Thomas Clozel, Owkin | What does AI Biotech really mean | E10

In an online special, we’re interviewing Thomas Clozal, the co-founder and CEO of Owkin. We talk about scaling AI in biotech and how it feels to be the child of billionaire biotech parents.

Thomas founded Owkin in 2016 and I first met him in 2017. At that time, Owkin was still a baby and I wasn’t sure where it would go. Now it’s a unicorn following a $180M investment from Sanofi in 2021 and has over 300 employees.

🔗 Links mentioned in the interview:


Transcript

[00:00:00] Intro

Thomas Clozel: The thing today where the field is going is everybody agreed it needs to be more data driven, improve the variety of success of drugs, get more drugs out, is the better population, but you need to be able to take horizontal bets. It’s like finding a panda in a zoo. A baby panda in a zoo. Yeah, a baby panda in a zoo.

I mean venture capitalists ate way less bamboo, but you still really have to find them. It’s a rare species. 

Philip Hemme: How did it feel and how did it impact 

Thomas Clozel: you? Having a social impact in life gives you less pressure. Because you did your best. 

Philip Hemme: Hey, I’m your host, Philip, and welcome to the Flot.bio Show, where I interview the best Europeans in biotech to help you grow.

Today, I am online to talk with Thomas Clozel who is the co founder and the CEO of Owkin. Thomas founded the co founder company in 2016, and we talked just after, around 2017, for the first time. And it was just the beginning, and I have to say, I wasn’t sure where it’d go. But today it’s a different story.

It’s a billion dollar company following an investment from Sanofi in 2021. 118 million, one over billion valuation and has a team of over 300 employees. Amazing journey so far. And so we will talk about AI or scaling AI in biotech, as well as how it feels to be the son of two biotech billionaires. So let’s connect our line and talk with Thomas.

[00:01:26] You need a good pipeline

Philip Hemme: Welcome to us. Welcome to the show. 

Thomas Clozel: Hey Philippe, good to be here. 

Philip Hemme: Yeah, I mean, it’s good to have you as well. It’s good it worked out. We tried for, I think, four or five months. Pretty much. We made it before 2023. 

Thomas Clozel: It’s perfect. That was the goal. 

Philip Hemme: Alright, so, so, for today, I think, Before challenging you on like Orkin, like specifically, I would like to start with, let’s say, the bigger picture in AI.

I get this request a lot. I mean, it seems like they have a lot of things happening, obviously super hot topic, a lot of deals being announced. I think it’s one of the fields, maybe with obesity drugs in biotech, where you have like, outperformed everything else. But at the same time, I mean, I think it’s super complex to understand, both from a technological point of view.

Also kind of a bit of a fear of missing out, fear of being replaced, whatever, you hear this like big AGI and everyone’s job will be replaced. Can you maybe start on the, let’s say on the top level view, what’s above Orkin and from a field perspective of like, what’s Yeah, it’s a different category, especially as it is like today yeah, as we act and today, yeah.

Thomas Clozel: No, you know, so my, my vision is really like, you know, the, the pharma model has to be changed. The probability of success, as you know, is super low. The pharma has spent a lot of money to create great drugs. Most pipelines are really empty, you know many, many pharma depends on one drug, you know only.

And so there’s a lot of things to be done. And the thing that is even the, the, the, the most like, you know, probably. Doable as a, as a, as a challenge to, to, to have the pharma space is, is so many steps are based on intuition and that could be replaced by data. And you want to go from an intuition based model to a data driven model to improve the priority of success and the number of drugs that can be out.

I think that’s really where, where, where, where the field has, has to stand. Another question is how do you actually do it? And so for us, and for me, the answer is using really good data and artificial intelligence. I think that today where the field is going is to be able to replace the farmer. A lot of people are taking very vertical bets.

vertical on just doing drug discovery, or doing clinical trials, or doing diagnostics. My vision is things cross fertilize way more. You can find biomarkers that can help you find new targets, or find the right population, or the right indication, or help you for clinical trial, and even do diagnostics with the same biomarkers and the same biology.

And my vision is like, biology can help, you know, for the full life cycle of the drug. And therefore beds should be more horizontal than actually vertical. And it’s not only vertical by the, which part of the farm I want to help, it’s also vertical by modality. People are just doing, you know, AI for pathology, or AI for omics, or foundation model for proteins, but, you know only the multi modal aspect of it can capture the right biology and the right way to help.

I think today we are, the field is going is everybody agreed needs to be more data driven, improve the variety of success of drugs, get more drugs out with the better population and better, but, but you need to, to really like being able to take horizontal bets and never go to the vertical side of it.

I don’t know if that resonates with you, but. 

Philip Hemme: Yeah, definitely. I mean, and I think that’s one of the big selling points of, of Orkin. You should have written to Orkin. Yep. Where should we go with it? I think, but what you mentioned, what I, what I like also, maybe you can comment on there. It’s like, I mean, you mentioned.

optimizing drug discovery, optimizing clinical trials, optimizing, let’s say, I mean, in there you have patient certification, optimizing, looking for biomarkers, all of that. And then even then optimizing, let’s say, really in the healthcare and how other drugs are prescribed or given. And let’s say in a better setting, you can optimize a lot of things.

But maybe there to understand a bit more, because I mean, you guys, you are trying to go For the horizontal, yeah, I think if you take these different categories, I mean, how can you are going very horizontal and try to cross except wet lab? If I’m, if I understood, okay, so you’re really going full horizontal, but if you go into the categories, I mean, if I look, there is.

In the drug, yes, I’m really on drug discovery doing whatever, like virtual cells. It’s the first one that comes to my mind, or like really, like let’s say specific step of the drug discovery. You mentioned pathology, I think like path. ai is probably one of your biggest competitors, really on like data.

pathology, but it’s pretty big also, but on the, and then the bigger differentiation is, is the ones who have their own pipeline versus the one we do deals with biotech and pharma companies with enough pipelines. 

Thomas Clozel: So can I, can I comment on that? 

Philip Hemme: Yeah, yeah, yeah, go for it. I’m trying, I’m really trying for a long time.

I try to 

Thomas Clozel: be, we try to decipher the categories. So, I mean, like the question of the pipeline is a bit different. First, the big question is what is the problem you’re solving, okay? For Okin, we are trying to solve the precision medicine problem. Why precision medicine does not happen? I was a clinician, and you know, when you treat patients the same way, pretty much, it’s frustrating.

So for us, like, the view of precision medicine was, we don’t have the right diagnostic tools to be able to augment the care in predicting response to treatment or survival or whatever. And we just don’t have enough drugs. And when we have drugs, they’re not matched to the right population. So the matching of population to biology is a real problem for the pharma.

So we believe that the problem we wanted to solve was precision medicine and therefore we had to bring diagnostic tools, AI enabled New therapeutics. In the right subgroups. So this is really the ion. We are, we, we, we take, so this is, I mean, we are trying to solve this problem. The only solution is to rebuild a full end-to-end pharma.

That’s really the thing we do. So I think like, how to defer the field in general. I think it’s, it’s first it’s not about creating a pipeline. Not creating a pipeline. Everybody has intake bio to create a pipeline. The big question is our pipeline is diagnostics and therapeutics other people is therapeutics.

The question is. Is your pipeline really data driven? A lot of companies, you know, have built a really amazing platform. For example, Knowledge Graph platform, but then had taken experts to choose the pipeline. So first the question is, is your pipeline completely driven by your technology and not by a few experts?

Because if you do a great platform, great technology, but then you go with a few expert driven decision, you’re not really making a revolution in the model. That’s the first point. And the second point is, Is your pipeline better? I know that the real pharma, have you found things that the pharma would not have found itself?

Not only faster, but better. In indications that failed in the pharma, glioblastoma, mesothelioma. And, and if you want to really be there, you have to wait for clinical proof of concepts, which is a phase two B in oncology, for example. And you really have to show clinical proof points. I mean, so the question is, is your pipeline linked to the platform?

And is your pipeline proven to be better? And there is no really other option to show that that. And that’s the, and the third point is, then this is what the first, the second thing you can prove. So the first thing you can prove, but then if you want to prove like even bigger metrics, like probability of success, can you, for example, double probability of success?

What we want to do at Token, go from 10 to 20%, for example. You really have you need a lot of trials. You need a lot of clinical trials to be able to show it. I think the problem as well is nobody can do the claim today that they have been able to improve productivity success by better chemistry, by better clinical trials, by better targets because you need so many trials to show it that there is no big claims that can happen.

And this is the big problem. Acceleration has been shown. Augmentation still has to be proven. And we are all here in the same game. And to finish on that, the solutions part of it, the solutions, the services. You know, for me, the services is only a way to cross fund your own pipeline. You have to get your own pipeline, otherwise you have no IP.

You cannot be a big business being a solution provider. So for us, we are doing like, you know, a lot of big contract with Pharma to learn from them, but also to get enough cash to cross fund our diagnostic and therapeutic pipeline. For me, this is how I view it. But you have seen a lot, so you can tell you may have other.

[00:09:24] How Owkin is scaling AI in biotech

Philip Hemme: Yeah, no, it’s great.

Can you, and there’s a, at least a few questions I wanted to jump in, a few points I want to jump in, but Maybe, can you be a bit more specific or mention, let’s say, the other players to see how you, how they, how they compare to you? I mean, without like trash talking them, but just more like about 

Thomas Clozel: I think first, so we’re very different.

