Rapid generation of antibody conformational ensembles using machine learning

Cadense generates diverse antibody Fv conformational ensembles from sequence in seconds, matching MD-informed conformational and biophysical property distributions for both apo and holo states.

• Why it matters: Static antibody structure prediction misses dynamic motions that affect binding and other properties, especially in flexible CDR loops that are hard to model.
• What they did: Cadense is a generative, dynamics-informed framework trained to produce ensembles whose conformational and biophysical property distributions match those from molecular dynamics, reducing generation time from days to seconds.
• The result: The generated ensembles recover apo and holo conformations of the same antibody and improve physics-based docking by capturing flexibility needed for antibody–antigen interaction modeling.

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