Recursion’s Khan: AI will be judged by the medicines it makes

AI’s clearest near-term gains, and its next frontier in drug development: The BioCentury Show

As AI becomes embedded across biopharma, the label of techbio may lose some of its meaning. But even now, having the shiniest new tools matters less than whether a company can use them to create better medicines, says Recursion CEO Najat Khan.

“It’s not going to be how you use AI. It’s going to be what value you’ve created using AI,”  Khan told The BioCentury Show. 

What will set companies apart, Khan argued, is not simply having AI models or large datasets, but deploying those tools selectively across drug R&D — always with a clear line of sight to the medicine that needs to be created.

Khan, president and CEO of  Recursion Pharmaceuticals Inc. (NASDAQ:RXRX), took the helm in November, at a time when public techbio companies are under growing pressure to move beyond platform promise and deliver clinical proof points, even as private peers still have more room to be valued on technology potential.

Generating those proof points is among her top priorities for Recursion, including through programs aimed at known but “unconquered” targets where AI may help solve therapeutic-index challenges.

More broadly, she sees AI’s most measurable early advantages emerging in chemistry and trial execution, where gains in compound optimization and patient recruitment can be quantified more readily, while biology remains the larger but slower-burning prize. The next frontier, she thinks, is “out-of-domain” prediction — the ability of models to make useful inferences in less-charted biological terrain, reducing the amount of experimentation needed to open new areas of discovery.

Khan also made the case for precompetitive collaboration, pointing to a recently launched effort around ADMET data, where fragmented datasets and privacy concerns have made it difficult for any one company to build the strongest possible models alone.

Late last year, Recursion posted positive Phase Ib/II data for REC-4881, a MEK1/2 inhibitor in familial adenomatous polyposis (FAP). Khan cited the program as an example of using different parts of the platform where they matter most: in REC-4881, she said, AI helped identify the biology, while in other programs the value may come from improving therapeutic index.

Recursion is building its broader validation in part through partnerships. The company has generated more than $500 million in upfront and progress-based milestone payments to date. With Roche (SIX:ROG; OTCQB:RHHBY) and Genentech, it has delivered six AI-derived maps of biology, including two in neuroscience; with Sanofi (Euronext:SAN; NASDAQ:SNY), it recently reached a fifth program milestone, helping build a joint portfolio of five novel AI-driven small molecules in immunology and oncology.