We’re in Budapest with Szabolcs Nagy, the founder of Turbine, one of the leading companies for AI in biotech in the world.
We talked about AI in biotech at large. We also cover the differences between opening a platform versus doing proprietary drug development and building the company from central Europe.
__
This episode is sponsored by Turbine, which can improve your drug development. Learn more at https://turbine.ai/reengineered?utm_source=flotbio&utm_medium=podcast&utm_id=ba
__
⭐️ ABOUT THE SPEAKER
I’ve known Szabolcs since 2017 when he joined Bayer’s accelerator in Berlin. I found him incredibly inspiring back then and still to this day. Szabi began his serial entrepreneurship with the creation of a cybersecurity startup. Once that was acquired, he launched himself into the biotech scene with the founding of Turbine in 2015.
🔗 LINKS MENTIONED
- Accenture’s Investment in Turbine Enhances AI-Powered Cell Simulations for Biopharma R&D: https://www.biopharmatrend.com/post/806-accentures-investment-in-turbine-enhances-ai-powered-cell-simulations-for-biopharma-rd/
- After a tough year, Exscientia folds into Recursion to create an AI superpower: https://www.fiercebiotech.com/biotech/after-tough-year-exscientia-folds-recursion-create-ai-super-power
- Werner Lanthaler, Evotec | Leading AI in Biotech | E02: https://flot.bio/episode/werner-lanthaler-evotec-leading-ai-in-biotech-and-how-to-be-authentic/
Transcript
[00:00:00] Intro
Szabolcs Nagy: AI in biopharma has been a thing for like seven to eight years. It was called bioinformatics. Yeah, of course, fair point, like 20, 30 years ago, AI was a thing, but it was called bioinformatics and then data science and then AI. What we are currently showing in the current partnerships and what our focus for the next 18 months is to demonstrate simulations in a variety of different use cases and show that the same underlying technology, the same underlying model of cellular biology can be applied across all of these different problems.
I
Philip Hemme: mean, you’re living here in Budapest and I see that you’re traveling a lot to European hubs, either Boston. How do you prioritize all of this traveling?
Szabolcs Nagy: I don’t think we’ve proven that it can work like this, that you don’t have to move to the UK or to Boston to build a successful AI and biotech company.
I think we’re putting it a good case.
Philip Hemme: I’m your host, Philippe, and on this show, I’m interviewing the best Europeans in biotech to help you grow. AI is changing the world, including the biotech world, but it’s hard to know what’s hype and what’s real. One of the best startups in Europe in this space is Turbine. So I went to Budapest to meet with its founder, Tzabi, who I’ve known actually since 2017 when they joined the Bayer Accelerator here in Berlin.
We talked about AI and biotech together. We also talked about opening up the platform versus keeping it closed for proprietary drug development, as well as building the company in Central Europe. For transparency, this episode has been sponsored by Turbine. So, here’s my conversation with Xabi, and please hit the like follow button if you’re enjoying it.
Welcome Xabi. Thanks for having me on, Philippe, great to be here. Yeah, it’s great to be here.
[00:01:47] AI in Biotech
Philip Hemme: I’m very excited to do this discussion and I want to start with, say, the hype around AI for drug development. And as you are really active in the space, I’m wondering what’s, what’s your take on it?
Szabolcs Nagy: Yeah. Who better to criticize the field than somebody that actually is part of the field.
I’ll go ahead and do that. Well definitely, you know, I think I’m not saying anything new when I say that AI is for sure overhyped. And it has been overhyped for years, right? I think actually we’re getting to. Like the moment of truth for a lot of the leaders of the space who are, of course many of them are in clinical trials with drugs that they use the AI to help develop and design.
And of course, we’re learning more about, you know, some claims that were bogus when somebody said, yeah, sure, we’ll, you know, just markedly improve the success rates of clinical trials by designing better molecules using AI. Clearly, that’s not necessarily going to happen. Using AI for some use cases has been impactful and is already standard practice across biopharma.
But but it hasn’t, you know, changed medicine yet. I mean, these technologies haven’t changed medicine. And likely we didn’t do ourselves a service in the field by over hyping the whole thing. However, I do think that you know, below the hype, the standard of practices, I said, like the day to day life of drug discovery scientists has been shifting and have been shifting.
So I think that’s cool to see actually.
Philip Hemme: And how much do you think the hype comes from the AI drug? discovery company, drug development companies versus the whole AI in general, I mean, the whole GPT and everything that there, there was a moment and
Szabolcs Nagy: the attraction surely feed off each other. Yeah. Because if you think about it, I mean, of course, AI in, in biopharma has been a thing for like, Maybe eight, you know, seven to eight years really when companies like, yeah, of course, Fairpoint, like 20, 30 years ago, AI was a thing, but it was called bioinformatics and then data science and then AI.
But I think like mid 2010s, of course, companies like, recursion, Accenture, you know, Atomvise, these guys have started raising their first major rounds. And so. You know, people started paying attention, you know, even Y Combinator started paying attention to you know, Y Combinator, of course, the epitome of you know, technology and you know, sort of West Coast bully bullies of course, bullishness rather bullishness.
Yeah. On technology, even they started paying attention. And so the hype started growing around AI and biotech around then, but but now with ChaiGPT around, suddenly everybody, of course, is a generative AI company, even the companies that have been around for 10 years and had nothing to do with the technology.
So, you know, clearly there’s a sort of a back and forth cycling of ideas, which is, it’s fine because clearly we’re one of the areas as in, you know, AI and biopharm is one of the areas of error really were largely. Reusing technological progress in other application areas of AI to our own use cases.
And we’re learning how to deal with the specialty nature of the data that we train to and the problems that we’re trying to predict on which of course is, is, you know, very difficult. So there’s plenty of. novel AI being done in biopharma, but really we’re, we’re learning from other fields. And of course, they have an impact on the hype cycle.
Philip Hemme: Yeah, that’s fair. And maybe just to stay on the hype cycle as well. I mean, you mentioned recursion, Excentia, they, they did a merger, but I think at the market cap for Excentia, that was lower than everything they raised, which At least if I look at it from an observer point of view, it’s a sign that the valuation that it comprised before, or they were higher before, I think benevolent AI also didn’t work.
What do you, like, like, what’s your take on those recent setback? Was it just a correction? Yeah. Was it set too thin?
Szabolcs Nagy: If you think about it, two things compounded in the valuations of a lot of the fraud runners of our field. One was the general. like crazy overheating of the biotech stocks. And of course, many of these companies really are ultimately biotech companies.
