A novel computational strategy for DNA methylation imputation using mixture regression model (MRM).

Area of Science:

• Epigenetics
• Genomics
• Computational Biology

Background:

• DNA methylation is a critical epigenetic mark influencing gene expression and disease.
• Current methods like array-based and RRBS assays cover limited CpG sites, restricting comprehensive analysis.
• This limitation hinders the full understanding of DNA methylation's role in human health and disorders.

Purpose of the Study:

• To develop and validate a novel computational strategy for imputing DNA methylation values at unmeasured CpG sites.
• To improve the scope and accuracy of DNA methylation analysis by leveraging existing data.
• To identify novel differentially methylated CpG sites associated with specific traits.

Main Methods:

• A mixture of regression model (MRM) using radial basis functions was employed.
• The method integrates information from neighboring CpGs and methylation pattern similarities across samples and genomic regions.
• Imputation accuracy was evaluated using simulated and empirical datasets, particularly under high missingness rates.

Main Results:

• The proposed MRM strategy demonstrated superior imputation accuracy compared to existing methods.
• Application to a reduced representation bisulfite sequencing (RRBS) dataset identified ~300,000 CpG sites in promoter regions.
• Novel differentially methylated CpG sites associated with bone mineral density (BMD) were discovered.

Conclusions:

• The MRM method is highly applicable to diverse DNA methylation studies.
• Expanding methylation data coverage to unmeasured sites significantly boosts the discovery of differential methylation signals.
• This approach facilitates a deeper understanding of the epigenetic mechanisms underlying human disorders and traits.