diffMeanVar

diffMeanVar detects gene probes exhibiting differential means or variances between two groups in (epi)-genomic datasets to identify differential methylation and differential variability associated with complex human diseases.

Key Features:

• Detection of differential mean and variance: Identifies probes with differential means or variances between two groups in (epi)-genomic datasets.
• Variance tests: Implements classical F test and Levene's test to assess equality of variances.
• Likelihood ratio testing: Uses likelihood ratio tests to evaluate equality of means and variances.
• Joint score tests: Provides three refined joint score tests—iAW.Lev, iAW.BF, and iAW.TM—for simultaneous detection of differential methylation and variability.
• Comparative benchmarking: Compares performance against joint likelihood ratio test (jointLRT), Kolmogorov–Smirnov (KS) test, and the original AW test.
• Simulation performance: Demonstrates increased power while maintaining nominal Type I error rates in simulation studies, particularly under outliers or non-normality.
• Empirical validation: Validated on Illumina HumanMethylation27 datasets (GSE37020, GSE20080) and an Illumina Infinium MethylationEPIC dataset (GSE107080).
• Robust test variant: Identifies iAW.BF as particularly robust and effective across simulated scenarios and empirical analyses.

Scientific Applications:

• DNA methylation studies: Detects differential variability and differential methylation at CpG probes to study associations with complex human diseases.
• Simultaneous mean/variance analysis: Enables joint assessment of differential methylation and differential variability in (epi)-genomic research.
• Method benchmarking and validation: Supports benchmarking of statistical methods using simulation studies and validation on Illumina methylation datasets (GSE37020, GSE20080, GSE107080).

Methodology:

Computational methods explicitly include F test and Levene's test for variance equality, likelihood ratio tests for mean and variance equality, joint score tests iAW.Lev, iAW.BF, iAW.TM, comparisons to jointLRT, Kolmogorov–Smirnov (KS) and AW tests, and assessment via simulation studies and empirical analyses of Illumina HumanMethylation27 (GSE37020, GSE20080) and Illumina Infinium MethylationEPIC (GSE107080) datasets.