diffdock

You can predict protein-ligand binding poses using diffusion models that calculate the 3D binding structure of small molecules to protein targets. This automates structure-based drug design and chemical biology research.

Yes, batch virtual screening allows you to evaluate large libraries of chemical compounds against a specific target protein. The system predicts binding poses for multiple ligands to identify and optimize lead compounds.

You do not need a pre-existing PDB file; the system can predict protein structures directly from amino acid sequences via ESMFold. Providing an existing protein structure file is also fully supported for molecular docking tasks.

Running diffusion models for molecular docking requires PyTorch, PyTorch Lightning, RDKit, and fair-esm. These frameworks respectively handle tensor computations, training loops, cheminformatics processing, and protein language modeling.

A diffusion model improves computational chemistry predictions by iteratively refining ligand positions to generate accurate 3D binding poses. This mechanism provides robust insights into protein-ligand interactions during virtual screening workflows.