Robertium hypothesis catalog: cross-domain drug repurposing candidates from biomedical literature

Robertium is an open-source platform for literature-based drug discovery that identifies cross-domain drug repurposing candidates by applying Don R. Swanson's ABC reasoning model to a knowledge graph extracted from biomedical literature. This dataset contains 672 outreach-quality cross-domain hypotheses (62 high-confidence, 610 medium-confidence) discovered across six therapeutic domains: glioblastoma, epilepsy, ALS, Alzheimer's disease, pancreatic cancer, and major depressive disorder. Each hypothesis describes a candidate where drug A (in domain X) may modulate outcome C (in domain Y) through a shared molecular mediator B. All hypotheses include:- Full evidence chain (PubMed IDs for both A→B and B→C legs)- Mediator type and predicate vocabulary- Outreach quality classification (high/mid)- Clinical trial status (active/completed/preclinical/none) via ClinicalTrials.gov v2 API- Literature status (novel signal/partial evidence/reproduces known) via PubMed E-utilities Source corpus: 99,132 papers, 47,255 passing biomedical relevance filter (PubTator), 140,178 structured claims extracted via DeepSeek V4 Flash. The catalog is intended as a starting point for researcher-led investigation. Hypotheses are not validated experimentally and must not influence clinical care. See robertium.com/method for methodology and limitations. Schema version: 1.2Format: JSONGenerated: 2026-05-10 Source code: https://github.com/routewise96/robertiumWeb interface: https://robertium.com