IRB Barcelona uses generative AI to design cell-selective molecules

A new IRB Barcelona study points to a more phenotypic style of AI drug discovery. EurekAlert reported that Patrick Aloy's Structural Bioinformatics and Network Biology lab developed a computational framework that designs molecules based on the biological effect researchers want in specific cell types, rather than starting with a known protein target. The team tested more than 11,000 chemical compounds across eight cell models, including six pancreatic-cancer lines and two controls, then used the data to train predictive models and guide a generative AI system. Several AI-designed molecules experimentally matched the intended selective activity, and the work was published in Communications Chemistry. This is early-stage compound discovery, not a validated therapy. Its significance is methodological: AI is being used to propose structurally novel molecules around desired cellular behavior when the target biology is poorly characterized.

Key details: June 2, 2026, IRB Barcelona, Communications Chemistry, Patrick Aloy, 11,000+ chemical compounds, Eight cell models, Six pancreatic-cancer lines, Phenotypic discovery.

Continue swiping for more AI Brief stories.