AI DRIVEN DIGITAL TWINS FOR DRUG REPURPOSING: CURRENT EVIDENCES, APPLICATIONS, TRANSLATIONAL CHALLENGES AND FUTURE DIRECTIONS

Conventional drug discovery and development is associated with substantial financial costs, lengthy development timelines, and high attrition rates during preclinical and clinical evaluation. Drug repurposing provides an alternative strategy by identifying new therapeutic applications for existing drugs with previously characterized pharmacological and safety profiles. Recent advances in artificial intelligence, multimodal biomedical data integration, mechanistic modelling, and digital twin technology have created new opportunities to improve computational drug repurposing and patient-specific treatment prediction. This review examines the emerging integration of AI-driven digital twins with drug repurposing, focusing on digital twin architectures, artificial intelligence methodologies multi modal data integration, drug target prediction, pharmacokinetic or pharmacodynamic modelling, treatment-response simulation, and personalized therapeutic decision-making. The review also discusses applications across cancer, neurodegenerative, cardiovascular, infectious, metabolic, and rare diseases. Particular emphasis is placed on the development of patient-specific digital twins, integration of multi-omics and clinical data, virtual treatment simulation, and validation through computational, experimental, and clinical approaches. Major challenges include data heterogeneity, interoperability, model reliability, uncertainty quantification, interpretability, algorithmic bias, cybersecurity, privacy, clinical workflow integration, and regulatory acceptance. Emerging approaches, including causal AI, federated learning, multimodal foundation models, virtual clinical trials, and real-time adaptive digital twins, may further enhance the translational potential of AI-enabled drug repurposing. However, prospective clinical validation and standardized regulatory frameworks remain necessary before widespread clinical implementation.