Finding Adverse Drug Reaction Side Effects using Graph Neural

Abstract

Adverse drug responses (ADRs) caused by drug-drug interactions are a major health concern. Graph Neural Networks (GNNs) are effective at modeling these relationships, but their onedimensional processing limits their ability to extract complex features. This work offers a novel expansion that incorporates a two-dimensional Convolutional Neural Network (CNN2D) to enhance ADR prediction. By converting drug interaction data into 2D matrices and capturing intricate spatial connections, CNN2D improves the graph-based insights of the GNN. Our hybrid model performs better than more traditional methods like KNN and Decision Trees, with a prediction accuracy of 99.87%. The extension shows how deep learning might improve the evaluation of drug safety.