PDB

PDB is the legacy text file format of the Protein Data Bank, introduced in 1972 to let researchers exchange the three-dimensional atomic coordinates of proteins, nucleic acids, and their bound ligands. Its structure is a direct legacy of punch-card computing: each line is an 80-column, fixed-width "record" whose meaning is determined by a six-character record name in columns 1-6. Files carry the extensions .pdb, .ent, or .brk and the media type chemical/x-pdb. Today the format is maintained by the Worldwide Protein Data Bank (wwPDB), which since 2014 has used PDBx/mmCIF as its master archive format and froze the legacy PDB specification at version 3.30 (November 2012).

A PDB file is a sequence of typed records read top to bottom. Metadata records such as HEADER, TITLE, AUTHOR, REMARK, and SEQRES (per-chain residue sequences) precede the coordinate block. The core payload is the ATOM records (and HETATM for heteroatoms: waters, ligands, ions), each giving an atom serial number, atom and residue names, chain identifier, residue number, orthogonal x/y/z coordinates in Angstroms, occupancy, and a B-factor. TER marks the end of a chain, MODEL/ENDMDL wrap alternative models (as in NMR ensembles), and HELIX/SHEET annotate secondary structure. Bonds are largely implicit, inferred from residue templates and interatomic distances; explicit connectivity appears only in optional CONECT records, typically for non-standard groups.

In graph terms a PDB entry is naturally a molecular graph: atoms are nodes, covalent bonds are edges, and residue/chain groupings form a hierarchy. This makes PDB comparable to other chemistry molecular-structure formats such as Molfile, MOL2, SDF, CML, and XYZ, though PDB is distinguished by its macromolecular, residue-and-chain organization rather than small-molecule focus. Toolkits including Biopython, Open Babel, and RDKit parse PDB into atom/bond graphs and interconvert with those formats.

Its strengths are decades of institutional adoption, human readability, and near-universal support in visualizers (PyMOL, VMD, Jmol, UCSF Chimera). Its weaknesses stem from the rigid 80-column layout: PDB IDs are limited to four characters and atom serials to 99,999, chain identifiers to a single character, and connectivity and chemistry are underspecified. These constraints are why large structures and modern annotation have migrated to the extensible mmCIF/PDBx representation, with the legacy PDB form retained mainly for compatibility.