Using AI & Virtual Patients to Make Clinical Trials Safe, Affordable, and Scalable with François-Henri Boissel CEO of Novadiscovery

 

 

Kelly Stanton: 0:16

Hey everyone. Thanks for joining the show today. I'm Kelly from Qualio and I'm your host here at From Lab to Launch. If you haven't already please subscribe and give us a review on Apple or Spotify. We'd love that. If you want to be on the show, please see the application link to the show notes. We've had a lot of people reach out and it's been a pleasure to connect with you today. Joining us from France is François-Henri Boissel co-founder and CEO of Novadiscovery. You can read the full bio in the show notes, but a quick summary François has been the CEO of Nova since its founding in 2010. Prior to that, he worked in investment banking at Lehmann Brothers in London and Tokyo, where he developed an expertise in financial engineering deal, structuring and execute. Nova is a leading health tech company, which has developed a hybrid approach, combining mechanistic models and artificial intelligence to de-risk research programs and optimize clinical development, essentially simulating clinical trials with virtual patients called in silico, which has really taken off in the last few years. JINKO is Nova's unified silico clinical trial simulation platform. We'll dive into the details and applications with our guest François. Let's bring him in. Good afternoon, I think for you. we love hearing how companies launch to tell us the story behind Novadiscovery.

François-Henri Boissel: 1:40

So it's, it's your usual father and son story, I guess, back in, in 2008 I, I left investment banking or rather investment banking and left me. I was working for Lehman, the company, as everyone knows went belly up I was working in Tokyo at the time for their structured finance group and decided that at that juncture in my career at the time was probably ripe for a bit of a fun and entrepreneurial experience. So. Relocated back home and started to look for something that would combine a couple of things, a strong social impact and potentially, and that's probably from my early banking years higher risk and high return in case of success. At the very same time a father was retiring from decades in academic research and especially in clinical pharmacology. So he pretty much spent the majority of his career thinking about the right methodology to evaluate new drugs. On on patient populations and he arrived to the conclusion during the nineties that there was so much knowledge out there in biology and medicine that we had reached a point where it was literally impossible for the brightest human brains to make sense. Of the entirety of available knowledge on any given disease. And therefore the only way to ensure that a drug or Randy would be more efficient, I would pretty much be to apply mathematical modeling to this knowledge in biology and medicine, and to build virtual patients in order to optimize R and D. So. We spent some time figuring out a business out of this realization eventually started Nova in 2010 and happy to expand on the journey if you will.

Kelly Stanton: 3:36

Oh, absolutely love to hear more about that too.

François-Henri Boissel: 3:39

So it's been the sort of a situation where, and I think that's pretty much the universal when you're looking at innovative startups. The key question, the first question that one needs to ask especially investors is, is the timing. You can have the best idea around the block. If the ecosystem is not ready for change then you'll be bouncing your head against the wall. So we've operated as, as a science experiment for the first, almost decade of existence before one inflection point actually from the USFDA helped us start to build. You know, sort of tangible use cases with customers that eventually led to our first round of institutional investments. And now the pretty aggressive scale-up trajectory that we've been going through in the past two years or so.

Kelly Stanton: 4:28

Tell us about how your forum helps biotech and pharma companies from discovery to clinical development to market access.

François-Henri Boissel: 4:36

So, in, in the most sort of, fundamental terms the issue, plaguing research and development for new drugs today, and it's been an issue for the past at least two decades, I would say is that. The paradigm is largely driven by experimental sciences which essentially means that it's a succession of trial and error attempts to test biological assumptions. And this comes with at least three detrimental consequences for patients. The first one is that success rate is extremely low. Second one is that it is a very capital intensive type of a. And the third one is that it takes a very long time to get innovative products, innovative drugs to patients in need and in a simplest terms and most fundamental way to put it. I guess Nova's mission really is to instill an engineering mindset into this whole drug R and D paradigm. So therefore, Give biotech and pharma companies with our GENCO clinical trial simulation platform, the ability to. Muddle simulate and predict in a much more efficient way than would otherwise be the case with more conventional experiments. So I'm not saying that we intend to replace trials on human subjects. This is not desirable for many different reasons, but what we want to achieve is thanks to the predictive capability of the platform, make sure that we help those companies design. More targeted trials, focusing on patients who are likely to respond to the product and thereby XR racing the timelines so that patients can have access to those innovative drugs in a much faster way.

