Robust joint score tests in the application of DNA methylation data analysis

Area of Science:

• Genomics
• Biostatistics
• Computational Biology

Background:

• Differential variability analysis is crucial for understanding DNA methylation's role in complex diseases.
• Joint statistical tests incorporating both methylation level and variability offer increased power over methods using only methylation level.
• The existing Anh and Wang (AW) joint score test is conservative and lacks comprehensive comparison.

Purpose of the Study:

• To propose and evaluate novel joint score tests for detecting differential DNA methylation and variability.
• To compare the performance of the proposed tests against existing methods like jointLRT, KS, and AW.

Main Methods:

• Development of three improved joint score tests: iAW.Lev, iAW.BF, and iAW.TM.
• Extensive simulation studies to assess performance under various data conditions (outliers, unequal variances, normality violations).
• Validation using real-world Illumina HumanMethylation27 and MethylationEPIC datasets (GSE37020, GSE20080, GSE107080).

Main Results:

• The three improved tests (iAW.Lev, iAW.BF, iAW.TM) demonstrated superior power and maintained Type I error rates in simulations with outliers and unequal variances compared to jointLRT, KS, and AW.
• For normally distributed data, the improved tests showed slightly lower power than jointLRT and AW.
• Real data analyses confirmed higher true validation rates for the proposed improved tests over existing methods.

Conclusions:

• The proposed joint score tests are robust to non-normality and outliers, offering advantages over existing methods.
• iAW.BF emerged as the most robust and effective test across simulated and real data scenarios.
• These improved tests enhance the analysis of DNA methylation data for disease association studies.