ReMeDy: A Flexible Statistical Framework For Region-based Detection of DNA Methylation Dysregulation

Area of Science:

• Epigenetics
• Genomics
• Statistical Bioinformatics

Background:

• Region-based epigenome-wide association studies (eGWAS) offer advantages over probe-wise analyses for DNA methylation data.
• Current region-based methods primarily focus on mean methylation changes, neglecting alterations in methylation variability.

Purpose of the Study:

• To develop a statistical framework that jointly models mean and variance changes in DNA methylation within regions.
• To identify differentially methylated regions (DMRs), variably methylated regions (VMRs), and regions with joint differential and variable methylation.

Main Methods:

• Proposed ReMeDy, a flexible framework using a hierarchical likelihood approach within a generalized linear model.
• Operates directly on biologically defined co-methylated regions to capture spatial correlation.
• Avoids heuristic tuning parameters, reducing subjectivity.

Main Results:

• ReMeDy maintains false discovery and type-I error rates at nominal levels in simulations.
• Consistently achieves higher statistical power across various realistic scenarios compared to existing models.
• Identifies biologically meaningful regions and pathways implicated in complex human diseases from population-level data.

Conclusions:

• ReMeDy provides a robust method for analyzing DNA methylation variability and mean changes in a region-based manner.
• Offers improved biological interpretability and statistical power for eGWAS.
• Identifies novel disease-associated epigenetic markers missed by conventional mean-based analyses.