In order to further refine gene modules (GM) in the context of alcohol dependence, GM network structure was correlated with phenotypic variation for lifetime consumption of alcohol across individual subjects. Establishing correspondence with a particular phenotypic trait (i.e., alcohol consumption), rather than a clinical diagnosis alone, is biologically relevant to substance abuse and experimentally tractable in future studies. Correlation with lifetime alcohol consumption ranged from |r = 0.01 to r = 0.73|, with ten individual GMs being in the upper and lower quartiles of the distribution (Fig. 2a). The first principal component of GM expression within the upper quartile on average accounted for 18.9% of the variance for lifetime consumption versus 1.5% for the lower quartile, a 12.6 fold-difference between the GM groups. Intermodular correlation is significantly higher among the upper quartile GMs associated with alcohol drinking compared to GMs residing in the lower quartile (t(88)=4.56, P = 1.65 e-05) (Fig. 2b). This correlation structure implies coherent biological function for a set of modules, comprising 2,330 genes in PFC, directly related to alcohol drinking behavior in humans. GMs within the
