罗氏推进与Recursion合作的首个AI发现神经科学靶点

Roche Holding AG advanced the first project from its AI drug-discovery partnership with Recursion Pharmaceuticals, marking an early test of whether artificial intelligence can uncover new medicines.

The drug target, aimed at treating an undisclosed neurodegenerative disease, emerged from AI analysis of Recursion’s proprietary datasets, and was then validated in extensive laboratory experiments by the two partners. It’s the first of as many as 40 potential programs under the collaboration, said Recursion Chief Executive Officer Najat Khan.

“This is the first example of taking proprietary large-scale data sets and converting that into an unexplored neuroscience target,” Khan said in an interview. “Once you have the dataset, you can mine it again and again and again.”

Developing drugs for diseases such as Alzheimer’s and Parkinson’s has proved notoriously difficult because scientists still don’t fully understand what causes them. Roche and Recursion aim to use AI to identify new biological targets that traditional research has missed.

Doing so requires a different approach, according to Khan. Where many AI drug-discovery efforts seek to optimize compounds against known biological targets, the collaboration aims to use AI to discover entirely new ones. Recursion grows large numbers of human brain cells, uses AI to search them for hidden patterns linked to disease, and then checks the computer’s predictions with laboratory experiments.

“In neurodegeneration neuroscience, everybody is working on the same 30, 40 targets,” said Khan. “If you want to go broader than that, that requires a fundamentally different way of doing biology.”

Search for Patterns

Basel-based Roche and its US biotechnology unit, Genentech, partnered with Recursion in 2021 to use AI to identify and validate new drug targets in neuroscience and cancer. The companies jointly develop successful discoveries, while Recursion, based in Salt Lake City, can earn more than $300 million in milestone payments plus royalties for each program that reaches the market.

Artificial intelligence is already generating more experimental medicines at Roche. The Swiss drugmaker has seen “a significant increase in new molecules” in its early-stage pipeline as AI helps speed drug development, improve the probability of success and increase research output, Chief Executive Officer Thomas Schinecker said at the half-year results in July. Roche’s pipeline currently includes 43 Phase I drug candidates.

The Recursion option, which gives Roche the right to co-develop the newly identified target as a potential medicine, follows several years of work building what Recursion calls “maps of biology“ from more than one trillion lab-grown human neurons whose genes were systematically switched off one by one.

Rather than supporting a single project, the companies say those datasets can be mined repeatedly to generate additional drug targets, making this the first of what they hope will be many programs.

“We’ve built a map,” said Christopher Winrow, Recursion’s vice president of neuroscience. Unlike traditional drug discovery, where “most of that data gets thrown away,” Recursion “can continue to go back to that map” to search for additional drug targets, he said.