The Protein Data Bank (PDB) collects experimentally determined 3D structures of proteins and other molecules. How many did it hold in 2023?
As of 2023, experimentally determined 3D molecular structures in the Protein Data Bank (PDB), thousand entries is 212.6 thousand: from 0.5 thousand to 212.6 thousand (1990–2023).
Why it changed
Two things are true at once: the open Protein Data Bank, founded in 1971, collects structures determined by experiment: mostly X-ray crystallography, also NMR and, since 2013, increasingly cryo-electron microscopy. And between 1990 and 2023 its entries grew more than 400-fold, from 507 to 212,599. These experiments are expensive and slow, so structures are now also predicted: the AlphaFold database already held 214.7 million models in 2022. But a prediction is not a measurement: AlphaFold models single protein chains rather than the complexes they form, and parts of them with low confidence.
Source · Protein Data Bank (Epoch AI data), via Our World in Data CC BY 4.0, Our World in Data; underlying: Epoch AI (2024) Last reviewed 2026-10-30
See one like this on every new tab.
Progress Tab replaces your new tab with sourced numbers about the world, on topics you pick. Some cards ask you to guess first.
Add to Chrome Free · no account · opens on every new tabAlso available for Firefox →Common questions
What is the source for this number?
Protein Data Bank (Epoch AI data), via Our World in Data. Last reviewed 2026-10-30. CC BY 4.0, Our World in Data; underlying: Epoch AI (2024).
How has this changed since 1990?
From 0.5 thousand to 212.6 thousand (1990–2023). As of 2023, experimentally determined 3D molecular structures in the Protein Data Bank (PDB), thousand entries stands at 212.6 thousand.
What still isn’t solved?
Between 1990 and 2023 its entries grew more than 400-fold, from 507 to 212,599. These experiments are expensive and slow, so structures are now also predicted: the AlphaFold database already held 214.7 million models in 2022. But a prediction is not a measurement: AlphaFold models single protein chains rather than the complexes they form, and parts of them with low confidence.