Drug repurposing — identifying new therapeutic uses for existing compounds — offers a dramatically shorter path to clinical proof of concept than de novo drug discovery. The compound's safety profile in humans is already characterised, formulation development is largely complete, and manufacturing processes are established. The challenge is identifying repurposing hypotheses that are biologically plausible, clinically significant, and not already being pursued by a competitor. Indication knowledge graphs enable systematic hypothesis generation that is simply not possible through literature review alone.
The Logic of Repurposing Hypotheses
Repurposing hypotheses are generated by reasoning over relationships in the indication knowledge graph: if drug A modulates target T, and target T is genetically associated with disease D, then drug A is a candidate for disease D — particularly if drug A has not previously been studied in disease D, and if the genetic evidence for T in D is strong and directional. More complex hypotheses involve multi-hop reasoning: drug A is approved for disease X; disease X shares a pathological mechanism with disease Y (as represented in the disease ontology); drug A's mechanism is expected to be active in that shared pathway; therefore drug A is a hypothesis for disease Y. Systematic evaluation of thousands of such hypotheses across a large drug portfolio and disease ontology can surface prioritised candidates in hours.
Evidence Triangulation for Hypothesis Validation
Computationally generated repurposing hypotheses must be validated against multiple independent evidence sources before resources are committed to experimental follow-up. A knowledge graph that integrates genetic association, pathway membership, protein interaction network proximity, literature co-occurrence, and existing clinical use data enables multi-evidence triangulation: a hypothesis supported by convergent evidence from genetic, mechanistic, and observational sources is substantially more likely to succeed than one supported by a single evidence type. This structured validation process replaces the largely informal "do we believe this?" discussion that currently gates repurposing decisions in most discovery organisations.
Portfolio-Level Repurposing Strategy
At the portfolio level, repurposing knowledge graphs enable strategic questions that individual hypothesis evaluation does not address: which of our clinical-stage compounds have the broadest repurposing opportunity based on their mechanism? Which disease areas have the highest density of validated target-disease associations where we have compounds with relevant mechanisms? Which repurposing hypotheses in our portfolio have the strongest competitive position based on the depth of our existing clinical evidence? These portfolio-level queries require the same knowledge graph infrastructure as individual hypothesis generation — but the strategic value of the portfolio view typically exceeds the individual opportunity analysis.
