Fig 1.
(A) Simulated Z-scores for SNPs at a quantitative trait locus in two different populations: East Asians (top) and Europeans (bottom), shown by their −log10(p-value). LD matrices for these populations were derived using data from the 1000 Genomes project. These are the input files to MsCAVIAR. (B) Meta-analysis results for this locus, showing many significant SNPs. Also displayed are the SNPs that are in the causal set that MsCAVIAR returns (red stars) and the truly causal SNPs (black stars).
Fig 2.
Comparison of sensitivity, precision, and set sizes using simulated data.
We compare MsCAVIAR, PAINTOR, and CAVIAR with c ∈ {1, 2, 3} causal variants implanted with results averaged over 20 replicates for 3 loci and 5 levels of heritability for all 3 values of c. (A) Bar graph indicating the sensitivity of each method with a dashed line to reflect the expected posterior probability, ρ, of recovering all causal SNPs (B) Box plots showing the average set sizes returned by the methods. Each box is the interquartile range of causal set sizes with the middle black line representing the median, and the white crosses showing the mean. (C) Bar graph displaying the average the number of SNPs in descending order of posterior inclusion probability (PIP) until 1, 2, or 3 causal SNPs is identified. Stacked bars represent increasing numbers of causal SNPs identified, until the true number of causal SNPs (x-axis) are identified.
Fig 3.
Comparing fine mapping resolution in trans-ethnic HDL analysis.
Comparison of the results of MsCAVIAR when applied to 185 loci from two high-density lipoprotein (HDL) GWAS, White European people from the UK Biobank [23, 24] and Japanese people from Biobank Japan [27, 28], versus trans-ethnic PAINTOR [15] and applying CAVIAR [8] to each population individually. The y-axis is the size of the causal set for each locus. The boxes represent the interquartile range of causal set sizes identified by each tool, the lines inside the boxes represent the median, and the whiskers extend to the non-outlier extremes. Outliers are represented as dots above or below the whiskers.
Fig 4.
Comparison of methods’ set sizes for each locus in the trans-ethnic HDL analysis.
Comparison of the returned causal set sizes of MsCAVIAR when applied to two high-density lipoprotein (HDL) GWAS, White European people from the UK Biobank [23, 24] and Japanese people from Biobank Japan [27, 28], versus trans-ethnic PAINTOR [15] and applying CAVIAR [8] to each population individually. In each scatter plot, each point reflects a specific locus, and the x-coordinate is MsCAVIAR’s returned causal set size, while the y-coordinate is a different method’s causal set size. Diagonal lines representing equal set sizes were plotted for each scatter plot. Points above the line represent loci where the alternate method had a larger causal set size than MsCAVIAR, while points below the line indicate the opposite.