MethylPipeR

MethylPipeR predicts 10-year type 2 diabetes (T2D) risk from blood-based DNA methylation data using statistical and machine-learning models to enhance incidence prediction and risk stratification.

Key Features:

• Multi-Omics Integration: Integrates blood-based DNA methylation with other omics layers to refine prediction of T2D incidence and improve predictive accuracy over standard risk factors.
• Flexible Modeling Approach: Implements linear and tree-ensemble models with incorporation of time-to-event data to model incident T2D.
• Improved Predictive Performance: In the Generation Scotland cohort it reported AUC=0.872 and PRAUC=0.302 for 10-year T2D onset versus standard risk factors with AUC=0.839 and PRAUC=0.227.
• Reproducibility: Predictive models were validated in independent cohorts, including replication in the KORA study (p = 1.6×10^-5).

Scientific Applications:

• Complex trait prediction: Generation of predictive models for T2D incidence using DNA methylation and other omics data.
• Biomarker discovery: Identification of methylation signatures associated with future T2D risk to inform biological insights into disease mechanisms.
• Translational risk stratification: Development of risk stratification approaches to support earlier diagnosis and personalized intervention strategies for T2D and other complex traits.

Methodology:

Processes blood-based DNA methylation data alongside other omics information and applies linear and tree-ensemble models with time-to-event (survival) modeling to predict incident T2D.