Source: https://github.com/aipoch/medical-research-skills

When to Use

• You need to fetch KEGG pathway, gene, compound, enzyme, disease, or drug records directly from the KEGG REST API.
• You want to perform gene ↔ pathway mapping (e.g., building inputs for pathway enrichment or reporting).
• You need cross-references between KEGG databases (e.g., pathway → genes, gene → KO, pathway → compounds).
• You must convert identifiers between KEGG and external databases (e.g., KEGG gene → NCBI Gene ID / UniProt; KEGG compound → PubChem).
• You need drug–drug interaction (DDI) lookups for KEGG drug IDs.

Note: KEGG REST access is intended for academic use. Non-academic/commercial use may require a separate KEGG license.

Key Features

• Full coverage of core KEGG REST operations via Python helpers:
  • kegg_info (database metadata)
  • kegg_list (catalog listing)
  • kegg_find (keyword/property search)
  • kegg_get (entry retrieval; sequences/structures/images)
  • kegg_conv (ID conversion)
  • kegg_link (cross-database linking)
  • kegg_ddi (drug–drug interactions)
• Supports common KEGG identifiers and formats:
  • Pathways: map00010, hsa00010
  • Genes: hsa:10458
  • Compounds: cpd:C00002
  • Drugs: dr:D00001
  • Enzymes: ec:1.1.1.1
  • KO: ko:K00001
• Output format options for kegg_get: aaseq, ntseq, mol, kcf, image, kgml, json (some formats are single-entry only).

Dependencies

• Python >=3.9
• requests >=2.31.0

Example Usage

"""
End-to-end example:
1) Find a human gene by keyword
2) Link the gene to pathways
3) Retrieve one pathway entry
4) Convert the gene ID to UniProt
"""

from scripts.kegg_api import kegg_find, kegg_link, kegg_get, kegg_conv

# 1) Search for a gene keyword in KEGG GENES
hits = kegg_find("genes", "p53")
print("FIND results (first lines):")
print("\n".join(hits.splitlines()[:5]), "\n")

# Choose a known KEGG gene ID for TP53 (human)
gene_id = "hsa:7157"

# 2) Link gene -> pathways
pathway_links = kegg_link("pathway", gene_id)
print("LINK gene -> pathways (first lines):")
print("\n".join(pathway_links.splitlines()[:5]), "\n")

# Parse the first pathway ID from the link output
# Typical line format: path:hsaXXXXX<TAB>hsa:7157
first_line = next((ln for ln in pathway_links.splitlines() if ln.strip()), None)
if not first_line:
    raise RuntimeError("No pathways returned for the gene ID.")

path_id = first_line.split("\t")[0].replace("path:", "")
print("Selected pathway:", path_id, "\n")

# 3) Retrieve the pathway entry (flat text)
pathway_entry = kegg_get(path_id)
print("GET pathway entry (first 30 lines):")
print("\n".join(pathway_entry.splitlines()[:30]), "\n")

# 4) Convert KEGG gene ID -> UniProt
uniprot_map = kegg_conv("uniprot", gene_id)
print("CONV KEGG -> UniProt:")
print(uniprot_map)

Implementation Details

API-to-function mapping

This skill wraps KEGG REST endpoints into Python functions (see scripts/kegg_api.py):

• kegg_info(database_or_org)
  Retrieves database or organism metadata (release info, counts, etc.).

• kegg_list(database, organism=None)
  Lists entries in a database; optionally scoped to an organism (e.g., ("pathway", "hsa")).
  Also supports listing explicit IDs (batch-style) when passed as a single string.

• kegg_find(database, query, option=None)
  Searches by keyword or by chemical properties. Common option values:

  • formula (exact match)
  • exact_mass (range like 300-310)
  • mol_weight (range)

• kegg_get(entry_ids, option=None)
  Retrieves full entries or specific formats:

  • Sequences: aaseq, ntseq
  • Structures: mol, kcf
  • Pathway assets: image (PNG), kgml (XML), json (Pathway JSON)

  Batching rules:

  • Most operations allow up to 10 entries per request.
  • image, kgml, and json typically allow only 1 entry per request.

• kegg_conv(target_db, source)
  Converts IDs between KEGG and external databases (e.g., uniprot, ncbi-geneid, pubchem, chebi).
  Output is tab-delimited pairs: source_id<TAB>target_id.

• kegg_link(target_db, source)
  Cross-references entries across KEGG databases (e.g., gene → pathway, pathway → compound, gene → KO).

• kegg_ddi(drug_ids)
  Returns known drug–drug interactions for one or more KEGG drug IDs (up to typical batch limits).

Practical constraints and error handling

• Entry limits: Prefer chunking lists into batches of ≤10 IDs; enforce single-entry calls for image/kgml/json.
• HTTP status codes: Treat non-200 responses as failures; common issues include:
  • 400 (bad request / malformed parameters)
  • 404 (unknown database or entry ID)
• Rate behavior: KEGG does not publish strict rate limits; avoid high-frequency polling and add backoff/retry for robustness.

Reference documentation

For detailed endpoint syntax, database lists, and species codes, consult:

• references/kegg_reference.md