We AI for biology because we believe that biology is the background to discover new biomarkers that can feed the full life cycle. What we don’t do is chemistry, okay? Why are we not doing chemistry today? Because I think chemistry space has improved incredibly with AI the space of molecules, but there are still things that are unsolved, like protein to protein interaction and others.

And today, if you ask me, should I, should I use AI to help chemistry? Yes, absolutely. But I still need the real, the real good chemists in the lab. Can be from Germany, can be Switzerland, and I still believe that, you know, you really have still to get really good chemistry in the wet lab. And, and AI is a help, you know, so today this is why we choose like to, to not do chemistry ourselves for now.

The rest, finding new biomarkers, new biology, new targets, match it to the right population, use these biomarkers to find the right subgroup and indication, improve clinical trials, synthetic control arms, co write adjustments, and develop a dynasty tools to be able to deploy the subgroup is really where the rest, but, but what we are doing and, and I think nobody’s really doing that today.

So first we, I think the only ones that do the end to end pharma, we really want to recreate a mini Rush, if we can. Very very, very mini. Very small. But, you know, I think this is what we wanted. The best step. Other Yes, the best step. Other players are A lot of other players are in the chemistry space.

That’s the first one. Other players can be in the biology space as well. Try to discover new biology. If you take, you know, like in situ or Or, you know, recursion. They’re amazing, but they’re using cell lines. For me, as a clinician, you know, if you don’t have the macro environments even if you’re using salines or organoids, you still miss the microenvironment, the immunity.

And I, I don’t have a huge belief in finding new biology in mechanisms with such a translational gap. And for me, the only way to reduce translational gap is using patient data. You still have to validate in the lab, but you know, as my, my CEO is always saying like you know, cancer was treated in mice 30 years ago.

So we have to find a new model. And for me, the model is patient data as well. And biology has to be discovered in very large patient data sets, deeply phenotype using new modalities, such as partialized transcriptomics. It’s a very important one. We’re doing this big mosaic project. And you have to get with an angle of finding a biology that is not saline driven.

Philip Hemme: That’s Isn’t then the a bit challenging, but even today, I mean, you said you had a wet lab as well, but you don’t do the chemistry, isn’t It’s the optimal to combine, let’s say, in situ with you guys. So you have the really patient, stratification, micro environment, bigger like patient picture at the same time with the biology, cell line, really the chemistry, biology, and you can kind of combine and cross 

facilitize.

Thomas Clozel: So we, we doing it too. So we we, we created the Okinaw and in Okinaw, what we are doing is using organoids with Sun Bioscience in Los Angeles, they’re amazing persons. You should be interviewed like the, the, the founders. And so we, we’re using organoids in the lab as well to, to test our hypothesis, but first hypothesis are being formed in patients.

That’s the difference, you know? And then we, we use a really big lab in the loop approach where we are creating perturbation screens to improve our models. So I’m not saying like, you know, the lab is still a very central place, but I still believe the first type of biology has to come from the patients and we show it in the Mesothelioma paper we published in Nature Medicine four years ago, we could discover a new mechanisms using a prediction.

From patient’s digital pathology and reverse engineering the model to understand new cell biology or gene biology that could explain different phenotypes. And there was a new discovery in, in, in, in, in mesothelioma. But the first step was having a lot of patient data access, multi model, making a prediction and reverse engineering.

And then you go to the cell lines and the guideline model to validate. But for me, the first thing has to be having 10 times more data and data. And when I say data, it’s also private data in the best hospitals that are from the patients with cubioniders. I don’t believe in public datasets. I mean, you know, like if you build your company on public datasets and some have done that, you know, that’s, 

Philip Hemme: I mean.

The quality of the data influence massive. Is the quality of your models or like the, and it’s directly linked, so you need really high quality data 

Thomas Clozel: and diverse and Yeah. No, because, because dive no, but diver. The thing about diversity, everybody believe, everybody always remembers that patient diversity matters of course, but doctors, diversity matters too.

Patients are being treated very differently in Europe and in and and in, in the us even in France. You go to Paris and you go to Y, you have a different treatment probably. And for the same tumor, and so, you know, if you don’t capture the type of heterogeneity of, of biopsy, of treatment and everything, you have a model that is not robust.

And by definition, AI is not robust. And this is why you need to have European data, as we do at Okin, and not only US patients. This is why data from Fat Tyrant, for example, have no use to do research. And also just because the clinical data, they’re not multi modal, but also just US based. in community centers.

We don’t believe in community access so much. We believe in best academic centers where you have QOP non leaders that help you curating really high quality datasets to go to the single cell level, you know. 

Philip Hemme: Do you think Rush agrees with that? 

Thomas Clozel: I think Rush agreed. I mean, Rush, I mean, I mean, come on, like today, 

Philip Hemme: it was very big. 3 billion, the deal with Flatiron or something. 

Thomas Clozel: So the deal was huge. The deal was huge. And Rush wanted to, the story of Rush is simple. Rush wanted to make the first move with Flatiron. They really wanted to do, to be the first. Then about Flatiron, Flatiron was super useful for market access, useless for research, very, very little useful for synthetic control arms.

Synthetic control arms using just Flatiron data is not proved to be very efficient. Then what they did is about Foundation Medicine, which was great, they tried to reconciliate both. There was only like 30, 000 patients, you know, in common. What, what has, you know, and, and, and, and they just showed that it was very hard for them to get multi modal data this way to do really great research.

I think now with Aviv Regev, they’re really trying to rebuild. It’s. You know Spatial Transcriptomic Data Sets, as Okene has done before. You know, I’m trying to go to building a new data lakes, and I think they’re taking the right strategy. But I think that the Flatiron, Flatiron was great for market access, not good for research, not very useful in clinical trials, and the reconciliation was very hard.

Tempus is doing better in taking, you know, the multimodal data from scratch. What is different from us is we, we are embedded within hospitals, so we can continuously enrich the data sets. We can add new things, Spatial Transcriptomic, long reads, single cell. And so we can always like, you know, enrich this, and we don’t have a static view like, you know, this type of companies do have.

And I think like Rush, Rush has been a bit jealous of Tempest’s approach in the past, and I think now they’re really trying to take things from scratch on the Genentech side. 

Philip Hemme: Okay, I, I read the, I 

will link it below, but there’s a good article in Endpoints about the, the head of AI from Genentech slash Roche and what he, what he’s building.

And it’s, 

Thomas Clozel: it’s. I think it’s, I mean, yeah, I think it’s Regev I think Regev is trying to build something amazing. I mean, I think what they’re trying to build, but they started, I mean, they started from scratch and they do exactly like us, patient data, special asymptomatic. And I’m so happy that we have inspired Rush.

Maybe they won’t admit it, but we’re doing this for them. 

[00:16:46] Augmenting AI in biotech

Philip Hemme: Get this quote there. Now 

that’s, that’s, that’s, that’s a lesson. On the I want to know a bit on, on, you said we, you haven’t, our AI in Biotech hasn’t augmented, or hasn’t shown the augmentation yet. Can you go a bit into what’s the, I mean, differentiating basically the vision of, let’s say, what could be done in five years versus what has been done really today and let’s say in the last five years and where do we really stand?

Thomas Clozel: Yeah, I mean, I think, I think there’s lots of great companies around, you know, I can name them, Xantia, Insilico. You know Path AI, Page AI, Recursion, I mean, I think they’re all great, and we’re all doing quite different things, and we’re all a bit playing with the pharma. I think what happened with the pharma today, the pharma has first, first, the story of the pharma and AI.

At the beginning, they wanted to do a lot of internal innovation, build the things internally. Then they started to realize that the problem is they cannot do it for two reasons. They have a very hard time building their own data lakes, okay? Very hot, very hard time to regroup their own data. And second, they’re not hiring the best data scientists.

Why are not the best data scientists going to the pharma? When you’re really good data scientists, you want to work for Google when you’re young, not for, you know, rush on a valley. And it, it’s sad, but it’s true. I think, I hope it’s gonna change. I really believe the best in the future will work for the big pharma.

But today the, the truth is the young data scientists super brilliant. They wanna work for, you know, like probably a, a, you know, like a a, a metaphor or Google. I, I don’t know, but it not really, it’s not still not super attractive. And, and I hope it’s gonna change. And then they really understood they have to do more external innovation.

And they tried to play with a lot of different players around. And I think we, they all make proof of concepts and we all make proof of concepts with them. Sometime we did not bring pipeline value. Sometime we were bringing things that were incremental, but then the CEO was coming and saying, how my pipelines changed thanks to this company.

I want to see my pipeline changed. And so you had the height of discoveries that was amazing and tell you, Hey, we did that. But the CEO was, I want to see difference in my pipeline. And I think the, the, how you can bring direct pipeline value really mattered, you know, I think in the for the pharma. And I think all of us, maybe we went sometime to, to, to upstream to show pipeline value.

And this is why, you know, we did a lot of clinical trials improvements to really show things that were more direct than very early discovery or diagnostic tools to improve, you know, but I think, yeah, this was a problem for all of us. And I think today, the pharma. Wants to, to, to have less partners, a few partners, and there is, they will have one chemistry partner.

