An assessment of compositional methods for the analysis of DNA methylation-based deconvolution estimates

Area of Science:

• Epigenetics and computational biology.
• Statistical methods for biological data analysis.

Background:

• DNA methylation (DNAm)-based deconvolution estimates yield relative data, forming a compositional dataset.
• Standard statistical methods are often unsuitable for analyzing compositional data directly.
• This necessitates specialized approaches for accurate interpretation of DNAm deconvolution results.

Purpose of the Study:

• To evaluate the performance of the Analysis of Compositions of Microbiomes (ANCOM) method for DNAm-based deconvolution estimates.
• To compare ANCOM's effectiveness against standard statistical approaches in handling compositional data.
• To assess ANCOM's applicability to real-world biological datasets.

Main Methods:

• Conducted two simulation studies to compare ANCOM with a standard two-sample t-test on cell proportions.
• Applied ANCOM to analyze DNAm-based deconvolution estimates from the Women's Health Initiative study.
• Evaluated ANCOM's ability to manage the compositional nature of DNAm deconvolution data.

Main Results:

• ANCOM effectively accounts for the compositional characteristics inherent in DNAm deconvolution estimates.
• The ANCOM method demonstrated adequate control over the false discovery rate.
• Statistical power using ANCOM was found to be comparable to traditional methods.

Conclusions:

• ANCOM is a suitable and effective method for analyzing compositional DNAm deconvolution data.
• The findings support the use of ANCOM for robust statistical inference in epigenetics studies.
• ANCOM offers a reliable alternative for researchers working with relative DNAm proportions.