track-lab-publications

name	track-lab-publications
description	Track recent publications from a specific research lab or PI. Given a PI name and/or institution, fetches their recent papers from PubMed and Europe PMC, summarizes research themes, citation impact, and generates a formatted publication report.

Track Lab Publications

Purpose

Monitor and analyze a Principal Investigator's (PI) recent publication output, research themes, and academic impact. This skill automatically fetches publications from PubMed, extracts metadata, analyzes research keywords and trends, and generates comprehensive publication reports with citation metrics and collaboration networks.

Use when / Not when

Use when

• You need to monitor a PI's recent publication activity (within last 1-5 years)
• You want to identify research themes and topics a lab focuses on
• You need to quantify citation impact and journal diversity for a research group
• You're building a collaborator network analysis or funding proposal
• You want to generate a formatted report for grant applications or lab websites
• You need to identify co-author patterns and collaboration networks

Not when

• You need real-time trending topics (PubMed data has 2-4 week lag)
• You require proprietary databases (preprints, conference abstracts)
• You need full-text PDF analysis (this skill uses only metadata)
• The PI has very common name without disambiguation (institution required)
• You need complete citation graphs (Europe PMC has limited historical coverage)

Expected inputs / outputs

Inputs

• CSV/TSV (optional): Pre-compiled PI and institution list
• Command-line arguments: PI name, institution, year range, result limits
• API calls: PubMed eUtils (free, no key required) and Europe PMC (free)

Outputs

• publications.tsv: Tab-separated table with PMID, title, authors, journal, year, DOI, citation count, abstract
• lab_report.md: Formatted markdown report with PI summary, yearly trends, top journals, top keywords, collaborators
• publications_timeline.png: Bar chart of publication count per year
• keyword_distribution.png: Horizontal bar chart of top 15 keywords/MeSH terms
• coauthor_network.png (optional): Network graph of co-author collaborations
• Summary to console: PI name, total papers found, date range, top 5 journals

Procedure

Step 1: Query PubMed for publications

1. Construct search query: <PI_name>[Author] AND <institution>[Affiliation]
2. Call PubMed eSearch API with retmax (default 100), mindate and maxdate filters
3. Retrieve list of PMIDs matching criteria
4. If no results, retry with relaxed query (PI name only) with warning

Step 2: Fetch publication details

1. Batch PMIDs into groups of 10 for eFetch API calls
2. Parse XML response to extract:
   • PubMed ID (PMID)
   • Title, authors (last name + initials), author affiliations
   • Journal name and impact factor (if available)
   • Publication year, DOI
   • Abstract (truncate to 500 chars for report)
   • MeSH descriptors (primary research keywords)
3. Handle missing fields gracefully (e.g., abstract may not exist)

Step 3: Fetch citation counts (optional)

1. For each PMID, query Europe PMC API: /search?query=PMID:<pmid>&format=json
2. Extract citedByCount field
3. Handle rate limiting (max 25 requests/second)

Step 4: Analyze research themes

1. Keyword extraction:
   • Combine all MeSH terms from all papers
   • Add high-frequency words from abstracts (exclude stopwords)
   • Rank by frequency across entire publication set
2. Yearly trends:
   • Count papers per year
   • Calculate growth rate
3. Journal analysis:
   • Identify top 5-10 journals by paper count
   • Track journal diversity (Shannon entropy)

Step 5: Build co-author network (optional)

1. Extract all author names and affiliations from papers
2. Count author co-occurrence frequencies
3. Identify top collaborators (co-authors with 2+ joint papers)
4. Rank by collaboration frequency

Step 6: Generate outputs

1. Write TSV table of all publications
2. Generate markdown report with:
   • PI profile header
   • Summary statistics
   • Year-by-year breakdown
   • Research theme summary
   • Top journals and keywords
   • Top collaborators
   • Full publication list with abstracts
3. Create matplotlib visualizations:
   • Time-series bar chart
   • Keyword frequency distribution
   • Co-author network graph (if requested)

Key execution patterns

# Basic: Track a single PI by name and institution
python scripts/track_lab_publications.py \
  --pi-name "Jennifer Doudna" \
  --institution "UC Berkeley" \
  --years-back 5 \
  --outdir ./doudna_publications

# Track PI with citation counts and save as JSON
python scripts/track_lab_publications.py \
  --pi-name "Feng Zhang" \
  --institution "Broad Institute" \
  --fetch-citations \
  --output-format json \
  --outdir ./zhang_pubs

# Build co-author network for a PI
python scripts/track_lab_publications.py \
  --pi-name "Edith Heard" \
  --institution "Institut Curie" \
  --co-author-network \
  --max-results 200 \
  --outdir ./heard_collab

# Track publications and limit to 50 papers
python scripts/track_lab_publications.py \
  --pi-name "David Baker" \
  --years-back 3 \
  --max-results 50 \
  --outdir ./baker_recent

# Relaxed search (PI name only, no institution)
python scripts/track_lab_publications.py \
  --pi-name "Doudna" \
  --max-results 100

Parameter decision guide

Parameter	Value	When to use	Rationale
--pi-name	Last name only (e.g., "Doudna")	PI has very unique surname	Broader search, may include unrelated authors
--pi-name	Full name (e.g., "Jennifer Doudna")	PI has common last name	Better specificity, fewer false positives
--institution	"UC Berkeley" or "UC San Francisco"	PI may work at multiple institutions	Disambiguate authors with same name
--years-back	3 (default)	Current publication trend	Recent impact and research direction
--years-back	5-10	Track research evolution over time	Historical perspective on research trajectory
--max-results	100 (default)	Most research groups	Captures 95% of active researcher output
--max-results	200+	Very prolific PIs (>20 papers/year)	Ensures complete publication list
--fetch-citations	True	Grant proposals or impact assessment	Demonstrates research influence
--fetch-citations	False	Quick turnaround needed	Saves 30-60 seconds API time
--co-author-network	True	Collaboration analysis or network studies	Visualize collaboration ecosystem
--co-author-network	False	Publication list only	Simpler output, faster execution
--output-format	markdown (default)	Human review or lab website	Best readability
--output-format	json	Data pipeline or downstream processing	Machine-readable, structured
--output-format	tsv	Import to spreadsheet or analysis	Excel/R-compatible format

Failure modes

Failure	Cause	Solution
No papers found	PI name misspelled or very common	Use full name + institution; check PubMed directly
Too many results (1000+)	PI name is common, institution too broad	Add institution, narrow years-back, or use initials
Missing abstracts	Older papers or non-indexed journals	Script handles gracefully; continues with available data
API rate limit exceeded	Too many citation fetch requests	Reduce --max-results or use --fetch-citations=false
Europe PMC timeout	Network issue or service unavailable	Retry; continue without citation counts
Garbled author affiliations	XML encoding issues from PubMed	Script sanitizes UTF-8; may lose diacritics
Empty co-author network	Few papers or no co-authors	Script skips network visualization if <5 papers
Network graph too dense	Many collaborators with many papers	Filter to top-50 edges; use --co-author-network sparingly