They have a lot of great companies around, you know, in Silico, Exentia, amazing. And they, they might have a one for biology, you know, us, and another. But I think in the future, they, they, they, they have less partners. Another question is why the pharma has not really acquired, but BioNTech was great team.

They were really amazing team. And they were, I think they made a really good move. I think today, you know, I think the pharma is still a bit scared of the cultural difference. They would be scared that if they acquire a company, people would leave. And I don’t think it’s true. I think, I think like people understand that it’s complimentary and I think people will, will stay.

But I still think this is maybe the reason why the big M& A has not really happened, like the Flatiron level. I mean, maybe, what do you think? What do you think? Why, why is this M& A not, you know? I mean, 

let’s say I mean, I can see that they are trying to wrap their head, their, their head around as well and probably still like going one foot at a time, like touching the ground, seeing, okay, and testing here and there.

And I think the culture thing is probably like, it seems like a really big shift. I mean, you guys, you call, I mean, you always have like biotech versus tech bio and we did this tech aspect. I can imagine, let’s say, data scientists or AI engineer, it’s not just the salary, it’s the package, but it’s also really from a culture point of view.

If you work for a big tech firm, or even for you guys, it will probably fit, or it will be much closer to that kind of culture that the person needs, or where the person will strive than, let’s say, a really big farmer. I can imagine, so. This, I mean, to your point, this, I can imagine, it’s a big shift, I mean, Even, I mean, for me personally coming from, let’s say, more tech, I mean, half tech, half biotech, but I can see in the industry there’s really a big mindset shift.

And I worked for, for some, like, a bit larger organization, biotech or pharma, and just the way they’re working and the culture, they’re great cultures, but it’s, I think, very hard to fit with someone who’s really like an AI or in tech. It’s really from a, just from an operation thing. I mean, I’m thinking about, okay, they’re still working, you know, on email and PowerPoints.

And they don’t have a project management software and some basic things, which for me was like, wow, it’s insane. And you’re a billion, you’re a billion dollar company, but you have no like project managers whatsoever, which I can imagine for 

some. Yeah. And I think things are a bit changing, you know, taking the part of Sanofi, like, you know, the head part of Sanofi looks more like a startup type of office.

And I think people are a bit less corporate. The mindset is a bit changing. You see that? And and but as soon as in the pharmatics, we, we start to have like people like CEOs or our head of BD global that has just had to be really like interested in the topic, like profoundly or even chairman. Take Sanofi, you know, like in Poland, really love AI and like the new CEO, the new chairman, sorry, for the good AI is really has a deep, you know, interest for it.

You know, take BMS as well. You know the team at BMS is really interested in, at the C level role. I mean, I think at the same time, it’s adding C levels that truly believe not, don’t only want to have an AI bread, but really want to understand with companies like. Like all of us, where, where there will be impacts.

And I think this is also, like, where we need to go. There is other company, the Pharma, that have way more classical CEO ship, 

Philip Hemme: Do they, do they really get it? Or are they still in the more curiosity, like

Thomas Clozel: I don’t know. I think some are really, I still, I still did diving in it and I think it’s not just only about being an AI company, having a brand, they really want to show things, but, but they want to show that we can bring pipeline value once again.

And this is where I think they want to say to the world, this is pipeline is AI driven and that, but, and, and this pipeline has shown success where we failed. This is where the hottest spot is for him. 

[00:23:05] How AI can work for small/medium biotech companies

Philip Hemme: I don’t know. So, I want to switch gears a bit from, I mean, we talked a lot about big pharma and pharmas, but, I mean, you have also all the, like, biotech companies, let’s say small and medium size who are, who are pushing a lot of, like, innovative, innovative drugs, and I’m curious in their case how, like, how they can really use it in biotech.

I mean, from what you said first was like, Oh, like basically Big Pharma or anyone has to be like AI first and have an AI pipeline, which means even the biotech company would have to internalize some stuff or should they like work with some partners for, let’s say, for, for discovery, for the clinical trials.

What’s your, or like, what’s your view there, really, for biotech companies, let’s say, or whatever, select this I mean like, clinical stage biotech company. 

Thomas Clozel: First, I think some are very involved in it, like, take names like GeneMab, I’d also have been very driven in AI and, and have a lot of different stuff.

Of course, the prime of, you know, I think like the, for, for AI companies like us, the, the opportunity is amazing to be able to, to get in licensing drugs from those, and, and maybe also just to, you know maybe have revenues, royalties. It’s very hard to get royalties with the pharma, you know, you get clinical milestones.

But royalties maybe, like, the biotech will be more open to it. Even if you prove your value. And I think so, this is it. The problem, of course, is biotech has less money to spend and doesn’t participate in the business as much. So, you know, for companies like, like the, the, like the TakeBio ones, the big pharma is a bit like, you know, can invest in the, in the equity rounds.

So it’s a bit more structural in a way. But I think they both are very important at different levels to work with. 

Philip Hemme: But not just like partner with you guys, but more how you perceive it. Basically, what should they do? I specifically, I got like, what do you think they should do? Because I got the question from, let’s say, VP Discovery at the biotech company, had an employee with billion having next gen biologics platform.

And she was like, Oh, what’s, what do you think about AI in biotech? Like, what’s basically kind of, what should I do? How can I use it in biotech and how can I use AI? Like, what do you, what do you, like, what would you think? 

Thomas Clozel: If you ask me for, for like a single asset biotech, you know, midsize, it depends on the fields.

I wouldn’t go for what brings the most direct value, which is for me finding the right subgroup. Everybody knows that finding the right subgroup for your drug can improve the probability of success. But matching biology to the right population, finding the right subgroup, or maybe the right indication, is not easy for a startup.

Most oncologists, they never know which indication they want to go to, and then they have to build very expensive basket trials and take a lot of risks. I would go, if I was a single asset, to take a company that can help me on identifying the right subgroup and the right indication for my drug. Before I go, help me positioning the drug really well.

I mean, of course I’m taking Akin use case, but you know, I, I would, I would start with this use case very simply. And , I can do that. 

Philip Hemme: And what about a non, a non OIN use case? 

Thomas Clozel: no, but about new case? No, I think, no, because, because the problem of a single asset is not having a better pipeline, a bigger pipeline is, is even more having the success of your first drug.

Right? In

Philip Hemme: this case was a, was a platform use case more, it was a platform company actually with, with research. So. 

Thomas Clozel: So in the platform Nobody, I would, I could also work within Silico, you know, or on, on, on the improve, on, on augmenting my pipeline with some of the new drugs, you know, why not, you know, I think it both are complimentary.

I mean, I think I would, I, at least I would choose one use case I don’t want to explore but I wouldn’t try different things at the parallel and I would really create a. A little common lab where people can meet. I mean, I love this idea of really having a common lab where people can exchange a lot.

Or, I would merge. Why not merging with a mid size? I think mid size biotech and AI scale apps should merge together in the future. Because this is where I probably could work as well. Why not, you know?

[00:27:18] Always looking for more data

Philip Hemme: It might happen actually. Let’s see, let’s see. No, that’s, that’s, that’s great. I mean, we, we touched the, we touched a lot of things already.

Let me just let me just check. Let’s talk a bit about, about, I mean, on the competition, I’m still like, one thing, I’m still wrapping a bit my head around is like, you, you said something, I think a podcast that you mentioned, you wanted to have 10 times more data, or let’s say 10 times more and better data than anybody else, like, where do.

Where like, this is a, a wish, but where do you stand today? Like compared to the biggest? 

Thomas Clozel: No, we are, I mean, we have, we, we look, look in a look in our papers. We publish a court in Otium on 3000 patients, you know to find some new biomarkers in new biology was in nature medicine. Today we have another thing called Mosaic Mosa is the largest special for systemic efforts.

We are getting towards 7,000 patients dataset with five sensors. This is the goal. And this is really to build, you know, a huge effect on the, I think it’s going to be huge for the, for the field, because we’re going to be able to show that we can have huge data sets. It’s a 15 million investment, 5 0, you know, and we are building the right pipeline to analyze it.

Spatial transcriptomics is the only way to see visually genes expression and, you know and and, tissue organization. It’s, it’s an amazing play for computer vision. And I think this is something, this is the type of modalities that can also capture the macro environments. It’s the only way to capture the microparameter effects.

And so, you know, we’re working with like, you know, Nanostream, TEDx, and we’re really building this and five amazing centers to build this thing. And we were first, and we were really happy to see some competitors climbing on our feet. Like, no HH, no formal alpha coming afterwards. Once again, happy that we inspired people.

You know, it’s, it’s, it’s we are really building something that that I think, but it has to be big, because if it’s not big, if you don’t have a size effect, you cannot validate things in the right way test it train things, test things, find the right hypothesis, and you’ll never know if it’s a data number thing or just a, you know, and, and you need to also capture the heterogeneity, so this is the type of efforts we are making, very big bets, gross products and we, we really aim to go super high.

Philip Hemme: Okay. Right. In the, I mean, that’s, sounds high, high enough, yeah, high enough, I mean, in the thousands of patients, it’s already high. And I guess what you said is, the quality of the last quote, 

Thomas Clozel: the court published. Yeah, I mean, everything is about the quality and the multi modality. And I just want to just re explain why multi modality matters.

