disease-analysis

Disease Analysis (multi-scale, TogoMCP-driven)

Systematic analysis of any disease from molecular defect to clinical symptom, built from structured queries (not model recall). The driver is the TogoMCP MCP tools (run_sparql, togoid_convertId, togoid_countId) plus OLS4 and PubMed.

The full prompt template, output format, and per-disease-category customizations live in references/disease_analysis.md — read it for the deliverable structure. This SKILL.md is the operational path: verified queries, the exact tool calls, and the traps.

Files under references/ are fetched with get_workflow(name="disease-analysis", path="references/...") (when this skill is installed locally, read them as files instead).

Prerequisites

• The TogoMCP tools (run_sparql, togoid_*, search_*, get_MIE_file) drive this skill.
• The OLS4 and PubMed MCP tools (used in Phase 1 and Phase 6).
• Network access from those servers to rdfportal.org and api.togoid.dbcls.jp.

Warm-up check (confirm connectivity before trusting the workflow)

Before a full analysis, send one Phase-2 query and one TogoID conversion through the MCP tools and confirm both return rows — that proves the SPARQL endpoint and the TogoID API are both reachable and the patterns still resolve:

run_sparql(database="uniprot", sparql_query="""
PREFIX up: <http://purl.uniprot.org/core/>
PREFIX rdfs: <http://www.w3.org/2000/01/rdf-schema#>
SELECT DISTINCT ?protein WHERE {
  ?protein a up:Protein ; up:reviewed 1 ;
           up:organism <http://purl.uniprot.org/taxonomy/9606> ;
           up:annotation ?annot .
  ?annot a up:Disease_Annotation ; rdfs:comment ?c .
  ?c bif:contains "'osteoarthritis'"
} LIMIT 5""")
togoid_convertId(ids="P45452", route="uniprot,pdb")     → structures for MMP13

If either returns nothing, that step is broken (endpoint down, schema drift, or full-text index changed) — diagnose it (re-read the database's MIE via get_MIE_file) before running the analysis.