# Query Alphafold

> Query AlphaFold protein structure predictions. Use when user asks about protein structure, 3D structure, protein folding, or structure prediction. Triggers on "alphafold", "protein structure", "3D structure", "folding", "pLDDT", "structure prediction".

- Skill: `stanfish06/query-alphafold` (Agent Skill)
- Install (CLI): `npx skillmds@latest add stanfish06/query-alphafold`
- Raw SKILL.md: https://api.skillmd.com/api/skills/stanfish06/query-alphafold/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: stanfish06 (https://skillmd.com/u/stanfish06)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/stanfish06/query-alphafold

---


# AlphaFold Structure Database Query

Query the AlphaFold EBI API for predicted protein structures.

## When to Use

- User asks about a protein's predicted 3D structure
- User wants to download PDB/CIF structure files
- User asks about structure confidence (pLDDT scores)
- User wants to visualize protein structure

## How to Execute

```python
import requests
import json

BASE_URL = "https://alphafold.ebi.ac.uk/api"

# 1. Get prediction info
def get_alphafold_prediction(uniprot_id):
    url = f"{BASE_URL}/prediction/{uniprot_id}"
    r = requests.get(url)
    r.raise_for_status()
    return r.json()

# 2. Download structure file
# take the URL from the API: AlphaFold DB serves only the latest version per entry,
# so a filename built from a pinned version number 404s after every release
def download_structure(uniprot_id, output_dir="/workspace/group", fmt="pdb"):
    entry = get_alphafold_prediction(uniprot_id)[0]
    url = entry[{"pdb": "pdbUrl", "cif": "cifUrl", "bcif": "bcifUrl"}[fmt]]
    r = requests.get(url)
    r.raise_for_status()
    filepath = f"{output_dir}/{url.rsplit('/', 1)[-1]}"
    with open(filepath, 'wb') as f:
        f.write(r.content)
    return filepath

# 3. Get per-residue confidence (pLDDT)
def get_plddt(uniprot_id):
    url = f"{BASE_URL}/prediction/{uniprot_id}"
    r = requests.get(url)
    data = r.json()
    if isinstance(data, list) and data:
        entry = data[0]
        return {"cif_url": entry.get("cifUrl", ""), "pae_url": entry.get("paeDocUrl", ""), "data": entry}
    return data

# Example
data = get_alphafold_prediction("P04637")  # TP53
if isinstance(data, list) and data:
    entry = data[0]
    print(f"UniProt: {entry.get('uniprotAccession')}")
    print(f"Gene: {entry.get('gene', 'N/A')}")
    print(f"Organism: {entry.get('organismScientificName', 'N/A')}")
    print(f"Model confidence (mean pLDDT): {entry.get('globalMetricValue', 'N/A')}")
    print(f"PDB URL: {entry.get('pdbUrl', 'N/A')}")
    print(f"CIF URL: {entry.get('cifUrl', 'N/A')}")
```

## Endpoints

| Endpoint | URL | Use |
|----------|-----|-----|
| Prediction | `/api/prediction/{uniprot_id}` | Get model info & download URLs |
| Summary | `/api/uniprot/summary/{uniprot_id}.json` | Brief summary |
| Annotations | `/api/annotations/{uniprot_id}.json?type=MUTAGEN` | Per-residue AlphaMissense annotations. Both the `.json` suffix and `type` are required; `MUTAGEN` is the only value the schema accepts |

## Download Formats

Read `pdbUrl`, `cifUrl`, or `bcifUrl` off the prediction response rather than building a filename. AlphaFold DB serves only the latest version per entry -- `AF-{UNIPROT_ID}-F1-model_v6.*` today, with v4 and v5 both 404 -- so any pinned version breaks at the next release.

- PAE image: `paeImageUrl`; PAE matrix JSON: `paeDocUrl`

## Follow-up Suggestions

- "Want me to analyze the structure confidence by region?"
- "Should I compare this to the experimental PDB structure?"
- "Want me to identify disordered regions?"