So you understand special asymptotic matters to capture the microenvironment. But in general, when you work in cancer and in oncologies, right? So when you work in cancer, you never know which modality will give signal to the prediction. So if you do only pathology, like PATH AI, you know, PAGE, sometimes pathology in lymphoma does not bring a lot of predictive signal.

You really need other modalities. Even if you anchor your model of pathology, you need to add multi omics. Sometimes, you know, CD scan or MRI are, or proteomics, you need to add other modalities because you never know where the signal will coming from. And so the multimodal approach is, is really complicated.

But it’s the only way to capture how you can understand the different skills of biology. So I mean, for us it’s not only about being multimodal, it’s also being multiscale. How you go from a gene to a cell cell to a tissue a tissue to a disease capturing causal signal. And this is the very thing.

So we, we really want to go to, to, to try to understand causality more than correlation. And the only way to do that is being multi modal and multi scale. We’re not there, but it’s a big work. And this is really the goal. 

Philip Hemme: But that’s, that’s where also AI makes total sense, because you can process tons of data very rapidly and figure out.

the relationship, the causality, like, really well and much, much better than any human can do. 

Thomas Clozel: I mean, yeah, but we haven’t proved, I mean, nobody, even us, we are still trying to prove a good explanation. I mean, before causality, you have other, like, success levels, like good explanation of a disease or something like that.

We have to go there, too. But, you know yeah, I mean, and what’s also interesting about data is today, you have to get to foundational models on every data modality. So here you need like millions of patients, you know, if you want to build a foundation. And Okin is, is spinoffing his foundation model for biology.

The first one that is going to be multi scale on our dataset. But we are really, you know, being able to work in millions of patients on pathology, millions of proteomic patients. And so this is where you need like, and then to connect the scales together, you need like multi modal longitudinal datasets.

So you need different types of data and numbers when, for the tasks you want to optimize. But you really like, you know, we, we for you, the, the, the number of data annotated, you know, not annotated, sorry, unlabeled, because you don’t need that. It’s going to be big as well. 

Philip Hemme: I looked into one company, actually, Valo in, in Boston.

I know quite well, David Berry from, from Flagship for, for almost 10 years, actually. Yeah, me too. And, To my understanding, I mean, they’re going a bit in a similar space as in having a lot of human data and then building a pipeline on it. And from what I see or from their presentation, they are also in the thousands of patients scale.

So how does that, how does that work? 

Thomas Clozel: Yeah, so I, I, I, so I’ve never read, I, So first, like, I think like the difference between that, I mean, the, the, we, we have discovered like, you know, very reference ways to analyze pathologies that we publish, you know, people are using and the multi modal, I, I haven’t really read so many papers in the, in the literature, in nature medicine, like us, or about vital technologies.

So I don’t know much about it. Okay. I haven’t read like science, so I don’t know. And then the thing is like, we are building like, you know we are building like, but I think David is great. He’s probably have an amazing team. No. I mean, the idea of in licensing drugs is a bit similar.

We’re also building dynasty tools. And we are partnering with 41 of the largest hospitals in the world. I’m not sure which partner For us, it’s really a win win partnership with the hospitals. We try to publish, we try to work with them. We are generating data against access. We also try to build a federated ecosystem of hospitals.

180 of the best Filipino leaders. I, I don’t really know about the, the, the, the, the partnership and the quality of the data. I can say, you know, so I, I really don’t know much about it to comment. 

Philip Hemme: Okay. I don’t know exactly either, but I mean, they mentioned acquisition of data, so I guess, and I guess they must, I mean, the data has to come from somewhere and probably the leading hospitals in 

Thomas Clozel: when you acquired that, so.

[00:33:58] Anonymization of data and federated learning

Thomas Clozel: So the, the thing about tech first, I’m a doctor. I like data to stay in the hospitals where they’ve been created. I don’t like to acquire, I like to have a remote access. I don’t need them to acquire them. I don’t acquire, we don’t acquire data. We have remote access to it because it’s privacy preserving. I believe in privacy.

Yeah. If you, if you wonder, data is c is the new oil. Try to wonder for every company, what is their policy? What is their vision on data access? Is it a sustainable way to access data? Is this, is this good for the patients? I mean. For Okin, it’s easy to privacy preserving access, you know, federated, to break the research silos.

We really want to break research silos and competitive silos between centers. You know, I think that at least it’s clear. We were the first one, and we’re still number one in federated learning for clinical data. This is why Google Venture came in years ago. I think like today, today I mean the broking of data, it’s something that I believe will happen and everywhere, but it’s not a very amazing vision on the, on, on, on data to, I mean, as a doctor, it’s not mine.

Yeah. 

Philip Hemme: You, you switch to the next topic on federated learning. It sounds amazing, honestly, and, and you’ve proved that it’s worked, I mean, the, the Melody project was, was, I think it was 10 Pharma, you proved also that it worked with, with hospitals. Yeah. But the other model of acquiring the data also seems to quite work, especially in the US.

I think it’s more a question of also of price, I mean. If a hospital, I don’t know what kind of deal we’re talking about for this, but if it’s in the millions to acquire, I mean, I think a hospital could be interested and I, like, I guess there’s a question of price and then there’s probably maybe some workarounds around privacy.

Let’s say if you anonymize the data, then you cannot identify the patients. Then it’s, it’s, so 

Thomas Clozel: you cannot, that’s never true. Philip, that’s not true. No, it’s not true. If you take, I, I put, I put 40 hackers on any data set. I come back to the identity of anyone. Today, there is no anonymization of data that works out.

We have to be very clear. None. 

Philip Hemme: None? Why do, why do Apple claim that that anonymization of the 

Thomas Clozel: Well, I’m pretty sure to try to give us data and put like 40 hackers. Why? Take a, take a pathology slide, okay? Pathology slide, by definition, if you take the name out, whatever, you can anonymize, absolutely anonymize it.

However, at Token, we’re able to predict genes based on this slide. We can predict signatures of genes on the pathology slide. If you predict for the signature of genes, you’re coming back to the genome of people, and you can really get back, and you can cross with genealogy database, 23andMe, whatever.

Today, I, I, I, I personally don’t believe that there is 

Philip Hemme: You predict the gene on the top level, you don’t predict the sequence to then identify the patient, like 

Thomas Clozel: You sell predict, you still predict signatures of t-cell or B cells, which is quite, you know, identifying. However, I mean, I’m just telling like the, the, the roots that could, like, predict more of this is the tumor.

But you know, you can still, I, I I don’t believe that, I mean, we have to prove, the only way to prove that something is really, is to put a dozen hackers on the, and see if they cannot come back to it. I mean, for me, I mean, I mean I’m not saying that federated or aggregation and I, I’m still think that it’s fine to buy data sets.

But I think both have to happen. In Germany, very hard to do data brooding, right? So that works for the U. S., but in the global scale, correct. 

Philip Hemme: I know the German problem, but I am curious on this, like, where is the model because, I mean, when you talk, it sounds like, yeah, it’s good to hear you both, because when you talk, it sounds always like the only solution is federated learning, but that’s from the reality, I think, yeah, okay.

Thomas Clozel: Yeah, reality is between both. However, just can I say that, however, there have been deals in the past that were exclusive deals on data generation with hospitals and scale up, and this is not okay. We have to all be okay that we are treat, we are trying everybody of us everything by is trying to treat patients and help medicine.

And if you’re making exclusive deals that your data cannot be even used for internal research within the hospital. You know, of course there is no commercial, right? When you generate data, that’s okay, but you still want things. You wanna, you don’t wanna have a hold on data sets for anybody else.

And we’ve seen it in the past. I don’t wanna name people, but you know, we know. That there’s been deals where the people couldn’t do research, they couldn’t participate with anybody else on even the, the, the, the, the, the data sets that were not generated before, and I think this is not okay, I mean, we have to see that if you put a hold on data sets and a key lock that nobody is able to use, you’re not helping research.

And so we have to find, we have to promote companies that are really trying to, to, to help open this system. 

Philip Hemme: Well, I guess this would be a, I mean, this would be a very challenging It may be a very challenging problem because I mean, data is basically new gold or whatever. And your model also, I mean, quality of the data improves.

So if you increase the quality of the data or you decrease the quality of the competitor’s data, maybe your model can be better or whatever. I mean, basically what Google is doing with YouTube, they’re not really sharing all the YouTube data. They’re Gemini. I mean, 

Thomas Clozel: Of course they’re not. Of course they’re not.

No, no, but, but I mean, yeah, I think you agree. I mean, but, and no, no, no, they wouldn’t. But I mean, just take the example of telemetry. I mean, take nice example of telemetry. Melody was the first time people were collaborating and it was the largest database of small molecules virtually. But take the example of glioblastoma, very important to my heart, a brain tumor.

You know, glioblastoma is 35 clinical trials. There is only one positive. I think it’s Novocure, an amazing company really. And so if you take there is so many silos of negative trials. Happening in different pharma anywhere, and, and they will never share it. Sharing, nobody will, they will never share the negative Klingon trial.

If you do federated learning on, on that, you could really capture new, you could have enough data to find some, maybe some new biomarkers for glioblastoma. It doesn’t mean you will find a new molecule, because there’s still a long way to do, on the chemistry side. But however, it’s very frustrating to know that people have silos of negative data, not collaborating, and that there is a technical solution.

