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

Figure 4

Power of MLC test based on LDSelect clustering (MLC-LD: blue) and CLQ clustering (MLC-CL: red) with threshold for the cases where LDSelect and CLQ clusters are different. Each point represents one case of all possible causal SNP assignments within a gene. The variance of the error term was adjusted for each of 1000 genes such that the Wald test power is exactly 0.6 (Grey) for Models A, B, and C and fixed as 1 for Model D.

(a) Model A (one causal SNP)

(b) Model B (two causal SNPs, both deleterious)

(c) Model C (two causal SNPs, one deleterious, one protective)

(d) Model D (1~4 random causal SNPs, random number of deleterious and protective SNPs)