Graph Attention Site Prediction (GrASP) HOLO4K Dataset

The modified version of the HOLO4K dataset that was used for GrASP evaluation. unprocessed_pdb.zip: PDB structures from the original HOLO4k dataset. split_pdb.zip: PDB structures split into individual chains, except in the case of interfacial binders, where the full interface is retained. ready_to_parse_mol2.zip: Protein and ligand structures after our additional processing was applied. raw.zip: NumPy arrays of the features used to construct PyTorch Geometric graphs. processed.zip: Processed protein graphs used as graph neural network inputs. mol2.zip: Protein(s) with hydrogens removed and atoms renumbered accordingly. Indices match the node feature order in the NumPy and PyTorch files. holo4k(mlig)_uniprot.pkl: Pickle containing UniProt ID(s) for each receptor, used to define train/test splits.