M3D

M3D identifies statistically significant differentially methylated regions (DMRs) of CpG sites from high-resolution DNA methylation data generated by sequencing technologies to assess epigenetic modifications.

Key Features:

• Kernel-Based Methodology: M3D employs a non-parametric, kernel-based Maximum Mean Discrepancy (MMD) approach to measure differences in methylation profiles and detect higher-order changes such as alterations in the shape of methylation patterns across predefined genomic regions.
• Accounting for Coverage Variation: M3D explicitly incorporates differences in coverage levels between samples into its statistical testing procedure to control for confounding due to variable read coverage.
• Spatial Correlation Consideration: M3D integrates spatial correlations among CpG sites directly into the test statistic to improve accuracy in identifying DMRs.

Scientific Applications:

• Cancer biology: Detection of DMRs to study epigenetic alterations associated with tumorigenesis and cancer progression.
• Developmental biology: Analysis of methylation changes during development to investigate epigenetic regulation of gene expression.
• Epigenetic disease mechanisms: Investigation of disease-associated epigenetic mechanisms by linking methylation changes to phenotypic variation and gene expression.

Methodology:

The computational approach uses a non-parametric kernel-based Maximum Mean Discrepancy (MMD) test statistic, explicitly accounts for coverage variation, integrates spatial correlations among CpG sites, and has been evaluated with empirical tests on real and simulated datasets demonstrating increased sensitivity and robustness to coverage and replication variability.