Researched
Protein Data Bank
Founded in 1971 with 7 structures, the free archive of 3D molecule structures now holds over 250,000 and supplied the training data for AlphaFold.
Open in the interactive tree →The PDB was announced in October 1971 by Brookhaven National Laboratory and the Cambridge Crystallographic Data Centre, and began with 7 structures under Walter Hamilton. Researchers deposit each solved structure, and since 2003 the worldwide PDB (United States, Europe, Japan) has kept one open archive under a public-domain licence; it held 254,149 entries in May 2026. DeepMind trained AlphaFold 2 on over 170,000 of these structures, and in 2020 it predicted protein shapes nearly as well as experiments.
Prerequisites
- Programmable Computer (Zuse Z3)1941An electronic archive that scientists could search from afar.
- First Protein Structure1958Every entry is a molecular structure solved by experiment, first by X-ray.
Unlocks
- AI Protein Structure Prediction2020AlphaFold 2 was trained on over 170,000 structures from the archive.