Access to AlphaFold Protein Structure Database via FastAPI
AlphaFold (Protein Structure Database) developed by Google DeepMind and EMBL-EBI is the first Ai that wins the 20024 Chemistry Nobel solving a 50-year-old grand challenge in biology — predicting a protein’s 3D structure directly from its amino acid sequence. What once took years of laboratory experiments can now be done in minutes, with atomic-level accuracy.
Technically, AlphaFold2 works by combining:
- Massive databases of protein sequences and structures https://alphafold.ebi.ac.uk/
- Deep neural networks (transformers) that learn how amino acids co-evolve
- Geometry-aware modules that predict how residues fold into a stable 3D form
An AI system that accurately predicts the structure of nearly every known protein on Earth, opening new frontiers in drug design, enzyme engineering, and synthetic biology.
A lightweight FastAPI microservice that wraps the public AlphaFold DB and UniProt REST APIs.
It lets you:
- Search by gene or protein name (e.g.
SLC6A4,TP53,ACE2) - Automatically resolve the UniProt accession
- Fetch AlphaFold 3D structure metadata
- Optionally download the predicted PDB/mmCIF file
/alphafold/search/{query}→ search by gene/protein name (UniProt + AlphaFold)/alphafold/prediction/{accession}→ direct AlphaFold prediction lookup/alphafold/structure/{accession}→ download structure file (.pdbor.cif)- Built-in retry logic and JSON error handling
- CORS enabled for front-end integration
- Simple, async architecture using
httpx
- Python ≥ 3.9
- Works on Linux, macOS, or GCP VM
python3 -m venv .venv source .venv/bin/activate pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 8000
An example to call the Insuline protein generating file JSON and PDB :
curl -s http://127.0.0.1:8000/alphafold/search/INS -o INS.json &&
( jq -r '.pdbUrl // empty' INS.json | xargs -r -I{} sh -c 'curl -L "{}" -o "$(basename "{}")"' )