We expect to combine traditional clinical risk factors and UK Biobank multi-omics data to accurately predict disease risk, diagnosis, and inform treatment of complex traits/diseases.
Construct a cross-omics data integration framework for biomarker discovery and risk prediction, validate the “gene-protein-metabolite-imaging phenotype” cascade regulatory pathway and identify key nodes in common chronic diseases, and develop algorithms to integrate omics and imaging data for early chronic disease screening and warning.
Scientific rationale
1. The complexity of disease mechanisms requires multidimensional analysis
The combination of multi omics integration and multimodal imaging can reveal the cross-scale association between gene regulatory networks, metabolic pathway abnormalities, and organ dysfunction. Systems biology models based on multi omics can identify cross omics “driving factors” and map molecular features to imaging phenotypes, thereby resolving disease subtype heterogeneity. For example, in predicting lung cancer recurrence, the combined analysis of EGFR gene mutations and tumor metabolic activity can improve the accuracy of prognostic models.
2. Modeling requirements for dynamic interactive networks
Disease occurrence is a process of dynamic imbalance between molecules, cells, and tissues. Multi omics time-series data can construct time-dependent causal networks to identify early warning markers. For example, metabolomics data can predict the decline track of renal function in patients with diabetes nephropathy 6-12 months in advance, and renal diffusion tensor imaging can verify the spatial-temporal relationship between metabolic abnormalities and renal fibrosis.
Radiomics can reveal the spatial heterogeneity mechanism of tumor microenvironment by high-throughput extraction of lesion texture and shape features, integrated with genomic copy number variations or epigenetic modifications such as methylation.
