Fig. 2: Comparisons of fine-mapping approaches in the absence of confounding bias. | Nature Communications

Fig. 2: Comparisons of fine-mapping approaches in the absence of confounding bias.

From: XMAP: Cross-population fine-mapping by leveraging genetic diversity and accounting for confounding bias

Fig. 2

a Manhattan plots of a simulated GWAS data in EAS (left) and EUR (right). All p-values are obtained from marginal regression. b Heat maps showing the absolute correlations between the three causal SNPs (highlighted with rectangles) and their nearby SNPs in EAS and EUR populations. c Comparisons of FDR control with n1 = n2 = 20, 000 and Ktrue = 3. We include an ad-hoc method by merging the discoveries from EAS and EUR (cyan triangles), which is equivalent to controlling the FDR by using the largest PIP across populations. This post-selection procedure introduces false positives. By contrast, XMAP is well-calibrated by modeling cross-population GWASs through an integrated statistical framework. d CPU timings of XMAP, MsCAVIAR, PAINTOR, FINEMAP, and DAP-G are shown for increasing K with p = 100. Solid lines are CPU time recorded in our experiments and dashed lines represent predicted CPU time based on the time complexity of corresponding approaches. e CPU timings are shown for increasing p with K = 2. f Comparisons of statistical power with n1 = n2 = 20, 000 and Ktrue = 3. Computational time is presented as the mean value +/- standard deviation. Results and error bars are summarized from 50 replications.

Back to article page