PocketZn: Zn-containing small-molecule ligand binding sites

Small-molecule ligand binding sites from BioLiP2 that have a zinc ion sitting in the same pocket: 5,183 binding sites over 2,817 PDB entries and 1,618 distinct ligand chemical components, of which 5,178 ship with coordinates rebuilt from RCSB PDB mmCIF entries in deposited author numbering. Each record carries the receptor chain, the ligand, the zinc and a concatenated complex, with the annotated binding site independently recomputed and verified. What the dataset is. The unit of PocketZn is the ligand binding site, not the zinc site. The zinc is a filter applied to an otherwise ordinary cofactor pocket. The distinction matters: a zinc-site dataset would be built around the metal and would include structural zincs with nothing else nearby, whereas PocketZn is built around druggable pockets and keeps only those that happen to contain a zinc. The 5,183 sites span 597 UniProt accessions. Of these, 274 survive into BioLiP's 90 %-identity non-redundant subset — a low fraction, because zinc enzymes such as carbonic anhydrase are represented by very many near-identical structures that collapse into one cluster. Files. The four dataset archives all unpack into pocketzn/ and reconstruct the dataset directory exactly when extracted together. pocketzn_tables holds the three tables, the FASTA and the documentation; pocketzn_ligands the ligand and zinc coordinates; pocketzn_receptors the receptor chains; and pocketzn_complexes the concatenated complexes. README.md, LICENSE and CITATION.cff are also included on their own so the record can be read without downloading anything. Licence. The Creative Commons Attribution 4.0 International licence covers what originates here, not everything in the archive: the selection, the columns computed rather than copied, the structure files as assembled, and the documentation. The annotation fields reproduced from BioLiP2 keep their own terms — BioLiP2 ships no licence file and states no terms beyond a request that it be cited (Zhang, Zhang, Freddolino and Zhang, Nucleic Acids Research 52:D404–D412, 2024) — and every coordinate is taken from RCSB PDB mmCIF entries, which the PDB places in the public domain. Cutting and reassembling those coordinates does not withdraw them from the public domain; what is licensed is the selection and assembly, not the atoms. Section 8 of the README states the scope in full.