UniProt, KEGG and RCSB PDB APIs

The Ensembl REST API taught you one shape of programmatic access; this lesson gives you three more, so that "fetch BRCA1 from UniProt / KEGG / PDB" stops meaning "open a browser tab" and starts meaning "run a script."

UniProt REST: one accession, any format you want

UniProt's current REST API lives at rest.uniprot.org, and its simplest trick is content negotiation by URL suffix: append a file extension to an accession and get that format back, no query parameters needed. For BRCA1's entry P38398 (BRCA1_HUMAN, the Swiss-Prot record you met in UniProt: Swiss-Prot vs TrEMBL):

https://rest.uniprot.org/uniprotkb/P38398.fasta   # sequence only
https://rest.uniprot.org/uniprotkb/P38398.txt     # full flat file
https://rest.uniprot.org/uniprotkb/P38398.gff     # feature table (domains, sites)
https://rest.uniprot.org/uniprotkb/P38398.json    # structured, script-friendly
https://rest.uniprot.org/uniprotkb/P38398.xml     # same data, XML shape

Every one of those is a plain, bookmarkable URL — paste it into a browser and it works exactly as a script sees it, which makes debugging trivial. In Python, a GET request is all you need:

import requests

entry = requests.get("https://rest.uniprot.org/uniprotkb/P38398.json").json()
name = entry["proteinDescription"]["recommendedName"]["fullName"]["value"]
print(name)  # Breast cancer type 1 susceptibility protein

For anything bigger than one accession — say every reviewed human DNA-repair protein — you switch to the /uniprotkb/search endpoint with a query string and format=tsv, which streams results as a flat table instead of one JSON blob per hit.

KEGG REST: three verbs, one strict rule

KEGG (Kyoto Encyclopedia of Genes and Genomes) exposes pathway, gene, and compound data through rest.kegg.jp, a deliberately minimal API: a handful of operations, no API key, plain text back. The three you will use constantly are get (retrieve a full record), list (enumerate entries in a category), and link (find cross-references between databases). BRCA1's KEGG gene identifier is hsa:672 — hsa for Homo sapiens, 672 matching the same NCBI Gene ID you met when you looked up the gene record directly:

from Bio.KEGG import REST

gene_record = REST.kegg_get("hsa:672").read()
print(gene_record[:200])  # ENTRY, NAME, DEFINITION, ORTHOLOGY...

pathways = REST.kegg_link("pathway", "hsa:672").read()
print(pathways)  # tab-separated hsa:672 <-> path:hsa* rows

Biopython's Bio.KEGG.REST module wraps these calls directly — no manual URL building needed. This pairs naturally with Pathways: KEGG and Reactome, where you saw which pathway maps BRCA1 turns up in; now you have the code to pull that list yourself.

RCSB PDB: two APIs for two different questions

The Protein Data Bank splits its programmatic access into two purposes. The Data API, at data.rcsb.org, answers "give me everything about a structure I already know the ID for" — a GET to /rest/v1/core/entry/{id} (REST) or a query to /graphql (GraphQL, letting you request only the fields you need in one round trip). The Search API, at search.rcsb.org, answers the opposite question — "which structures match this criterion?" — accepting a JSON query body describing sequence, resolution, organism, or text filters.

For the BRCA1 structures introduced in PDB: 3D Structures at RCSB and PDBe — 1JNX, the BRCT-repeat region, and 1JM7, the BRCA1/BARD1 RING-domain heterodimer — the Data API call is a single line:

curl -s "https://data.rcsb.org/rest/v1/core/entry/1JM7" | head -c 300

That returns the deposition's resolution, experimental method, and citation as JSON. To instead find structures — every human BRCA1 entry solved by X-ray at 2.5 Å or better, say — you would POST a query to the Search API rather than guess IDs by hand.

Comparing the three at a glance

The pattern underneath all three

Every one of these APIs, like the Ensembl endpoint before it, is a thin wrapper over the same lookups the website performs — a URL suffix instead of a format dropdown, a JSON body instead of a search form. The one habit worth carrying forward is reading each provider's own rate-limit and terms-of-use page before you write a loop, not after it gets your IP blocked. UniProt, KEGG, and RCSB each publish theirs separately, and they do not agree with each other or with NCBI's 3 req/s E-utilities limit from the earlier lesson. Whichever endpoint you call, the payload that comes back is still just FASTA, JSON, or a flat file waiting to be parsed — the same handful of shapes you have already learned to read by eye.