AlphaFold Protein Structure Database
ResearchEUalphafold.ebi.ac.uk
EMBL-EBI and Google DeepMind's AlphaFold Protein Structure Database: predicted three-dimensional structures for over 200 million proteins, searchable by UniProt accession, AlphaFold model id, amino-acid sequence or sequence MD5, returning per-model confidence (pLDDT) and interface scores, downloadable mmCIF/PDB/BCIF/PAE files, predicted protein complexes, and AlphaMissense pathogenicity predictions for every possible missense variant of a human protein.
Capabilities (6)
GET /v1/data/alphafold/alphamissense1 creditReturns the AlphaMissense annotation for one human protein: the accession, AlphaFold model id, full amino-acid sequence, and a single MUTAGEN annotation block whose regions array carries one mean pathogenicity score (annotation_value, 0-1) per residue position, with source_url linking the CSV of the individual substitution scores. Per-substitution scores and the benign/ambiguous/pathogenic class are NOT in this JSON, only the per-residue mean.
GET /v1/data/alphafold/checksum-summary1 creditThe same sequence lookup keyed on the MD5 checksum of the sequence instead of the sequence itself: returns the entry block (sequence, checksum, checksum_type) and the structures array of models built for it.
GET /v1/data/alphafold/complexes1 creditReturns an array of predicted multi-chain complex models that include the given protein, each with modelEntityId, globalMetricValue, the interface accuracy scores complexPredictionAccuracy_ipTM, _ipSAE, _pDockQ, _pDockQ2 and _LIS, the providerId that generated it, and complexName, complexComposition, oligomericState and assemblyType describing the chains; HTTP 404 with an empty object when the protein is in no complex.
GET /v1/data/alphafold/prediction1 creditReturns an array of AlphaFold model records for one UniProt accession or model id, each with modelEntityId, globalMetricValue (mean pLDDT), the fraction of residues at each pLDDT confidence band, latestVersion and allVersions, gene, uniprotDescription, organismScientificName, taxId, the full uniprotSequence, and download URLs for the mmCIF, BCIF, PDB, PAE image, MSA and AlphaMissense annotation files.
GET /v1/data/alphafold/sequence-summary1 creditMatches a raw amino-acid sequence against the database and returns an entry block with the sequence and its MD5 checksum, plus a structures array of every model built for that exact sequence, each with model_identifier, model_url, model_format, provider, confidence scores and coverage.
GET /v1/data/alphafold/uniprot-summary1 creditReturns a compact 3D-Beacons-format summary for one UniProt accession: a uniprot_entry block with the accession, entry id, sequence checksum, sequence length and covered segment, plus a structures array giving each model's identifier, model_category, model_url, model_format, provider, confidence type and avg confidence score, coverage and created date.
Licence
The AlphaFold Protein Structure Database states under "Licence and attributions" on alphafold.ebi.ac.uk: "Data is available for academic and commercial use, under a CC-BY-4.0 licence." (linking https://creativecommons.org/licenses/by/4.0/). The same section continues: "EMBL-EBI expects attribution (e.g., in publications, services, or products) for any of its online services, databases, or software in accordance with good scientific practice." and "If you use this resource, please cite the following papers: Bertoni D. et al. AlphaFold Protein Structure Database 2025: a redesigned interface and updated structural coverage, Nucleic Acids Research, (2025)", with the caveat "Note: A given structure may be associated with further publications. For authoritative information regarding the relevant publication(s), please consult the dataset collection or the structure metadata provided on the entry page." Copyright notices listed there: "AlphaFold Data Copyright (2022) DeepMind Technologies Limited.", "AlphaMissense Copyright (2023) DeepMind Technologies Limited." and, for third-party contributed sets served through the same API, "Kinetoplastid data copyright (2025) Wheeler Lab", "AllTheBacteria data copyright (2025) AllTheBacteria Consortium", "A BFVD Data Copyright (2025) The BFVD Development Team", "Viro3D dataset (2025)" and "VAD dataset (2026)" - all of which the page places under the same blanket CC-BY-4.0 statement, each asking for its own citation. AlphaMissense predictions additionally ask: "If you use data from AlphaMissense in your work, please cite the following paper: Cheng, J et al. Accurate proteome-wide missense variant effect prediction with AlphaMissense. Science (2023)." CC BY 4.0 permits commercial use and redistribution provided attribution is given, so both are set true.
https://creativecommons.org/licenses/by/4.0/Attribution
Display this wherever you show the data. It arrives on every response as provenance.attribution_text.
AlphaFold Protein Structure Database (EMBL-EBI and Google DeepMind), CC BY 4.0. AlphaFold Data Copyright (2022) DeepMind Technologies Limited; AlphaMissense Copyright (2023) DeepMind Technologies Limited.
Cache: up to 3600s