Why, why are we not doing it at Okin? They have customized a small market, so we will have very low support from the pharma. We won’t have enough money to do it, it costs a lot of money, and it’s a big headache to work with multiple pharma with federated learning. So, you know, we’re just not doing it because we don’t have a business case.

But it should happen, I mean, the data are here. 

Philip Hemme: I mean, yeah, yeah. Feels a pity. Yeah, let’s see. I want to jump on what you said about the, that it’s mostly a European problem. And I think that’s, you’re touching a point where I think it’s probably one of the biggest value propositions of Orkin as in, I mean, GDPR and let’s say Germany, even Switzerland are super strict on, on data privacy.

So I guess, and also even for you as a Orkin, as a business, I mean, I think that most of your employees are based in, let’s say Paris or Europe. And I think also most of your pharma partners are Europeans as well, even though you have some in the U. S. as well, so. No. Not true? 

Thomas Clozel: No, no, no, no, no. On this side of the, I mean, no.

First, I don’t think what is a pharma or pharma is that all global. But but no, no, no, I think today we have more U. S. partners than Pharma. In the Pharma space, we have more U. S. partners. 

Philip Hemme: Is the, are the U. S. one partnering with you for the European data or as a, from a global perspective?

Thomas Clozel: Everybody’s, everybody’s, people are partnering with us for two reasons. We have global data. We have as many hospitals in the U. S. even more than in Europe. But we can, we have both. The second is the fact that we can enrich the data. When you acquire data, you cannot enrich it. You cannot go to a new analysis, accessing the samples.

We can, theoretically, access samples from the patients. Of course, in a, in a very strict ecosystem, and with the but, you know, we work with the patients, we are really relation, we have a relationship with hospitals. And so people work with this, this, this federated access data worldwide. Then they work with us because we, we can bring the AI layer and the biomarker, and then we can implement this within the life cycle of the drug.

We are a one stop shop for the pharma in biology. This is why they work with us. We can do clinical trials, diagnostics new therapeutics, indication discoveries, so we’re discovering new targets. So I think today they would like to work with us for that. And and of course I work on the chemistry with great players in Silico, Exentia, whatever.

I think today this is where, where we stand but we don’t give access to our data anytime. So nobody has access to our data. We are in the middle. 

[00:42:18] Becoming a biotech company

Philip Hemme: Yeah, that’s good. Maybe last point on Oken is the, the transition to being let’s say full fledged or, or full biotech. I mean, you You were saying, I think you, I heard you you, you want to reinvent the pharma biotech full stack.

You mentioned Tesla building everything end to end. 

I mean, 

and I mean, you hired some, some, some big people. I mean, this is Andreas, the CFO of Merkur Partners who joined you and he built Merkur Partners, co founded Merkur Partners, co founded. That’s a biotech company. You had the global SVP BD from Sanofi.

a really like heavy, really biotech, biotech people, I would call it biotech or drug development people. I’m curious on, on where, how, how did, I mean, how are things going there on like building your own pipeline? I mean, I think it’s been, these people joined a bit more 12 to 18 months ago. So how, how did it go there on the Like on the pipe.

Thomas Clozel: Yeah. I mean, so, so first we, we build a tech. Yeah. Thank you for mentioning them. You know, we build a tech bio and it’s led by Jean Philippe Vert, who is one of the very top specialists in the world of AI for biology. You know, he used to lead a Google brand in Paris and, and professor at Berkeley and then Paris at Le Muin.

So on the, you know, we have an amazing data science lab and people that we also hire like a really great medical team, you know, Vassily Sumilis, who is a professor of immunology and, and immunology. To have a really medical expertise because we believe it’s important. On the pharma side, what’s, what’s hard is to find people from the pharma that are platform minds, you know, that have this that are, they have the sense of urgency and really want to accelerate things.

It’s not that easy. And I think, yeah, Albon de Sablier, our Chief Operating Officer, came from Santa Fe, and he has this sense of urgency. And Andreas brings this, like, you know, four IPOs. So I think we are building up. We’re hiring SVP to discovery. You know, we are hiring people in the labs that have been through you know, so I think this is the hardest part, for sure.

Finding the right people that can bring this this value. Still, understanding the transactional side of the tech is very hard. But I mean, there are people in the pharma space that do that. In the venture capital space, You know, venture capital is another story. Very few people get both, you know, understanding what representation learning means, how you find the right data and understanding, you know, the, the, the, what is a really good pipeline or the diagnostic.

I mean, it’s pretty much a very rare species. It’s like finding a panda in a zoo. You maybe have one every a hundred zoo. I think like a mix of baby pandas around. Yeah, baby panda in a zoo. And, and and I think like they eat. I mean, venture capitalists ate way less bamboo, but you know, we, we, we still really have to find them.

It’s a rare species. 

Philip Hemme: Yeah, that’s, no, that’s, and so specifically, where do you stand now in the, in the pipe? Can you, do you, can you share anything? 

Thomas Clozel: Yeah, so we, we’re going to have, yeah, I cannot share with anyone else, but we have two dynasty tools approved, you know, the first one is being deployed. It’s called MSI Intuit.

The second one is gonna come very soon. We have the first phase one coming in. And we really, you know, are working with the pharma. We are on 20 years of chemistry. A really, like, best in class drug. Everything we built in diagnostics is first in class. Everything we do in diagnostics, in therapeutic, is best in class.

But once again, we are really trying to show, we, the big thing for us is try to show the world that it’s not just another pipeline. Because a pipeline is a pipeline, you know, and you can have a hundred molecules if it doesn’t show Really great success. And I think for me, for everybody in Silico, we all have the same problem, you know, it’s showing that our pipeline or value of health, showing the pipeline is different.

Philip Hemme: Everyone wants the hard one. 

And 

I’m curious on the type of drug, is it like small molecules only or are you agnostic of the type of technology? 

Thomas Clozel: No, we’re, we’re agnostic today. We, we, you know, I think we, we, we, we did this thing today. We, we’re going to announce a big, a big partnership with a very famous CRO on small molecules.

But we are open, you know, because we, we want to discover biology first. If you find a protein at the surface, they can be only targeted by, you know, by biology. And there is good companies doing AI and, and good CRO on the, on the, on the, you know, the biology side. Biologics is also important. We, we, we couldn’t restrain us because we want to discover new biology first.

So today we open. We don’t have any internal expertise on one or the other. We externalize this expertise, working with the best CRO there. It costs a lot of money, of course, but you know, we have to get there. 

Philip Hemme: I think, I mean, from my view, it’s always something that’s not puzzled me, but I was a bit cautious about that.

Most of the AI biotech, especially the ones doing their own drug development and having, they’re mostly doing small molecules. Which I think it is great and there’s still need for some amazing small molecules. But if you look at the pipes, most of the pharma, like 90 percent of the pipe is biological, biologicals or advanced therapeutics.

And I haven’t seen that many biotech. We have like breakthrough biologicals thing. And the first that came to mind, where I was like, oh, wow, it was exciting to, I think, mid of this year, they announced that they have, Now they, they can, they can generate antibodies, I think, or biologicals. And I was like, okay, now, 

[00:47:29] Small molecules

Philip Hemme: now it starts to be like really, I mean, probably really interesting.

Thomas Clozel: Very smart companies. Yeah, yeah. No, I, I, and there’s so many mechanisms, especially in the microenvironment, small molecules is not the solution. You can’t. First, it’s biological first, but we don’t, you know, once again, it’s working with CROs, but the problem of biologics is, of course, like, in a way, small molecules are a bit, a bit simpler, probably because you have less optimization, you have less stuff to do, but the patent is shorter, there’s a patent problem as well creating a bit less value.

I agree with you. 

Philip Hemme: Do you think there is a, I mean, or at least from my view, was a bit like, maybe the, the biological complexity? At the moment, it was just too complex for Gen. AI or AI in general to really, to, to, like, manage. Is that one of 

Thomas Clozel: the I think so, but also, also, also Yeah, probably, probably the, the, and this is, this is the first part.

I think the second part is accessing privacy data sets. It was the idea of Melody, accessing small molecules, libraries from the farmer, not public ones. And, you know, I don’t think there’s enough like access to really high quality private data of really high quality with once again, the idea of continuously, I really insist on the idea of being able to continuously enrich your data sets.

What happens again, we can, we can add new analysis if needed for the same patients. And this is really what matters, because there is so many new technologies that are coming in, you want to be able to do that. And I think this is what’s missing in the chemistry space. A lot of people are using, like, public data sets to work on the chemistry side.

And, you know, there will be a limitation too. But I think, I mean, come on, Insilico, Exemshar, they’re all doing a great job. They’re doing great stuff. It’s crazy. I, well, 

[00:49:10] Limitations

Philip Hemme: I, really, I’m playing the devil’s advocate here and trying to challenge you. I’m trying to understand, really, like Where are the limitations?

What’s hype? What’s reality? Where is it standing? So I think, but I think you did a really good job of presenting, 

Thomas Clozel: yeah. Thank you. Thank you. First limitation is data. Technology is the second, but data is always the biggest limitation. First. Always. And then I think, yeah. But I think, and then the other limitation is time to show proof points.

Time to show that, you know And this is the terrible thing about this field. And once again, the terrible thing is waiting four, five years. to get your face to be showing that you, you should prove something. Of course, you can accelerate and whatever. How? This is so frustrating. I mean, it’s, it’s, but it’s true for every biotech, right?

