Causal Reasoning in Medical Digital Twins: Methods and Architectures

Overview

Medical digital twins (MDTs) offer a promising paradigm for personalized medicine. The proposed framework, DeepCARES-DT, integrates genetic, clinical, and knowledge-based components within a unified multi-layered architecture. It combines risk aggregation from genetic data, causal discovery from clinical biomarkers, and large-scale knowledge graph integration, enabling structured representation of biomedical evidence. A causal-oriented reasoning agent further supports interpretable exploration and explanation of these relationships, while graph-based models incorporate causal structure into clinical prediction.

Validation on real-world clinical data demonstrates improved predictive performance. Overall, this work provides a computational blueprint for developing interpretable, causally informed AI systems for personalized risk assessment and precision medicine.