And that’s also how they want to position themselves. They, they develop, you know, pipelines of drug drugs and and they hope to be valued as, as a biotech company. So there was that sort of boost as we know, you know, sort of around until the end of 21, of course, everybody was inflated or everybody’s price was inflated, but that was, I think, further compounded by the fact that, you know, Of course, many of us were also looked at, I think almost everybody was looked at as also like a tech company.
And so a lot of us front runners got investments from tech investors, which priced us very differently than, than more traditional biotech investors usually do the platforms that they invest into. And so those two in fact combine and really drove the valuation, you know, recursion was 8 billion plus at one point after the IPO.
And so, of course that has been. Most likely corrected to a good extent and of course it’s also being depressed just because of the general macro cycling on the market, which is, which is fine.
Philip Hemme: Yeah.
Szabolcs Nagy: We’ll have some of that, but definitely the, the actual, I think what really matters is not the, the valuations, but more the, the results of the companies, especially in clinical trials.
And also in discovery and generally the progress of the pipelines of these companies. And, and those of course have. seen some, some, you know, promising results, some good face to face fun stuff. And there is to be results now, I think. First efficacy data is coming out. Yeah, some efficacy data. Of course, I think, especially within Silicon Recursion, they do have the, they have reported recent efficacy data.
And sort of, you can, you know, you can read into that, whatever you wish, because do you want to put a silver lining on that, but it’s not yet sort of compelling enough. But of course, also we have to keep in mind that, that the drug discovery programs that are in the clinical trials of these companies now really came from almost their very first sets of ideas, right?
And all those years ago, when they started out. They made some bets, they chose some programs, I believe the, the front running program of recursion is the one that actually Chris Gibson’s PhD was about just to a great extent. Yeah. So it was a repurposing opportunity identified during his PhD work as far as I know.
And so, you know, clearly, you know, Chris is a wonderful scientist and a brilliant entrepreneur, but. Like the idea that PhD may not be the ultimate output of recursion. So even though there is, there are some setbacks, clearly there’s more promising stuff earlier in their pipelines and we’ll have to wait for that.
But I think also the setbacks do sort of like are a warning sign or like a cautionary sign that we have to, you know, be mindful of what will come down the pipe
Philip Hemme: later. I like that also of like, what’s coming from a platform, I think. It’s very similar in biotech, you know, it was a biological platform where usually your first product idea
Szabolcs Nagy: is usually not the best.
It’s more like a POC usually, right? You want to show that this thing even makes sense and you can get a drug, you know, out of it or a drug candidate or a drug like molecule out of it. And then of course you spend more time developing something that is really, really meaningful. That’s good.
[00:09:03] Open platform vs proprietary drug development
Philip Hemme: And transition, a very connected topic as well is, and I think for, I mean, what’s interesting with you guys is that.
You are, you have your own platform, the R& D platform, but you are more open than let’s say recursion, which has their own platform, but it’s only for their own internal pipelines. So can you talk a bit about this, of like the difference of models, why you went to that?
Szabolcs Nagy: Yeah, sure. I think it’s, it’s been a debate and it’s still an ongoing debate very much so in the field.
So yeah, again, when we, I think when the field started out again, 2010s, late 2010s, initially, I mean, most everybody came in, in almost like from a tech angle. Everybody said to investors, okay, you know, I’ll scale drug discovery to sort of the heavens in a way. I mean, I’ll, I’ll run a hundred programs for every program that pharma can run.
I’ll do something in days that they can do in a year. So those claims were very much. It’s tough. Like, Oh, it’s 2 billion to develop a drug. So we would cut it down by X percent. And exactly, exactly. We’ll do it for a hundred million or like 10 million or whatever. So there definitely, there were very aggressive claims about scalability.
And and those also led to again, the valuations, the tech angles, et cetera. And, and, and in a way, I think also because tech investors invested in the space, they kind of thought, okay, how are we going to make these companies extremely scalable? Like you know, B2B, You know, tech startup and the business model for that, of course, is to service, you know, companies and large enterprises, which are big pharma in our space.
And so most everybody started working with buggy big pharma. We’d hope that, okay, they’ll just pay me well for my cool AI technology. And what everybody found in the early years was that pharma itself was just very much warming up to the idea that that AI could be a thing that applies more widely across its, its processes.
And so almost nobody in biopharma was really willing to pay significant amounts to access technology early on. And also, of course, a lot of the technologies out there in the AI field in the late 2010s were not really validated. I mean, maybe you had some in vitro data to back up your claims, but nobody had more advanced programs, let alone clinical trial data.
And so that was like a catch 22 of, okay, I need to sell to biopharma, but I don’t have validation that shows that my cool AI gizmo will give them something really meaningful that translates better to patients than, than what they already have. And so they’re not really going to pay me and now my business model doesn’t work.
So that was definitely the the thing that happened early on. And so these learnings actually led to. almost everybody essentially questioning, okay, what is the right business model? Do I really want to sell to these guys that are not really willing to pay me enough for my technology? And then that turned into the, the massive shift of the AI field towards mostly developing pharmaceutical products essentially, because clearly everybody understood, okay, value accrues to the assets.
And so if my technology, my platform gets you, you know, essentially new assets or I develop new assets myself. Then then of course Farmer will pay me whatever they pay for the others in the field because they will not be buying unvalidated technology. They’ll be why, you know, buying certain stage assets.
And that looked like a, like a cool shift that made a perfect sense to everybody at the time. But of course. I’m pretty sure that a lot a lot of us actually, like ourselves included, really definitely underestimated the complexity of the actual drug discovery and development process and the insane number of things that can go wrong or delay you.
And so if you really look at the front runners, really the most successful companies today the recursions, Accenture’s. You know, benevolence and I’m, I’m, I’m saying success, not in the sense that they’ve, you know, delivered on their promises, but in the sense that’s how, how, how much do you raise?
For example, like you could take that it’s a super bad proxy, but it’s a good proxy of how excited everybody was about the technology. And so if you look at the most, the companies that raised the most they really Mostly went into the drug discovery direction and then ended up not really being able to deliver on those massive, massive promises of you know, massive pipelines, you know, hundreds of programs or dozens of programs.
And however, It’s clear that some of them were smarter, slash also potentially to some extent luckier, but definitely, you know, overall, I think smarter as well in, in how they played it. Because the ones that, Raised a lot, but didn’t really put enough of it towards actual drug discovery and development to have a wide enough pipeline and enough shots on goal to kind of get a chance to really get some successes in the clinic.