Philip Hemme: I mean, health, yeah. Drug development. I mean, you can bend time or bend a bit time, but once you know a bit the complexity of it, it’s, you cannot cut too many corners. I liked what, what Mark, the guy that actually, we, we talked about it, that you watched the episode, but he said that. Yeah. It’s very hard.

Great episode, congratulations. Thanks. Thanks. It’s very hard to find actually the balance between speed and quality in a clinical trial because you’re, and you have so much pressure to get the faster in terms of capital, capital limitation and you want to help patients, but at the same time, if you compromise too much on the quality of the trial, It can backfire like pretty, pretty big times.

So I like that. I mean, it’s, it’s, it’s not, it’s not like a, I mean, in tech you can cut a bit more corners and probably you can scale maybe faster. But in healthcare and with patients, I mean, we saw Theranos, we saw even 23andMe. I mean, when you cut too many corners. Yeah, but it’s 

Thomas Clozel: a great, I mean, but I think 23andMe is better than Theranos.

But still, 

Philip Hemme: there are some No, 

Thomas Clozel: no, no, it’s discussable. Yeah, but I like it because, yeah, but I like the fact that 23andMe takes patient data and finds molecules and stuff, and I think they’re doing a better job. I mean, one thing that is also very limiting is the Phase 1. Phase 1 is so in oncology. It’s so limiting.

They all look the same. You never know. I think phase 1 is a big topic. We are trying to improve phase 1. You know, trying to, not finding evidence, you can’t, but insights, with counterfactual outcome, building like, trying to compare the response, the actual response to a predicted response. Sorry. But I feel like phase 1 in oncology especially is a big limitation because you never know what to think about it.

Yeah. And we try to work on it. That’s a good point. 

[00:51:51] Growing up with billionaire biotech parents

Philip Hemme: I want to switch a bit topic more to the personal side and, and talking about more your background and, and kind of your mindset. And the first thing that comes to mind is, is talking more about your, like, let’s say your parents and where, where you came from.

I mean, I mean, of course, I mean, your parents built, I tell you, I mean, super successful. One of the few, let’s say billionaires in biotech, at least in Europe. And must be, how did it like, feel and how did it like, impact you, growing up that way? 

Thomas Clozel: Thank you. So first my parents, my parents are, first my parents are doctors.

You know, they’re not like lawyers, like some of the CEOs are doctors, they really think about their patients. And they’ve been like putting up drugs that are amazing. Like for example, take the example of Q Vivic, it’s their last drug and it’s building up. It’s an amazing drug. It’s revolutionizing sleep. I take it all the time.

You dream more. You sleep so much better. And of course, it’s not the success of, of, of of WigoV today. However, I believe that sleep and obesity are both very important for your life. And I hope it will come. And I think my patients are just like, extremely like, idealistic also very, sometimes maybe too much.

Doctors, I think that my, my, my parents give me the first, the, the, the patient first thing because they’re doctors. And I’m a doctor too. And, and, and they teach me that patients matter first, and when you bring patients value, there is a business. At the beginning of Actelion, they were building something for, you know, like, for, for for permanent hypertension.

People were telling them the market is too small, whatever, whatever. They were like, this is the first one. And, and the second thing is, I think they also, like, show me that you need to, to trust in what you’re doing because don’t, don’t listen too much about people, you know? I mean and, and my, my father is very, like, go forward.

But I think that I listen to, to trust in my, my instincts as well. Listening to advice, but I’m And the last thing is not being afraid to fail. If you do something good in health, you know, really, and I really promise, I don’t feel the pressure to fail. And I think they haven’t either. It will give you some strength because when you’re a doctor, you see people having very severe things, diseases, you know, that you can cure.

And it makes you very humble about life and really help you to, to say that if you fail what you’re doing, there is one thing in life, which is your health. Health matters more than anything else. So my parents have been very cool about failing, even when I was a kid at school. If I was failing anything, they were always super cool about it.

Not really pressurizing me anything and giving me the right values, and I think it really helped me. And even when I stopped being a doctor to to to they were like super supportive, and and some parents could have said, you should stay with your job. You know but but however, they’re not really helping me day to day, you know?

I’m I’m they haven’t helped me on deals. Ha ha ha, still. Because they want me to succeed, and that’s helping me. I mean, the American way, really, that’s, and, and this gives me, this gives me the, I wanna make my own money. I wanna make my own job. I wanna make my own success, and I wanna do better than them.

That’s kind of what driving me and the show. I’m that doing, I’m not working. I never work with Act Docia. I’m making my own stuff. And and yeah, I, and I my father always made the joke saying, I’m the, I’m the, I’m the father of, of ma Arts. This is how he present me. Just for the joke. And I hope, I hope one day will mean something.

Today, not that much. Today, like, okay, who is this guy? Well, you know, I don’t know. But it’s, you know, but like, my parents always take huge bets.

Philip Hemme: I like a lot the, I like a lot the, the, inspiration lessons part, but also very hands off and like, giving you the, the freedom to explore, to fail, to, To build yourself and to build yourself differently.

I think it’s very good. A good mix. 

Thomas Clozel: Very. And being single-minded. My, my fa my father, my, my mother, they refused all the boards. They, there are not in any boards. They’re only in my father, only in I board, and my mother say no to hundreds of boards. You know, honestly, just because they’re very, they also teach me to be single-minded, single about one thing you wanna do in life well and just do it.

Don’t, you know, there’s so many people that love to be bored people, being in the X, they don’t care. You wouldn’t, I mean, they’re really out of the radar at every point of view. 

Philip Hemme: Yeah, I can tell you, your dad is very focused also on PR, or at least he, at 

Thomas Clozel: least I’ve never managed to do it. Yeah, he never do anything, he doesn’t care.

Yeah, he never do anything. He doesn’t, he doesn’t invest, he doesn’t do anything else, you know. 

Philip Hemme: It’s pretty impressive, I mean, I love it’s pretty impressive actually, I have to say. 

Thomas Clozel: It works, I can’t lie. No, no, I mean, but, but they are doctors. Always remember, they’re just doctors. They really don’t see, they really believe that all the patients giving their drugs is their own patients.

They really are. 

[00:56:37] Becoming unfocused and having fun

Philip Hemme: On, on on focus, it’s, it’s funny because, I mean, I mean, she, that is very focused and you apply this, but at the same time I heard you in an interview on, on Fast Company about that you want to be more unfocused and Being open to every opportunity versus being ultra focused and that lot of VCs challenged you on that, of where is the focus 14?

Thomas Clozel: Yeah. No, I think it’s, it’s exactly, so at the beginning I had so many board members, a few members say, you have to do one thing, take one modality pathology, just do one use case. And I was like, I, I want to discover, I wanna build a net of technology that will I help to understand where the right use case.

Because once again, this is new technologies for the healthcare, we don’t really know. What we’re really discovering and how much it’s going to impact patients tomorrow. I want to be way more open and a very more unsupervised mindset. And, and they were all, I know I need to be focused. And I, I’ve never agreed to be focused in Farrokhine.

I’ve always agreed to go large and to be ambitious and try to cross fertilize things. Because it does cross fertilize. Of course, like I got one board member that fire me once and you know and 

Philip Hemme: I see, I guess you, you learn, you learn from your dad there. No, he had some lessons to share.

Thomas Clozel: The lessons from my dad, which is like Was an activist 

Philip Hemme: investor. I heard about this. Chloé was like, whoa. 

Thomas Clozel: Yeah, yeah, yeah, yeah, yeah. Oh, yeah, yeah, yeah. Oh, yeah. No, but you, yeah. I mean, I mean, I mean, the board, I mean, the board is a big topic, you know, for a company like us. Because, you know, it’s very hard to get board members that get to the right level of understanding of what you do.

We have some very, very but it’s not an easy topic. So, no, no, no. And, yeah, I mean, I, I’ve always fight for what I, what I believe. And if this company is all of us, we are moonshot companies. It’s in situ in Silico Exentia. I mean, we, we want to try to change big stuff and at the same time, it’s, but you know, as, as of course, all the big tech people have we’re not like an operation.

Hey, I mean, we have to be in some ways and we, most of us will fail. Maybe me too, but I think it’s, it’s fine. We, we all have to, to be in this mindset and, and. Out of focus is the, is the way to go. Try to 

Philip Hemme: I like, I like that choosing the case for when to be ultra focused and when to be less focused.

And I can see for me personally, it’s something that I learned recently and I try to to be much more like Malleable or how you say it in English, more like flexible rather than, I was like very rigid, like always ultra focused, always deep work and like very rigid. And I find that in some cases, whether it’s with a company or whether it’s even personal or whatever, or, or in the time of project, sometimes like choosing the right setting of focus works actually better.

Thomas Clozel: I, I like it exactly. That’s good. And I think also it’s the same when you, when you, when you go out for a walk and when you are sitting at your desk. When you’re sitting at your desk, you’re in a mode, it’s a concentration mode. You know, you really focus. When you go out to walk, you’re more attentive. It’s called attention versus concentration.

You’re more, you know, ideas coming and flowing and everything. And I think to be successful, you need both. You need to walk out and be sitting at your desk. You need to focus on the mission and you need to have very precise QP key objective and key results. You need to have. You need to know what you’re solving.

