PREDICTION OF SIDE EFFECTS OF DRUGS USING GRAPH NEURAL NETWORKS

This project focuses on predicting adverse drug reactions (side effects) using a Graph Neural Network (GNN) approach. Traditional methods for identifying drug side effects are time-consuming and expensive, and they often fail to detect rare or complex reactions early. To address this, the proposed system uses computational techniques to analyze biomedical data efficiently. A knowledge graph is constructed by integrating multiple datasets, including drug names, side effect data (MedDRA), drug–disease relationships, drug–symptom associations, and drug action labels. In this graph, drugs, diseases, and side effects are represented as nodes, while their relationships are modeled as edges. A Graph Convolutional Network (GCN) is applied to learn patterns from this graph structure. The model generates node embeddings and performs link prediction to estimate the probability of potential side effects for a given drug. Additionally, the system classifies drugs based on their biological action (e.g., bacteria, virus, fungi, or human cell targets). The model demonstrates improved performance compared to traditional machine learning approaches by effectively capturing complex relationships in biomedical data. This approach can support early detection of drug risks, improve patient safety, and assist in more efficient drug development.