Beyond Prediction: Mechanism-Guided Artificial Intelligence for Drug Repurposing and Translational Pharmacology

Drug repurposing offers an efficient strategy for identifying new therapeutic applications for existing medicines. Advances in artificial intelligence (AI), machine learning, and systems pharmacology have enabled the analysis of complex biological and clinical datasets to identify potential drug–disease associations. However, conventional predictive models often lack mechanistic interpretability, limiting their translational reliability. Mechanism-guided AI integrates computational predictions with molecular interactions, signaling pathways, multi-omics data, causal inference, and biological networks to improve the identification and prioritization of repurposable drugs. This review examines key computational approaches, including knowledge graphs, explainable AI, molecular docking, network-based drug discovery, and causal modeling, highlighting their applications across major disease categories. It also discusses the integration of computational findings with experimental pharmacology, pharmacokinetic–pharmacodynamic assessment, biomarker discovery, patient stratification, and clinical trial development. Furthermore, the review evaluates challenges associated with data heterogeneity, model bias, causal validation, drug safety, reproducibility, and regulatory acceptance. Integrating real-world evidence, pharmacovigilance data, and patient-specific molecular profiles may further strengthen therapeutic predictions and support precision medicine. Future progress requires transparent reporting, standardized benchmarking, rigorous experimental validation, and interdisciplinary collaboration. Mechanism-guided AI may bridge computational drug discovery and clinical pharmacology by generating biologically plausible, experimentally testable, and clinically relevant therapeutic hypotheses, ultimately supporting more efficient and evidence-based drug repurposing.