Artificial intelligence in pharmacogenomics: From data integration to clinical decision support

Artificial intelligence (AI) has shown potential to support pharmacogenomics (PGx), advancing the broader precision-medicine research agenda. This narrative review provides a comprehensive overview of the latest progress in AI technologies applied to PGx, including traditional machine learning, deep learning, graph neural networks, and large language models. We analyze how AI integrates genomic and multi-omics data to predict drug efficacy and toxicity and to support genotype-guided clinical decision-making; work on de novo drug discovery is cited only as methodological analogy and lies outside the scope of this review. These technologies are positioned to support the transition of PGx from traditional reactive single-gene studies into a systems-based precision medicine paradigm. However, challenges remain, such as data scarcity, algorithmic bias, limited model interpretability, and difficulties in clinical integration. Future efforts should focus on developing interpretable and ethically aligned AI frameworks, rigorously validated through prospective clinical trials, to promote their safe and effective clinical adoption. The review also positions Clinical Pharmacogenetics Implementation Consortium (CPIC), Dutch Pharmacogenetics Working Group (DPWG), Pharmacogenomics Knowledgebase (PharmGKB), and Pharmacogenomics Clinical Annotation Tool (PharmCAT) as the operational infrastructure of clinical PGx and directly examines star-allele calling, phenoconversion, electronic health record (EHR) integration, and alert fatigue as barriers to implementation.