A Pleiotropy-Informed Bayesian False Discovery Rate Adapted to a Shared Control Design Finds New Disease Associations From GWAS Summary Statistics
Figure 3
L is the locus of a set of points with .
The FDR is the ratio of null SNPs to total SNPs in L. If all the non-null SNPs were concentrated in the lower left corner, then the number of non-null SNPs in L would be equal to that in any individual rectangle with vertices at the origin and on L, but the number of null SNPs would be greater, meaning that the expected FDR of all SNPs in L would be greater than α. M* is the largest rectangle by area contained within L. The false discovery rate within M* is less than α*, the value of at the upper right vertex, which is usually equal to α, as in this case. The FDR of L is bounded by α*v(L)/v(M), where v(L) and v(M*) are the expected number of null SNPs in L and M* respectively (S1 Text, section B).