Fig. 2

Simulated GWAS pairs with overlapping samples, after correction for sample overlap using the decor relation transform. Data before correction is presented in Fig. 1. Data was simulated for two quantitative trait GWAS with no genetic effects but overlap in samples ((each with n=12,500 including 5,000 overlapping samples). d=100,000 SNPs were simulated under the null model (phenotype is simulated independent from genotype). The decor relation transformation proposed here was applied to the simulated summary statistics. Panel a: the p-value distribution for trait 1; Panel b: the p-value distribution for trait 2; Panel c: The p-value distribution for trait 2 given that the p-value in study 1 was less than 0.1; Panel d: quantile-quantile plot for the p-values in study 2, stratified by the p-value in study 1