out. We next filtered out probe sets that were not detectable above background in our samples,; this has been shown to reduce noise in microarray experiments (McClintick and Edenberg, 2003) Probe sets that did not have at least 25% of samples with normalized scaled expression greater than 64 were not analyzed. Linear modeling to calculate gene-wise p-values for the contrasts of the ethanol group versus water group was performed using the package Limma (Smyth, 2004). For each region, models including treatment, time of sacrifice after last drinking session (time), labeling batch, and hybridization batch, including interaction effects for all factors, were constructed. For each region, the most parsimonious model was chosen. For the Acb-shell, the model included only treatment; for the CeA the model included both treatment and labeling/hybridization batch. Probe sets were considered to be statistically significant at p < 0.01, which, because of differences in the p-value distribution, was equivalent to a false discovery rate (FDR; calculated according to Storey et al., 2004) less than 0.25 for the ACB-shell and 0.10 for the CeA. These FDR values were selected to provide sufficient numbers of ‘named’ genes to conduct Gene Ontology (GO) and Ingenuity pathways analyses. Any false positives
