🧬 PubMed Research Normalizer

Search 37M+ biomedical papers from PubMed with smart normalization — drug names mapped to generics, authors deduplicated, MeSH terms categorized, institutions cleaned. Built on the free NCBI E-utilities API.

✨ What makes this different?

Most PubMed scrapers just dump raw API output. This actor enriches and normalizes:

• 💊 Drug name normalization: 150+ brand names → generic names (Keytruda → Pembrolizumab)
• 👥 Author deduplication: "Smith JA" and "Smith J" normalized to "Smith, J."
• 🏥 Institution cleaning: "Harvard Med School" → "Harvard Medical School"
• 🏷️ MeSH categorization: Raw MeSH terms grouped into clinical categories (Oncology, Cardiology, etc.)
• 📊 Human-readable output: Formatted summary table in the run log

📥 Input

Field	Type	Required	Description
searchQuery	string	✅	PubMed search query. Supports field tags: [tiab], [au], [dp], [mh]
maxResults	integer	❌	Max papers to return (1–500). Default: 20
sortBy	string	❌	relevance, date, or pub_date. Default: relevance
dateFrom	string	❌	Earliest date (YYYY/MM/DD or YYYY)
dateTo	string	❌	Latest date (YYYY/MM/DD or YYYY)
includeAbstracts	boolean	❌	Fetch full abstracts. Default: true
normalizeDrugNames	boolean	❌	Map brand→generic drug names. Default: true
normalizeAuthors	boolean	❌	Deduplicate author names. Default: true

Example Input

{
  "searchQuery": "pembrolizumab melanoma[tiab] AND 2025[dp]",
  "maxResults": 25,
  "sortBy": "pub_date",
  "includeAbstracts": true,
  "normalizeDrugNames": true
}

📤 Output

Each paper is returned as a normalized JSON object:

{
  "pmid": "42560690",
  "doi": "10.1001/jamaoto.2026.2124",
  "title": "Combined Positive Score and Cost-Effectiveness of Perioperative Pembrolizumab...",
  "journal": "JAMA Otolaryngology--Head & Neck Surgery",
  "publicationDate": "2026 Aug 6",
  "authors": ["Coyle, A.H.", "Hutton, D.W.", "Buchakjian, M.R.", ...],
  "authorCount": 17,
  "drugMentions": ["Pembrolizumab"],
  "meshCategories": ["Oncology", "Immunotherapy", "Health Policy"],
  "abstract": "Importance: Head and neck cancer...",
  "pubmedUrl": "https://pubmed.ncbi.nlm.nih.gov/42560690/",
  "doiUrl": "https://doi.org/10.1001/jamaoto.2026.2124"
}

Output Schema

Three endpoints for different consumers:

• Dataset: Full paper records as JSON array
• OUTPUT (KVS): Human-readable formatted table
• STATS (KVS): Machine-readable run statistics

🎯 Use Cases

• Pharma competitive intelligence: Track competitor drug publications
• Systematic reviews: Search and normalize papers for meta-analysis
• Medical affairs: Monitor literature for key drugs and conditions
• Biotech R&D: Identify emerging research trends
• AI/ML training data: Curated biomedical text corpus

🔌 Data Source

NCBI PubMed via the E-utilities API. Free, no API key required. Rate limit: ~3 requests/second.

💰 Pricing

Pay per result — only charged when papers are returned. Empty searches are free.

---

Built by Chris The Dev