Kelly Stanton: 6:24

That's interesting. So in that modeling and such, then I would imagine you would want to use maybe existing clinical trial data and things to sort of teach. The the model then how does that, I guess, without getting too far into proprietary questioning I'm I'm curious, how, how does the model know? What, what are you, what sort of existing data do you leverage for that sort of

François-Henri Boissel: 6:48

thing? So that's a very good question. And that's that there's actually a catch we're not using data as an input to build those predictive models. We are. Extracting and manually curating knowledge from scientific articles. What we want to build when we build a disease model is effectively a map of all biological entities that are known to be involved in a particular biological process or series of processes and the functional relationships between those entities. So it's. Paradigm, we call this mechanistic modeling or Kozol modeling as opposed to AI, which as sophisticated as it may be at the end of the day, you rely on correlations and they tend to break down in in biology. This is not to say that we're not using roll days or whether preclinical or clinical at any point in the process, we do use them. But once the disease model has been built, that it is sort of scaffolding, which is derived from the scientific literature. We will use data to refine the calibration of the model, and then to validate the models, predictive power by essentially reproducing past experiments. So. Not using at any point, the data from the validation sets to inform the model design, but rather really try and super-imposed simulation outputs with what has been otherwise observed. And if there's a satisfactory overlap, then the model is deemed to be properly validated and we can run a simulated trials.

Kelly Stanton: 8:20

Wow. Wow. That's exciting. Goodness. Your website has some interesting case studies and deep dives like the one for hepatitis B. Can you tell us more about how your team is applying this technology to problems like Hep-B, which has no cure currently?

François-Henri Boissel: 8:36

So that's, that's a very good illustration of, of the points I was making about the value proposition earlier on we've helped a biotech company. Optimize the design of a phase two on human subjects by running for them. Approximately 700 different trial designs. So a simulated face to explore. Treatment combinations. So they wanted to see if their product would fare better with patients by combining it with any available standard of care as well as optimizing the regimen of those treatment combination. So looking at different doses, different administration scans. So at the end of the day, the in silico phase two was performed on 78,000 digital patients with a very interesting characteristic of those in silico clinical trials, which is. Your digital patients from one scenario to another are absolutely identical, which is obviously impossible to achieve in real life. You do have differences and most of the time that are not accounted for between the, the treated and the control group, for instance with trials on students' subjects. So, we help them essentially identify. Avenue to continue their clinical development program by pointing towards what according to the simulation outputs was the best combination and the best regimen. So I guess the key value we created in this particular instance was the ability to explore multiple scenarios in, in a, you know, dimensional space that would be unachievable in real.

Kelly Stanton: 10:11

Definitely. Yeah. And certainly the, the, the modeling I've been involved in, in my, in my time supporting clinical trial work as a quality specialist it just felt like there was always one more thing you didn't think of. And when biology is involved, there's always one more thing. Right. So I love that you can. Get it and see it, 78,000 that's, you know, nobody can run a trial that size you can't afford to let alone finding. I mean, gosh, that would, that would be aids of, of patient recruitment. So, wow. Talk about accelerating. How has the COVID-19 pandemic accelerated, acceptance and adoption of in silico trials?

François-Henri Boissel: 10:55

So that's a really good question. I think there's a short term and a longterm impact in the short term, what we've seen in, in 2020 and 2021, where. A lot of interest from pharma companies, inbound inquiries, which quite literally never happened before that. We we've started to bootstrap the company and the entire scale-up phase, essentially supporting smaller biotechs rather than large pharma companies. But to the extent that all of their ongoing trials were being disrupted and you know, how has those large companies operate? You have the budget, you know, that's as simple as that. And so they figured out, okay, I need to continue to explore some of my key assumptions about my program. Well, let's try clinical trial simulation. So that was very beneficial from a sort of short-term impact standpoint. And we progressively started to. And now actually the majority of our deal flow is with large pharma companies rather than smaller biotech launch. And I think the second interesting impact the longer term one is that the COVID pandemic has changed expectations about the duration of clinical developments. We have demonstrated to the world that we can. Much faster, certainly in dire conditions. And we need to be very careful because, you know, obviously we're dealing with the, the safety of patients. So I'm not suggesting that we should be cutting corners going forward, but we can do much better and much faster.

Kelly Stanton: 12:21

absolutely agree with that. One as a founder and leader of this company, what would you say are your biggest challenges or opportunities to this year?

