EDGAR uses statistical enrichment analysis over biomedical knowledge graphs to predict missing links. Given a query entity and relationship type, the pipeline operates in three stages:
- Lookup Stage — Retrieve all known entities connected to the input (e.g., all chemicals that treat a disease).
- Enrichment Analysis — Identify features (genes, phenotypes, pathways, etc.) that are statistically over-represented among the lookup set using hypergeometric/Poisson tests. Each enrichment rule is scored by p-value.
- Enrichment-based Results — Use the significant enrichment rules to infer novel entities that share those enriched features but were not in the original lookup set.
EDGAR supports a range of biomedical link prediction tasks, including:
- Drug Repurposing — drugs that may treat a disease
- Gene Discovery — genes associated with conditions
- Function Prediction — biological processes for genes
- Pathway Analysis — pathways involving specific genes
- Target Identification — gene targets for chemical compounds
O. Olasunkanmi, E. Morris, Y. Kebede, H. Lee, S. C. Ahalt, A. Tropsha, and C. Bizon, "Explainable Enrichment-Driven GrAph Reasoner (EDGAR) for Large Knowledge Graphs with Applications in Drug Repurposing," 2024 IEEE International Conference on Big Data (BigData), Washington, DC, USA, 2024, pp. 777-782.
BibTeX
@inproceedings{olasunkanmi2024explainable,
title={Explainable enrichment-driven graph reasoner (edgar) for large knowledge graphs with applications in drug repurposing},
author={Olasunkanmi, Olawumi and Morris, Evan and Kebede, Yaphet and Lee, Harlin and Ahalt, Stanley C and Tropsha, Alexander and Bizon, Chris},
booktitle={2024 IEEE International Conference on Big Data (BigData)},
pages={777--782},
year={2024},
organization={IEEE}
}A web application for biomedical link prediction using enrichment analysis over knowledge graphs, built on the TRAPI (Translator Reasoner API) standard.
Live instance: edgar.apps.renci.org
EDGAR submits TRAPI queries to AnswerCoalesce (AC) via its /asyncquery endpoint with a callback URL, and surfaces the results through a React UI.
- Backend — FastAPI (async) that submits queries to AC's
/asyncquerywith a callback URL. AC processes in the background and POSTs the full TRAPI result back when done. Falls back to polling if the callback approach is unavailable. See backend/README.md. - Frontend — React + TypeScript + Vite with real-time job status updates via WebSocket. See frontend/README.md.
- Auth — JWT-based, demo user in development.
Job state is currently held in-process (no database). That has implications for deployment — see DEPLOYMENT.md.
Prerequisites: Python 3.9+, Node.js 18+.
1. Install once:
cd backend && pip install -r requirements.txt && cd ..
cd frontend && npm install && cd ..2. Start both services:
./start.shThat's it — start.sh boots the backend on :8000 and the frontend on :5173 in the same terminal and cleans up on Ctrl+C.
Prefer two terminals? Run them manually:
# terminal 1 — backend
cd backend && python3 -m uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
# terminal 2 — frontend
cd frontend && npm run devOpen http://localhost:5173 and log in with demo / demo (dev-only credentials, defined in the backend).
After merging to main, the release workflow pushes new container images. To roll the pods in the cluster:
kubectl rollout restart deployment edgar-ui-backend -n <your-namespace>
kubectl rollout restart deployment edgar-ui-frontend -n <your-namespace>First-time install (helm chart, namespace, ingress) is covered in DEPLOYMENT.md.
- backend/README.md — backend setup, API endpoints, TRAPI format
- frontend/README.md — frontend features, project structure, build
- DEPLOYMENT.md — Docker images, CI, production caveats
Part of the NIH Biomedical Data Translator Consortium.
- NIH Biomedical Data Translator Consortium
- Renaissance Computing Institute (RENCI)
- TRAPI standard contributors