OncoGraph: An Open-Source Framework for Federated Cancer Data Discovery Across 800+ Cancer Types and 29 Repositories
Cancer research data are scattered across dozens of repositories with incompatible interfaces, forcing researchers to navigate each one independently before any analysis can begin. The problem is especially burdensome in neuro-oncology, where the 2021 WHO reclassification of CNS tumors around molecular markers like IDH1 and MGMT means that a single glioblastoma study may need genomic profiles from TCGA, segmentation benchmarks from BraTS, drug response data from DepMap, and survival statistics from SEER. We present OncoGraph, an open-source discovery engine that accepts a cancer type and research need and returns the relevant datasets, biomarkers, therapies, survival data, signaling pathways, and repositories across 29 sources. OncoGraph is implemented as eleven composable TypeScript packages totaling over 8,000 lines of code: a core query engine built on an OncoTree-derived taxonomy of 802 cancer types with ICD-O-3 codes, domain packages mapping genomic (TCGA, ICGC, GENIE) with 104 biomarkers, imaging (TCIA, IDC, BraTS) with 53 datasets, pharmacological (DepMap, GDSC) with 66 drugs across 17 classes, and clinical (SEER, UK Biobank, ClinicalTrials.gov) data sources across 29 repositories, a TF-IDF search index with field boosting and synonym expansion, structured export with data completeness scoring, 20 cancer signaling pathways with gene-pathway analysis, 41 survival data entries with Kaplan-Meier curve generation, a REST API server with 10 endpoints, and React hooks and components for web integration. Federated API clients connect directly to GDC GraphQL, IDC, CDA, DepMap, and ClinicalTrials.gov. The system includes 47 automated tests via Vitest and a FAIR compliance checker. We walk through representative queries for glioblastoma, NSCLC, breast cancer, and AML, and report a structured coverage audit of the system's completeness against published repository catalogs. OncoGraph is available under the MIT license at https://github.com/kingofmit/oncograph (DOI: 10.5281/zenodo.24268993).
Authors
- Ahmed Allem (ORCID: https://orcid.org/0009-0003-4937-5858)
- Andrea Dykes (ORCID: https://orcid.org/0009-0008-2170-2331)
- Freedom Cheteni-Mandela (ORCID: https://orcid.org/0009-0006-3507-0094)
- Maria Paulina Valenzuela Bonillas (ORCID: https://orcid.org/0009-0002-4010-427X)
Institutions
- Universidad Autónoma de Baja California (MX)
- Stanford University (US)
Publication Details
- Published
- 2026-11-18
- DOI
- https://doi.org/10.68956/journalos.jrn-2026-neural-6b0534873d0e4dcd8cb5d9ff2d583ff3
- Primary Topic
- Research Data Management Practices
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- University of Pennsylvania
- Memorial Sloan-Kettering Cancer Center
- Broad Institute
- National Cancer Institute