gcorso/DiffDock
DiffDock uses score-based diffusion models to predict how small molecules bind to proteins, achieving state-of-the-art molecular docking performance.
Velocity · 7d
+1.1
★ / day
Trend
→steady
star history
DiffDock applies diffusion model techniques to molecular docking, predicting binding conformations between proteins and small molecules. It leverages score-based generative models operating on non-Euclidean geometric representations of molecular structures. The 2024 release (DiffDock-L) improves performance and generalization through enhanced model architecture, and is available via Docker and a HuggingFace web interface.