Fig 1.
The detection power of each GWAS method.
Boxplot of the detection power evaluated by −log10(p) and −log10(pa). a-d: The results for the “coupling” scenario. e-h: The results for the “repulsion” scenario. a,b,e,f: The results evaluated by −log10(p) with the scale on the vertical axis aligned in these four figures. c,d,g,h: The results evaluated by −log10(pa) with the scale on the vertical axis aligned in these four figures. a,c,e,g: The results evaluated by the unit of the causal SNP or haplotype block itself. b,d,f,h: The results evaluated by the unit of the regions within the extent of LD. The abbreviation of each method is as follows. R: RAINBOW. SS: Single-SNP GWAS. H2k-H4p: HGF methods. The numbers in the method names correspond to the numbers of the groups they assume. The last letters of the methods are “k” or “p”. “k” corresponds to the k-medoids method and “p” corresponds to UPGMA method for the grouping method. SK: SKAT.
Fig 2.
Recall, precision and F-measure of each GWAS method.
Bar plot of the mean of each summary statistic for 100 simulation results. The red bars show the results for recall, the green bars show the results for precision, and the blue bars show the results for F-measure. a: Results for the coupling scenario. b: Results for the repulsion scenario. The abbreviations of each method are the same as those of Fig 1.
Fig 3.
AUC for regions around causals.
Bar plot of the mean of AUC for regions around causals. This summary statistic indicates the extent to which the causal itself can be detected by GWAS without relying on the LD. The red bars show the results for the means of 100 simulation results and the blue bars show the results for the means of the simulation results whose QTN1 and QTN2 were detected. a: Results for the coupling scenario. b: Results for the repulsion scenario. The abbreviations of each method are the same as those of Fig 1.
Fig 4.
An example of GWAS results for the repulsion scenario.
Manhattan plots of 4 GWAS methods (RAINBOW, Single-SNP GWAS, HGF (the number of groups is 2, the grouping method is UPGMA), and SKAT) for one simulation result of the Repulsion model. The black horizontal dashed lines represent the thresholds determined by the Benjamini-Hochberg method (FDR = 0.01) for each result of the Repulsion model. The red vertical dashed lines show the positions of QTN1, QTN2, and the purple ones show the position of QTN3. The red points show −log10(p) of causal SNPs or haplotypes including QTN1 and QTN2, and the purple ones show −log10(p) of QTN3 or haplotypes including QTN3.