Two primary points have become apparent over the last several years from GWAS: (1) the effect sizes associated with individual genetic variants are very small, usually with odds ratios (ORs) on the order of 1.1, and (2) our ability to select a priori which genes are viable candidates for psychiatric and substance use disorders has been poor (K. S. Kendler, 2013; P. F. Sullivan, Daly, & O'Donovan, 2012). There are rare exceptions, such as the role of alcohol dehydrogenase genes in alcohol dependence (Shen et al., 1997). We now realize that early candidate gene studies, as well as early atheoretical systematic gene finding efforts (such as linkage studies), were underpowered to detect genes with the small effect sizes that more recent studies suggest are likely to be realistic. Although GWAS were more successful for some conditions (Crohn’s, diabetes, macular degeneration; Manolio & Collins, 2009), like linkage studies, early GWAS were largely unsuccessful in the area of psychiatric and substance use disorders (K. S. Kendler, 2013). Few SNPs were detected that met genome-wide levels of significance. As increasingly large sample sizes