Those are the ones that are really struggling now. Like Benevolent, of course you know, it was, you know a super, super, super cool company. And they’re really, you know, front running pioneering company in the, in the, you are in the UK, of course, but but the, they sort of did a whole bunch of development or like discovery.
They tried to figure out what they wanted to do. They likely went too broad at one point, had to cut back significantly, and that ended up with a very, very few advanced stage assets. And then what happens to any biotech platform, it happened to them as well. They essentially became slaves to the the most advanced assets they have.
And as soon as they had a setback in the clinic, you know, now everything is falling apart to some extent where everything is, is, is, you know, a bit wonky over there. And so, you know, clearly looking back 2020 hindsight allows me to say this stuff. Of course, when you were there. the decisions they made made much more sense.
But but if you look at recursion, I think as really the best example here, they were able to raise so much over the years and they committed enough of it to drug discovery to an actual pipeline build that You know, now, especially after the Accenture merger, they’ll have so many shots on goal, so many stuff in the clinic, so many readouts, I think over like 10 or even 20 readouts coming in the next sort of 18 months, I think is what Chris mentioned recently.
So with all of those shots on goal, they’re likely going to get a few of them right, right? Even if they just hit the industry standard success rate, and then of course they’ll have some validation. They’ll, they’ll be able to. you know, sort of pull through. And I think ultimately what that means is that the business model of becoming a biotech is viable, but you either have to raise it off to really make it a portfolio play and have a reasonable chance of success, or you need to do insanely good drug discovery just as any other biotech platform to be able to succeed.
And that is a viable route. But essentially it really makes your company again, sort of completely subject to a lot of risks that your core platform has nothing to do with, right? It cannot help you really control all the risks. Maybe it helps you be faster in drug discovery, come up with a new molecule.
Maybe in our case, for example, it helps you predict biology a bit better, find something out that others couldn’t. But but none of these platforms can fully de risk the drug discovery process. And so ultimately, if you go the biotech route. you have to manage risks that are far outside of your platform.
And ultimately, as the field was learning about this we just felt so strongly that this really means that if our mission or our vision as founders is to build the best simulation of biology, which is, you know, harsh tech, then then we have to fully commit to that and we have to find a way to make a business model.
That, that focuses on making that technology available to the industry versus becoming a biotech that again, has all the risks that I mentioned already.
Philip Hemme: Yeah. Makes sense. And that’s, I remember when we talked, when was it, 2018, 2019, I think. I made an intro to you for you to biotech VC and you were more on the like developing assets.
Absolutely. And now much more on the
Szabolcs Nagy: open platform. Yeah. We’ve definitely done our share. So I will say this, right just to be completely transparent that it’s not like we always knew, like when we started, you know, spinning out of our Hungarian university, we kind of were, we started with the vision that we have today, but that vision was let’s build the best computational simulation of biology, because if you have a simulation of biology.
Surely that will help you run better experiments at certain stages of the process. That was the initial vision. That’s still a vision today, but in terms of what the business model is early on, we also thought, okay, we’re going to sell this to biopharma. Why not? That’s going to be the right play. And then we also found, well, the simulations are not really validated enough at their initial stage to really command enough of a price from the pharma world.
And so then we said, okay, well, this biotech business model sounds better and better. And so we tried to. Go for that. And we actually did, I think, in a way, we made the right bet because we said, we’ll, we’ll develop a very targeted pipeline to demonstrate the simulations capabilities in finding novel targets in high MFD patient populations underrepresented patients.
When you look at how well represented a certain patient population is in the available biological models. We said, well, let’s show the simulations are not just a point solution. So you can not just find the target in a, in a simulation, but you can also use it every time you run an experiment to sort of increase the chance that you run the right experiment.
to, to guide your program. And we did these two major things in our own pipeline build. And that generated plenty of in vitro and in vivo proof points, lots of good benchmarks that we today can use to actually convince biopharma much better than if we haven’t done it. But but we, we did, as you say, Definitely sort of court that biotech business model at the time, but luckily then the, the stock markets collapsed end of 21.
And just when I was ready to kick off my series, a, we had to completely rethink everything, which I think ultimately likely benefited the company. But Yeah. It was a learning journey, as you say.
[00:19:10] Turbine’s platform and offering
Philip Hemme: And can you talk about that, about, about today, about the platform, about the offering, about, yeah, who, who is the main benefits, benefactors, like?
Szabolcs Nagy: Sure. So as I said, really, the, the idea is that we’ve developed a computational simulation of human cells and tissue. And we use machine learning to, to learn certain parameters about how human cells function. molecules that act on proteins interact and how all of these interactions, molecular interactions sort of overall drive cellular decision making.
And that’s what we essentially use machine learning to do. And that allows us to predict an experiment we have never run yet. Essentially that’s the core idea that a simulation can predict something that is outside of the training set. So there’s no clear data point in the training set that we use that actually allows us to that allow us to predict something, but we can still do it because of the, the way we’ve architected the system.
Because of that, the, the simulations allow us to predict experiments at the earliest stages of discovery. So when you’re looking at. You know, finding novel targets, but even just trying to assess which is the right target to select and invest in.
Philip Hemme: Yeah.
Szabolcs Nagy: We can, we use simulations to then help select the right indication for a therapy program.
We can use simulations to identify potential biomarkers, design the right combinations, and even select the right molecule out of a potentially super large, you know, millions set of molecules. And at each of these stages, really, really ultimately. What the simulations do are essentially just provide a means for scientists to test ideas before they have to actually commit to the experiment.
And if you think about it, in any engineering discipline, that’s actually what people do, right? Clearly an engineer or designer doesn’t actually do that. build a small version of a car or a jet engine or whatever before they actually pick the right design and tinker with it. They do a lot of the iteration in computational simulations and that’s what we believed back in 2015 when we spun out of university that we, we had a way to build.
And we were also quite convinced that eventually the field has to, has to develop this because of course today we don’t have such simulations because we do not understand the underlying systems of our, you know, guiding our biological behavior and decision making on the cellular level, let alone on the tissue or the organ or the body.
level. But if we had such a simulation, of course, it would allow us to just come up with better ideas, hopefully, and get through the process faster and in a more informed fashion. So ideally, ultimately your chance of success also grows.
