Analyzing Data Breach Factors Through Modern Crime Theory: A Structural Equation Modeling Approach

Adverse Drug Reactions (ADRs) caused by drug-drug interactions (DDIs) pose a serious global health problem and are a determinant of high morbidity, mortality, and healthcare costs. With the complexity of therapeutics escalating in tandem with population aging, ADR identification before availability of drugs for sale is an ongoing concern. Traditional methods, such as clinical trials, rarely identify sparse ADRs due to their small sample sizes and heterogeneity in patients. Filling this lacuna, we present an innovative method that leverages the power of Graph Neural Networks (GNNs) and Self-Supervised Learning (SSL) to model drug-drug interactions at the molecular level. Our method maps drugs into molecular graphs and incorporates spatial and physical attributes and predicts potential adverse interactions.We compare our approach on the TwoSIDES Polypharmacy dataset of Therapeutic Data Commons (TDC) and achieve state-of-the-art performance with 90% accuracy and 75% precision. We also conduct a case study on DrugBank dataset, for which the model does best in the prediction of interaction types, with 99% accuracy,precision, and F1-score. Our results establish a solid and scalable framework for predicting ADRs, paving the way for future applications of GNNs in pharmacovigilance and molecular biology