Although our sample size is substantial relative to others that have been the subject of alcohol-related GWAS (e.g., [14, 15, 21, 22]), a subset of the data was imputed which could have introduced error into the alcohol problems phenotype. It is evident that quite large sample sizes are necessary to detect the small effect sizes expected for variants influencing complex traits. Thus, it is critical that the current findings be replicated in additional, and larger, samples. In addition, the findings presented here reflect a particular socio-cultural context of a predominantly white sample in southwest England, and these results might not generalize to other ethnicities or cultures. Although our use of a continuous phenotype is more statistically powerful than a dichotomous phenotype, the factor score is comprised of heterogeneous components (e.g., social consequences of drinking, development of tolerance, frequency of drinking, etc.), making it potentially less powerful than a homogeneous phenotype. We note that the number of SNPs for which data were available for some cell lines/tissue types limited our analyses examining the localization of implicated SNPs to genomic regions of