Sometimes people don’t know the problems they’re solving. You need to have a clear vision on data. You need to know the technology you wanna be number one on. You need to be always number one. We are number one on multi scale AI for biology. But then you need to be also attentive to what the world, to, to do ideas and come in, you know, need to be both.

I think to do a bit of both. And, and, and usually people try to, to make you only go for the focus side. Yeah, that’s, 

Philip Hemme: that’s, that’s, I like that. On the, maybe a bit on the, on the, on the other personal topic is on, on fun. I like a lot, like that, at least it seems, but even from our discussion that you’re having fun in what you do.

And yeah, I remember you, you were like whatever doing interview and wearing like a pink elephant sweater and I messaged you about this and then you just posted out the JPM blue bike. It’s, it’s small things, but first it’s about having fun a day to day and even making a bit more public and, and, and have it sharing that fun.

Can you like, first, I mean, I like that a lot. I can connect to it and relate to it, but can you elaborate a bit more on like, maybe what’s, what’s behind like it? 

Thomas Clozel: First, yeah, thank you. At first, I think like, once again, I, I know what’s important. I mean the, the, being a doctor and oncologist for, for not that many years, but enough to understand what that health is, the only thing that matters, your health, your family, your friends, the rest is a, is a coup d’etat.

You know being a billionaire you’re still going to die. Not going to absolutely be more happy. I mean, I know that I, I feel, I think I understood my own way to things. It’s like health really matters. My family, my friends, the rest is a great top on and, and it’s just an adventure. That’s the first one.

Then I, I think, you know my way, you know, I think I love to have challenges in life. It makes you, makes everybody happy, makes me happy. And there are doable challenges. And, and Okin is an amazing challenge. Very hard company. You know I need two aspirants after every day, or each board four aspirants probably, maybe 20 and, and and but it’s also super doable, maybe doable, maybe we can achieve it, and it’s just so amazing.

And the last thing is, I’m surrounded by so cool people we, we keep the culture, I’m talking very childish culture. And that is, you know about about about being, you know, together we are playing games with the Pictionary yesterday, you know, together. And I think I love the people and it makes me so happy to have people that I actually really do love.

I think we are 320 people. There’s nobody I could not have a lunch with. There is none. And that makes me so fulfilling, you know, in a way. But, but, but the most side of it is, you know, if it fails, it fails. I promise I won’t be, you know, because life is bigger than that. And my, my, my, my human adventure with my colleagues is the best thing I’ve created in the company.

And I really mean it. And I love so many of them, like from the bottom of my heart. So I think it’s something that makes me just giving a bit lighter and lighter hearts because, you know, and, and this is where social doing something like, you know, we are all in the same space, right? Having a social impact in life gives, give you less pressure.

Because you did your best, you know, like, you’re trying to help, and people that just work in the big tech but don’t understand what the impact they want to create socially, I think they’re missing something at least it gives them more pressure. And if I was building an NFT for, if I was building an NFT company for whatever, I would feel way more pressure.

I haven’t succeeded, but I absorbed. NFT 

Philip Hemme: for what? Well, NFT for, 

Thomas Clozel: I don’t know, for us. For 

Philip Hemme: biotech, I don’t know. 

Thomas Clozel: I don’t know. I mean, there’s great companies around that as well. It’s just not my stuff. I like that. But you invited, I hope you’re going to, I hope you’re going to come on our, please, at J. P. Morgan, we have to do our tuk tuk tour.

We have a tuk tuk at J. P. Morgan. It’s going to be, it’s going to be, because I want to be outside. I hate to be inside. So we’re going to have hot chocolates. We’re going to be outside. We’re going to go by the water, by the Embarcadero, come back to Union Square. So I’m, I really hope we can come and everybody can enjoy our tuk tuk 

Philip Hemme: meetings.

I will show the, I will, I will show the picture post it. It looks, it looks really cool. I like, yeah, it’s, it’s, I like this, this, I like this, this mindset. Because on, on pressure, I mean, I like your view that because you are doing something with high impact, this relieves pressure. My feeling also is that quite a lot of people, let’s say, in the industry, because it’s such a serious topic and high pressure, I mean, you’re talking about people with, like, cancer or, like, really heavy stuff, sometimes they become, I feel like, very serious about it and, like, a bit, let’s say, not boring, but it’s a bit judgment, but a bit, like, too serious, too stiff.

Which, I quite like your view of it, because basically it’s the same, like, whatever, the same cause, but the interpretation, how you go about it is completely different, and I feel like, 

Thomas Clozel: yeah, I like that. I mean, I think both, both make sense. Don’t get me wrong. Both make sense. I mean, I would, I’m, I’m, I mean, we also need people that feel the pressure.

I mean, it’s not because you have no, you feel less pressure that you don’t feel the sense of urgency. It’s different, right? I do feel the sense of urgency for the patients. I mean, it’s, it’s unbelievable that we still have cancer with this level of understanding. Take, take cholangiocarcinoma. It’s increasing.

We don’t get it. Where is it coming from? It did not exist 20 years ago. What’s the food effects on it? You know, whatever, we, we are so humble You have to be super humble about what you, what you’re building Because we, the, the, the The biology makes you humble But I feel like we also need people that are stressed out.

I mean, it’s just not my way to go. And I’m not saying I’m never stressed. I am sometimes, but, you know, at the end, it’s so hard. Biologists really are, I mean, and I’m so, I just want to, I’m so grateful for people that work there as well. Everybody in the tech, in the biotech, in the big pharma, in the labs, in academics, even more in academics where people don’t make so much money.

I’m so grateful because you can see so much money around like the new large language model, the, the, the, the, the, the new me too of open AI raising a lot of money. Cool. But it’s amazing. But I mean, I mean, and, and, and having like small biotech startup having so much hard time to raise money. It gives me like you know, sometimes a little bit and I’m a bit angry, but it’s like, I mean and you have to be, you know, so I’m really grateful for, for the people that work in health, wherever.

Philip Hemme: I like that. Still on the fun topic what, what I, what I question. What I like is that I think even if it’s serious, if you need the urgency, I think having fun at work or having fun with your colleague is just something so like fundamental from a human level, or at least for some personality of people, it is so fundamental that I think if you don’t have it, you cannot do like really good work.

Thomas Clozel: Do you see that as well? I, so, yeah. I mean, if you don’t come, if you don’t like to come to the office in the morning, you know you know, it’s, it’s, and it’s hard because, you know, I think also the work environment is the more you grow, the more you have to be closer to, to some type of work principles.

Work is being like, tried to create new, new, mid. And I think we resisted quite well for now. It’s not easy, you know. But you, yeah, you have to keep fresh, you have to, to keep and I think, like, also being in your, being more European company, it’s easier, you know, a little bit. Why, 

Philip Hemme: why is it, why is it easier?

You’re less exposed to the, how do you call it, the woke virus in the, in 

Thomas Clozel: the U. S., so. Yes. Yeah. I don’t want to be political, but No, no, no, no. But we, I mean, we’re still very careful about these values and American has also brought really important values, and we are very respectful of diversity and everything.

I, I just say like we, it’s a bit like, it’s easy. It’s harder to talk about religion or, or things like this, for example, or politics in, in the US you have to be a bit more careful and I think it, I mean, we also have a big US company, but. Our French spirits, we quite free, we very we have really revolutionary company.

I mean, we, our company is like, we know, we very like my, my, my co founder was probably the smartest people I’ve ever met in my life. And, and one of the coolest, he’s so French, you know, he was like, when, when we had Google Venture coming in and the company was like, we have to say no. We have to be the first company, no, saying no to Google Venture.

And he was like, and he really meant it, I mean, he meant it. And Gilles has always been saying revolutionary. And, and very revolutionary. It’s the coolest guy around. But you 

Philip Hemme: still said yes at the end. 

Thomas Clozel: Yeah, I believe. So we got a bit of American pragmatism. American pragmatism coming again. 

[01:09:02] Getting the best investors

Philip Hemme: When you saw the money on the table as well, I mean, the money and the value, I guess.

How is it actually to have Google Ventures as investors? Because they invested quite a lot. I mean, I think they’re in past AI as well. And then obviously with Google background, I mean, they’re crazy view on, on AI in general, and they have also their own, let’s say, bio efforts. So how, 

Thomas Clozel: like, how was it? Yeah, no, I have nothing specific to say.

I mean they’ve been supportive and nice. I mean, I, I mean, we were the first, I think, in Europe when they invested at the time after they made more investments. So we were the first deal. I think what happened at Google is, it’s not really about Google Venture. It’s, it’s not about the funds that invest, it’s about the partner.

And at the time we had somebody called Adam Gobara that has created top Harvest Capital. Which is one of the coolest people I met in my, in my life in my, in my work life. He’s, he’s been always super fond friendly, who was supporting us with a real vision for the company. You know, and he knew why he was investing.

He, I mean, you want people that are not in what we call momentum investing, which is a lot for a lot of VCs investing because other people are investing. You want people that invest because they have a very, very precise theory about what you want to do. And, and and you know, you really want, and I, I also took momentum money and I probably shouldn’t have, you know because the people are coming because all the people are here and not the people by earning your value. You want people that we had the same with Gwendolya from First Capital that invested at the beginning he, because he believed in us, in the vision, you know. It’s more about the partners and the funds, really.

Okay. Can 

Philip Hemme: you, can you share a bit more about this momentum money and like slash the fit with investors? 

