A knowledge graph represents entities and typed relationships as a queryable structure. In drug discovery, it combines biological, clinical, and literature evidence so computational methods can surface connections for expert review. section 13.2.3 section 13.4.1
A practical drug-discovery knowledge graph can support:
- Target discovery by linking genes, diseases, proteins, pathways, patient measurements, genetic associations, and publications. section 13.2.3
- Relationship extraction by applying natural language processing to identify gene–disease associations, protein interactions, and pathway dysregulation. section 13.2.3
- Hypothesis generation by querying large collections of biological and clinical relationships, then ranking candidate explanations or repurposing opportunities. section 13.4.1
Repurposing baricitinib
BenevolentAI used a knowledge graph containing 350 million biological and clinical relationships to identify baricitinib as a potential COVID-19 treatment. The graph accelerated discovery of a connection that would be difficult to find through literature review alone, while expert analysis remained part of the process. section 13.4.1














