Companies designing drugs with AI could shave years, millions off development

IBM’s Watson might not be your new oncologist, but that doesn't mean artificial intelligence (AI) is in decline. In fact, its role is growing rapidly in healthcare. Machine learning—a form of artificial intelligence that uses algorithms and data to adapt and predict outcomes—is taking off not only within hospitals, but also in the pharmaceutical industry.

The industry is increasingly turning to machine learning to discover and develop new medications with the hope of expediting the yearslong slog to get new drugs to market—and saving substantial cash. An Insider Intelligence report found AI could decrease the cost of discovering a new drug by as much as 70%, where it currently takes about $2.8 billion and 12+ years, on average, to bring one drug to market.

That promise has attracted growing investment in early-stage startups employing machine learning to discover and develop new drugs, said Jonathan Norris, managing director for business at Silicon Valley Bank’s healthcare unit. He put the amount plowed into seed and Series A funding for companies in this niche at up to $1 billion in the first half of 2022 alone, according to SVB research. That’s slightly under the pace set in 2021, when $2.4 billion was invested for the calendar year, but Norris said it’s still multiples of the $600 to $700 million invested in 2020—and he doesn’t expect the investment pace to slow.

Exscientia, an AI-driven pharmatech company that went public in October 2021 after raising about $500 million, was the first to have an AI-designed drug in human clinical trials. Its immuno-oncology medicine is designed to treat certain types of solid brain tumors. The company now has three drugs in human clinical trials, according to Exscientia’s chief financial officer, Ben Taylor, including one drug that would treat obsessive-compulsive disorder and another for Alzheimer’s disease psychosis.

The focus with that first cancer drug and others is not only to take years off the development process, but to also use machine learning to try out more chemical compounds with the aim of finding more-effective drugs with fewer side effects—all before human trials, Taylor said. The lightning speed of AI allows companies to test not just thousands or even hundreds of thousands of compounds, but billions, he said.

“The estimated number of different possible chemical compounds is 10 to the 60th [power]. That’s effectively infinite,” Taylor said.

Navigating that vast landscape, AI can be used to generate compounds virtually and test them with different models designed to predict how they might work in people before the clinical trial stage, he said.

To date, no AI-designed drug has yet been approved by the FDA (though a bunch of AI-designed devices have gotten the OK). So far, the furthest any AI-designed drug has gone is in human trials. But investors, like the tech-focused private equity firm SoftBank Investment Advisers, which has invested in Exscientia, seem to feel as if there’s at least enough early promise to warrant making their play.

“Whether AI can truly design a drug…from the beginning to the end, and how quickly, how efficiently it can do that—there’s some early signs, but the end result is yet to be seen,” said Joanne Xu, a partner at SoftBank based in Hong Kong. “But we definitely have seen some very good progress made by Exscientia and [other] companies alike.”