Contrastive learning of adverse events to provide effective and interpretable vector representations for machine-assisted pharmacovigilance.

Contrastive learning produces interpretable adverse event vectors that outperform existing methods in drug-event association prediction with an AUC of 0.88.

• Why it matters: Effective pharmacovigilance relies on analyzing real-world drug safety data, yet current tools often depend on text-based models like large language models, which may lack interpretability and real-world relevance.
• What they did: The authors adapted contrastive learning algorithms to generate numerical representations from spontaneous adverse event reports, enabling better capture of functional and causal relationships in drug safety data.
• The result: These vector representations improve signal detection and causality assessment, outperforming traditional methods and LLM-based features, thus supporting more accurate and interpretable pharmacovigilance decision-making.