Figure 1
From: Identification of genes and haplotypes that predict rheumatoid arthritis using random forests
Application of the gene-permutation method to investigate SNP and interaction importance in simulated data. Labels along the x-axis identify the permuted SNP. Darker shades of green represent a bigger DVI. The first column of each plot shows the changes in variable importance of all SNPs after permuting SNP1 (DVI1), the second column shows the change in importance after permuting SNP2 (DVI2), etc. Thus, the diagonal shows 
for k = g, while off diagonal k ≠g. A, SNP1 and SNP2 have marginal effects but no interaction effect. B, SNPs 1 and 2 interact to influence the probability of disease.