RareGraph provides an AI‑driven knowledge graph that consolidates fragmented biomedical data on rare diseases into a searchable, evidence‑backed network. By continuously ingesting papers, trial data, and registries, its reader, curator, and linker agents extract entities such as genes, therapies, and biomarkers, assign confidence scores, and present answers with full provenance—currently demonstrated with Duchenne Muscular Dystrophy. This enables researchers and clinicians to navigate relationships, compare interventions, and make data‑driven decisions without manual literature mining.
Funding
Funding not disclosed
Founders
Product
Problem
Rare disease research suffers from fragmented data spread across publications, trial registries, and disparate databases, requiring extensive manual effort to locate and verify information. This dispersion leads to low confidence in findings, duplicated work, and slow discovery of therapeutic insights.
Solution
RareGraph provides an AI‑driven knowledge graph that continuously ingests papers, clinical trial records, and registry data to create a unified, searchable network of entities such as genes, therapies, biomarkers, pathways, and outcomes. Specialized AI agents extract and standardize entities, map them to ontologies, and link them with evidence‑backed relationships, each assigned a confidence score. Users can query the graph in natural language and receive answers that include citations, evidence snippets, and exportable tables, enabling rapid, transparent navigation of rare disease knowledge with full provenance.
Target Audience
Primary users are researchers, clinicians, and drug development teams focused on rare diseases who need a reliable, up‑to‑date evidence base for hypothesis generation and therapeutic decision‑making.
Features
- Continuous ingestion of biomedical literature, trial data, and registry entries with automated entity extraction
- Mapping of entities to standard ontologies (e.g., HPO, MONDO) for consistent terminology
- Construction of a knowledge graph where every node and edge carries evidence citations and confidence scores
- AI agents that assess study quality, flag confounders, and generate reliability metrics for each relationship
- Natural‑language query interface delivering cited answers, evidence excerpts, and downloadable tables
- Interactive graph visualization allowing users to explore, filter, and rearrange entities and their connections