Pharmacogenomics and Personalized Medicine: Leveraging AI and Machine Learning for Optimized Therapies

Abstract

To examine how artificial intelligence (AI) and machine learning (ML) can advance pharmacogenomics and personalized medicine by improving drug-response prediction, treatment selection, dosing, and clinical decision support, while identifying the major barriers to clinical implementation. This narrative review critically synthesizes current applications of AI-enabled pharmacogenomics across oncology, cardiovascular medicine, psychiatry, and other therapeutic areas. It evaluates classical ML, deep learning, transformer models, unsupervised learning, federated learning, explainable AI, knowledge graphs, multimodal data integration, and electronic health record–based clinical decision support. Clinical applications, implementation requirements, ethical concerns, regulatory issues, and future research priorities are comparatively examined. AI and ML can integrate genomic, multi-omic, clinical, environmental, and longitudinal data to identify gene–drug associations, predict therapeutic efficacy and adverse drug reactions, optimize individualized dosing, and support genotype-guided prescribing. Evidence from oncology, cardiology, and psychiatry indicates that these approaches may improve patient stratification and treatment selection. Their integration into electronic health records and clinical decision support systems can enable timely, actionable recommendations. However, clinical translation remains constrained by limited and ancestry-biased datasets, inconsistent phenotype definitions, missing data, overfitting, class imbalance, weak external validation, limited interpretability, workflow incompatibility, clinician training gaps, privacy risks, algorithmic bias, and evolving regulatory requirements. AI-enabled pharmacogenomics has potential to support safer, more effective, and individualized therapy. Broad clinical adoption requires diverse standardized datasets, externally validated and explainable models, interoperable health information systems, ethical and regulatory governance, clinician education, and interdisciplinary collaboration. Federated learning, multimodal architectures, and continuous-learning systems may enable equitable, real-time pharmacogenomic decision-making.