François-Henri Boissel: 12:31

So challenges and opportunities. That's an interesting one opportunities. I see opportunities everywhere. There's so many areas and companies that are not yet into, in silico clinical trials. So there's You know, therapeutic areas that we need to invest in and get those conversations going with pharma companies that are still looking into the technology rather than research investing in it. Hopefully there are less and less of of those those companies. You, you now see within Gerawan pharma companies Groups modeling groups being being set up. So that's very positive for us. I guess the one key challenge that we're faced with this year is the which is a very good challenge to have is the, the scale-up scaling up a company is incredibly difficult. It's incredibly difficult because you know, you're imposing a cadence on, on the team. And, and that can be fairly brutal. You need to be extremely adaptable and comfortable with ambiguity and that that's not necessarily as straightforward. You need to ensure that those sort of early joiners and team members rise up to the challenge of evolving towards managerial functions. And you need to continue to recruit as fast as possible while upholding your standards. So that's, that's a key challenge for us.

Kelly Stanton: 13:51

Definitely. No. And I'm hearing a, it's interesting as a quality professional in the space I would imagine the adoption and that being comfortable with this idea of, you know, modeling. And computers in places we can't see, it's not paper, I can't touch it. Right. Like how you, how do you overcome that? The clinical QA folks in my experience are among the most conservative among us quality folks. And so I would imagine the idea of that technology is a little scary for them.

François-Henri Boissel: 14:24

Excellent question. Fortunately, I have an answer to it. So it's actually a combination of having an established regulatory framework for model validation purposes. So that's, that's something that the FDA has been working on since quite a number of years, actually historically starting in the medical devices, clinical trial space but gradually shifting towards drug developments. The agency has launched in 2018, a pilot program called model informed drug development. MITD which as far as I understand is one of the most successful pilot programs in the agency's history based on the number of applications and what will come out of this program is essentially the. You know, using the equivalent of a GCPS of good clinical practices a set of good simulation practices. So that's number one in order to build trust and confidence. Then the other driver, which really is specific to Nova and our GENCO platform is that the entire model and framework is fully white box. So even a non subject matter experts in modeling applied to biology, someone who's essentially in charge of designing a protocol for a future trial, but does not know anything about mathematical modeling applied to biology. That person can trace from the output of the simulation on the platform to each primary source of knowledge and literally extracts of those articles that went into building the models. So in other words, there's a blueprint of this complex mathematical series of equations, which is in texts and graphical form that can be understood with someone who knows something about biology and medicine. That's

Kelly Stanton: 16:18

incredible. And the perfect answer to those questions. How do you see this technology evolving in the next five to 10 years?

François-Henri Boissel: 16:28

So I, you know, I'm, I'm now a hundred percent certain that the, the most difficult in terms of adoption is, is probably behind us. So in other words, it's not really a matter of if it's just a matter of when, in terms of widespread adoption within the next five. Or potentially, if we're being a bit more reasonable within the next 10 years, it will become part and parcel of any dose that is submitted to the FDA or the EMA.

Kelly Stanton: 16:56

Awesome. Excellent. If you could go back and tell yourself something at the start of your career, what would that be?

François-Henri Boissel: 17:02

Okay. That's an interesting one. Not necessarily. In hindsight, I wish I would've known, but rather something that on me a couple of years ago. And maybe it's, it's very specific to the way business is being taught in a French business schools. It's very theoretical and it's very focused on how skills especially talking about finance, obviously but not, not necessarily only finance and in order to be successful in your early days, what would I've I've sort of recognized and realized is that you need to be on top of the game from a hard skills. But as a curator progresses over time, as you start to have more exposure to senior responsibilities, then there's a shift that starts to occur towards soft skills. And I have to confess nowadays for lack of practice my heart scales Negligible. I sort of, you know, I, I, I still know how to use a spreadsheet, but that's, that's probably the most I can give and sub skills really become a lot more prevalent in terms of ensuring success and success measured by, you know, how you, you, you need to listen to your team. You need to. Work on aligning everyone so that the entire organization shoots in the same direction, which obviously gets increasingly difficult as the headcount extends over time. But certainly the balance between hard skills and soft skills is, is changing as as one progresses through through the career path.

Kelly Stanton: 18:33

Definitely, the, where can people go to learn more follow along and connect with.

François-Henri Boissel: 18:37

So I guess the easiest way is to reach out to me on LinkedIn. And I'll just see on the website novadiscovery.com. There's an entire page dedicated to resources and sort of educational content. So you'll find blood posts. You'll find an ebook on in silico clinical trials. As well as references to the articles that we have published over the past couple of years.

Kelly Stanton: 18:59

Thank you so much for your time this morning, France. Well, it's been a pleasure chatting with you.

François-Henri Boissel: 19:03

Thank you guys. Had a great time myself.