Non-communicable chronic diseases (NCDs) drive 74% of global mortality through dynamic gene-environment (G×E) interactions that existing models inadequately capture. Three critical gaps persist: Polygenic risk scores (PRS) ignoring lifespan-modulated G×E (e.g., deprivation exacerbating inflammation); static biomarkers missing metabolomic drift signaling transitions; fragmented multimorbidity studies. Leveraging UK Biobank’s longitudinal multi-omics, imaging, and health data, this study pioneers a lifespan exposome approach to decode NCDs.
Aims:
1.How do temporal exposures (e.g., pollution, stress) interact with genetic risks via shared pathways (e.g., epigenetics) to drive multimorbidity?
2.Can longitudinal omics trajectories (e.g., lipidome) + digital phenotypes (e.g., actigraphy) outperform static models in predicting preclinical transitions?
3.Do critical windows exist where lifestyle interventions disrupt genetic risk in high-risk subgroups?
Objectives:
1.Lifespan Exposure Cartography: Apply functional survival analysis to UKB multimodal data to quantify time-sensitive G×E and identify modifiable factors with nonlinear exposure-response curves.
2.Multimodal Predictive Architecture: Develop temporal graph neural networks integrating longitudinal omics, imaging, and behavioral data. Validate against Cox models using time-dependent AUC-ROC and multimorbidity incidence.
3.Equity-Aware Validation: Test biomarker generalizability via stratified cross-validation (e.g., ethnic subgroups). Simulate precision interventions using counterfactual Mendelian randomization (e.g., LDL-PRS × diet effects).