Philip Hemme: Yeah. Okay. And then now you’re on a partnering model or you have, I’ve seen you have a lot of partners.
So, and across different stages. So can you talk a bit on, I’ve seen you recently. Maybe can you
Szabolcs Nagy: start talking about this one? Sure. Absolutely. Absolutely. So what we are currently showing in the current partnerships and what, what’s sort of our focus for the next 18 months really is to demonstrate simulations in a variety of different use cases and show that the same underlying technology, the same underlying model of cellular biology can be applied across all of these different problems.
As you mentioned, we have an ongoing collaboration with Ono Pharmaceutical on essentially identifying targets in a very complex biological mechanism, which, which has all the hallmarks of like being a perfect fit to simulations, because it’s a very complex mechanism with potentially any human gene or any human protein being a target for patients with this specific sort of subtype of cancer.
But but you’re, there’s no way for you to actually run all of the experiments to really test all the targets in all the various subtypes of of patients where this particular mechanism plays a role. And you can do this in simulations and that’s what we’ve done about 350 million simulated experiments to actually test almost fully.
the combination of patient subtypes or cells representing patient subtypes and and the potential targets. And we were able to then not just provide a, you know, a ridiculously long list of ideas to the partner, because of course that’s not helpful from their discovery perspective. Coming up with even more ideas doesn’t really help scientists, right, who often have enough as they are.
It’s more about once you run the simulations and test all the potential outcomes. Hypotheses that you want to look into. You can actually in the simulations or using the simulation data, you can really try to understand what went on in each of those experiments and the ones that you found the most interesting can be, can be really deconvoluted and analyzed almost as if you had run a real experiment and actually to an even greater degree because a simulation is of course perfectly known in the computational, in the clouds and you know, sort of set up that we have.
So you know everything about the cell that you simulate. And of course you can never know everything about even a cell that you experiment on. And so we know what happens to each and every protein in our simulation. We know how each and every protein and their interactions and other molecules that act on them respond to a treatment by a drug or the knockout of a certain target or the modulation of a certain sort of interaction between two proteins.
And so a biologists who uses our digital lab can, once they run the simulations, look at every one of the interesting experiments and really try to understand what went on in there. And that mechanistic understanding is extremely key if you are a biologist, a translational scientist or anybody that, that draws work in, in drug discovery, because of course, lots of things can go wrong.
You know, look interesting in a cell. Like lots of things kill a cell, lots of things help mice with a xenograft. But but most of these things of course in the end do not really help the actual human patients that we try to help. And oftentimes the reason for that is because Let’s say a drug killed the cell through a mechanism that actually doesn’t occur in a patient.
Yeah. But because you only saw that the cell’s viability decreased,
Philip Hemme: reminds you Exactly.
Szabolcs Nagy: You can, you can cure cancer in a Petri dish. Exactly. And so the problem is that if you don’t know why something occurred, especially in a simulation, which is of course not even a real biological experiment, is just a simulation, just a model.
And even greater abstraction than even an in vitro Petri dish, like a, you know, in vitro model Petri dish. Because of that, you really need to understand why something occurs so that you can take that information and then design the right experiment to confirm it. And that’s something we’ve spent an insane amount of time working on to make the model interpretable.
To a trained biologist so they can glean new mechanistic insights or glean new ideas about biology and biological mechanisms from these simulations and then they can take that to an experiment and confirm it directly. So essentially they know everything they need to know about the essay that they need to run, the experiment, the perturbation, the model that they need to pick in a in vitro or in vivo or potentially even in a patient setting.
And that information helps them to lots of ideas that come out of a simulation or ideas that they may have already had and actually prioritize and select a feasible number to then experiment on much better than if they didn’t or if they haven’t run simulations. And so that’s actually one of the key benefits.
It’s not just finding new stuff, like in the auto collaboration, new targets. It’s more about, it’s, it’s just as importantly, the fact that we can provide an interpretable hypothesis that can be derived from the simulations and observing them.
Philip Hemme: And at the end of the day, if you’re improving or like pointing down or reducing the number of experiments or the number of ideas, you’re also improving.
making the whole thing discover faster, I guess, for, for the partner.
Szabolcs Nagy: Yeah. You would assume that, that really three things happen ultimately. One is that you You do reduce a number of experiments and that has a time, a time component or a time benefit, right? Hopefully you run fewer experiments, maybe fewer iterations of the experiment, and that can cut down on the time.
Of course, you always have to keep in mind that adding any technology to a process means that you’re adding a new step and that does take time, right? Even though we’ve optimized our simulation to be really fast, so you can be sort of up and running in like as little as three months, essentially you can get a specific insight into your problem, you can do that.
But even though that’s initial three months, that has to be kept in mind. But I think over like a couple of steps of the discovery process, you’re likely going to cut down time as well. You’re clearly cutting costs because fewer iterations, especially with certain experiments, it can be very costly. And we do hope that ultimately it does benefit the probability of success.
Not for any magical reasons of the but I think more because With the simulations, actually, we’ve routinely found in our own drug discovery work and also in partners that if you have this level of insight into the biology, or at least predictions on it, that actually sometimes reveals or opens up new questions or new lines of inquiry.
It tells you more about really what is the right type of patient. You are really more thoughtful about how you pick the models that you test your drugs on, what that means for the patient selection. The clinical trial design combination potential. And so this kind of insight we’ve seen, it actually helps you better prepare essentially your data package for that IND and hopefully better prepare your discovery effort for, for the trials.
And I think that ultimately does have a real impact on the likelihood of success.
Philip Hemme: That’s good.
[00:29:09] Pricing and value of AI technology in biotech
Philip Hemme: And you mentioned before on the, on the pricing that. was one of the challenge was to price the technology. Yeah. How does it look like now, now that you have more information, more partnerships and you have a bit sounds like your old platform is a bit more proven as well.
Szabolcs Nagy: Oh yeah. It was in 2018. I would hope so. Yes. But for sure, for sure. We have, you know, many more like novel targets, biomarkers, combinations have been identified and have been validated in various stages of discovery. So for surely more validated, the validation definitely helps. I mean, what we’ve shown, for example, in the collaboration with Ohno and in some of the other work that we’ve announced, for example, with AstraZeneca is that the simulations are seen as a much more likely source of new IP.
Then before, so there is you know, the more standard components of pricing like milestones, for example, for success. But really what we’re working on now and the new collaborations that we’re spinning up are following this format is that we, we actually don’t want to make simulations and partnering with us.
