Predicts 3D protein-ligand binding poses using diffusion-based deep learning for structure-based drug design.
DiffDock is a state-of-the-art computational chemistry skill that enables Claude to perform diffusion-based molecular docking for drug discovery. It allows researchers to predict high-accuracy 3D binding poses for small molecule ligands against protein targets using either PDB files or amino acid sequences. With features for batch virtual screening, confidence scoring, and support for multiple ligand formats like SMILES and SDF, this skill bridges the gap between AI-driven structural biology and practical lead optimization workflows.
Key Features
01Supports protein structures via PDB files or sequences via ESMFold
02Predicts 3D ligand binding poses with high structural accuracy
0381 GitHub stars
04Generates confidence scores to rank and validate prediction reliability
05Processes diverse ligand inputs including SMILES, SDF, and MOL2
06Performs batch virtual screening for large compound libraries
Use Cases
01Predicting the binding orientation of a drug candidate to a target protein
02Identifying potential binding sites for proteins with unknown experimental structures
03Running virtual screening campaigns on libraries of SMILES strings
