A Pleiotropy-Informed Bayesian False Discovery Rate Adapted to a Shared Control Design Finds New Disease Associations From GWAS Summary Statistics
Figure 1
Correction for shared controls.
Simulation of GWAS summary statistics for 20000 SNPs, all null for phenotype i and variably null or non-null for phenotype j, with association tested using a shared control group. Black dots show p values for phenotype i at all SNPs with p value for phenotype j less than 0.05, with evident downward bias. Blue dots show our adjustment to expected quantile of p values. The red dots show the expected quantile we would compute if we were to assume incorrectly that all SNPs were null for the conditional phenotype i. We see that this quantity overestimates the true quantile.