Assessing Differential Expression in Two-Color Microarrays: A Resampling-Based Empirical Bayes Approach

Abstract

Microarrays are widely used for examining differential gene expression, identifying single nucleotide polymorphisms, and detecting methylation loci. Multiple testing methods in microarray data analysis aim at controlling both Type I and Type II error rates; however, real microarray data do not always fit their distribution assumptions. Smyth's ubiquitous parametric method, for example, inadequately accommodates violations of normality assumptions, resulting in inflated Type I error rates. The Significance Analysis of Microarrays, another widely used microarray data analysis method, is based on a permutation test and is robust to non-normally distributed data; however, the Significance Analysis of Microarrays method fold change criteria are problematic, and can critically alter the conclusion of a study, as a result of compositional changes of the control data set in the analysis. We propose a novel approach, combining resampling with empirical Bayes methods: the Resampling-based empirical Bayes Methods. This approach not only reduces false discovery rates for non-normally distributed microarray data, but it is also impervious to fold change threshold since no control data set selection is needed. Through simulation studies, sensitivities, specificities, total rejections, and false discovery rates are compared across the Smyth's parametric method, the Significance Analysis of Microarrays, and the Resampling-based empirical Bayes Methods. Differences in false discovery rates controls between each approach are illustrated through a preterm delivery methylation study. The results show that the Resampling-based empirical Bayes Methods offer significantly higher specificity and lower false discovery rates compared to Smyth's parametric method when data are not normally distributed. The Resampling-based empirical Bayes Methods also offers higher statistical power than the Significance Analysis of Microarrays method when the proportion of significantly differentially expressed genes is large for both normally and non-normally distributed data. Finally, the Resampling-based empirical Bayes Methods are generalizable to next generation sequencing RNA-seq data analysis.

Epigenetic Delay in the Neurodevelopmental Trajectory of DNA Methylation States in Autism Spectrum Disorders

“…We next determined genome-wide differences in DNA methylation between autism-diagnosed and typically developing individuals by identifying CpG sites from the 450k array dataset that exhibited significant differences ( P < 0.05) in the levels of DNA methylation between ASD cases and age-matched controls using a resampling-based empirical Bayes method statistical approach with a defined cutoff parameter (mean delta β value of ≥10%) and designated these as differentially methylated loci (DML) (Li et al, 2013). We identified 1,079, 1,072, and 897 DML from young, middle, and old age groups, respectively ( Supplementary Dataset 1 ).…”

“…Filtered beta values obtained from the RnBeads pipeline analysis package for all SVZ samples ( n = 34) were loaded into the R statistical environment 3.1.2. Three planned comparisons of young, middle, and old age groups were analyzed comparing autism and control samples using the resampling-based empirical Bayes methods permutation approach (Li et al, 2013). This approach reduces the false discovery rates for non-normally distributed array-based data and offers higher statistical power.…”

Comparative DNA methylomic analyses reveal potential origins of novel epigenetic biomarkers of insulin resistance in monocytes from virally suppressed HIV-infected adults

“…Of the ~ 450,000 CpG sites that were quantified for methylation, a total of ~ 15,000 probes were removed after filtering for missing probes, SNP-enriched probes, and probes with statistically insignificant low intensity detections ( P > 0.05), leaving 435,267 CpGs. From those remaining, we sought to identify sites with significant DNA methylation differences ( δ of the β value) between the IS and IR groups ( P < 0.05) using the resampling-based empirical Bayes Methods permutation approach with P < 0.05 [42] and filtered for sites with absolute average methylation differences greater than 10% ( δ ) between the IR and IS groups. This approach reduces the false discovery rates for non-normally distributed array-based data and offers higher statistical power.…”

“…Illumina Infinium HumanMethylation450 BeadChip data was downloaded from GEO Accession: GSE35069 [45]. Cell type-specific methylation data for fluorescence-activated cell sorted (FACS) monocytes ( n = 6) and PBMCs ( n = 6) were used to determine cell type-specific DNA methylation sites using the resampling-based empirical Bayes Methods permutation approach as performed above but with an average methylation δ > 30% between monocytes and PBMCs [42]. This produced 5,124 CpGs whose methylation states appeared robustly cell type-specific, allowing distinction between monocytes and PBMCs derived from DNA methylation.…”

“…Leukocytes were obtained by means of leukapheresis from two non-infected, healthy volunteers, and monocyte subsets were collected using FACS performed by the Flow Cytometry Services of the University of Hawaii at Manoa, John A. Burns School of Medicine (University of Hawaii, Honolulu, HI, USA). Monocyte subsets were then subjected to the 450 K microarray for methylation profiling and differentially methylated sites between each subset was determined by the resampling-based empirical Bayes Methods permutation approach as performed above with an average δ > 10% [42]. This revealed 394 CpGs, 128 CpGs, and 496 CpGs whose methylation states robustly stratified classical, intermediate, and non-classical monocytes, respectively.…”