can go and fundraise for those drugs. So now I think we’re significantly de-risked from the technology perspective and many people understand that. So we try to be very cost efficient.

We’ll look at the most, efficient way to run those clinical trials and to make it to market well to the certain point where we can license the asset as quickly as possible. And for some of the assets you actually do need to get to a certain stage within the clinical trial in order for it to become licensable because to, to, to realize the maximum value.

We have a few molecules that we have on the shelf that are actually not on the pipeline, and they used to be on the pipeline and they just have not reached the level of interest from the pharmaceutical companies in order for them to pull the trigger and and buy them. So one of the main decisions for keeping or.

You know, progressing the asset or putting it to the shelf or even like starting the program is commercial attractability. So whenever you are dev deciding on a target, you basically need to think about three main verticals, and those are novelty, confidence and commercial attractability. So if the target is too novel.

Naturally it’ll become less commercially tractable in the short term because people would want to see phase two data. Yeah. And but at the same time, long term, if you’re phase two is successful people are willing to pay a lot of money for it because there are very few drugs with broad indication expansion potential or broad indications.

That have demonstrated efficacy and that are novel. But even then, pharma would do very deep due diligence on on the assets. So the more data you have in the data room, the better it is and you can realize more value from it. Because at the end of the day, pharmaceutical companies like to buy really old established targets.

So if you are in works, 

Philip Hemme: if you have efficacy data. 

Alex Zhavoronkov: That’s why I’m saying established targets, established meaning that it’s been already even on the market. So if you look at the recent deals that pharmaceutical companies are making, like some are buying PI three K, or you know, or like EGFR or you know, KRAS, if you demonstrate molecular differentiation and just slight difference, a slight benefit compared to current best in class and you become best in class or you are basically targeting best in class, or you are safer or there is some differentiation that would provide a nice avenue for sales, pharma companies would be interested.

But what we’ve experienced is that pharma companies, when it comes to noble targets, they like to partner very, very early. And those programs usually never succeed or succeed in somebody else’s hands. When some of those pharma com people who worked with you, they migrate to another company, to another company, then to biotech, and they start another, company around this novel target and they might actually progress and then sell it to somebody. Or they in-license the IP from the big pharma company that was working on that fundraise and take it forward. Also viable strategy. But within big pharmaceutical companies, most of the you know, really advanced IP that actually makes it to market has been unlicensed around, you know, phase one, phase two, phase three.

Yep. 

Philip Hemme: I like that. 

[00:36:34] Small molecules vs biologics

Philip Hemme: The, I’m wondering on the, like most of your pipe is small molecules, right? Or even exclusively. 

Alex Zhavoronkov: We focus on small molecules because we have the luxury to do so. They’re much more difficult to make and develop them biologics because you need to take care of a lot of safety.

Issues that biologics just simply do not have. Okay. And for us we focus on small molecule oral mo molecules. So all of what we are developing is oral, so no subq now iv. And to be able to do that, you need to have really advanced AI for predicting. Liver, kidney, lung doc, neurotox, and many other everything.

Okay. 

Philip Hemme: Well, and it’s, it’s not the limit, it’s not the limit of the ai that’s biologics to engineer. A bit more complicated than small molecules from just compound and, and like compound generation.

Alex Zhavoronkov: No, no. Small molecules are much more difficult. And we actually do have biologics software that a lot of people use. So we have generative biologics and it delivers very good heat trait for peptides. For antibodies. For antibodies, you can do ADCs using the software as well. So other people, including CROs, use my software.

We do have the capability to go after biologics. And I cannot say now, if you say I going to, you know, ever do biologics, maybe because it’s temping, because now big pharma companies really want biologics. Again, for those reasons that again, safety is taken fear of usually. I mean, what is it like top on target?

Philip Hemme: What is it 80 or 90% of the top 10 selling drugs are biologics.

Alex Zhavoronkov: But it’s not it’s because small molecules are much more difficult to make. And usually if you are, let’s say if I were to go after a novel target as a biotech, which or as a big pharma, which discovered that target is like super convinced about it, but doesn’t have the capabilities that we have on a small molecule front.

