Zero-Shot Protein–Ligand Binding Site Prediction Dataset

This record contains the original and modified datasets associated with the study “Zero-Shot Protein-Ligand Binding Site Prediction from Protein Sequence and SMILES.” The benchmark was derived from BioLiP2 and contains data for residue-level protein-ligand binding-site prediction across overrepresented, underrepresented, and zero-shot ligand regimes. The archive includes both the original datasets and the modified datasets used for the revised Stage 3 evaluation. The modified evaluation contains 45 overrepresented, 121 underrepresented, and 5,583 zero-shot ligand IDs. The accompanying README provides details on dataset construction, splitting, chemical composition, molecular size and input length, SMILES compatibility changes, receptor sequence diversity and overlap, and differences between the original and modified releases. Dataset version: 1.0. Source code is available from the associated GitHub repository.