ComBat-met: adjusting batch effects in DNA methylation data

Area of Science:

• Genomics
• Bioinformatics
• Statistical genetics

Background:

• Genomics data integration is often impeded by technical variations known as batch effects.
• Current batch correction methods struggle to accurately address the unique characteristics of DNA methylation data.

Purpose of the Study:

• To introduce ComBat-met, a new beta regression framework designed to correct batch effects in DNA methylation studies.
• To evaluate the performance of ComBat-met compared to existing methods.

Main Methods:

• ComBat-met employs beta regression models to analyze DNA methylation data.
• The framework calculates batch-free data distributions by mapping estimated distribution quantiles to their batch-free equivalents.
• The method was tested using simulated data and The Cancer Genome Atlas (TCGA) datasets.

Main Results:

• ComBat-met demonstrated improved statistical power in differential methylation analysis when compared to traditional methods.
• The method effectively removed cross-batch variations and successfully recovered biological signals in TCGA data.
• False positive rates were not compromised by using ComBat-met.

Conclusions:

• ComBat-met offers a robust solution for batch effect correction in DNA methylation studies.
• The framework enhances the reliability and statistical power of downstream analyses, such as differential methylation analysis.
• ComBat-met is effective in real-world applications, improving the quality of large-scale genomics datasets.