Although the results of our analysis are consistent with the conclusion from twin studies that SNPs cluster according to hierarchical genetic factors, this study only provides a groundwork for further studies as a proof of principle. Simulations and replications are needed to fully validate these results. We focused our interpretation on 2- and 4-class models based on twin findings, but the optimal number of classes was not clear based upon commonly accepted criteria. Determining the optimal number of classes depends not only on the underlying mechanism being continuous or categorical, but also on adequacy of assumptions, parameter specifications, and characteristics of data, such as sample size and the scales of responses (Lubke and Miller 2015; Lubke and Neale 2006). Further simulations with varying genetic architecture would be needed to validate the clusters of SNPs correspond to the structure of comorbid phenotypes observed in our study. In addition, while previous studies on genetic architecture of comorbid psychiatric disorders are based on general population samples (Kendler et al. 2003; Kendler et al. 2011a; Krueger et al. 2002), our sample was ascertained based
