Fig. 5: Replication analysis of XMAP and related methods on height GWASs.
a–d Overview of replication analyses of high-PIP fine-mapped SNPs across populations: bar charts showing the fraction and number of fine-mapped SNPs with p-value < 5 × 10−8 in the replication cohorts of EUR Sibship GWAS and BBJ cohorts and bar charts showing the distribution of PIP for fine-mapped SNPs computed by SuSiE in the replication cohorts of EUR Sibship GWAS and BBJ. Values for counts < 4 are not shown in the figure. e–i Fine-mapping of locus 66.55 Mbp–66.85 Mbp in chromosome 15. The SNP rs2053005 is significant (p-value < 1 × 10−6) in UKBB, but not significant in Chinese GWAS and EUR Sibship GWAS. When UKBB and Chinese cohorts were combined for cross-population fine-mapping (h), the PIP of rs2053005 was computed to be > 0.8 by PAINTOR, MsCAVIAR, SuSiEx, and XMAP when we set c1 = c2 = 1 (XMAP C=I). XMAP estimated the inflation constants of UKBB and BBJ as 1.66 and 1.39, suggesting they are influenced by confounding bias. After correcting for confounding bias, this signal was excluded in XMAP with a PIP < 0.05, which suggests that the high PIP of the SNP could have been induced by uncorrected population stratification. To test our assumption, we combined Chinese and EUR Sibship GWASs, which are both less influenced by confounding factors (both with inflation constant estimated as 1.07). As expected, all methods consistently produced a low PIP for rs2053005 (i), which confirmed our assumption and suggested XMAP can reduce spurious signals.
