Machine Learning Algorithms for Predicting Drug Efficacy and Adverse Effects Based on Individual Genetic Profiles

ADRs, or adverse drug reactions, are a chronic problem in drug development and patient safety, frequently resulting in expensive clinical failures and serious health outcomes. Conventional detection techniques have limited scalability and accuracy, requiring sophisticated computational methods. This research examines Machine learning (ML) models' effectiveness in forecasting drug responses and linked ADRs by comparing three popular models: Deep Neural Network, Random Forest, and Support Vector Machine (SVM). A systematic data collection process was adopted, including clinical and molecular datasets. Preprocessing involved the use of SMOTE to overcome class imbalance, and statistically relevant features were selected before model training. Performance measures showed that DNN was the best performer, with maximum accuracy (92.5%) and AUC-ROC (95.2%), though RF provided an optimal trade-off between accuracy and computational overhead. SVM showed high precision but with low recall. Deep models took longer to train, though they performed better in detecting intricate patterns.