Fig 1. The projected number of replicated loci and proportion of shared causative variants among the replicated loci.

A total of 2,500 association peaks from well-powered GWAS studies (n = 150,000) were tested for the shared effect in simulated discovery cohorts (n = 10,000), and then select loci were tested for replication in simulated validation cohorts of the same genetic ancestry (n = 10,000–35,000). The candidate loci were identified by conditional false discovery rate (cFDR), colocalization tests (JLIM and eCAVIAR), or the intersection of cFDR and colocalization test. The p-value cutoff was set to 0.01 for cFDR and JLIM, and for eCAVIAR, the equivalent posterior threshold was calibrated using H₀ simulation data. The 2,500 GWAS peaks consist of the loci simulating no causal effect for underpowered traits (H₀) and those simulating the same causal effect between two traits (H₁) or distinct causal effects (H₂). The H₁:H₂ ratio was set to 1:19 (A and C) or 1:4 (B and D). The effect sizes of causative variants are correlated (ρ = 0.7) under H₁ but uncorrelated under H₂. Bonferroni correction was applied on replication tests. In Panel E, we contrast the number of replicated loci expected in Panel A (dashed line) with a more extreme scenario (10,000 GWAS peaks, consisting of 2,500 with H₁:H₂ of 1:19 and 7,500 with H₁:H₂ of 1:39; solid lines). In all, the proportion of H₀ was set to 30% of all examined loci. The shaded area denotes the 95% CIs.