Clique-Based Clustering of Correlated SNPs in a Gene Can Improve Performance of Gene-Based Multi-Bin Linear Combination Test

Figure 3

Averages of (a) the size of the largest cluster and (b) the number of singleton clusters produced in each gene by LDSelect and CLQ for a fixed number of clusters per gene. For each gene, the number of clusters produced by each clustering method was found at threshold values within a grid from 0.1 to 0.9 by 0.01. Excluding results at the extremes (i.e., including cluster numbers that fell between 20% and 90% of the number of SNPs), the LDSelect and CLQ cluster numbers were matched for each gene and the maximum cluster size for each was averaged across genes at a fixed value of the number of clusters.

(a) The size of the largest cluster

(b) The number of singleton clusters