DiffDock-L

Diffusion-based blind protein-ligand docking

DiffDock-L performs blind protein-ligand docking with a diffusion generative model and confidence bootstrapping, sampling ligand poses across the whole protein without pocket definition. It outputs ranked poses with calibrated confidence scores and strong PoseBusters validity. Suited to large-scale virtual screening pose generation.

DiffDock-L is being onboarded — request access and be first in line.