An Overview of Unsupervised Machine Learning in Drug Discovery, Pharmacogenomics, and Pharmacovigilance

As biomedical data continue to grow in complexity and often lack annotation, unsupervised machine learning (UML) has emerged as a powerful approach for uncovering hidden patterns and enabling data-driven discovery in pharmacology. This review examines the expanding role of UML in drug discovery, pharmacogenomics, and pharmacovigilance. Core techniques such as clustering, dimensionality reduction, and generative modeling support hypothesis-free analysis and contribute to translational insight. In drug discovery, UML enhances molecular representation, facilitates lead optimization, and guides scaffold development. Graph neural networks (GNNs) are emphasized for their ability to capture complex structural and chemical features that improve drug–target interaction prediction and de novo molecular generation. In clinical contexts, UML enables stratification of pharmacogenomic profiles and supports early detection of adverse drug events through anomaly detection and natural language processing applied to real-world data. By integrating insights from both algorithmic development and real-world applications, this review underscores the growing value of UML as a pivotal tool in advancing contemporary pharmacological research and data-driven decision-making.