DiffPocket: Dataset and Pre-trained Models for Protein–Ligand Binding-Site Prediction

DiffPocket is a research dataset and model release for protein–ligand binding-site prediction. This release provides data and pre-trained model checkpoints associated with the DiffPocket framework, along with documentation describing the package structure and contents. The release includes model checkpoints for the documented evaluation settings, supporting data, a README, a data dictionary, and file-integrity information to facilitate inspection and reproducibility. The package is organized to help researchers identify the available checkpoints and supporting resources. Users should consult the accompanying README and data dictionary for details about the files, intended usage, and evaluation configurations. This release is version 1.0.0 of the DiffPocket dataset and pre-trained model package.