The long road to a future of AI-driven drug development

Pharmaceutical and biotech companies are investing billions today to get a future in which new AI drives drug discovery, trials and development, bring treatment to patients faster, safer and more affordably. Merck struck a deal with Google Cloud to leverage the tech giant’s agentic platform. Novartis, Takeda and Eli Lilly have their own big AI deals for research and drug discovery.  

But investment is just one piece of the puzzle. Pharmaceutical and biotech industry players still have to figure out how the billions being spent will translate into AI capabilities that drive true change to their drug development pipelines.  

“There’s not a single molecule that has been created by AI and has made it into humans yet,”  Rahul Gupta, MD, former director of the White House’s Office of National Drug Control Policy and current president of drug discovery company GATC Health, said.   

DHI spoke to Gupta and Satty Chandrashekhar, healthcare and life sciences AI leader at consulting firm Bain & Company, about how AI is being used in the drug industry today, how it could change the way drugs are discovered and made and the work that lies ahead for enabling that change.   

The hurdles

For AI to become a normal, functional element of drug R&D, a number of things have to happen first. Pharmaceutical and biotech companies need to figure out how to actually deploy and scale AI, and they haven’t gotten there quite yet.   

“They have embraced AI but primarily as disconnected pilots,” Chandrashekhar said. “There is no clear leader in the end-to-end transformation that connects this fragmented system together at scale.”  

Just a small sliver of pharma and medtech companies (5%) have gotten to a point where they could say genAI is a competitive differentiator with financial value, according to a McKinsey & Company report published last year.   

Data readiness is among the challenges pharmaceutical companies face as they integrate AI into their workflows. AI cannot be an effective tool if the underlying data isn’t quality, clean and governed.   

Companies also need defined strategies for the use of AI, not scattered pilots. Three-quarters of companies that responded to the McKinsey survey said they did not have “a comprehensive vision for genAI.”  

Building a comprehensive, scalable strategy for AI is not as simple as layering the technology on top of existing manual workflows, according to Chandrashekhar.   

“It is thinking about the workflow of drug development end-to-end and then simplifying or reimagining the work and then applying AI,” he said. “When companies don’t think of it that way — they think of AI as a tool that just layers on top of how they currently develop drugs — that’s a risk that leads to [suboptimal] outcomes or failed pilots that don’t scale. “  

As companies find ways to extract value from AI, they also have to identify and manage the risks that come with using these tools.   

“Figure out how those guardrails of AI will apply so that when we do have the molecules that are ultimately into humans, we can be relatively confident that they’re at least as…safe and efficacious as the traditional way of doing things,” Gupta said.   

The possibilities

What could AI do for the industry as pharma and biotech companies figure out how to use it effectively? We have some early glimpse into the possibilities.  

For one, the technology offers the tantalizing possibility of discovering novel drugs on a dramatically reduced timeline. Work being done at the U.S. Department of Energy’s Argonne National Laboratory is an example of AI’s powerful potential in this arena.   

Researchers at Argonne are making use of Aurora, an exascale supercomputer, and AI to screen billions of molecules for their potential as cancer inhibitors. In collaboration with the University of Chicago Medicine Comprehensive Cancer Center, researchers at the lab are focusing on “undruggable targets,” proteins that are difficult to treat due to their structure.   

Advances in research are exciting but an early step in the long journey necessary to bring drugs into clinical use.   

“It takes hundreds of millions of dollars and years and years of trial and error from a laboratory standpoint: from theoretical frameworks to lab to animal testing to then get to clinical testing,” Gupta said.  

Even after all of that time and money, 90% of clinical drug development ends up failing. AI has the potential to reduce that failure in a number of ways. For one, pharmaceutical companies can use it to optimize clinical trial site selection.  

“The ability to predict which sites are going to be high enrollers of those patients for that trial and which sites are likely not going to enroll a single patient and then using AI to optimize the enrollment rate per site, the number of patients per site per month…these are very classical predictive model machine learning applications of AI,” Chandrashekhar said.   

AI can also enhance pharmaceutical companies’ in silico testing capabilities.   

“What the AI has done now is be able to understand that human biology, create digital twins and find those…druggable targets and experiment with that in terms of trials with in silico [methods] in a matter of days and weeks instead of years and decades,” Gupta explained.   

The competitive landscape will be influenced by the ability to use AI effectively, whether by building internal programs or acquiring other companies. “For the pharma companies, I think those that adapt AI effectively and early will replace those that don’t. I think that goes the same way for the biotech industry,” Gupta, said.   

AI has the potential to deliver a future where new drugs can be developed and go to market faster, where new treatments for rare diseases are possible, where patients receive more personalized therapies. As the pharmaceutical companies pushed toward those possibilities, the industry could start to look very different.