Figure 3
From: Recent natural selection conferred protection against schizophrenia by non-antagonistic pleiotropy
Schizophrenia Polygenic score predictions (PGSSCZ) in an independent cohort stratified by the probability of the SNPs to be subject to RNS. (A) PGSSCZ predictions (% pseudo-R2) on an independent schizophrenia case–control sample (CIBERSAM; NSCZ = 1927; NHC = 1561) using 20 subsets of SNPs ranked by their probability to be subject to RNS (pRNS) are displayed. For this analysis, we used the most significant schizophrenia threshold in the PGSSCZ analysis using whole genome data (pSCZ < 0.2; NSNPs = 61,040). LD-clumping based on pSCZ (r2 < 0.1; 500 kb) was performed for PGS estimations. (B) PGSSCZ performance (% pseudo-R2) using variants from the first RNS quantile (SNPs with the highest probability to be subject to RNS; NSNPs = 3052) was compared against the distribution of PGSSCZ Naggelkerke's pseudo-R2 obtained by selecting the same number of matched SNPs from the remaining 95%. A total of 1000 sets of matched SNPs were randomly drawn from SNPs matched to the variants from the first RNS quantile based on MAF, number of SNPs in LD, distance to nearest gene, and gene density. See Supplementary Data 6 for detailed results.