Such a pricey thing for pharma. that, you know, only the biggest companies can afford it, where we’re one of the very few bets that a large pharmaceutical makes on AI for a very simple reason, because we think that we’re the best used, not as a particular partner to solve one specific problem. So, you know, find me targets in this difficult area, or, you know, take my drug and give me one specific biomarker for the drug.
We think we’re much more of a capability to a biopharma. And we were best used like all the time, essentially. So if you say, okay, here’s my program is per clinical I really want to know which indication to go for, which patient subtype, you know, which trucks to combine with potentially what’s my likely combination effect with standard of care, et cetera, we can provide simulations essentially for us to answer all of those questions, help you run the actual experiments to confirm them.
And then once you’ve done that. You’re going to enter the next phase of your discovery or development effort and you’ll have even more questions as always, right? You’ll learn more stuff. You’ll see responses that you didn’t expect, especially in the trials. And those questions should be plugged back into a simulation.
New simulations should be launched to answer those questions. You know, why are certain patients responding? Why are the others not? What could be the differentiating biomarkers pattern between those two? For example, how do I expand beyond my initial indication where I’ve shown efficacy? These questions can also be answered and have been actually by various pharma companies using simulations.
And we really want to become essentially, I mean, I, I don’t like to use this word, but you could say like a co pilot for the programs, right? Different steps, essentially the simulations can be there. And this is also held by the simple fact that simulations do get better every time you predict and then validate a simulation result because the data that’s generated in the ensuing validation experiment can be fed back into the simulations training set.
And we’ve shown this with our own lab where we’ve sort of closed this loop, labbing the loop, quote unquote around the simulations that You predict, you validate, and then the data, if you run the right assay, and we’ve learned a lot about what is the right assay of the years, then that data can be fed back to improve specifically the simulation’s ability to predict around that target or that mechanism that’s of interest, or in that indication, for example, or in that model that you’re using.
And that improvement compounds over time. And of course, should benefit you more and more as you, with your drug discovery program, go from in visro to in vivo to patient.
Philip Hemme: Okay. And picking on what you just said, also, you want to be accessible to, let’s say, a lot of partners versus trying to optimize the biggest upfront possible, I guess.
So like, I guess what’s coming down also, and that’s the power of AI or digital simulations that you could scale it. Yeah. More than a biology.
Szabolcs Nagy: It’s essentially, absolutely. I mean, ultimately the simulations, the digital lab is, you know, it runs on the cloud, right? I mean, running on Microsoft’s Azure and also to some extent, like now on Amazon’s AWS and what essentially you, you get when you or what happens when you run a simulated experiment partner one or partner two is essentially just more essentially more cloud capacity being spun up to help answer that question.
So ultimately, it’s infinitely more scalable than any vet lab, as you said. That’s one of the key benefits as well, if you think about it. Again, I mentioned 350 million experiments for Ohno and that experiment actually took us, let’s say a week essentially to run, less than a week. And I’m training the model getting the right data in, et cetera.
That was actually almost the biggest amount of time to make sure that we’re predictive on, on what matters. So once our, as our models, predictivity improves over time and it becomes more and more predictive on more and more biological mechanisms, and then hopefully also beyond cancer cells. In other disease areas and more and more complex sort of tissue level effects.
Our plan is to have essentially a single model serving all of the various partnerships and that should allow us to make it very, very quick to get started and to get an answer. No sort of weeks or months needed to set up. You’re essentially using the best version of the simulated cells, simulated tissue that we can provide today.
Yeah. And we’re just a start, you know, a standard part of your practice as you sort of think about the right experiment to run, the right way to take a program. You’re always sort of running a few simulations, playing around with ideas, and then you know which experiment to actually invest months and hundreds of thousands into.
Philip Hemme: Yeah. Okay. That sounds good. And yeah.
[00:35:05] ADCs
Philip Hemme: I’ve seen you’re also active in the you have a models for ADCs, which is a super hot field at the moment. Yes. Especially in biotech. How, how, how specifically do you, like, how do you specifically, are you active in this field or
Szabolcs Nagy: like? Good question. What do you bring here?
Good question. When, when the immune oncology wave sort of hit us all those years ago, we were very, very early on. And so we could not really build. Simulations that were predictable, such a complex mechanism, and that was sort of a huge opportunity missed in a way. But now with, of course, the ADC wave, you know, rolling everybody over, we’re actually with this much more mature platform.
We’ve shown that we can predict a number of things that, that are sort of thorns in the side of a lot of the pharma companies developing ADC platforms or ADC programs. Because obviously, ultimately ADCs again are a really good fit for simulations for a simple reason. There’s all of these moving pieces, you know, what is your payload?
What is your linker? What is your antibody? What target do you hit? In what tissue? Are you looking for a biomarker? Etcetera. All of these together make up a too complex problem to try to experimentally really deconvolute. And so of course, without the simulation, you do end up cutting corners, you do have to make assumptions, sort of lean on the best.
information that’s available publicly, et cetera. Whereas in the simulations, what you can do is sort of test all the potential targets in all the potential biological contexts test all the various payloads that you’re thinking about in all of these contexts, and then run that whole search screen, the full grid, essentially.
And the benefit of that is that in a word, we don’t want to help you design a better linker. For example, we don’t think that that’s the simulations value at here. But what we can tell you is what is the right payload, for example, in what patient type to go for, how do you maybe want to cycle payloads at one after the other.
How do you overcome resistance that’s being observed in the clinic for the more, you know, front running crop two, et cetera, ADCs. How do you build the next generation of the ADCs and how do you expand the more advanced ADC programs beyond those initial sort of beachheads where they’ve been successful?
What could be a good biomarker, for example, to, to use to select patients if you, you, you’re not succeeding with more like an all comer approach or not all comer, but of course you are more going for some target expression. It’s like, what’s the biomarker beyond the target expression? These are all clearly questions that the ADC field needs to struggle, struggle with or grapple with because they’re leading to.
Sort of patient benefit being left on the table and trial failures. And this is something that we’ve now really put a lot of effort into generating. Targeted data sets to help train our simulated cells to better predict ADC effects and payload effects and and helping them essentially find the right payload subtype to target combinations essentially.
Yeah.
Philip Hemme: And I mean, ADCs, there’s some big pharma, I mean, there’s definitely some, some big pharma active in the field, but you have also a lot of biotechs or smaller companies. Yes.
