Using AI to design small molecule drugs is tough. Here are five companies trying to crack the code

Anne Thompson

akstanwyck

More stories by Anne

‘Nosferatu’: How to Make a Robert Eggers Movie, with Help from Mel Brooks and Chris Columbus

In the Photorealistic

September 5 Director Tim Fehlbaum Didn

Time：2024-12-03 09:12

Photo Illustration: Christine Kao/STAT; Photos: Iambic Therapeutics, Isomorphic Labs, Genesis Therapeutics, Terray Therapeutics

By Brittany Trang

Dec. 3, 2024

Health Tech Reporter

Ian Quigley was employee #13 at AI drug company Recursion until 2021, when he quit and turned the basement of his Utah home into a lab to develop AI models for predicting which proteins small molecule drugs might target — a tool he and co-founder Andrew Blevins found was missing during their time at Recursion.

When he was setting up the lab, Quigley ordered an Illumina DNA sequencer that the delivery person had wanted to leave on his porch. Freaked out by that idea, he instead picked it up in his Toyota Yaris, folding down the back seats to fit it next to his dog’s kennel.

In that basement lab, Quigley and his small team at Leash Bio produced about 133 million data points of small molecules binding to protein targets. Earlier this year, they released that data as training fodder in a competition aimed at determining whether any AI models for predicting small molecule interactions with proteins were really learning anything about biology, or just memorizing what they had seen before.

STAT+ Exclusive Story