Research Questions:
1. What are the interactive effects of polygenic risk, environmental exposures, and metabolic abnormalities on the onset and progression of cardiovascular disease, cancer, diabetes, and respiratory diseases?
2. Can time-aware machine learning models, integrating longitudinal clinical, environmental, and multi-omics data, improve early prediction and risk stratification for multiple chronic diseases?
3. What are the common and disease-specific biological pathways mediating shared risk exposures across these disease domains?
Research Objectives:
1. Data Integration & Risk Characterization:
a. Integrate multimodal data from UK Biobank: genetic (polygenic risk scores), environmental (pollution, diet, lifestyle), and metabolic (glycemic variability, lipidomics).
b. Characterize independent and joint effects of these exposures on disease-specific and multimorbid outcomes using statistical and interaction models.
2. Predictive Model Development:
a. Develop deep learning models to predict disease onset and multimorbidity progression.
b. Use time-series clinical data, omics, and imaging to train RNNs, survival forests, and multimodal models.
c. Compare model performance with traditional risk stratification tools.
3. Mechanistic Exploration:
a. Identify shared biological mechanisms through integrated multi-omics analysis.
b. Discover modifiable molecular targets relevant across disease types.
Scientific Rationale:
Chronic diseases share common biological and environmental drivers yet are often studied in isolation. Air pollution, metabolic dysfunction, and genetic risk contribute across multiple diseases, but their combined, time-dependent effects remain poorly understood. Traditional models overlook dynamic risk trajectories. This study leverages longitudinal, high-dimensional data and machine learning to uncover shared mechanisms, predict early disease onset, and support precision prevention.
