decrease in DPW. The predicted gene expression results from AUD and DPW GWAS were compared with observed expression differences in the brains of AUD subjects and controls to validate the results. Our differential expression analysis of alcohol consumption in the human brain indeed showed that the mRNA expression of MAPT was associated with increased alcohol consumption (Fig. 4c). The association between MAPT expression and alcohol consumption did not pass multiple test correction, most likely due to the small sample size of the brain dataset from people with AUD (P = 7.4 × 10−3; PBonferroni = 0.46). We did not observe any association between the expression of LRR37A and the level of alcohol intake in this brain dataset. The co-localized SNPs within the 17q.21.31 locus were also compared with the promoter (H3K27ac, H3K4me3), enhancer (ATAC-Seq), and promoter–enhancer interactome (PLAC-Seq) data from four specific brain cell types (microglia, neuron, astrocytes, and oligodendrocytes) to elucidate the functional significance of these variants. The co-localized mQTLs (rs3785884 and rs17651887) overlapped with the chromatin interaction region specifically in oligodendrocytes and these interactions looped at the MAPT promoter (Fig. 4b). This observation combined with differential expression data in the human brain provides strong supporting evidence that MAPT
