How generative AI is shaping drug discovery

Generative AI is making waves in drug discovery.

Drug companies are increasingly interested in how AI could potentially help speed up the drug discovery process, which can take around a decade and an average of $2.7 billion, the Financial Times reported. In Q1 2024, the global pharmaceutical industry saw a 45% increase in how often AI was mentioned in company filings from Q4 2023, according to Pharmaceutical Technology.

This time and money savings has been a “holy grail” for decades, said Anthony Gotzis, senior managing director at consulting firm FTI consulting, who told Healthcare Brew that “when AI can start to provide real intelligence around that discovery, trial and error—that’s going to be a huge shot in the arm.”

Zoom out. Drug discovery can be broken down into three steps, according to Alex Snyder, head of research and development at biotherapeutics company Generate: Biomedicines. Researchers identify a “target,” such as a disease or virus; they build a molecular structure for a drug; then they conduct clinical trials to see if the drug works on the target.

While automation has long been used for data collection and spotting data patterns, Snyder said generative AI is now making suggestions for new drugs in a fraction of the time.

Her team used the AI tech on Generate’s platform while developing a drug intended to prevent Covid-19 from spreading in the body. According to Snyder, AI was able to create a drug that binds to an area of the virus that researchers hadn’t been able to identify previously. While many drugs can typically take three to five years to move from idea to first dose, she told Healthcare Brew, AI condensed the time to 1.5 years. (Generate has not yet published its research for this development process.)

What about the risks? Snyder stressed the importance of providing the highest-quality data that AI systems analyze.

AI is already being utilized to speed up pharmaceutical clinical trials, which can move slowly due to the challenge of patient recruitment, according to Gotzis.

“If we can reduce our reliance on human beings in the clinical trial process and…for AI basically to simulate those humans—simulating in a sense of their [genetic] makeup and how a particular drug will interact with them—then it’s a huge win,” Gotzis said.

When researchers test drugs, especially for rare diseases, he added, there’s a “very limited global population of individuals for a terrible disease,” so it’s a “big game changer” if AI can simulate it instead.

Future forecast. Snyder said generative AI has changed how she works. “I have ChatGPT open at all times. When I’m like, ‘Oh, remind me what this paper was,’ or ‘Write me a quick summary of the following bullets,’” she said. “We use it fundamentally for drug discovery, but I also just use it as a tool as a day-to-day support.”

Regulation is the biggest hurdle AI will face in this industry, Gotzis said, because drug development is a “highly regulated space,” and “just because it’s feasible, doesn’t mean you can do it.” If generative AI were to impact this industry in the next five to 10 years, he added, that would be quick for researchers and scientists.

“It’s going to take awhile, I think, for the health authorities—FDA being the largest but not the only one—out there around the globe to really feel comfortable that this is doing what it’s intended to do,” he said.