[00:38:19] Working with the average biotech
Philip Hemme: So how do you work with. Let’s say ADCs are even beyond ADCs. How do you work with smaller biotech that has raised typically, let’s say the average European biotech has raised 50 million has maybe a program or early stage program in, in their, in ADCs.
How do you work with them?
Szabolcs Nagy: This comes back to the question about pricing that you mentioned. And I mentioned that. What we really want to do is find a way to work with any company, because again, if we, if we want to go for a business model that focuses on partnering and then making this technology available essentially to others and being valued for the utilization and revenue drive from this technology, then we have to find a way to go beyond the biggest 10, 20 pharma companies.
because that is not a scalable business model. And for that reason, we’re working a lot on developing a generally predictive model of cancer types and certain cell types and certain tissues, because we believe that if we have well performing model that doesn’t have to be fine tuned, we don’t have to generate more training data, for example, for it, then we can make that accessible for somebody that’s as much a company that’s much smaller.
to run simulations on without them having to actually pay us to generate even more training data and fine tune the model, which is actually the costliest part. It’s not the simulations, it’s more the, the data generation in a very targeted fashion and also the training of the models that takes up most of our compute resources.
And then of course, interpretation. When our trained biologists and translational scientists work with the other team on the pharma side to help them understand simulations and help them derive hypotheses. And if we can develop in a way that sort of strips out a lot of these resources from a, an interaction with a small biotech that really wants an answer to a very specific question or wants to just support simulate simulations, a single program, then, then we can make it available.
At a price that’s sort of feasible to pay because it’s a technology and a capability and you’re getting your money back because you’re cutting time iteration and failure out of your process.
Philip Hemme: Yeah.
Szabolcs Nagy: That’s good.
Philip Hemme: And that’s,
Szabolcs Nagy: you’re already at that stage now where you have a general model. We’re talking to, we’re talking.
Yes, yes, yes. That’s it. Something that I didn’t mention, but just as a sort of an interesting data point, really, when we started, we, we sort of set some foundational rules for how we develop our technology. And one of those rules was If we want to predict cells in the more complex biology we need to build a model that is a singular model underlying every, every different cell model that we have, for example.
There’s a simple reason for that. Ultimately, if you think about it, human biology has a set of foundational rules that really guides how it behaves and how it makes decisions and how it reacts to outside stimuli. A healthy lung cell is very different from a colorectal cancer cell of a certain patient.
But we believe that there is foundational set of sort of rules and guideposts that you can figure out that help you then create from a singular representation of biology, a model that predicts on the cancer cell and also a model that predicts on the healthy cell. And we think that that has to do with, again, how proteins interact and how other molecules act on proteins.
And essentially learning that foundational wiring diagram is what we call it, allows us already actually to take one basic trained model and add specific data coming from a certain cancer cell type coming from a certain cell line or coming from a patient’s tumor and then have differential behavior and accurate behavior when, when simulating the effects of drugs or knockouts on, on these different models.
So we already have that and of course it’s a, it’s a working progress. Current model has about 8, 500 proteins and their various interactions and various effects acting on them in it. And we’ve shown that it’s predictive on the cancer cell level and we’re going to go beyond cancer cells. Now we’ve shown some productivity in immune cells as well, but it’s very much sort of a long, a long, long, long road map that we have to continue to travel on.
So it’s not done for sure.
Philip Hemme: That’s good. That’s really good.
[00:42:42] History of Turbine and Szabolcs Nagy
Philip Hemme: And maybe now stepping down to back to the history of the company, I mean, you mentioned a bit, you’re a spell from the university, which I think I’ve seen the building like on the right around the corner, really cool building. Can you talk a bit on the history from, from there to now?
I know you, you went through Berlin and through the biograms, that’s where we, where we met. And yeah, can you walk through the main steps of the history of the company?
Szabolcs Nagy: Sure. Sure. Sure. Sure. So as you say, it all actually funnily enough started right around the corner at Semmelweis University where two of my co founders, Donnie and Krish were CSO and CTO respectively.
They were collaborating to essentially build the very foundations, the seeds of the simulated cell. Like one of them is an, as an electrical engineer, software engineer. Math whiz turned biochemist, and the other is a, as a medical doctor that wanted to, you know, find a more rational way of, of treating and diagnosing disease.
And they essentially put two and two together during their PhD work and figured out that you can represent. Cellular biology as a network that can be simulated and and eventually also figured out how machine learning can be used to train and parametrize and then grow and expand this network.
And we took this, I essentially came along to help canalize the spin out. Funnily enough, I had nothing to do with any, anything that we essentially work in. I had nothing to do with oncology, drug discovery. Let alone AI, but but I did actually have some experience spinning out cool IP and cybersecurity actually from university and making that successful.
And I wanted to recreate that, I wanted to do that, but in a more meaningful field. And of course that meant years of learning for, for all of us. I mean, for me about everything, essentially. To be even remotely useful, but for the guys also there had to be a lot of learning done to better understand ultimately the problems that we could use with this cool technology because we were very much a science first team where, you know, really cool idea, cool early results, but what is really the problem that you’re solving and what is the product?
And that’s something that I hope I helped figure it out, figure out just based on my previous experience doing this in a different field. And what really catalyzed us was. The ability to work with Bayer very closely during as you said, the grants for apps programs that he used to, you know, run at that time, a shout out to the entire team because of course they really transformed us because they gave us a chance to sort of roll around Bayer HQ, talk to everybody from the C level down to bench scientists.
And that really revealed some of the core use cases they’re running experiments based off of the, the best knowledge of a trained scientist falls short. They’re just too many experiments need to be run to really understand what’s going on. And we started developing simulations to help tackle those core use case around identifying biomarkers, overcoming resistance, designing combinations, and understanding mechanism.
So these were very early inputs and they allowed us to really develop the, the simulations in a direction that actually is useful and, and can be deployed sort of in the day to day life of a drug discovery scientist. Because we, we wanted to make sure that very early on, we don’t spend that, you know, 10 years, you know, raise, you know, tens of millions and not really have anything that’s impactful.
So that, that’s what we focus on for first years. And then around 2020, 2021 is when we started fundraising, when we had enough data from the early pharma collaborations that we were predictive and we’re finding interesting things and actually impacting decisions across the, the R& D process. And as you mentioned earlier, we then very, in a very targeted fashion, validated core simulation use cases and solutions we have and in our own discovery work.
And then we raised our Series A about two years ago from Merck or QS to actually really start commercializing essentially the simulations.
