paper-search

name	paper-search
description	Search and retrieve academic papers about genes, targets, and diseases using paper-search-mcp tools. Use when the user asks for literature, papers, publications, PubMed, arXiv, citations, or evidence from scientific articles.

Paper Search

Overview

This skill uses paper-search-mcp tools to find and read academic papers relevant to gene and target research.

Search tools return metadata + abstract, not full text. Each hit includes paper_id, title, authors, abstract, doi, published_date, url, and optionally pdf_url. PubMed abstracts come from NCBI E-utilities; arXiv / bioRxiv / medRxiv include preprint abstracts. Full text requires a separate download/read step (PubMed does not support direct PDF read).

Preferred tools

Tool	When to use
search_papers	Broad multi-source search with deduplication
search_pubmed	Biomedical and clinical literature
search_arxiv	Preprints and computational biology
search_semantic_scholar	Citation-aware metadata and open-access links
download_with_fallback	Retrieve PDF when full text is needed
read_paper	Extract text from a downloaded paper

Search strategy

Literature search is required for every gene/target research run as step 5 of the unified workflow (after Open Targets and ClinPGx). Run PubMed even when the user does not explicitly ask for papers.

1. Build queries from resolved gene symbols and disease/context terms:

   • "{gene_symbol}" AND (disease OR target OR association)
   • Example: "BRCA2" AND ("breast cancer" OR "DNA repair")

2. For biomedical topics, prioritize PubMed and Europe PMC sources.

3. Start with search_papers for a broad sweep; use source-specific tools when you need finer control.

4. Use max_results=3 by default for PubMed; add arXiv when relevant.

Full-text workflow

When the user needs paper content or detailed evidence:

1. Pick the most relevant result with a DOI or open-access link.
2. Call download_with_fallback with the paper metadata.
3. Call read_paper on the downloaded file to extract text.
4. Quote or summarize only the sections relevant to the user's question.

Configuration notes

• PAPER_SEARCH_MCP_UNPAYWALL_EMAIL is required for Unpaywall DOI lookups.
• PAPER_SEARCH_MCP_SEMANTIC_SCHOLAR_API_KEY is optional but improves rate limits.

Output format

For each paper, include:

• Title
• Authors (first author et al. is fine)
• Year (from published_date)
• Source / journal (source field: pubmed, arxiv, etc.)
• DOI or URL
• Brief summary from the returned abstract (do not claim full-text access unless you ran read_*_paper)
• One-line relevance note

End with a brief synthesis of how the papers support or contradict the gene/target findings from Open Targets data.