Artificial intelligence won’t replace people in biopharma, but it is infiltrating every step of drug development, including in some ways that aren’t so obvious.
AI is revolutionizing the drug development process, from finding new targets and designing novel drugs to streamlining clinical trials, regulatory filings and manufacturing. Beyond these direct applications, however, the technology may even have a role in how researchers work together, Sanofi’s Matt Truppo told BioSpace.
Think of the old adage about two scientists bumping into each other in the middle of the lab and spontaneously coming up with an idea for a new drug in the midst of conversation. In an effort to engineer such creativity, Sanofi has designed its workspaces to provide “really cool lab spaces” with common areas to encourage people to chat over the water cooler, said Truppo, global head of research platforms and computational R&D at the pharma company.
“You never want to underestimate serendipity. That’s absolutely a thing in discovery,” Truppo told BioSpace.
But now, he added, AI is starting to play a role—even at the level of creative insights that historically have come from these sorts of human interactions. “[AI is] uniquely good at making connections between diverse data sets that maybe you haven’t found before. That’s serendipity. What I like to refer to it as is accelerated serendipity.”
AI can’t do everything, and it certainly still needs a human touch to guide it through the drug development process, biopharma leaders agree. But with people at the helm, the technology is accelerating human creativity, chipping away at challenges and making the industry more efficient and faster.
“The idea that AI is going to solve all of it, and it’s going to change the probability of success from mid- to mid-single digits to perfect, I think is an unrealistic expectation,” Mike Nally, CEO of Flagship Pioneering’s Generate:Biomedicines, told BioSpace. “But if we could make small changes to the probability of success, the cascading impact on patients and humanity is extraordinary, and I think certainly these technologies are poised to do that.”
‘Spoiled for Choice’
One reason AI is so invaluable is that humans simply can’t understand the whole of how life works, Nally argued. “The complexities of biology are extraordinarily humbling and anyone that’s been in this industry recognizes that from the outset,” he explained. “Using these tools to understand it at a more deep level than the human mind alone could is extraordinarily powerful.”
Many of the drugs on the market today were randomly discovered or developed through lines of inquiry into the same targets that previously proved beneficial. Now, thanks to AI, biopharma has more targets to interrogate than could ever have been imagined for drug discovery, according to Sara Choi, a partner with Wing Venture Capital. At the same time, pharmas are putting billions into licensing deals in the hopes of discovering even more new targets.
“We’re actually evolving at a speed that’s so fast that we’re forgetting how far we’ve exactly come,” Choi told BioSpace. “We’re in the beginning of a super cycle, and there’s just more and more and more innovation that’s to come.”
Many biotechs, like Nally’s Generate, say they have more than they could ever deal with. This isn’t a bad thing, Truppo noted, but it does make portfolio prioritization ever more important: “To be spoiled for choice is rare in our industry.”
And where AI helps generate such a surplus of options, it can also help with that prioritization, Truppo continued. But humans can still bring crucial context to the table, such as examining the potential market, patient need, competition, payer dynamics and how the drug is administered—calculations that Sanofi is considering earlier and earlier in the drug development process, according to Truppo.
But, Nally warned, “AI is not a panacea. If you pick bad biology, go after it with an amazing molecule that is AI designed, you’re going to get a bad answer.”
The key, he said, is integrating AI into what have always been human processes. “What we’re in is an era where we have to bring the best of traditional drug discovery together with these extraordinary technologies,” Nally said. “We need people who understand the nuances of the regulatory environment and of clinical development.”
From Discovery to Market
Perhaps one of the most important things to understand is safety. Even known targets that have spawned approved drugs we use today have safety issues that an AI may have been able to scan for—if only it existed at the time, Choi said. Determining which targets are predisposed to work in the human body has been science fiction until now. “Our industry hasn’t done that before.”
Now, pharma is starting to incorporate AI into protocols to scan for off-target effects when designing a drug. Sanofi, for example, is using AI to design molecules specifically to avoid potential side effects. On the clinical side, the company is also applying this technology to population selection for trials by scanning patient backgrounds for any possible risk factors.
This is “just a completely different paradigm,” Choi noted, predicting that more safe and effective molecules are headed to the clinic soon.
AI can also replace crucial paperwork as a drug approaches regulators. Sanofi is using generative AI to write early drafts of the clinical study reports that are eventually sent to the FDA for a drug review. This process can take up to 35 weeks of human labor, Truppo said. Sanofi writes about 50 of these reports every single year, but the work is expected to grow as the company’s Phase III trials increase by 50% over the next two years, meaning even more paperwork to come.
With AI, Sanofi has automated about 60% of the reports, which then go to a human being for editing and fine tuning, Truppo said. The new process has saved about 40% of the overall time. The goal is to make it a five- to seven-week process, speeding the time from Phase III readout to when an application gets into the FDA’s hands, he added. Once a molecule has been selected, AI can also help streamline manufacturing processes.
All told, it can cost nearly $3 billion to develop a new drug, according to a 2020 study published in the National Institutes of Health’s library, but other estimates have put that much higher. And the timeline for getting a drug from discovery to approval is typically 12 years or more. Shaving down those numbers by even a fraction of a percent could have a major impact on patients, experts told BioSpace.
Although it would be nearly impossible for a single company, especially a young one, to assemble all pieces of the AI-driven drug development puzzle, Choi said she believes that the technology is already enabling biotechs to develop drugs with less money than the traditional biopharma model.
Truppo noted that the more parts of the process that can be supplemented with AI, the more significant the returns. “With AI tools and insights embedded throughout our entire workflow from beginning to end, perhaps we can take a bite out of those timelines, a bite out of that cost to bring a therapeutic to market.”
