diffdock

To predict protein-ligand binding poses, you can provide a protein PDB file or amino-acid sequence alongside a ligand SMILES string. DiffDock uses diffusion models to generate ranked 3D binding geometries with per-pose confidence scores for structure-based drug design.

Yes, you can run batch virtual screening for multiple ligands against a single protein target. DiffDock automates pose prediction using diffusion models and outputs ranked pose files with confidence data, enabling rapid evaluation of numerous ligand candidates.

Yes, DiffDock provides per-pose confidence scores for molecular docking pose predictions. It uses diffusion models to generate multiple binding geometries, assigning confidence values to each pose to help prioritize results for structure-based drug design.

Supported input formats include PDB files or amino-acid sequences for proteins, and SMILES strings or ligand structure files for ligands. This flexibility allows you to run diffusion-based docking predictions using standard bioinformatics data formats.

Diffusion models offer a generative approach to binding pose prediction, producing multiple plausible 3D ligand geometries with confidence scores. This differs from traditional molecular docking by automating pose generation through diffusion processes rather than relying solely on scoring function optimizations.

Yes, you can integrate docking results with downstream scoring and refinement tools. DiffDock outputs ranked pose files and confidence data, which can be used directly by downstream tools for affinity estimation or lead optimization in structure-based drug design.