We calculated false discovery rate (FDR) q-values (Storey and Tibshirani, 2003) for all SNPs to select loci for further study. For replication and secondary analysis of discovery data, we used a threshold q<0.3 (p<1×10−5 for our p-value distribution, 30% of results are false discoveries) to maximize discovery potential. For functional studies, we included any loci achieving genomewide significant statistical signals in discovery or replication (COL6A3, LOC339975) and any loci with both q<0.1 (10% of results are false discoveries) and prior evidence of involvement in ethanol phenotypes from human and/or MO studies (KLF12, RYR3).
