gene-set-provenance

name	gene-set-provenance
description	Use this skill for gene set provenance, source datasets, gene membership, overlap searches, Pigean phenotype associations, graph neighborhoods, and evidence-backed interpretation from the gene set MCP server.

gene-set-provenance

Use this skill when the user asks about gene set provenance, source datasets, gene membership, overlap, Pigean phenotype associations, graph neighborhoods, or evidence behind a gene set or gene.

Goals

• Ground answers in MCP data instead of assumptions.
• Keep tool calls compact because the remote server is rate-limited.
• Return provenance in a form that is easy to inspect and cite.
• Use configured MCP tools instead of manual HTTP or handcrafted JSON-RPC whenever possible.

Recommended Tool Order

1. search_gene_sets_semantic Use first when the user is not asking for a specific gene set standard_name and needs concept-based discovery.
2. search_gene_sets Use when the user gives or is trying to confirm an exact or partial gene set standard_name.
3. get_gene_set Use after identifying the target gene set.
4. get_pigean_gene_set Use after identifying a gene set when the user wants phenotype associations for that gene set; query it by resolved gene set standard_name.
5. get_provenance Use for provenance graph and source metadata.
6. find_gene_sets_by_gene Use when the user starts from genes instead of a gene set.
7. get_pigean_gene Use when the user starts from a gene and wants phenotype associations for that gene.
8. get_graph_neighborhood Use for adjacent entities and relationship summaries.

Workflow

1. If the request is concept-based and not for a specific gene set standard_name, start with search_gene_sets_semantic.
2. If needed, refine or confirm with search_gene_sets using the candidate standard_name.
3. Resolve the exact gene set identifier or standard name.
4. Fetch the gene set itself before discussing provenance.
5. If phenotype-association context is relevant, fetch Pigean associations for the resolved gene set using its standard_name.
6. Fetch provenance separately and distinguish:
   • core gene set fields
   • Pigean phenotype associations
   • provenance graph
   • provenance metadata
7. If the user starts from genes, use find_gene_sets_by_gene before broader neighborhood exploration.
8. If the user asks for gene-level phenotype associations, use get_pigean_gene.
9. If the user asks for overlap or neighborhood context, run the smallest relevant follow-up query.
10. If configured MCP tools are unavailable in the session, stop and tell the user that MCP is unavailable.
11. Do not fall back to curl, repo test scripts, or manual JSON-RPC unless the task is explicitly to test or debug the MCP server itself.

Output Pattern

• Gene set: standard name, identifier, organism, library if present
• Pigean gene set associations: phenotype, beta, beta_uncorrected, rs_score if returned Query key: resolved gene set standard_name
• Provenance: source dataset, method, contrast, tissue, pipeline, or other linked entities if returned
• Pigean gene associations: phenotype, phenotype_name, combined, label if returned
• Evidence: direct fields returned by the server
• Interpretation: concise explanation of what the provenance implies

Guardrails

• Do not infer causal biology from provenance alone.
• Do not infer causality or disease relevance from Pigean scores alone.
• Do not claim fields exist if the tool response omits them.
• Do not invoke the MCP HTTP endpoint directly when the configured MCP tools can answer the request.
• Do not use local scripts or README examples as a substitute for configured MCP tools during normal data queries.
• If search_gene_sets_semantic returns candidates, confirm the final target with exact-name lookup before deeper analysis.
• If multiple gene sets match, present candidates before continuing.
• If the tool output is truncated, say so and avoid overclaiming.