Thomas Clozel: Yeah, I mean, for example, at the beginning, you know, we had like people like Nicole Druckerman, she’s a German amazing investor. You know, you’re more of like first Google. I think they are Adam Gobera.

This is really people more than funds. That Jacky Abitbol from Quetta is that really believe in us. You know, they were really like believing in the stories. They didn’t want, they didn’t care if an American fund was coming or whatever. Or another fund was here. A lot of funds are just waiting for leads to come, for the diligence to be done.

And when they see good names, they’re just coming in. And it’s, it’s part of the problem of venture capital. Is they really work by FOMO, right? We all know that. Some of us, but, but there is amazing stuff. I, you know, take the example of Sofinova. Sofinova has been an amazing biotech fund, as you know.

And now they’re creating this great digital fund. And I think like Ed and Simon, the two partners, have really, really strong view on things. And, and they understand the, the, the intersection. That, that we start to have in this field, people that have, you know, that have balls and that have ideas. But, but you also have people that just want to invest because whatever, blah, blah, blah, blah.

Philip Hemme: Mm. I was about to mention Sofinova. It’s, it’s, it’s, it’s great. But I think. Talk about this maybe from your perspective on like, because I mean, there is always a level of FOMO that is, let’s say, healthy or that, that, that, that works. I mean, especially when you talk about venture rounds. I mean, I don’t know if you’re talking about angels coming at the beginning.

I mean, angels, there is a bit more believing and really the fit was a person and really long term investing. But let’s say from a VC, I mean, you need to be in this like sweet spot of, okay, it’s not too early. It’s not too late. And what will be my return kind of. having a decent return possibility. And this is also very dependent on, on let’s say, the markets, other investors, pharma partners who follow, they quite like momentum and follow also a lot of others.

And the public market, which is very momentum based as well. So How do you, like, wrap all of this around and then you take the decision of which investor to take, like, you build an AI model for this? 

Thomas Clozel: Yeah, no, no, I think it’s a personal thing as well. You want to, I think it’s like buddies. You want to take people that you want to work with.

I mean, I, I took people that I did not want to work with in the past and, and in I’m, I’m proving my, today, I’m really choosing, it’s the people, it’s not the name, it’s not the real thing because people make the difference. So I think, you know, you have to take people and you have to understand what part of momentum investing or a part of like real theory, you have to ask them, why are you here?

Why do you want to work with me? Well, you know, well, what’s, and, and, you know, but, but the problem is what is the actual value in our company, in TechBio in general? I mean, there is people that bring you great insights, but I have to say the truth, money is money. Because nobody has done this before, right?

Ha, ha, ha, ha, ha. So it’s very rare that somebody’s going to bring you, I mean, at the earliest stage, nobody’s going to bring you something dramatic. Then, when you have to structure, when you need to understand better, when you have to position the dra Then the, the, the, they’re going to come and help a lot.

At the beginning, you know, I think they’re more functional than just exponential, you know, in what they can bring to you. Mm, mm, I like that. And actually, I’m going to be loved by the VC community. Oh yeah, it’s good, I mean. You will have it. They’re not going to like me, but I don’t care. 

Philip Hemme: Yeah, you work with Sophie Noble already as Venture Partner or something, so it’s okay.

Yeah. No, 

Thomas Clozel: yeah, I have to mention, true. But, but because I really believe in Ed and Simon and they proved me right. I mean, it’s a great company. They’re very fun guys. They’re fun guys. So it’s fun as well. But I think just something about investing. I, I, something as, sorry, just about that. I think it was like, we, we, we have taken as both observers people from Salafi from Nestle and I mean Venkat from BMS.

And I think like Venkat leading, like, you know, a big part of data science from the discovery. They’ve been amazing to help us, you know, and I would really, I really recommend for people to take also people from the industry in as board observers or board directors, you know, I just like to help you build things because I have an industry view that is also super interesting and for me, they’ve been extremely helpful.

Yeah, 

Philip Hemme: especially as you said, on certain aspects of your business or your case where it is kind of industry, where there is some industry part and industry knowledge needed. I mean, I wanna ask you one, one question also on the investor side, because, and act, I think you, you’ve billed it more as a, as a tech company as well as in like raising money, but with not too much dilution.

And even, I mean, when Sanofi invested, I think was a 180 million, but at a 1 billion valuation, I think they got 10, 15% of equity, which is, quite low and a very big multiple, much more tech multiple than a biotech multiple, where probably they would have whatever, have 30 to 50% of the company, and I guess what, what translate behind as well is that you guys as founders instead have much more, say, control or things to say on the actual company versus the typical model in biotech where a VC fund will come in and, and take a lot of control.

Sorry. 

Thomas Clozel: Can you hear me? Yeah. Yeah. No, no, no. So I think we, we, yeah, yeah. So 

Philip Hemme: can you, can you, yeah, the question is more like is that the case? And can you elaborate a bit on it? 

Thomas Clozel: Yeah. No, I mean, we, we, we took, we, we, we, we were not very delusion sensitive with my founder, with my co founder, because I always prefer to have a very small part of a huge company, which is not huge today.

So I hope Okin will become huge. So we were not really, we, we know we took the money when we had a really great fund that came in and we took the money where it’s coming from. And so we took maybe too many rounds. So we got delusion. However, we don’t control the company, my co founder and I. But we’re careful, of course, you know, we’re careful to, to still try to get the right dilution and everything.

But, you know, we are not in the, in the, in the position where we have, we control the company. Which maybe is healthy as well, you know because it’s such a complex thing, you want more people to control it. But I think, you know, in the, you know, the more you grow, it’s my first company, right? So you know, I learned that you know, of course, like dilution matters for the control of it a lot.

So, you know, 

Philip Hemme: But, I mean, you, you don’t I thought, control maybe You don’t control more than 50 percent with your co founder, I guess, but you still can’t, you still have, you still have more shares than in a typical biotech setup was a founder CEO ends up with whatever before percent or two, three, or one, two, 3 percent or something.

Thomas Clozel: Maybe a, yeah, for now. For now. More than that. More than that. But you know, not 50 percent either. Yeah, yeah, yeah. But you know, I think like, you know, we, we, we, when there is good money coming in and because we have to do, we have to build. You know, technology, we have to find data, we have to build molecules, so much to do.

It’s a costly project. So, you know, we’re taking the money and you can’t really stay, you cannot really like do like bootstrap it and the money and program your own thing. It doesn’t work out because in biology, all the weight lab thing also works a lot. This is the problem of biology. You always need funding.

Philip Hemme: You need a lot of it. You need a lot of cash. And as we discussed on the whole field as well, I mean, the field is moving so fast and you have quite some competitors as well. And like, Yeah. You need to move also fast to be relevant and to still be the, the best while keeping, while 

Thomas Clozel: keeping, yeah. While keeping always years of cash ahead of you, you know?

Mm-Hmm. Talking, we have at least three years of cash ahead of us. It power you because you, you have to be careful with that. That’s great. 

Philip Hemme: It’s great. Thomas. I think we are, we are about about time. Was was a great discussion. I hope the audio I think worked. I think it worked. Normally it records offline.

I think it worked. Our least. This will be, we, we will edit up. 

Thomas Clozel: Yeah, I think you were us and them. A little pixel ice. And and I love the hats. What is the, is your hat for this, this wing, this weekend or this winter? You gotta take it for skiing. 

Philip Hemme: It’s, it’s my Christmas hat. 

Thomas Clozel: It’s so cool. You have Christmas hats.

I love a Christmas hat. Actually. See you soon. 

[01:18:51] Thanks for listening

Philip Hemme: Wait, wait, wait. I forgot. We need to take a picture together so I can put the hat on. Can you just say, wave, say hi? Great. Thanks for listening to the end. I really enjoyed the conversation. I think this topic is, is really complex and there’s a lot of, like, hype and everyone talks of I’m doing AI and biotech, but I think Thomas really tried to challenge him and I think he’s really good at it.

Really good speaker, good at answering, but I think he nailed it pretty well, explaining also what’s the difference, the key differences between really native tech bio, AI biotechs, what Oaken is doing. I really like the approach of starting with the patient data and really like a huge amount of data and crazy data and then going backwards towards assets instead of having a pipeline versus a lot of other AI tools where It’s harder to capture value in biotech.

I hope you enjoyed as well. If you did, please hit the like, subscribe, review button, especially if you’re on Spotify or Apple. And I would be also curious to hear what you think. So please, if you’re on YouTube or Spotify, you can comment directly or you can also send me an email to philip at 

Thomas Clozel: float dot.

Philip Hemme: All right, catch you in the next one.

Further Episodes

Episode 27

Thomas Ybert, DNA Script 🇫🇷 | DNA Printing & Synthesis | E27

Thomas Ybert of DNA Script on the Flot.bio Show
Episode 26

Gunilla Osswald, BioArctic 🇸🇪 | Leqembi, Alzheimer’s disease | E26

Gunilla Osswald, BioArctic on the Flot.bio Show
                           

Never miss a new episode 💌

                           

Trusted by 250+ professionals from top companies like Argenx, BioNTech, or Genmab. No spam, no noise.

                                           

Watch & Subscribe

Flot.bio on Youtube                 Flot.bio on Spotify                                    Flot.bio on Apple Podcasts