Robust prediction of drug combination side effects in realistic settings.

DCSE predicts drug combination side effects with over 30% higher accuracy than existing methods in realistic, imbalanced settings, improving safety assessments in healthcare.

• Why it matters: Limited knowledge of polypharmacy side effects hampers safe drug prescribing, especially since clinical trials often miss rare or long-term adverse effects, risking patient safety.
• What they did: The team developed DCSE, a machine learning model that learns latent signatures for drugs, pairs, and side effects, and evaluated it using both traditional and realistic prospective datasets, including warm- and cold-start scenarios.
• The result: DCSE consistently outperforms current state-of-the-art approaches, demonstrating robustness and enabling more reliable prediction of side effects for uncharacterized drug combinations in real-world settings.