advertisement
advertisement

This pharma exec is using AI differently–and spotting effective treatments others miss

This pharma exec is using AI differently–and spotting effective treatments others miss
[Illustration: Artur Tenczyński]


Most AI-driven biopharma companies parse available data to find drug molecules that target single proteins believed to underlie a disease. This yields a lot of leads, but few pan out. Recursion Pharmaceuticals generates its own data by running 100,000 mini-experiments weekly, robotically dosing samples of “sick” cells with an array of treatments. Recursion then takes a microscopic snapshot of the reaction and uses machine-learning software trained to spot the sick and healthy cells to sort out what works. “It lets you operate without bias,” says CEO Chris Gibson. “The treatments aren’t what a human would have predicted.” Because scanning an image file is faster and cheaper than chemical-based tests, Recursion can collect exponentially more data per hour than a traditional lab. It has compiled one of the largest global databases of cell images, which it uses to improve its algorithms. The company’s approach was validated when, in 2018, it received FDA clearance to begin Phase I clinical trials for a compound it had identified to target cerebral cavernous malformation, which affects more people than cystic fibrosis. Ten other drug candidates are in preclinical studies, and Recursion is working on additional treatments with pharma partners like Sanofi and Takeda. The company raised $84 million in funding last year.

advertisement
advertisement