An AI-powered platform that takes a disease name and produces a ranked, evidence-weighted list of drug targets by aggregating data from Open Targets, PubMed, ClinicalTrials.gov, Protein Atlas, STRING-DB, Enrichr, and more — no manual curation required.
Prerequisites: Node.js 18+
-
Install dependencies:
npm install -
Set the
GEMINI_API_KEYin.envto your Gemini API key:GEMINI_API_KEY=your_key_here -
Run the app:
npx tsx server.ts
Important: Use
npx tsx server.ts, notnpm run dev. The app needs the Express server running alongside Vite — without it, all external API calls (ClinicalTrials, PubMed, Protein Atlas, PubTator) will return empty results.
| Variable | Description |
|---|---|
NOTION_TOKEN |
Notion integration token — enables Export → Notion |
NOTION_DATABASE_ID |
Target Notion database ID |
NVIDIA_API_KEY |
NVIDIA NIM — fallback AI for disease name correction and target summaries |
- Target table — ranks genes by GET Score (Genetic × 0.50 + Expression × 0.25 + Tractability × 0.25) with configurable visible columns, sortable headers, and dual-handle range filters
- Pathway enrichment — parallel KEGG, Reactome, and WikiPathways queries via Enrichr with FDR, gene overlap, and source filtering
- Literature tab — PubTator3 gene-mention mining with publication velocity scoring
- Drill-down panel — per-target ClinicalTrials data, PubMed stats, Europe PMC velocity, and AI-generated summary
- Network scoring — optional RWR (Random Walk with Restart) and WINNER scores via STRING-DB protein interaction network
- TAU scores — Protein Atlas tissue and single-cell specificity loaded in background after initial results
- Export — filtered table to CSV, DOCX, or Notion database
- Co-Pilot — context-aware Gemini AI assistant with full research state
Open Targets · Protein Atlas · PubTator3 / PubMed · Europe PMC · ClinicalTrials.gov · STRING-DB · Enrichr (KEGG / Reactome / WikiPathways) · Google Gemini