Research Questions:
1.What genetic, metabolic, and environmental factors drive the onset and progression of pancreatic cancer, cholangiocarcinoma, and complex biliary stones?
2.Can integrated multi-modal data improve early diagnosis and prognosis prediction for these diseases?
3.What molecular mechanisms underlie their pathogenesis?
Objectives:
1.Risk Identification: Identify demographic, genetic (GWAS-derived polygenic risk scores), and lifestyle factors (e.g., obesity, diabetes) linked to disease incidence and subtypes.
2.Prediction Models: Develop machine learning algorithms using clinical, genetic, and imaging data to predict disease risk, postoperative outcomes, and recurrence.
3.Mechanistic Insights: Uncover disease-associated pathways (e.g., bile acid metabolism, Wnt/!-catenin signaling) and microbiome interactions via multi-omics integration.
Methodology:
Data Sources: UK Biobank’s genetic data, serum biomarkers, imaging (MRI/ultrasound), and stool metagenomics.
Analytic Approaches:
Machine Learning: XGBoost/neural networks for risk stratification.
Genetic Analyses: GWAS, Mendelian randomization, and pathway enrichment.
Microbiome-Disease Associations: Multivariate models adjusting for diet/medication.
Validation: Temporal splitting (training/validation cohorts) and external replication.
Scientific Rationale:
These diseases are lethal due to late diagnosis, heterogeneous biology, and limited therapies. UK Biobank’s prospective design enables causal inference by analyzing pre-diagnostic samples, while its multi-modal data (genomics, imaging, microbiome) allows disentangling complex gene-environment interactions-unachievable in smaller cohorts. For example, longitudinal biomarker data will clarify whether metabolic dysregulation (e.g., hyperglycemia) precedes cancer development or results from it.
