AI Co-Scientist Accelerates Discovery of Liver Disease Treatment Pathways

Biomedical research generates more information than any individual scientist can feasibly process. To tackle this challenge, bioengineer Filippo Menolascina at the University of Edinburgh turned to Co-Scientist, an AI system developed by Google DeepMind designed to sift through scientific literature and surface novel hypotheses.

Menolascina's team focused on metabolic dysfunction-associated steatohepatitis (MASH), a common liver disease that has proven difficult to treat. The disease involves multiple intertwined biological processes—including liver inflammation and metabolism—making single-target drugs ineffective. This has pushed the field toward exploring combination therapies, but the number of possible drug pairings is staggering.

Finding Signal in the Noise#

Using Co-Scientist, Menolascina's team synthesized evidence across liver biology and pharmacology, narrowed the search space, and identified candidate combination treatments worth testing. The system highlighted mechanisms that deserved deeper investigation and prioritized the most promising therapeutic approaches.

In one striking case, the AI tackled a pressing clinical question: Why does resmetirom—a recently approved drug prescribed for specific stages of MASH—help only a small fraction of eligible patients?

Co-Scientist generated a hypothesis identifying the NLRP3 inflammasome as a molecular bridge coupling inflammation and metabolism in the disease. This connection had never been synthesized into a single, coherent explanation before. The team subsequently verified the hypothesis experimentally, revealing a pathway that could guide the development of targeted dual-therapies.

The work demonstrates how AI can accelerate biomedical discovery by connecting disparate threads across the scientific literature and converting them into testable hypotheses—speeding up the journey from hypothesis to intervention.