Multi-scale digital twins for personalized medicine

The future of personalized medicine requires moving beyond isolated data streams toward integrated, multi-scale representations of human health. Digital twins (DTs) have emerged as promising solutions, offering dynamic, individualized simulations of biological systems. However, current implementations often rely on narrow data sources, limiting predictive power, adaptability, and clinical utility. The next generation of digital twins must integrate molecular, cellular, tissue, organ, clinical, behavioral, and environmental data to accurately model health trajectories and disease evolution. This review synthesizes the conceptual foundations, technical architectures, clinical applications, and ethical challenges associated with multi-scale digital twins (MSDTs). Key enabling technologies include multimodal data fusion, graph neural networks, causal inference frameworks, reinforcement learning, and hybrid mechanistic-AI modeling approaches. Clinical applications can illustrate the potential of MSDTs to personalize interventions dynamically. Significant barriers persist regarding data integration, ethical governance, bias mitigation, and regulatory adaptation.