Pharmacology Knowledge Graphs: Do We Need Chemical Structure for Drug Repurposing?

Researchers developed a pharmacology knowledge graph for drug repurposing and found that removing chemical structure representations improved performance while dramatically reducing computational requirements. The study showed that drug behavior can be accurately predicted using only target protein information and network topology, with larger datasets proving more valuable than complex models.

• →Removing chemical structure encoders improved drug-protein prediction accuracy from 0.5631 to 0.5785 PR-AUC while reducing memory usage from 5.30 GB to 353 MB.
• →The study used rigorous temporal validation with training data through 2022 and testing on 2023-2025 data from a knowledge graph of 5,348 entities.
• →Increasing model size beyond 2.44 million parameters showed diminishing returns, while larger training datasets consistently improved performance.
• →External validation confirmed 6 of 14 top novel drug repurposing predictions as established therapeutic indications.
• →Target-centric information and network topology alone can accurately predict pharmacological behavior without explicit chemical structure data.