Philip Hemme: That’s nice. And then I saw you, you raised or Accenture invested, I
Szabolcs Nagy: think it was in May or? Yeah, yeah, yeah. We joined forces with a couple, a couple of new companies and sort of struck some interesting partnerships and we’re doing some more now.
We hope to announce some of that in the next quarter because one of the key questions around, like, how do you get simulation into the sort of day to day life of. Of scientists that have never used such a technology and just do not run their process, do not work like this yet. Is of course, like who talks to those scientists, who talks to maybe the leaders at those companies all the time and who helps sell technology to these types of enterprises and companies and companies like Accenture seem to be like, you know, sort of a left field out of left field collaboration.
But of course Accenture is one of the most successful companies on the, on the, you know, on the earth. Accenture. selling technology to, to Pharma. They definitely, you know, I know they’re more for IT. Fair point. I wanted to say actually, can they sell to R& D? That’s really, that remains to be seen. That’s definitely a work in progress, even on their end.
But I think what they, what their focus on, on our field reveals is, It’s that if you really think about it, pharma R and D does not yet really know how to buy technology, which you can completely understand because people in R and D are extremely smart, really good scientists, really good decision makers, and they know how to partner around therapeutics.
But some of these new technologies clearly sort of stretch the boundaries of how they used to work. value things that they get from partners. And as we said, first generation of drug discovery or AI and drug discovery companies clearly went to a business model that fits the way the industry works, but I think us and a lot of others that are, you know, sort of the next wave of technologies, hopefully we can start shifting the way that pharma we use and biotech we use technology and how they value it.
Philip Hemme: Okay.
Szabolcs Nagy: And that starts with learning how to, how to sell technology and educating our sort of buyers essentially on how to buy technology in R and D as well. And companies like Accenture are also trying to figure that out because as you say, it’s not ID, IT, sorry, that’s going to buy an AI technology to make better decisions in drug discovery and development, but but clearly somebody will need to sort of figure out how to do it.
Okay. the inside of these companies and partnering with. Much larger enterprises than us that sell to these sort of end users routinely helps us also learn how to do it better. And also hopefully get to more more, more entities and maybe be more integrated into their Workflows and other technologies that they purchase.
Philip Hemme: Okay. Yeah. So for example, it’s a step to, yeah, to see AI drug discovery or any kind of digital tool for drug discovery. Absolutely.
Szabolcs Nagy: As in like, I mean, if you really think of it, yeah, if you think about it, what they’re trying to do is, is, and of course, many others, McKinsey and the others are also doing this they’re trying to figure out how do we sell to essentially a new budget, a new type of customer.
But for doing that, or in order to do that, it’s not enough that they’re extremely. competent at the basic computational technologies of our time, like cloud and AI and gen AI, that’s not enough because they need to also be domain experts. And that’s what we, and a bunch of others provided are now sort of strategic partners over there.
That
Philip Hemme: makes sense.
[00:50:09] Building biotech in Budapest and travelling
Philip Hemme: And one, I think more personal question, but you, you, I mean, you’re living in Budapest and then I see that you’re traveling a lot to either European hubs, either Boston. How, how do you like prioritize, yeah, how do you prioritize all of this
Szabolcs Nagy: traveling? Yeah. I also have a small daughter now, so I also need to be, you know, dedicating enough time there.
It’s a good question. I think it’s, it’s also, you know, clearly that I will say that I don’t think we’ve sort of proven that it can work like this, right. That you don’t have to move to, the UK or to Boston to build a successful, you know, AI and biotech company. I think we’re putting in a good case, but I think what I miss the most is just the, the gossip essentially around the field.
Just more information about what people are trying, what doesn’t work, what works, what’s going on, who’s looking for what, you know, who’s interested in what. That’s clearly the, the really the most useful information as you think about business models, as you think about strategy, as you think about BD, these information, like this information is what you’re missing the most.
And and the, We’ve sort of found, found workarounds for that. Essentially we’ve built a really strong board, really strong SAB, not just strong in the sense that it’s really good and smart and well validated and credible people on there. It’s more about having very strong personal relationships with them that, that I think helped us individually actually.
To, to try to sort of stop ga that or try, try to learn the information that I would learn likely sort of day in and day out if I, I live somewhere else.
Philip Hemme: Yeah,
Szabolcs Nagy: of course. Develop, you have to compensate, essentially you can compensate. You do have to take a bit more travel, but I think it’s more about try to do enough travel to, to get this type of information and the reason that we stayed home.
was because I think sort of two reasons. What is, of course, there’s clearly sort of operational advantages to running such a company out of the center, you know, Central Eastern Europe versus even the UK, let alone the US, you know, clearly cost advantages. And you’re also being able to tap into really good talent, especially on the computational and AI side.
They’re also actually on the biology side and talent that doesn’t really have Sort of anybody that competes for its attention, if they want to work on some of the really cool problems around drug discovery and AI in biology. because we’re really almost the sole sort of more advanced or sort of scale up phase company in this region.
And so we get to really take the cream of the crop. That definitely helps. But the other advantage is that, or the reason we stayed home is because we just wanted to do something good for our ecosystem. And of course, when people usually hear Hungary, I don’t think they hear it in a very positive context.
And, you know, clearly there’s Not just us, but a lot of others that do good stuff, good work here, that don’t agree with the politics, but want to still do something good for our country and, you know, the region more widely. And clearly if we move somewhere else, we may have an easier time of doing certain things.
But I think being here allows us to have a different view on a lot of problems, maybe think about it, things in a more, sometimes in a bit different way that gives you a competitive advantage and also hopefully like pull others with us and, you know, give a better shot at the next generation of let’s say tech bio companies coming out of the region.
Philip Hemme: I mean, you did pretty well so far.
Szabolcs Nagy: Well, you know, fingers crossed.
Philip Hemme: Fingers crossed for the next steps.
[00:53:36] Data-driven drug development
Philip Hemme: Maybe last, last question before the quick fire is on. I had Thomas Clausell from Orkin on the show. I saw. And one thing that he, he mentioned pretty clearly at the beginning was, But the industry needs to be more data driven versus good anecdotal.
Yeah. I guess you must agree on this, but maybe what’s, what’s your take there? What’s your, yeah, what’s your view
Szabolcs Nagy: there? I think actually it’s funnily enough ties to like how pharma needs to eventually learn to value technology. Technology is like ours because I think currently people are like, you know, AI is useful if it gives me a new drug or, you know, new IP of certain sort.
