Validated skills are necessary but not sufficient for trustworthy agentic genomics

​Join Manuel Corpas online for a Crick AI Club talk on scientific trust in agentic genomics.

​AI agents can find datasets, run software, call tools and reason over results. Their outputs can look convincing even when they contain errors, use the wrong method or cannot be reproduced.

​Using pharmacogenomics as a test case, this talk examines how ClawBio executable skills support deterministic, auditable analysis, and why those properties alone do not guarantee accuracy. It asks what researchers must check before trusting an agent's scientific output.

​The session will be relevant to genomic scientists, bioinformaticians, AI researchers and developers building agents for scientific work.

​Speaker: Dr Manuel Corpas
Host: Crick AI Club, The Francis Crick Institute
When: Monday 14 September 2026, 17:30 to 19:00 BST (London).

​Attend online via Zoom. Register here to receive the webinar joining link. This registration is for online attendance only.

​Thank you to Fabian Frohlich and the Crick AI Club team for the invitation.