DMRScan

DMRScan implements an R/Bioconductor scan-statistic framework to detect differentially methylated regions (DMRs) in genome-wide DNA methylation studies while controlling genome-wide multiple testing and false discovery rates.

Key Features:

• R/Bioconductor package: Provides an implementation of the scan-statistic framework within the R/Bioconductor ecosystem.
• Scan-statistic framework: Evaluates contiguous genomic segments using a scan statistic for regional inference of methylation differences.
• Dynamic sliding-window: Applies a dynamic sliding-window that accounts for regional CpG density and variable methylation patterns without fixed window sizes or ad hoc smoothing.
• Genome-wide multiple-testing control: Explicitly controls multiple testing across the genome to maintain appropriate false discovery rates.
• Accounts for spatial correlation: Handles spatial correlation among CpG sites in regional inference.
• Modeling flexibility: Supports complex experimental designs, including multi-group comparisons and models with continuous covariates.
• Simulation-based performance: Demonstrates increased statistical power in simulation studies parameterized using real bisulfite sequencing data, especially for small effect sizes and subtle methylation shifts.
• Comparison to alternatives: Offers greater modeling flexibility than bumphunter and DMRcate.

Scientific Applications:

• Genome-wide DMR detection: Identification of differentially methylated regions in genome-wide DNA methylation studies, including bisulfite sequencing and array data.
• Comparative epigenomics: Analysis of multi-group comparisons and studies incorporating continuous covariates.
• Detection of subtle effects: Sensitive detection of small effect sizes and subtle methylation shifts while controlling false discovery rates.
• Regional epigenomic inference: Principled regional inference that accounts for CpG density and spatial correlation across the genome.

Methodology:

Implements a genome-wide scan-statistic that evaluates contiguous genomic segments via a dynamic sliding-window accounting for regional CpG density and variable methylation patterns, with explicit genome-wide multiple-testing correction and support for multi-group comparisons and continuous covariates.