AlphaFold Structure Viewer — 3D Prediction with Residue Highlighting
Look up a UniProt accession's AlphaFold prediction, view it in 3D with pLDDT confidence coloring, and highlight residue ranges.
🔒 Nothing you paste is logged or stored — an accession lookup resolves against AlphaFold DB, which also serves the structure coordinates; a sequence fold goes to the public ESM Metagenomic Atlas API
Look up a protein's AlphaFold Protein Structure Database prediction by UniProt accession and explore it in an interactive 3D viewer, powered by Mol* — the same viewer used by the PDB and AlphaFold DB websites. The structure loads with AlphaFold's standard per-residue confidence (pLDDT) coloring applied automatically, and you can add one or more residue ranges — a domain, a mutation, a binding site — each highlighted in its own color directly on the 3D structure, using the same 1-based inclusive {start, end} convention as the Protein Domain Annotator's output, so you can paste positions straight across. Not every UniProt accession has a prediction; the tool reports that plainly rather than as an error. Structure coordinates are fetched directly from AlphaFold's own CDN, and the accession you look up is never logged or stored.
Only have a gene symbol? Resolve it to a UniProt accession first →
For an accession lookup, only the UniProt accession is sent to the SeqBench API to fetch a small metadata record from AlphaFold DB, and structure coordinates are fetched directly by your browser from AlphaFold's own CDN. For a sequence fold, your pasted sequence is sent directly from your browser to the public ESM Metagenomic Atlas API, not to SeqBench's own server. The metadata lookup (alphafold_lookup) is also available via the REST API and the MCP server.
How to use the AlphaFold Structure Viewer tool
- 1Enter a UniProt accession (e.g. P04637) — use the Gene ID Mapper first if you only have a gene symbol.
- 2Click Look up. If a prediction exists, the 3D structure loads with pLDDT confidence coloring; if not, you'll see a clear "no prediction" message.
- 3Add one or more residue ranges (start–end, 1-based inclusive, plus an optional label) to highlight a domain, mutation or region of interest in its own color on the structure.
Frequently asked questions
Why do some accessions show "no prediction found"?
AlphaFold DB covers roughly 214 million structures across essentially all of UniProt, so a missing prediction usually isn't about obscurity — it genuinely means no model exists for that exact accession (this does happen). Double-check the accession is a canonical UniProt entry, e.g. via the Gene ID Mapper, rather than assuming the protein is simply too obscure to be covered.
What do the structure colors mean?
By default the structure is colored by AlphaFold's per-residue confidence (pLDDT), from dark blue (very high confidence) through yellow to orange (low confidence) — AlphaFold's own standard scheme, applied automatically. Any residue range you add is highlighted on top in its own distinct color, so you can see how a domain or mutation's confidence compares to the rest of the model.
What does the residue range use — 1-based or 0-based positions?
1-based inclusive, matching standard UniProt numbering and the Protein Domain Annotator's {start, end} output — so you can paste a domain's reported position directly into a highlight range without converting it.
Is my data stored?
No. Only the UniProt accession you look up is used — to pull a small metadata record (confidence score, model version, file URLs) from AlphaFold DB — and it is never logged or stored. For a sequence fold, the sequence goes to the public ESM Metagenomic Atlas API to run the prediction. AlphaFold DB data is CC-BY 4.0 licensed — cite AlphaFold DB when reusing a structure.