I would start with nucleic acid first, right? Like I would go RNA, then I would go after an antibody with the same target, and then I would develop a small molecule. Okay? So you can actually see right now in a GLP one world, so GLP ones. All of them are biologics out there, right? Clinically peptides.

And, and I mean that, that were approved. So you’ve got GLP one peptides. You’ve got GLP one antibodies. You’ve got GLP one Gs that are very popular. So all of them are injectable and some are oral peptides. Well, those are biologics so far. Nobody has gotten a small molecule approved, even though there is a huge race and you can really had the good results.

Yeah. Yeah. orAG, Liron went through phase three successfully. So we’re waiting for the approval and at the same time, Pfizer had Theron, which is a different scaffold, small molecule, and they failed in phase three. Just because one person one patient out of like 1,500 showed liver enzyme elevations.

Mm. So welcome to the world of small molecule. And that’s why people don’t like to pull the trigger on small molecule before they actually have a running biologics program. 

Philip Hemme: Okay. 

[00:40:57] Competition and benchmarks

Philip Hemme: I’m just wondering, you mentioned a bit your competition and I mentioned it a bit on the question as well. I’m just curious on like how you, how you look at it.

Like, let’s say the, the recursion, the init, the isomorphic labs now. I mean, how do you Yeah. How do you stand out? How do you differentiate? How do you look at them? 

Alex Zhavoronkov: Well, on. Three out of the, well, two out of the top three that you kind of named. I respect very much. And when you don’t res, I think they’re like, oh, I’m not gonna tell you that.

Ha ha. But I think it’s very important to either be like extremely good at AI and demonstrate that people are using your software open source. Or have drugs in clinical trials that came from your pipeline mobile platform. Right? And you, when you are an AI company I think the fundamentals and what I’m gonna say right now, it’s probably the most important thing I’m gonna say during the interview.

If you are an AI drug discovery company, you should be looking for several things, right? So you are basically targeting. Better than before, better than traditional approach. And if it’s better than traditional approach, it could be faster, cheaper, higher probability of success. And of course, maybe higher novelty.

Maybe you invent something new. So if you cannot demonstrate those four features, you are a traditional biotech or you actually don’t know what you’re doing. And you can tell all kinds of stories, right? You can say, I’ve got massive amounts of data. I’ve got massive you know, compute computational power.

I’ve got massive number of AI scientists. I’ve got you know, Nobel laureates. I’ve got big pharma executives on my board. So actually nothing matters as long as you do not have benchmarks. Yeah. And if you are in existence for, let’s say seven years or eight years, and you have not demonstrated that you have a pipeline and the pipeline is moving rapidly faster than in big pharma and also succeeding better and faster, and also you can sell those molecules.

Right, because there are two ways to prove whether the quality of the molecule is good or not. One is to get it approved and another one is to actually sell it to somebody. So somebody is willing to pay what they earn for, for a good price. Yeah. So if you don’t have that, you actually are just telling a good story.

Maybe you’re doing something great, maybe you’re doing science for science. But that’s not what the promise of AI is. Yep. In my opinion, the promise of AI is faster, cheaper higher probability of success, higher novelty. If you don’t have that, you are, I mean, the fifth one is actually basically anti-aging, right?

So the broadest population possible. That’s what I’m focusing on, those five facts, and we’ve delivered on all five so far, however. We, and we just established a set of benchmarks. So the companies that you mentioned some of them we don’t like, we don’t see most of them as competitors. Sure. I don’t even like like an investor’s minds.

People will be seeing us as competitors, but we don’t see them as competitors very often. I see some efficient Chinese biotechs. Those are the competitors. Okay, so because they actually have, ’cause they go 

Philip Hemme: after the same medication in the same target. 

Alex Zhavoronkov: Oh, not, not they. They might go after something that you actually don’t even have, but they have the benchmarks.

They can say that with a biologic, my time to developmental candidate is six to nine months. If I’m doing biologic, like a DC monoclonal antibody. So within six months they can actually start IND enabling. And if, if, if in six months they start IND enabling IND enabling is the same for everybody, right? Or plus minus.

