Fig. 3

Component associations estimated using GWAS summary (x-axis) are consistent with estimated using individual-level data (y-axis). Data were simulated as a trait that can be attributed to a randomly chosen eigen-gene component with 10% (left), 5% (middle) and 2% (right) heritability. Red squares indicate the z-score of the causal component. We confirmed by Kolmogorov–Smirnov test that the z-scores from non-causal components (blue points) followed a normal distribution (p = 0.1691, 0.5393 and 0.2542 for 10, 5 and 2% (right) heritability, respectively)