A dataset of protein–ligand binding-sites with matched ligand-based and geometry-based pocket representations

Pock-DB 2026 is a standardized dataset of protein-ligand binding-site entries derived from experimentally resolved source PDB structures. Each row in `Pock_DB.csv.gz` corresponds to one retained protein-ligand binding-site entry. Each binding-site entry is associated with one ligand-based pocket representation and one matched geometry-based pocket representation. Pocket files are provided at atom and residue levels. A web interface for this dataset is available at the following address: https://pockdb.rpbs.univ-paris-diderot.fr Binding-site identifier Binding-site-specific files use the following composite identifier:<PDB_ID>_<num_assembly>_<Ligand_ID>_<Chain_Asym_ID_-_ligand>_<num_ligand> Example: 1RM8_1_BAT_F_800 This composite identifier is reconstructed from the corresponding columns in `Pock_DB.csv.gz` and links each retained binding-site entry to its associated structural files through `binding_site_file_manifest.csv.gz`. Main annotation table `Pock_DB.csv.gz` is the main annotation table. It contains one row per retained protein-ligand binding-site entry. `Table1_additives_CCD_HET_codes.csv` contains the curated list of PDB chemical component identifiers, also referred to as CCD/HET codes, used to populate the `Additive` field in `Pock_DB.csv.gz`. This list was compiled from literature-reported crystallization additives and from BioLiP annotations of additives or non-biologically relevant ligands. The `Additive` field is provided as a filtering annotation to support user-defined subset selection and was not used to exclude entries from the full released dataset. `Table2_Pock_DB_column_dictionary.csv` provides the complete column dictionary for `Pock_DB.csv.gz`. For each column, it reports the column name, annotation category, description, and expected data type. Columns are grouped into structure-level, ligand-level, and pocket-level annotations. The PockDrug-derived pocket descriptors are described directly in `Table2_Pock_DB_column_dictionary.csv` together with all other columns of the main annotation table. Manifest file The repository includes one binding-site manifest file. `binding_site_file_manifest.csv.gz` contains one row per retained binding-site entry and reports the relative paths to the associated structural files:- cleaned assembly file;- ligand file;- ligand-environment file when available;- ligand-based atom-level pocket file;- ligand-based residue-level pocket file;- geometry-based atom-level pocket file;- geometry-based residue-level pocket file. This manifest provides the technical correspondence between the scientific annotation table (`Pock_DB.csv.gz`) and the distributed structural files, without duplicating file paths in the main annotation table. Dataset summary and validation files `dataset_summary.csv` summarizes the number and size of files by file category. `validation_report.csv` reports file-level validation checks, including expected and observed counts for each major file category and verification that each retained binding-site entry is associated with the expected structural files. `checksums_sha256.txt` provides SHA256 checksums for distributed metadata and archive files. `archives_manifest.csv` describes the compressed archives deposited in Zenodo, including archive name, file category, source directory, number of files, uncompressed size, compressed size, and SHA256 checksum. Compressed archives The structural files are distributed as one compressed archive per file category: - `PockDB2026_assemblies_cleaned.tar.gz`- `PockDB2026_ligand_files.tar.gz`- `PockDB2026_ligand_environments.tar.gz`- `PockDB2026_LB_pockets_atoms.tar.gz`- `PockDB2026_LB_pockets_residues.tar.gz`- `PockDB2026_GB_pockets_atoms.tar.gz`- `PockDB2026_GB_pockets_residues.tar.gz` Each archive preserves the internal repository paths under `Pock_DB_2026/`. PockDB2026_assemblies_cleaned.tar.gz: Pock_DB_2026/ assemblies_cleaned/ <PDB-prefix>/ <PDB_ID>_<num_assembly>_cleaned.cif PockDB2026_ligand_files.tar.gz: Pock_DB_2026/ ligands/ ligand_files/ <PDB-prefix>/ <binding_site_id>__ligand.pdb or <binding_site_id>__ligand.cif PockDB2026_ligand_environments.tar.gz: Pock_DB_2026/ ligand_environments/ <PDB-prefix>/ <binding_site_id>__ligand_environment.cif PockDB2026_LB_pockets_atoms.tar.gz: Pock_DB_2026/ ligand-based_pockets/ atoms/ <PDB-prefix>/ <binding_site_id>__LB_pocket_atm.pdb PockDB2026_LB_pockets_residues.tar.gz: Pock_DB_2026/ ligand-based_pockets/ residues/ <PDB-prefix>/ <binding_site_id>__LB_pocket_res.pdb PockDB2026_GB_pockets_atoms.tar.gz: Pock_DB_2026/ geometry-based_pockets/ atoms/ <PDB-prefix>/ <binding_site_id>__GB_pocket_atm.pdb PockDB2026_GB_pockets_residues.tar.gz: Pock_DB_2026/ geometry-based_pockets/ residues/ <PDB-prefix>/ <binding_site_id>__GB_pocket_res.pdb File types - `assemblies_cleaned`: cleaned biological assembly files.- `ligands/ligand_files`: ligand coordinate files in PDB or PDBx/mmCIF format.- `ligands/ligand_environments`: local ligand-environment files, provided when available.- `ligand-based_pockets/atoms`: atom-level ligand-based pocket files.- `ligand-based_pockets/residues`: residue-level ligand-based pocket files.- `geometry-based_pockets/atoms`: atom-level geometry-based pocket files.- `geometry-based_pockets/residues`: residue-level geometry-based pocket files. File correspondence The expected correspondence between `Pock_DB.csv.gz` and the distributed structural files is reported in `binding_site_file_manifest.csv.gz`. Users should use this manifest to retrieve the files associated with a given binding-site entry. For each retained binding-site entry, the required structural files are:- one cleaned biological assembly file;- one ligand file;- one ligand-based atom-level pocket file;- one ligand-based residue-level pocket file;- one geometry-based atom-level pocket file;- one geometry-based residue-level pocket file. The ligand-environment file is optional and is provided when available. Recommended use Users interested in tabular analyses can start from `Pock_DB.csv.gz` and `Table2_Pock_DB_column_dictionary.csv`. Users interested in structural analyses should use `binding_site_file_manifest.csv.gz` to identify and retrieve the structural files associated with selected binding-site entries. Users interested in filtering additive-like or non-primary ligands can use the `Additive` field in `Pock_DB.csv.gz` together with `Table1_additives_CCD_HET_codes.csv`. Notes on file formats Structural files are provided in PDB or PDBx/mmCIF format depending on identifier constraints. Ligand files may be distributed in either PDB or PDBx/mmCIF format. PDBx/mmCIF is used when required to preserve complete ligand identifiers that exceed the limitations of the legacy PDB format. Ligand-environment files are provided when available and are explicitly tracked in `binding_site_file_manifest.csv.gz`. Example binding-site entry For the binding-site entry: 1RM8_1_BAT_F_800 the corresponding files are expected to be:assemblies_cleaned/1R/1RM8_1_cleaned.cifligands/ligand_files/1R/1RM8_1_BAT_F_800__ligand.pdbligands/ligand_environments/1R/1RM8_1_BAT_F_800__ligand_environment.cifligand-based_pockets/atoms/1R/1RM8_1_BAT_F_800__LB_pocket_atm.pdbligand-based_pockets/residues/1R/1RM8_1_BAT_F_800__LB_pocket_res.pdbgeometry-based_pockets/atoms/1R/1RM8_1_BAT_F_800__GB_pocket_atm.pdbgeometry-based_pockets/residues/1R/1RM8_1_BAT_F_800__GB_pocket_res.pdb