Data on untreated cells from a previous microarray experiment that used the Affymetrix HGU133plus2 GeneChip™ analyzed with microarray suite 5.0 software (Affymetrix, San Clara, CA) were available as log2(MAS5 expression) (McClintick et al., 2014), GEO (series accession # GSE52553). These data were matched by gene symbol to the log2(counts per million reads) from the present RNAseq experiment and imported into Partek Genomics Suite version 6.6 (Partek, Inc. St. Louis, Mo). Both experiments had similar ranges of values (Supplemental Figure 2), so the most appropriate approach was to combine the log2 RMA (microarray) and log2 CPM (sequencing) values, with a term in the ANOVA for experiment type (study) but with no additional normalization. The ANOVA used phenotype (AD, control), study (array, RNAseq) and gender as factors plus the interaction term (phenotype × study). Genes with an absolute fold change <1.2 were removed and genes with p< 0.05 for the interaction were removed. FDR was calculated within Partek using the Storey and Tibshirani method (Storey & Tibshirani, 2003).
