Moving the dialogue forward: The need for better AI integration in drug discovery
There has never been more discussion about the usefulness, and proper use of, artificial intelligence (AI) in drug discovery. When Aria first started utilizing AI as a tool in the early drug discovery process, much of that conversation centered on AI as a panacea that will revolutionize healthcare. Not a useful conversation and not a practical use of AI.
Today, I do see that sentiment changing, and for the better. Last week, Endpoints, a preeminent pharmaceutical research publication, hosted a panel discussion about AI in drug discovery, aptly titled “AI faces its moment of truth.”
Two things struck me about the discussion. First is that the conversation around AI has become more pragmatic. There was a clear recognition that AI has become an important part of the R&D equation, and more education is needed around its limits and abilities. Don’t get me wrong, there is plenty of overpromise, but maybe we’re through the disillusionment phase of the hype cycle. I was pleasantly surprised with the overall tone.
My second observation was that better integration of AI into the pharmaceutical R&D process is still much needed. While the panel didn’t include traditional scientific R&D experts, the value of working alongside those with expertise in drug development was apparent. I have consistently advocated for close partnerships between computer scientists and biology and medical scientists, and it’s a significant gap that I have observed for some years now. The next phase of this discussion is evolving to focus around integration.
But what is integration?
I think it’s helpful to revisit the past decade of AI’s application in drug discovery. While AI has been around for years and has been used in drug discovery for some time, today we have access to much greater computational power making the capabilities of AI that much greater in drug discovery. As a result, I have seen hundreds of startups venture into the space and massive amounts of funding thrown at them. However, those investments have been made almost solely on the promise of technology and software creation. In a way, it’s a disservice because it has decoupled AI from the pharmaceutical industry. In other words, most AI drug discovery companies are squarely focused on selling their technology, and not fully committed to applying AI to the ecosystem that is research and development. As a result, there is little differentiation between any of them.
At Aria, we’re working on a couple of things that I think are important and add to this dialogue: cross-discipline collaboration and data-based results.
AI will undoubtedly become an integrated and natural part of the drug discovery process, not replacing traditional science but rather supporting it. It’s just a matter of time, and I’ve written before about the fact that AI does not drive drug discovery, science does. AI should be used as a component that supports biologists, pharmacologists, clinicians, and others vested in finding breakthroughs. For AI to take hold more broadly, there needs to be an equal seat at the table for science and technology.
At Aria we have invested in top pharmaceutical R&D talent to help guide our science. Most importantly, we’ve established a culture where our cross-disciplinary teams are fully integrated, and everyone has ideas to share. We’re fortunate because we built our company from the ground up to be fully integrated, and have set the tone for this style of collaboration. Other companies may need to revisit their culture and adapt accordingly.
In terms of results, we have made our benchmark clear: Positive safety and efficacy data within our pipeline. That’s what counts, and it is the benchmark for drug discovery. There’s no other success factor in early research. Too much of the conversation today places results on a means to discovering drugs — proprietary data sets, algorithms, target identification and so on. Those are just means to an end. The end is the result and that’s a pipeline with scientifically validated data.
Everyone agrees that getting medicines to patients is what matters. At Aria, we have purposefully focused on examining multiple candidates and narrowing those down to the ones we believe will have the best chance at becoming a treatment. We can achieve a 30x hit rate at in vivo milestones compared to traditional research by interrogating many different mechanisms of action against different disease targets at the same time using our technology. We can take multiple shots on goal. We will look at thousands of molecules against our data sets and narrow that down to around 10 different candidates to take into in vivo testing. We can do this in a matter of months versus years by applying technology to make things more efficient. However, perhaps the most important part is that our end goal is not to simply dissect biology for a better understanding, it’s to maximize our chances of finding a novel, safe and efficacious treatment.
The promising news is that AI is positively influencing many recent drug candidates that are progressing to the clinic. We are just one example with a pipeline across 18 therapeutic areas with several candidates passing early preclinical safety and efficacy milestones. I think as we see more and more drugs discovered with the help of AI achieve preclinical and clinical milestones, the industry will pay much closer attention to how AI can and should support scientists.
I still believe that AI alone is not the hero that will address the pharmaceutical industry’s productivity decline. It’s a tool that can help find efficiencies in parts of the process, most promising seems to be drug discovery. In any industry, hype only lasts so long. I have always held the same beliefs about AI in drug discovery and have delivered a consistent message over the years. Not only because I advocate for these fundamental principles, but I also truly believe that when the industry realizes the potential of integration, patients will be better served.
My hope is that we are out of the hype cycle and can truly begin a productive conversation where the pharmaceutical industry is better prepared to integrate AI into the drug discovery process. That’s the next and needed evolution.