Assessing the Differential Methylation Analysis Quality for Microarray and NGS Platforms

Area of Science:

• Epigenetics
• Genomics
• Bioinformatics

Background:

• Differential methylation (DM) analysis is crucial in biological and translational research.
• Microarray and next-generation sequencing (NGS) are common methods for methylation analysis, each with various statistical models.
• Benchmarking DM models is difficult due to the lack of gold standard datasets.

Purpose of the Study:

• To evaluate the performance of different statistical models for DM analysis using public NGS and microarray datasets.
• To assess the utility of the Hobotnica rank-statistic-based approach for quality evaluation of DM signatures.

Main Methods:

• Analysis of numerous public NGS and microarray datasets.
• Application of various widely used statistical models for DM signature extraction.
• Utilized the Hobotnica metric for quality assessment of DM results.

Main Results:

• Microarray-based methods yielded more robust and consistent DM results compared to NGS-based models.
• NGS-based models demonstrated significant heterogeneity in their findings.
• Simulated NGS data tests may overestimate DM method quality and should be used cautiously.
• Microarray data showed more stable results for evaluating top differentially methylated cytosines (DMCs).

Conclusions:

• The heterogeneity observed in NGS methylation data necessitates rigorous evaluation of newly generated DM signatures.
• The Hobotnica metric offers a robust, sensitive, and informative way to estimate DM method performance and signature quality.
• Hobotnica addresses a long-standing challenge in DM analysis by providing reliable evaluation without gold standard data.