improvement_plan.md

Bio-MCP Enhancement Plan: Research Capabilities

To transform this from a simple look-up tool into a Research-Grade Bio-AI Agent, we will add context and evidence layers.

1. Deep Evidence Retrieval (PubMed Integration)

Goal: Allow researchers to verify variant classifications by reading the underlying literature.

• New Tool: get_supporting_literature(variant_id)
• Logic:
  1. Extract PMIDs (PubMed IDs) from the ClinVar record.
  2. Query NCBI PubMed API (eutils/esummary.fcgi?db=pubmed).
  3. Return titles, journals, and publication dates of supporting papers.
• Benefit: Provides the source of truth behind a "Pathogenic" label.

2. Gene Context & Function (NCBI Gene)

Goal: Provide biological context for variants in less familiar genes.

• New Tool: get_gene_info(gene_symbol)
• Logic:
  1. Search NCBI Gene DB (db=gene).
  2. Retrieve "Summary" (function), "MapLocation", and "Aliases".
• Benefit: Helps the agent understand the mechanism of action (e.g., "This gene encodes a tumor suppressor...").

3. Phenotype-Driven Search

Goal: Find variants associated with a specific disease.

• Enhancement: Update search_clinvar to support explicit phenotype filtering.
• Logic: Add logic to append [dise] or [pdis] tags to the query term if the internal reasoning detects a disease name.

Implementation Steps

Phase 1: PubMed Integration

• Update src/clinvar.py to handle db=pubmed requests.
• Add parsing for citation_list in ClinVar responses.
• Create get_supporting_literature tool in src/main.py.

Phase 2: Gene Info Tool

• Create src/gene.py (or add to clinvar.py) for db=gene interactions.
• Create get_gene_info tool in src/main.py.

Phase 3: Verification

• Update test_integration.py to check for citations and gene info.