We propose to leverage the UK Biobank’s rich, multimodal data-combining genetic, imaging, proteomic, and traditional clinical/lifestyle information-to build integrative, individualized disease!risk models. Specifically, we will:
Genotype & Polygenic Risk Scores: Compute PRS for major complex diseases (e.g., coronary artery disease, type 2 diabetes) using imputed array data and published GWAS summary statistics.
Imaging Biomarkers: Extract quantitative features from brain and cardiac MRI (e.g., regional volumes, tissue composition) via deep!learning pipelines, linking structural and functional variation to genetic risk.
Proteomic Profiles: Incorporate plasma protein measurements (Olink, SomaScan) to capture circulating biomarkers, using penalized regression to integrate high!dimensional proteomic data with PRS and imaging features.
Clinical & Lifestyle Covariates: Adjust for age, sex, BMI, smoking, alcohol use, physical activity, dietary factors, and socioeconomic status.
We will fit Cox proportional hazards and logistic!regression models with main effects and interaction terms (e.g., PRS × imaging biomarker; PRS × protein level) to quantify how genetic predisposition is modified by molecular and imaging phenotypes. Model performance (AUC, C-index) will be assessed via nested cross!validation and tested in held!out subcohorts.
This integrative framework will reveal novel endophenotypes-where genetics shape tissue structure and circulating proteins-and identify high!risk subgroups who may benefit from tailored prevention or early interventions. By elucidating genotype × imaging × proteome interactions, our study aims to advance precision medicine and inform targeted screening strategies.