So it is 28 g day GLP toxicity studies and two species or three species, depending on what you wanna do. Hmm. And that takes you like nine months to a year. It actually is a long, lengthy process. That’s why it’s GLP. But before that you have the timeframe that it takes to you to go from A to B.

And of course then you have B2C C 2D but that timeframe is extremely important. And also the cost of the time or of, of, of this exercise going from zero to developmental candidates. And many of the companies in China now can do zero to developmental candidates in 12 months, but with low level of novelty in small molecules or biologics, six to nine months.

So that’s why I say it’s easier, like actually, even if you contract like Wishi Bio, WDC or biologics and you say, well, I actually want to get to developmental candidate or to the clinic in this amount of time, they will usually tell you how long is gonna take you. So if you are an AI company, and if you are doing it slower than a CRO would have done it for you, well, you’re probably doing it wrong.

So that’s why you need to actually go and compete with, well first of all, understand how people are doing this. So that’s why I think if you are not in China, you are already like a year behind. It’s extremely important. At least understand how people do it because then you are competing with a proper set of competitors.

You know what you’re competing against. Otherwise you can tell like great stories, but then you are basically a naked emperor walking around. In my case, I deeply respect isomorphic, for example, right? I think that Demi is going to deliver and you know, I have a picture of him in my office.

When I think that I’m smart, I look at him and I understand that I will do it. And on 

Philip Hemme: he’s crazy spot to say 

Alex Zhavoronkov: yeah. And then when, you know, I have a picture of Elon as well and I’m thinking, well, whenever I’m working too hard, I’m not working at all compared to, to, to this guy. Right. Or some of his guys.

And then, you know, recursion also managed to demonstrate reasonable progress in the pipeline. I like the management. So they have delivered something in, you know, the years that they, they have been out there. Also they actually had like late stage clinical assets. And then they had the courage to, which 

Philip Hemme: didn’t work very well, I think for cus, but yeah, they still have others in the 

Alex Zhavoronkov: well, a, a, a again, at least you kind of, you get there, right?

And then you have the ability to deprioritize the the question is that how you got there? How long? Right. So they repurposed originally right. And it might have worked, right? It’s just not commercially tractable. We, we don’t know, actually liked one of the assets. So if you are not following a set of benchmarks, so for example, if I give you right now I imagine that there was chat, GPT or like open ai in philanthropic and x.

Imagine that you didn’t have access to their tools, but they would be telling you really great stories that, look, I can do this and this and that. I can move mountains, I’m going to change the industry.

Yeah. But then like, where is your benchmarks, right? So how do you perform in you know, a language test? How do you perform an IQ test? How do you perform a math test? And until you can actually demonstrate some benchmarks or make people use it at scale, you cannot claim that you actually are better than stuff that came before.

So for me, I actually don’t look at ourself as an AI company anymore. I look at ourself as a company that delivers value our. And we deliver it within a certain timeframe with a certain probability of success, with a certain novelty criteria. And sometimes we have to de decrease the level of novelty on a target and actually increase the level of the, the timeframe that we take to deliver a molecule.

But then we significantly increase the novelty of the molecule. 

Philip Hemme: Right. Okay. 

Alex Zhavoronkov: And our, our program one, the one that is a demo currently, right, of our 2020 capabilities, that one has novel target, novel molecule, but the level of novelty, it’s not like the protein class that is undruggable or the molecule is not the molecule class that is like a molecular glue or something.

So we didn’t go out of the box for either, but we’ve managed to take. High level of novelty target that has never been on a clinic before and it didn’t have molecules for it, so we actually helped to develop all, everything to the clinic. Now we have more luxury to build up our benchmarks in order to reach what I call the pharmaceutical super intelligence.

If you want to get to the pharmaceutical super intelligence, you need to be able to out to, to have some benchmarks and outperform on those benchmarks. In every step of the pharmaceutical drug discovery and development. So for me, the pharmaceutical super intelligence is one AI model, one model that can outperform every human at every task of drug discovery and development.

So we are working on that for the past couple years. I’m not kind of advertising it wildly. Yeah. But you can see some of our publications and actually. Models that are accessible to others like NI zero one, for example, foundational natural and chemical languages model or Precious 3G PT is open source.

