There are several limitations to our study. The use of postmortem brain samples means that both pre-existing differences that increase risk for AUD and differences associated with the extended, excessive alcohol consumption characteristic of it are both present. We cannot clearly delineate which differences are consequences and which are causes of AUD, although our incorporation of GWAS and eQTL information allowed us to infer some of the latter. Experimental studies such as high-throughput CRISPR inhibition or activation of the genes identified herein in cell models could confirm some of the networks and regulatory pathways,86 but cellular studies will not allow confirmation of effects on behavior of individuals. Another limitation is that those who drink heavily are more likely to smoke. A recent study found that 63.3% of drinkers at risk of alcohol dependence were smokers compared with 18.2% among drinkers not at risk, and 19.2% among non-drinkers.87 Thus, some differences might be attributed in part to smoking. Finally, our results are primarily from males genetically closely related to the 1000 Genomes European samples, and therefore do not capture the transcriptional and epigenetic diversity across ancestry groups or sexes.88
