A Pleiotropy-Informed Bayesian False Discovery Rate Adapted to a Shared Control Design Finds New Disease Associations From GWAS Summary Statistics
Figure 2
Validation of the shared-control approach and the p-value adjustment due to shared controls.
Panel A shows the effect of splitting controls on power to detect association. The number of SNPs with p values less than a given cutoff are shown for split-control and shared-control approaches. For all p-value cutoffs, fewer SNPs reach significance when using a split-control design. Panel B shows the number of SNPs with values less than a given cutoff using the existing method on a split-control design, our extended method on a shared control design with the adjustment for shared controls, or using the split-control approach naively on the shared-control design; that is, assuming incorrectly that . The second figure shows that failing to correctly calculate leads to a subtle increase in the number of SNPs declared non-null at all cutoffs, due to the incorrect underestimation of .