---

name: literature-search description: Search PubMed for RNAi/dsRNA research on target genes or species (utility skill - use anytime) type: utility

Literature Search Skill

When to Use This Skill

This is a utility skill - use it at ANY point in the workflow when you need:

• Published RNAi/dsRNA studies for a pest species
• Evidence supporting gene essentiality
• Validation of candidate gene targets
• References for the final report

IMPORTANT: Tool Selection

ALWAYS use PubMed MCP tools for literature searches. DO NOT use WebSearch/Tavily.

PubMed provides peer-reviewed, citable scientific literature with structured metadata (PMIDs, DOIs, abstracts). Web search returns unstructured, potentially unreliable results.

Instructions

Step 1: Search PubMed Using MCP Server

Use the PubMed MCP tools.

For species-wide RNAi research:

Copypubmed_search_articles
query: "{species}" AND (RNAi OR dsRNA OR "RNA interference" OR "gene silencing")
max_results: 50

For specific gene targets:

Copypubmed_search_articles
query: "{gene_name}" AND (RNAi OR dsRNA) AND insect
max_results: 20

Step 2: Get Article Details

For relevant PMIDs, fetch full metadata:

Copypubmed_get_article_metadata
pmids: ["PMID1", "PMID2", ...]

Step 3: Extract Gene Names (CRITICAL)

You MUST extract gene names from each paper's title and abstract.

The match_essential.py script relies on the gene_names field to give literature support scores to candidate genes. If this field is empty or missing, literature support will be ignored.

Look for these gene patterns in titles and abstracts:

Step 4: Save Results in Required Format

Analysis outputs go in output/{run}/, NOT in data/.

Write to output/{run}/literature_search.json:

REQUIRED FORMAT:

Copy[
  {
    "pmid": "12345678",
    "doi": "10.1234/example",
    "title": "RNAi silencing of vATPase in Drosophila suzukii causes mortality",
    "authors": ["Smith J", "Jones K"],
    "journal": "Journal of Insect Physiology",
    "year": "2020",
    "gene_names": ["vATPase"],
    "abstract_snippet": "We demonstrate effective gene silencing..."
  },
  {
    "pmid": "12345679",
    "title": "Chitin synthase and acetylcholinesterase as RNAi targets",
    "gene_names": ["chitin synthase", "acetylcholinesterase"],
    ...
  }
]

CRITICAL FIELDS:

• gene_names - REQUIRED - Array of gene names found in title/abstract
• pmid - PubMed ID
• title - Article title

The downstream script match_essential.py checks paper.get('gene_names', []) for each paper. If gene_names is missing or empty, that paper won't contribute to literature support scores.

Step 5: Verify Format

After saving, verify the format is correct:

Copyjq '.[0:2] | .[] | {pmid, gene_names}' output/{run}/literature_search.json

Should show each paper with its extracted gene_names array.

Alternative: Use parse_pubmed.py Script

If you have raw PubMed XML, you can use the bundled script to extract genes:

Copypython dsrna_agent/skills/literature-search/scripts/parse_pubmed.py \
  --xml-file /tmp/pubmed_results.xml \
  --output output/{run}/literature_search.json

This automatically extracts gene names using pattern matching.

Available MCP Tools

Notes

• Always cite PubMed and include DOIs when reporting findings
• If no results for exact species, try related species or genus-level queries
• Gene mentions from literature boost candidate scores in the scoring step
• This skill runs AUTONOMOUSLY - no user confirmation needed
• Do NOT ask "Would you like me to search PubMed?" - just search when relevant
• Integrate results silently and continue with the workflow