Robust Multichip Average signals (RMA) (Irizarry et al., 2003) were generated for the core probe sets using the RMA background correction. Quantile normalization and summarization was done by Median Polish analysis using the Partek Genomics Suite. Summarized signals for each probe set were log2 transformed. These data are deposited in the NCBI Gene Expression Omnibus under series number GSE44456. The log2 transformed signals were used for principal components analysis, hierarchical clustering, and signal histograms to determine if there were any outlier arrays; none were found. We have previously shown that removing probe sets not reliably detected above background in any experimental condition improves analysis by reducing the multiple testing burden (McClintick & Edenberg, 2006). The signal histogram (not shown) indicated that probe sets with log2 values < 4 were at background level. Therefore, probe sets with mean log2 values < 4.0 in both alcoholics and controls were removed. The remaining probe sets were analyzed using a 3-way ANOVA with the factors of phenotype (control/alcoholic), sex (male/female), and processing batch (for potential technical variations). Interaction between sex and phenotype was not
