PDB-CCD

What PDB-CCD is

PDB-CCD is a tabular dataset containing the Chemical Component Dictionary used by the Protein Data Bank (PDB). It documents the monomers, residues, modified amino acids, ligands, and solvent molecules found within experimentally resolved macromolecular structures. Packaged by LiteFold into Apache Parquet format, it provides structured access to chemical definitions that are typically distributed across complex macromolecular Crystallographic Information File (mmCIF) dictionaries.

Structural biologists and computational chemists rely on consistent chemical references to interpret small molecules, cofactors, and unnatural residues bound to proteins and nucleic acids. The dataset organises 50,054 components into standardised records with explicit chemical graphs, identifiers, and stereochemical details, facilitating large-scale data loading for structural machine learning pipelines.

How it was made

The dataset extracts chemical definitions from the PDB Chemical Component Dictionary and structures them into tabular rows. Each record captures molecular formula properties, formal charges, synonyms, systematic nomenclature, and linear representations such as SMILES, InChI, and InChIKey. Topological structures are encoded via parallel arrays detailing atom identifiers, element symbols, formal charges, aromatic flags, leaving atom flags, stereochemical configurations, and bond connectivity with bond orders.

Data splits are assigned deterministically using the SHA-256 hash of each component identifier modulo 10. Components assigned to bucket zero form the test set of 5,009 rows, while components in buckets one to nine form the training set of 45,045 rows. Protein modification annotations and related parent component references are retained to trace non-standard monomers.

Key facts

What you can use it for

Structural machine learning

Train and evaluate deep learning models on small molecules and macromolecular building blocks using the predefined deterministic train and test splits.

Ligand property and identifier retrieval

Look up canonical SMILES, InChIKeys, chemical formulas, and formula weights for cofactors, metabolites, and ligands observed in the PDB.

Molecular graph construction

Reconstruct complete 2D molecular graphs from atom element arrays, bond connectivity lists, bond orders, and stereochemical annotations.

Residue and post-translational modification analysis

Identify modified residues and map them to their parent amino acids using non-standard monomer parent IDs and protein modification annotations.

Explore the data

The schema, real rows, full-text search and a SQL filter — read in your browser, nothing downloaded.

Strengths and limits

Questions

How to cite it

Westbrook, John D. and Shao, Chuming and Feng, Zukang and Zhuravleva, Maria and Velankar, Sameer and Young, Jasmine (2015). The chemical component dictionary: complete descriptions of constituent molecules in experimentally determined 3D macromolecules in the Protein Data Bank. Bioinformatics, 31(8), 1274–1278. doi:10.1093/bioinformatics/btu789