Figure 5

Representative examples of eQTL associations. (a-b) Non-linear associations. kruX identifies more non-linear relations where the gene expression level of the heterozygous samples lies outside the typical range of the homozygous samples (a) or where one allele has a dominant effect on the gene expression level (b). (c-d) Problematic associations. Parametric ANOVA gives high significance to spurious associations for genes with outlying expression samples that coincide with singleton genotype groups (c). Associations with skewed genotype group sizes where the model assumptions are difficult to ascertain achieve high significance using linear models (d).