A jewel in the crown of structural biology, the Protein Data Bank is an unparalleled archive of protein and nucleic acid structures. Its operation depends on information-rich relational databases, the schemas for which are efficiently expressed in the macromolecular CIF (mmCIF) formalism. Discrete entities that may be stored in the databases are characterised in CIF-format data dictionaries, which express ontologies widely used in areas as diverse as structural genomics, NMR, cryo-electron microscopy and protein production.