1.Research Questions: (1) What are the synergistic or antagonistic interactions between polygenic risk scores (e.g., T2D-GRS, CAD-PRS), neighborhood deprivation indices, dynamic clinical trajectories (e.g., HbA1c variability), and multi-organ imaging biomarkers (e.g., hepatic steatosis, epicardial fat thickness) in determining risks of heterogeneous cardio-metabolic-cerebro-renal outcomes, including but not limited to diabetes, NAFLD/NASH, chronic kidney disease, stroke, and vascular dementia? (2) Can a time-aware deep learning framework integrating baseline and longitudinal multi-modal data improve prediction of composite adverse metabolic events compared to traditional static models? 2. Objectives:! Integrate genetic, socioeconomic, clinical, and imaging data from the UK Biobank to establish a comprehensive dataset for studying metabolic health. Identify key factors and their interactions that contribute to different metabolic health states using advanced statistical and machine learning methods. Develop and validate a risk stratification model that incorporates multi-dimensional data to predict an individual’s likelihood of developing specific metabolic health states. Explore the potential of the developed model for guiding personalized prevention and intervention strategies for metabolic disorders. 3. Scientific Rationale:
The escalating burden of interconnected metabolic diseases demands a paradigm shift from organ-specific to systemic risk prediction. Current siloed approaches fail to capture shared mechanisms. UK Biobank’s unique combination of whole-body MRI , cognitive assessments, and geospatial pollution data enables modeling of pan-metabolic interactomes. By integrating diverse data types and leveraging advanced analytical methods, this study aims to develop a more comprehensive and accurate risk stratification model for metabolic health states.
