To identify the SNPs associated with the phenotypes, we adopted the commonly used significance level for GWAS of 10−6 to account for multiplicity. Based on the results of the marginal tests, the numbers of SNPs identified to be associated with the individual phenotypes are 917 for alcohol dependence symptoms, 0 for nicotine symptoms, 9 for marijuana symptoms, and 0 for cocaine symptoms. Using the proposed Fisher combination method (FC-Kendall), on the other hand, we identified 6 SNPs associated with the multivariate phenotype of addiction. Among them, 5 SNPs were also identified by the marginal test for alcohol symptoms. This implies that if we ignore the correlations among the 4 phenotypic traits and conduct the marginal tests, we would identify many SNPs that may be specific to alcohol dependence. Thus, if our goal is to identify the genes associated with the construct of addiction that contributes to the 4 types of substance dependence symptomatology, the proposed method is a better approach.
