The importance of balancing AI and people in drug development

VOICEOVER: The Bioworld Insider podcast.

Lynn: This is the Bioworld Insider podcast. I’m Lynn Yoffee, Bioworld‘s publisher. Today on the show we have Josep Bassaganya-Riera, who has created two companies. He’s the founder and CEO of Nimmune Biopharma, which is developing treatments for autoimmune diseases. And he was the founder and CEO of Landos Biopharma, which was acquired by Abbvie in 2024. He brought Nimmune’s lead asset, omilancor, with him from Landos, and he’s readying it for phase three studies in ulcerative colitis. He has more than 20 years in the industry, and he’s here to share his insights and experience with us about using artificial intelligence-enabled drug development. Thank you very much for joining us today, Josep.

Josep: Thank you for having me.

Lynn: He’s talking today with Lee Landenberger, a Bioworld staff writer and the Bioworld Insider podcast host. Lee?

Lee: Thanks, Lynn, and we greatly appreciate you being here today, Josep, to talk about the company and about AI and some broader topics. I’m really curious about your proprietary platform, your AI platform at Nimmune. Can you tell me about it, and would you also talk about how it fits in with the foundation that you created?

Josep: Absolutely. The proprietary AI platform at Nimmune is called the Titan-X platform, and it’s a combination of AI, bioinformatics, advanced computational modeling to really advance, accelerate drug development from target identification to identification of novel signatures for precision medicine to tailoring drug use to the right patient population, primarily in inflammatory and autoimmune diseases, but I would say that it’s an agnostic platform that can be applied to immuno-oncology, chronic diseases and other diseases that affect humans or animals. And the genesis, and I would say the inspiration, of the Titan-X platform is some foundational work we conducted on computational modeling of immune responses under a program called Modeling Immunity for Biodefense, which I led. Actually, it dates back to 2010. It’s a $12 million NIH-funded program that was designed to build large-scale computational and mathematical models of the immune system driven by HPC. And that’s, if you think about it from a chronology perspective, that’s before AI was popular, that was before there was this influx of capital in AI-based methods and scalable, but we are building centers, we are building our next-generation HPC systems, we are refining our AI algorithms and building models. And all that, even though it was applied to infectious disease, was really an inspiration when I started my entrepreneurial ventures to basically optimize and apply to finding solutions for very widespread and devastating diseases with unmet clinical needs.

Lee: So, it grew into something that you probably didn’t foresee it becoming.

Josep: In 2010, there were four nationwide centers of Modeling Immunity for Biodefense. I was successful in competing to secure one. When I started that program in 2010, I did not anticipate that the foundational work we did on the computational data science immunology, whether it’s preclinical or clinical immunology, would later on be a part of a platform for precision medicine. It was a cascade of events that, in a serendipitous kind of way, led to, you know, in 2017, when I founded Landos to thinking that, well, maybe we can develop drugs at a faster pace by utilizing leveraging computational models, utilizing advanced AI algorithms. And I think that there are some use cases that show that we could do that. I think omilancor, it’s a prime example, and X13, which was the asset, the earlier stage asset acquired by Abbvie, now called Abbvie113, is another case, both of them in the clinic. And then NIM-1324, a third clinical asset, in this case for lupus and RA, that have leveraged the AI capabilities, the computational modeling capabilities, at different stages of their development.

Lee: I’m curious, was Abbvie, obviously they must have been excited when they saw what you had. Did they have comments about AI and what they were doing? I assume you were probably further ahead, further along than they were.

Josep: Yeah. So look, Abbvie has had an interest in a range of our assets over the years, had conversations about omilancor, had conversations about NX-13 and NIM-1324. And it turns out that, you know, the final conversation that led to their acquisition happened after we had orchestrated an asset split transaction. As you know, a lot of these assets originated in Landos. In 2023, we designed an asset split transaction, which basically moved the most advanced assets, including omilancor, NIM-1324, into Nimmune. And it really allowed the creation of Nimmune. And then the earlier stage assets that remained behind in Landos, including NX-13, or primarily NX-13, were, you know, the focus of attention of Abbvie. You might be aware that Abbvie has acquired several early stage assets, and they are running clinical programs in IBD. And so NX-13 fit very well at that stage, in that strategy for Abbvie. And I think it’s in great hands. They have tremendous commercial capabilities, impressive clinical operations capabilities. I saw that they are moving to phase IIb. And I’m very happy to see that one of the assets that we developed is advancing under Abbvie’s umbrella. And I’m also very happy that omilancor is continuing to advance. Obviously, omilancor, it’s more advanced than NX-13. Phase III clinical development, oral, once daily, novel mechanism of action, targeting a very unique pathway, the LNCL2 pathway. And there’s elements of Titan-X contributing to the development of omilancor, as well as NX-13.

