Research Questions:
(1) Can integrating multi-modal data (e.g., genomics, proteomics, metabolomics, imaging, clinical records, diet, and environmental exposures) significantly improve sensitivity and specificity of risk prediction models compared to single-modality approaches?
(2) Are there distinct molecular and phenotypic subgroups (stratification) within complex diseases that correlate with differential progression or therapeutic response?
(3) Can we identify novel, robust biomarkers or multi-omics signatures enabling earlier disease detection and serving as actionable targets for drug discovery?
Objectives:
(1) Data Integration: To construct a harmonized framework integrating high-dimensional omics with phenotypic, demographic, lifestyle, and environmental data from UK Biobank.
(2) Biomarker Discovery & Stratification: To develop responsible ML/AI methods identifying multi-modal signatures linked to disease onset, progression, and severity, and stratifying individuals into clinically meaningful sub-populations (e.g., high-risk groups, responders) for precision diagnostic and therapeutic decision-making.
(3) Therapeutic Insight: To map stratified profiles against known drug targets to uncover novel therapeutic opportunities and repurposing candidates.
Scientific Rationale:
Complex diseases are rarely driven by a single factor. Traditional analyses often study genomic or clinical data in isolation, missing the interplay between genetics, metabolism, lifestyle, and environment. UK Biobank provides a unique resource to bridge this gap. By applying multi-modal integration, we aim to capture non-linear relationships between diverse data layers. This approach enables a holistic view of human health, moving beyond “one-size-fits-all” medicine toward precision care. The research will specifically focus on reducing false positives (specificity) and capturing early-stage disease (sensitivity).