Artificial Intelligence–based Prediction of Drug– Drug Interactions in Polypharmacy Patients: Current Advances and Future Perspectives

Polypharmacy, typically defined as the concurrent use of five or more medications, is increasingly prevalent among elderly and multi-morbid populations, substantially elevating the risk of drug–drug interactions (DDIs), adverse drug reactions, hospitalization, and therapeutic failure. Traditional DDI detection strategies, largely dependent on retrospective analyses and rule-based alert systems, are inadequate to address the combinatorial complexity and patient-specific variability inherent to modern pharmacotherapy. Artificial intelligence (AI) has emerged as a powerful paradigm for DDI prediction, leveraging large-scale, heterogeneous datasets including electronic health records, pharmacogenomic profiles, and drug interaction databases. This review critically examines current AI-driven methodologies—encompassing machine learning, deep learning, graph neural networks, and natural language processing—for the identification of both established and previously unrecognized DDIs in polypharmacy settings. These approaches enable high-dimensional data integration, facilitate real-time risk stratification, and support personalized therapeutic decision-making. In addition, the role of real-world evidence, explainable AI (XAI), and hybrid modeling frameworks in improving model interpretability, clinical trust, and translational applicability is discussed. Despite significant advances, key challenges persist, including data heterogeneity, algorithmic bias, limited modeling of multi-drug interactions, and regulatory and ethical constraints. Overall, AI-driven DDI prediction represents a transformative advancement in medication safety for polypharmacy patients. Future efforts should prioritize the development of robust, interpretable, and clinically validated AI systems integrated within decision support platforms to enable safe, scalable, and patient-centric healthcare delivery.