Fig. 5.

AUC values and their standard errors for PRS of schizophrenia when SNPs are imputed. PRSs were trained based on the PGC data excluding the MGS study; the MGS study was used as the validation data for calculating AUC and its variance. MGS samples were imputed using IMPUTE2 and the haplotypes in The 1000 Genome Project. AUC₀ and SE₀ were calculated assuming individual level data. AUC₁ and SE₁ were calculated using SummaryAUC based on GWAS summary data in MGS, where pairwise SNP correlation was estimated using the genotype in The 1000 Genome Project directly. AUC₂ and SE₂ were calculated using SummaryAUC based on the summary data in MGS, where pairwise correlation was estimated using the imputed genotypic dosage data illustrated in Figure 1. $r^2=0.01$: SNPs were LD-clumped using $r^2=0.01$ in PLINK. $r^2=0.2$: SNPs were LD-clumped using $r^2=0.2$ in PLINK