Lee: I’m curious about whether you’ve changed your business model strategy or anything you’re doing with AI at Nimmune based on your experience at Landos.

Josep: That’s a great question. And I would say the short answer is yes. In fact, even within Landos, the initial strategy that we implemented or that the vision that we had set in motion in 2017, when I founded the company, the company that I founded, I led, I built from the ground up, there were changes that we had to implement in 2017. There were changes that we had to implement in 2018, 2019, 2020. So change has been part of our DNA. If you are an entrepreneur, you always adapt to external pressures, make the modifications necessary to continue to, in our case, address the unmet clinical needs of patients with inflammatory and autoimmune diseases. And so this extends not only in my tenure as chairman, president and CEO of Landos, but also evidently in my role as chairman, president and CEO of Nimmune. And all the experience gained with those changes in Landos has contributed to shaping the strategy in Nimmune. Now, one important consideration when you adapt, you need to be extremely cautious on when you implement change and when there are certain occasions in which you need to count until 10 before you implement change. And those occasions, I would say, would be the cases in which shareholder value could be affected, right? So if the company, if the leadership, if the board is going to implement a change that has the potential to negatively affect shareholder value, then perhaps in that occasion, you do not implement that change. But change, other than that, it’s always about changing. It’s about improving. It’s about learning about previous mistakes, learning about feedback from advisors and so on. But what you don’t want to do is make a change that could be a fatal flaw, a change that could result in an implosion in market cap, a change that could result in losing shareholder value.

Lee: You believe that the overall impact of advanced computational methods and biopharma has largely been underreported. And I was curious, what makes you think that?

Josep: Again, the short answer is yes. And it has to do with some probably trade secret or proprietary algorithms. The fact of the matter is that all big pharma are developing their advanced computational shops. In fact, some of them have had those in place for a while. Some of them have strengthened them as a result of the popularity of AI lately. Whether they claim that drugs in their pipelines are the result of AI or other advanced computational methods or not, the reality is that each of the drugs in the pipeline will have gone at some stage in the R&D process through certain computational methods that will have made the tasks of those R&D teams more efficient, that will have accelerated it from target identification, maybe identifying the binding site for developing new drugs against a novel target, to identifying signatures that are predictive of response to treatment, to biomarkers. In some cases, I’m sure you are aware, there are simulations of clinical trials where you run what-if scenarios. Like if you choose this population, this patient population, then there’s a probability of success of X. If you choose this other population that has slightly different characteristics, then the probability of success is Y. And so I’m sure that a broad range of industry partners or competitors, friends, are utilizing advanced computational methods. Whether there’s a complete integration of those methods in developing a drug or a pipeline of drugs, whether it is disclosed or not, that’s another matter. But I know that those methods are being utilized. And then in some cases, if, you know, let’s say big pharma, if a big pharma company has a very unique computational capability, they have two options. They can patent it and the patent has certain patent life. So once that patent life has expired, it’s OK, goodbye to that protection. Or you can use some of that as a trade secret. And so if you are good at maintaining that as a trade secret, then you have basically indefinite protection of that unique capability that makes you more capable, more efficient, faster than your competitor. So I think that some of that is playing a role. But the trends are clear. And whether they disclose it or not, there’s a significant use of these methods in drug development.

Lee: Also, being clear, it’s interesting, your thoughts about AI. You believe, like so many do, that AI is there to make drug discovery cheaper, faster, better. And you said, though, that AI is not a replacement for human ingenuity. I’m curious what your experience is with that and why you believe that so strongly.

