Agentic cryo-EM structure determination with CryoPilot

CryoPilot improved cryo-EM reconstruction resolution for an alpha latrocrustotoxin dataset from 5.9 to 4.15 Å by changing the particle picking strategy it suggested.

• Why it matters: Cryo-EM structure determination still requires experts to inspect intermediates, choose processing routes, and decide when to revise earlier steps, which limits automation and traceability.
• What they did: CryoPilot uses two language-model agents: a reconstruction agent evaluates intermediate images and diagnostics, selects and parameterizes CryoSPARC jobs, and compares alternative strategies; a modeling agent determines handedness and builds and refines atomic models with CryoZeta.
• The result: Across nine EMPIAR datasets, eight reconstructions reached 2.61 to 3.51 Å resolution, and refined models from seven maps matched reference structures within 0.4 to 4.9 Å Cα RMSD.

AI summary of the abstract · may contain errors · not medical advice · read the paper.