Fig. 4

Simulation study showing false positive rate (a,b,c,d) and power (e,f,g,h) of two-component pleiotropy score. Top row shows performance on non-pleiotropic simulated variants (black line shows 5% false positive rate); bottom row shows performance on pleiotropic variants (black line shows 80% power). Simulations were run for both \( {P}_m^{\mathrm{LD}} \) (left) and \( {P}_n^{\mathrm{LD}} \) (right), and both without correction for polygenicity (a,c,e,g) and with the correction (b,f,d,h), with per-variant heritability ranging from 0.0002 to 0.2, proportion of non-pleiotropic causal loci ranging from 0 to 1%, and proportion of pleiotropic causal loci ranging from 0.1 to 1%. Our method has good power to detect pleiotropy for highly heritable traits, though its power is reduced by extreme polygenicity. Extreme polygenicity also increases the false positive rate, though this effect is corrected by our polygenicity correction