GNN – Enhanced Drug Interaction Detection and Adverse Reaction Analysis with Sentiment Insights

This study presents a novel approach for identifying Drug-Drug Interactions (DDIs) and forecasting Adverse Drug Reactions (ADRs) by integrating Graph Neural Networks (GNNs) with Sentiment Analysis. Unlike conventional models that primarily focus on molecular interactions, our method integrates molecular data with real-world patient-reported outcomes to provide a more comprehensive drug safety assessment. By leveraging GNNs to model complex drug relationships and analyzing sentiment from the FDA's Adverse Drug Reaction database, our framework enhances predictive accuracy. Experimental results demonstrate that our model outperforms existing approaches, achieving an AUC of 0.981, an AUPR of 0.983, and an accuracy of 0.997. This integration not only improves the detection of DDIs but also quantifies their severity more effectively, offering a robust and clinically relevant tool for improved drug safety and decision-making.