Josep: Yes, I would say that AI is a tool. It’s a very powerful adjunct. AI in combination with high-performance computing data centers can make a difference, a very important difference in the development path of a broad range of drugs. But those AI methods, without having a connection to the human minds, leveraging the ingenuity of the human minds, they are not going to deliver on what the expectations are. And so, for instance, an example, and that goes back to 2010 when I was leading modeling immunity for biodefense, a key differentiating feature of that center, which was then replicated in Landos and is part of the pedigree of Nimmune, and that’s a key comparison vs others, is that we were very deliberate in validating all key predictions from our computational models and AI systems, experimental. In other words, you have a computational model, you have an AI algorithm, you use a lot of data to get to a prediction, but then you don’t stop there when you get the prediction. You just don’t go and implement it and make a modification in the drug. What you do is you go to the lab or you go to the clinic and you validate, okay, is this prediction true? And if it turns out that you have that experimental validation and that requires the minds of scientists, clinicians, and so on, then you come up with, okay, the model is providing the right answers with regard to this specific question, which doesn’t mean that the model will be right about other questions. And this system becomes an iterative cycle from model creation, calibration, refinement, and validation, where human ingenuity synergizes with our AI and HPC capabilities, high-performance computing capabilities, to make important discoveries. And this concept is a fundamental tenet, a key pillar, if you wish, of how Titan-X operates in the context of the broader preclinical and clinical development efforts that we are leading at Nimmune, the NIMML Institute and biotherapeutics.

Lee: Lynn, did you have any questions you’d like to ask?

Lynn: I do have one. Looking ahead at AI and the use in drug development, it seems like very separate functions at this point. You have a technology staff running, creating large language models and the AI, and then you have the scientists. Do you see this evolving into one sort of position in the future as AI matures, or are they still very separate? Just look ahead for us a little bit and tell us what you see on the horizon in terms of sort of melding the technology of AI with drug development.

Josep: I think integration is the key. So for AI, you need computer scientists, data scientists, statisticians, you need software engineers. And then for the drug development, we are developing autoimmune disease, inflammatory disease drugs, and you need immunologists, inflammation experts, clinicians, preclinical experts, PK, toxicology experts. The key is to have these people in the same room and interacting. And I find that project managers are a great way to bring together these people that have had very different training, a very different vision of how the world looks like. But in the end, they have to work together to solve the same problem. And I started to see the need for integration in 2010 when I was leading the modeling immunity for biodefense program, where we had software engineers, immunologists, mathematicians, and then HPC experts working together with different languages. And I realized the value of transdisciplinary research. A lot of people talk about interdisciplinary, bringing disciplines together, but transdisciplinary is, I think, the key to integration. It’s like bringing together teams where the vision and the mission transcends the individual visions of what their trainings were. In other words, an immunologist will stop thinking just about immunology, but will think about immunology and the implications in the mathematical model. And the computer scientist is going to stop thinking just about computer science, zeros and ones and programming, and will think about how to better understand the immune system. And if you accomplish that marriage of disciplines, then you can accomplish quite a lot. You can accelerate drug development. You can identify novel biomarkers of response treatment. You can create a pipeline of 17 drugs in a very short period of time. And to be honest, that’s how the big projects of humankind have been managed. I can talk about the Manhattan Project. It had to bring together from physicists to mathematicians to engineers to be able to deliver on the goal. And they were able to deliver on that goal successfully.

Lynn: Super exciting and interesting. And we’ll be tracking the evolution of transdisciplinary approach with drug development and AI. Josep, thank you for your time today. We enjoyed having you, your time and your insight.

Josep: Thank you very much. It was a pleasure.

Lynn: That’s our show for today. As always, BioWorld will continue to keep you informed of all the most important scientific, clinical, regulatory and business updates. We’re a daily news service covering the development of the most important human therapeutics designed to improve the human condition. If you need to track the development of drugs, turn to BioWorld.com. You can follow us on LinkedIn or X. And if you would like to share news with us, drop us an email at NewsDesk@BioWorld.com. Also, if you’re enjoying this podcast, don’t forget to subscribe via your favorite platform. Thanks for joining us today.

VOICEOVER: BioWorld, published by Clarivate, is a subscription-based news service that delivers actionable intelligence on the most innovative therapeutics and medical technologies in development.