A generative model called Cadense produces antibody Fv conformational ensembles in seconds, standing in for days of molecular dynamics simulation
- Cadense is a generative model that samples varied antibody Fv structures from sequence alone, trained to reflect motion seen in molecular dynamics.
- Generated ensembles are compared against MD-derived conformational and biophysical property distributions, with generation taking seconds rather than days.
- The ensembles capture both unbound and antigen-bound states of the same antibody and work as input for physics-based docking; evidence is computational, not experimental.
💡 Points Based on the abstract
If the training data is MD, matching MD distributions is a low bar. The harder question is whether Cadense recovers CDR-H3 states that long simulations rarely visit, and how it handles antibodies far from its training set. With no experimental check reported, try it first on a sequence-held-out set of paired apo/holo structures before trusting ensemble docking.
Rapid generation of antibody conformational ensembles using machine learning ↗