ABSTRACT
Accumulating evidence from genome wide association studies (GWAS) suggests abundant presence of shared genetic influences among complex human traits and disorders. A major challenge that often limits our ability to detect and quantify shared genetic variation is that current methods of cross-trait analysis are not designed to work in scenarios with low or absent genetic correlation. Here we introduce a statistical tool BGMG (Bivariate Gaussian Mixture Model of GWAS) which can uncover various scenarios of genetic overlap regardless of genetic correlation, using GWAS summary statistics from studies with potentially shared participants. We perform extensive simulation on synthetic GWAS data to ensure that BGMG provides accurate estimates of model parameters in the presence of realistic linkage disequilibrium (LD) structure.