Yes. So that’s kind of the step towards biological super intelligence. Tiny, tiny step. But we’re getting there. 

Philip Hemme: Mm-hmm. Okay. Cool. 

[00:52:20] Alex Zhavoronkov on longevity myths

Philip Hemme: I seeing the time. Maybe just a last quick one. I want ask just on. What do you, what do you believe or what’s, what people believe about longevity that you think is not true?

What’s the widest belief?

Alex Zhavoronkov: Well, one, there are many common beliefs that are not true. Most common belief is that we’re not going to be able to significantly manipulate human lifespan at all. Mm-hmm. And that is the belief that is absolutely wrong because give it, you know, 150, 200 years and we will be you know, practically immortal.

And there will be different technologies getting us there. So, and people who do not believe, they do not try. And if you don’t try the pace of progress in this area. Is going to be very slow. So that’s why we had to actually build a very sustainable business model to go after more complex things, to have kind of incremental steps to go to more radical steps.

Kind of like what Elon did with SpaceX. If you want to go to Mars, you want to go to orbit multiple times, deliver the payload, and then go after the next ship, right? And then the neck. So that’s the way to innovate an aging. I think that people need to realize that they’re living in the timeframe when they could put all their resources at work and to work and achieve significantly longer and healthier lifespans.

And instead they chose to prioritize some things that are much less important. They are important, but they are much less important, like, you know, which bathroom to choose. Come on guys. Right? Choose is one of them. Like, why would you argue about that, right? Or who is able to you know, decide on the baby, right?

How many of those people are there who have to go through this challenge, right? It’s important question. But first question is like, can you live a decade longer without disease? The longer you live, 50 years longer? Without disease for everyone on the planet, how cool would that be? Right. And people who are not trying, they’re wasting their life.

Philip Hemme: WI, I saw the numbers of how much was invested in, in anti-aging, especially from like Silicon Valley, billionaire basically was like whatever, between five and 10 billion. But I didn’t see, although the results were quite, I mean, has not paid off yet, basically. Like, which believe did these guys have? Which was, let’s say one.

Alex Zhavoronkov: Yeah. So. I’ll put you in perspective currently where global scientific funding probably exceeds half a trillion dollars for one year. So if somebody invested you know, 10, 15 billion over a course of many, many years into something that they did, they did not even try. As a matter of fact, the US government alone.

Is the larger funder of science and we need to do a Deb bow and deeply bow to the US government and say thank you and thank you, the American people for driving innovation and funding innovation over so many years. Now, of course you know, many other countries are coming to to, to play in this molecular casino, but the institute, or National Institute of On Aging Alone, NIA.

The current annual budget, like $4 billion annually. Tell me about what, what came out of that? Right. So if you don’t have very specific focus, and if you don’t have special KPIs and benchmarks, you can spend a trillion dollars and nothing will happen. Right. Actually, the US also spends about a trillion dollars on, defense, right? Or now they’re purposing it as the Department of War. So we’re spending infinite ally more money on the weapons of death than on the weapons of life. And that is deeply concerning. So if we were to spend a trillion dollars on aging and actually go get nowhere. It’s still better than spending it on, you know, other things.

Yeah. 

Philip Hemme: Hmm. 

Alex Zhavoronkov: I like that. 

Philip Hemme: I’m seeing the time, Alex, I think it’s, it’s a good way to wrap it up on the, on the high level. Yeah, I really enjoyed the conversation. Thanks for taking the time and yeah, good luck with everything, with the whole pipe and executing and yeah. Thank you Philip. Yeah. Had a good one.

Bye, Alex. 

Alex Zhavoronkov: Bye bye. 

Philip Hemme: I’m impressed by Alex. Not only his passion and skills in AI and longevity, but also how globally thinks, especially on China in drug development. If you enjoyed this episode, please hit the like follow our view button. It really helps more people discover the podcast. And if you would like to support us even more, you can make a donation with a link in the description.

Hello. We also have plenty of similar videos on the channel, so please free to check it out. I will also be curious to hear what you think. So if you could just leave a comment wherever you are or shoot me an email@philipatflo.bio. Thanks so much for staying to the end, and I will see you in the next episode.