Scale-free co-expression networks were constructed using the weighted gene coexpression network analysis (WGCNA) package in R15. WGCNA provides a global perspective, emphasizing the correlation between genes to classify different molecular groupings, rather than focusing on individual genes. WGCNA defines modules using a dynamic tree-cutting algorithm based on hierarchical clustering of expression values (minimum module size = 100, cutting height = 0.99, deepSplit = TRUE). The networks were constructed at a soft power of 14 at which the scale-free topology fit index reached 0.90 (Supplementary Figure 2B). We further merged modules that had similar co-expression patterns by calculating the eigengenes and merging those having a correlation > 75% (Supplementary Figure 2C). Correlation of module eigengenes with alcohol dependence, alcohol consumption, AUDIT scores and number of years of drinking (module-trait correlation analysis) was evaluated using Spearman’s rank correlation analysis. We used the DSM4 criteria for alcohol dependence classification as provided by the New South Wales Tissue Resource Centre at the University of Sydney. For each individual in the RNA-Seq dataset a module eigen value was calculated for each module. This module eigen
