AF2RAVE protocol without prior system-specific information

Develop an AF2RAVE-based drug discovery protocol that identifies ligand-binding metastable states and selects structures for docking without requiring any prior system-specific information, such as residue pairwise distance cutoffs for activation loop position or Dunbrack DFG-type definitions.

Background

The AF2RAVE workflow, as demonstrated, leverages prior information (e.g., distance cutoffs and DFG-type definitions) to learn latent coordinates and enrich holo-like metastable structures before docking. For broader applicability to general protein targets (including GPCRs), the authors point to the need for removing such priors.

A fully prior-free protocol would make AF2RAVE-Glide more widely deployable across targets where reliable domain-specific annotations are unavailable.