alcoholics. RNA-Seq defined modules were significantly conserved in relation to previous network-based studies from human cortex (Table S1), forming distinctive functional categories that demonstrate consistency of networks with previous human brain studies. Alcoholic modules were further over-represented for a similar network analysis on alcohol dependence using microarray data (Fig. S3) 6; however, more modules were discerned in our current investigation, which may be attributed to the precision, larger dynamic range, and improved network characteristics involving RNA-Seq 38. The normalized read counts could also be further summarized for transcript-level and exon-level information (Fig. 1), providing an overall perspective of the transcriptome network structure. Neither controls or alcoholics showed any substantial clustering bias for brain weight, pH, PMI, RIN, age, or smoking (Fig. S4). To determine global differences in transcriptome architecture, we compared mean ranked expression and connectivity patterns between controls and alcoholics. Expression across all three levels, gene-, transcript-, and exon-level, was strongly correlated between controls and alcoholics (Fig. 1a). In contrast the global connectivity, a measure of interrelationship among all the features within a biological network, demonstrated progressively weaker correlation (Fig. 1a). The discrepancy between strong overall expression and weaker connectivity suggests the neurobiology of alcohol dependence in PFC is
