Detecting differentially methylated regions with multiple distinct associations.

Area of Science:

• Epigenetics
• Bioinformatics
• Statistical genomics

Background:

• Accurate detection of differentially methylated regions (DMRs) is crucial for understanding epigenetic regulation in various biological contexts.
• Existing computational methods for DMR detection vary in their performance and statistical rigor.

Purpose of the Study:

• To evaluate and compare the performance of five distinct methods for identifying DMRs: DMRcate, comb-p, seqlm, GlobalP, and dmrff.
• To assess the impact of using an ancestry-matched reference cohort for estimating CpG site correlations in cord blood data.

Main Methods:

• A simulation study was conducted to assess the statistical properties of the evaluated methods.
• Real-world data analysis was performed to complement simulation findings.
• The utility of an ancestry-matched reference cohort was investigated for correlation estimation.

Main Results:

• Several DMR detection methods demonstrated an inflated Type I error rate, particularly at stringent significance thresholds.
• The dmrff method consistently ranked among the most powerful approaches in power simulations involving 1-2 causal CpG sites with consistent effect directions.
• The study identified performance variations among the evaluated DMR detection tools.

Conclusions:

• The findings underscore the necessity for comprehensive simulation studies to rigorously evaluate novel bioinformatics methods.
• Further development is required for methods that offer well-controlled Type I error rates without necessitating individual-level data.