AlphaFold Protein Structure Database
The AlphaFold Protein Structure Database is the open archive of protein structures predicted by DeepMind's AlphaFold system, built and hosted jointly by EMBL's European Bioinformatics Institute (EMBL-EBI) and Google DeepMind. It launched in July 2021 with a few hundred thousand models for key organism proteomes, and has since grown to cover over 200 million protein sequences, close to the whole of UniProt, which means most catalogued proteins now have a predicted three-dimensional structure attached to them. Every entry carries per-residue confidence scores (pLDDT) and a predicted aligned error (PAE) map, so you can judge which parts of a model to trust.
There are several ways in. The website serves individual entries you can browse and search by UniProt accession, gene or organism; a programmatic API returns structures and metadata for a given accession; and for bulk work the full database is available by FTP and mirrored as a Google Cloud Public Dataset. Structure files come as PDB and mmCIF, with the confidence and PAE data as JSON.
For RAG and AI work this is less a text corpus than a structured reference to link against. You can ground a life-sciences assistant by resolving a protein identifier to its predicted fold, confidence and cross-references, or turn the structure metadata into short natural-language summaries before embedding rather than indexing raw coordinate files. It sits naturally alongside sources we already list: UniProt supplies the sequences and annotations these predictions are keyed to, and the Protein Data Bank holds the experimentally determined structures you should prefer wherever one exists, since AlphaFold entries are predictions rather than measurements.
The database is released under CC BY 4.0, so reuse, including in commercial products, is permitted as long as you credit the source; there is no share-alike obligation and no territory restriction. The caveat worth remembering is that these are computational predictions of varying confidence, so treat low-pLDDT or disordered regions with care.
Related sources
ClinicalTrials.gov
A US National Institutes of Health database of privately and publicly funded clinical studies. It holds structured records on more than 450,000 studies, including protocols, conditions, interventions, outcomes, and results, all in the public domain.
DrugBank
A freely accessible database that combines detailed drug data with drug target information, covering more than 15,000 drug entries. It links chemistry, pharmacology, and biology in one place, which makes it a strong knowledge source for medical and pharmaceutical RAG.
MIMIC-III / MIMIC-IV
Freely accessible critical care databases with de-identified health data from more than 40,000 intensive care patients at Beth Israel Deaconess Medical Center. They include vital signs, lab results, medications, and clinical notes. Access needs credentialing and a data-use agreement.
MIRIAD
A million-scale medical instruction and retrieval dataset of roughly 5.8 million question-answer pairs, each grounded in a passage from peer-reviewed biomedical literature. Built for medical RAG, retrieval, and hallucination detection.