As a comparison, we also used the “conditional FDR” approach proposed by Andreassen et al. [17] to prioritize SNPs in our simulations. The comparison results between GPA and the conditional FDR at and are shown in Figure 3 (other results are provided in Figures S77–S87 in Text S1). GPA significantly outperformed the conditional FDR approach in SNP prioritization. More importantly, in the absence of pleiotropy, the conditional FDR approach had worse accuracy than single-GWAS analysis using the standard FDR approach, whereas GPA achieved comparable accuracy with single-GWAS analysis in this scenario. This suggests that GPA was able to take advantage of pleiotropy when it exists while, in clear contrast to the conditional FDR approach, it does not sacrifice much statistical power than the conditional FDR approach when it is absent.
