How one startup is using artificial intelligence to search for new drugs

Gregory Bull / AP

Atomwise, a San Francisco-based biotech startup, is trying to use artificial intelligence to cut the development time for new drugs — and it's launching a new project today to help researchers do that. It's inviting scientists to apply for quick screenings using its AI technology. The company will test millions of molecules for up to 100 labs, find out which ones are most likely to work on a disease or protein, and then ship 72 customized compounds to each lab for testing.

What they're trying to solve: Drug companies often talk about the time and expense it takes to develop and win approval for a new drug — the commonly cited statistics are $2.6 billion and 15 years. Atomwise says it's trying to cut that time by using AI to speed up the beginning of the process. If it works, the time from the idea stage to clinical trials will be "significantly shorter" and "the success of the clinical trials will be greater," said Dr. Han Lim, who's in charge of academic partnerships for the startup.

How it works: The technology can screen millions of potential drug compounds, company officials say, by simulating different proteins associated with a disease and figuring out whether and where those compounds would attach to them — possibly preventing their role in triggering a disease. It will screen 10 million compounds for each lab, so by the end of the project, it will have analyzed up to 1 billion compounds, according to Alexander Levy, the company's co-founder and chief operating officer.

What they get out of it: The scientists will get the compounds for free, at a cost of up to $1 million for Atomwise. The company is paying for it out of its venture capital financing and corporate partnership revenues. They say they're doing it to support a good scientific cause, but there's a financial reward down the road: If their support helps lead to a new drug invention, they'll work with the researchers to patent it and bring it to market.