Quantifying the strength of association between DNA-methylation and gene expression

When investigating the association between each DNA-methylation and gene transcripts, we look at two scenarios, first each DNA-methylation against all possible gene transcripts in the analysis, and secondly each DNA-methylation against only the cis related genes. In both cases, we quantify the strength of an association by using the utility test called the “global test” [20]. The null hypothesis is that there’s no association between gene expressions and DNA-methylation, and the alternative hypothesis is that one or more gene transcript levels are associated with the DNA-methylation level. This test is valid for both high-dimensional data (p>n) as well as the case where there are more individuals than covariates. The association between DNA-methylations and gene expressions can be both positive and negative. See Additional file 2 for more details. We performed tests for each DNA-methylation, and got a corresponding p-value p_j for j=1,…,p. Since a large number of tests is performed, we adjust for multiple testing by controlling the false discovery rate (FDR) [21]. We use the adjusted p-value, called the q-value q_j, as a measure of the strength of the association between the particular DNA-methylation and the gene expressions. Moreover, we interpret low values of q_j, as a strong association.