TWIN-X: Multimodal Digital TWINs with Generative AI for eXplainable Precision Medicine

TWIN-X – Multimodal Digital TWINs with Generative AI for eXplainable Precision Medicine is a four-year Horizon Europe project developing trustworthy and explainable digital patient twins for precision medicine. The project brings together multimodal information from clinical narratives, medical imaging, laboratory data and pathology to create modular representations of individual patients. These digital twins will support researchers and clinicians in exploring disease trajectories, comparing possible interventions and performing “what-if” analyses, while keeping predictions traceable to the underlying clinical evidence.

At BHT, the Data Science +X research group takes a leading role in Work Package 4. Our work focuses on methods for building modular, clinically coherent and explainable AI components from heterogeneous health data. In particular, we investigate how generative AI and representation-learning methods can structure and combine clinical information while preserving its medical meaning, and how predictions can be accompanied by faithful explanations, calibrated uncertainty and transparent links to supporting evidence. The resulting methods will contribute to trustworthy digital-twin systems that can be evaluated in oncology and cardiovascular research.