We used the Bioconductor R package Maanova to identify expression probes whose expression differed significantly between pairs of methods [37]. We fitted a linear mixed model for each probe using the Fs distribution as the null distribution and we fitted method as fixed, and visit and individual as random effects. We considered probes as differentially expressed when significant at a 5% FDR. We tested for significant enrichment of GO terms among the set of differentially expressed probes relative to the overlapping detected probes of two methods. Because a large proportion of probes were significantly differentially expressed, we selected the 5% of top hits ranked by FDR p-value. Of these 5% of top probes, we used only these probes that showed a more than a three fold change between methods.
