The present study has identified several novel candidate genes that may influence alcohol consumption, and suggests that other SNPs among the top hits (p<10−4) may also be true causal factors. Using the GCTA program, we found that nearly 40% of the variance for maxdrinks can be explained by considering all SNPs simultaneously on the Illumina 1M chip. The strongest SNPs in the present study explain a very small proportion of this variance. This shows that part of the missing heritability for maxdrinks is tagged by SNPs on the 1M chip, but due to stringent p value thresholds for GWAS, we likely ignored some real signals, underscoring the need to find effective ways to extract meaningful genetic data from the noise. Another concern is that variants like rs1229984, which explain a relatively large fraction of the variance, are not present on most genotyping chips because of their relatively low frequency in European populations (MAF=3%). There may be other rare and novel variants like this that influence maxdrinks, which are not tagged by common SNPs and are difficult to impute. To identify