And. A lot of, a lot of these technologies, us included, can provide that, but I think ultimately the bigger impact of AI will not be necessarily just a new IP, it will more be the better decisions and as Thomas said, better IP, better, better IP, yeah, in a way, like I think the more systematic approach to drug discovery, just because you’re not limited to the number of experiments you can run at any stage.
You are limited to the number of ideas you can have, essentially, and and you can find the right experiment to run more like, you know, in a, in a higher chance or with a higher chance. I think that more systematic versus more anecdotal approach will really be ultimately the bigger impact. And I think already today, that’s the bigger impact AI has on the field.
I’m under the hype. And yeah. If you think about it, what we, what we usually say is that we want to turn biology into an engineering discipline, but I think this applies more broadly to human health, right? Of course, this is not going to happen tomorrow and you know, we ourselves are not the sole solution and we ourselves will likely make, you know, we hope.
A small, but overall significant dent in the overall problem. But if we don’t invest into these technologies today, if we don’t test them, if we don’t learn how to use them to make better decisions, then ultimately the field is not going to, to be sustainable as we know, it’s not going to go ahead and move, move forward.
So clearly we have to make these bets. if we are to turn, you know, biology and broadly health into engineering in the best possible way. Yeah. Yeah. Not to, not to, you know, take out the human touch, but to actually help, help put that into health.
Philip Hemme: And do you see some actors that are better students in the, in the growth?
I mean, just better at being data driven, some farmers that are more active, more embracing it. I mean. Clearly,
Szabolcs Nagy: clearly Roche slash Genentech, you know, made a massive bet with Aviv. Being named head of r and d and, you know, clearly, clearly a lot of them are investing into was about, but but I think say not just saying, okay, I, I really name like a CIO or versus I usually not a cio, but it’s a, it’s not just that I have a strong data science slash ai group within my teams, but it’s that, you know, really somebody that bridges biology drug discovery and ai.
He is the head of R& D for me that that is a huge sort of step forward and a big bet for sure. We’ve had really good, I’m not just saying this because we do collaborate with them, but I think AstraZeneca is, is very knowledgeable and actually has deployed various data science and machine learning technologies.
The real problems that they have. Yeah. So they’ve really invested there. You know, clearly Sanofi, of course Sanofi, I was would say, is going all out. And I think the Sanofi actually does a really good thing thing for definite
Philip Hemme: system. But is, is it PR or, or how much of it is PR versus, but I do, but I think they put really like hundreds of millions.
Yeah. I think that
Szabolcs Nagy: that. I’m sure that it’s PR to, to to some extent, but I will say, I think actually that’s exactly what, what this field and these applications need because they need somebody saying, listen, look at all this money I put into this particular technology. I, A made a bet. B gave this company a chance to actually progress.
C. I’m talking very openly about it. So actually that tells everybody else that they should be thinking about making these bets. So I think Sanofi really does actually do a good work across the field for just sort of advertising the use of AI better, like how impactful it is day to day. Yeah. I, I wouldn’t be able to tell, but but my assumption is that, yeah, we’ll see.
Well, we have to get our, you know, get our game up beyond oncology for that, most likely.
Philip Hemme: Cool.
[00:58:04] Quickfire
Philip Hemme: I think it’s a good, it’s a good way to wrap up and to, to finish on the, on the quick fire. So quick questions, you answer one sentence or yes or no first one, what’s, what’s on the top of your mind at the moment?
Szabolcs Nagy: Yes or no? Yes. What? One sentence. One sentence. Top of mind for me. Essentially pricing model. How do we. Get simulations in the door at a pharma and then convince them that they should use simulations over time, multiple times.
Philip Hemme: What’s one of your favorite biotech books?
Szabolcs Nagy: The Billion Dollar Molecule.
Philip Hemme: The biotech store, yep.
Szabolcs Nagy: I think it reveals how crazy this industry has always been in terms of its approach and how much it’s just sick of luck plus capable people like giving it their all.
Philip Hemme: Yeah. And how much you need to sell your story. Yeah. If I remember actually fair point from this, from this book.
Szabolcs Nagy: Yeah. Yeah.
Philip Hemme: How much selling matters point, even when you, yeah. One mistake you made in the past 12 months.
Szabolcs Nagy: One mistake. Not fixing my sleeve schedule before BB Maya’s arrival. I don’t have a chance anymore. About
Philip Hemme: how much do you sleep at night?
Szabolcs Nagy: I guess? Well, actually up to 6.5 hours. Which is, I think, historically, okay for me.
Philip Hemme: You’re still functional. Yeah. What’s the, oh, what’s the most impressive drug on the market at the moment?
Szabolcs Nagy: First of all, you have to say Keytruda for sure. I think just the story of how it got to where it got to, and also the approach of then, like, making it essentially a standard part of any trial, I think, you know, kudos to Merck.
Philip Hemme: And 20, what is it? 25 billion. Yeah. And you want to say something? Yeah, yeah, I’m saying One of your favorite European biotechs?
Favorite European biotech? The
Szabolcs Nagy: company? Yeah. I think Biontech. Hmm. Biontech for sure. Just because they also do, the founders also do a lot to give back to the industry and invest more. Hmm. Yes,
Philip Hemme: we had I just talked last week with the, the former CBO. So this episode will come up and I think they do a lot on AI as well.
Yeah, yeah. Last question. One of your biotech heroes, European or not European
Szabolcs Nagy: biotech heroes? I think I will say Aviva gv for the simple reason that I like, I’m really interested in, in seeing the. the outcome of that experiment at
Philip Hemme: Genentech. I
Szabolcs Nagy: think like you know, she’s, she’s making a lot of bets around, you Implementing AI into a lot of, some of the things that we also want to implement it, but also of course, others are working on.
And I hope that they share more about their learnings. Awesome.
[01:01:20] Thanks for listening
Szabolcs Nagy: Thank you. Thanks for having me. Great chat. It’s fun. Yeah. Thank you.
Philip Hemme: I’m convinced that AI and machine learning will change drug development and that simulation will be a big part of it. I’m also impressed by what. Zabi and his team was able to build especially out of Budapest. If you also enjoyed this episode, please hit the like, follow, review button. Any of these actions would help a lot to give more people access to the show.
If you want to see similar videos, please feel free to check the channel where we have many more. I would also be curious to hear what you think, so please leave a comment below or send me an email at philip at flot. bio. All right, thanks for staying until the end and see you in the next episode.
