Combining gene expression and DNA methylation profiles

To combine methylation and expression data, we established a two-step analysis strategy. First, significantly differentially methylated genes between NOA and OA patients were analyzed. Genes that were differentially methylated in NOA compared with OA were selected if the absolute value of Δβ was >0.2 and pvalue was <0.05. Second, for each differentially methylated gene, the Spearman's rank correlation for the median of the methylation level of CpGs in an amplicon against the expression probes was calculated using the function cor.test in R. Significantly negative correlations were considered if the correlation coefficient ρ was <0 and pvalue was <0.05, and significantly positive correlations were considered if ρ was >0 and pvalue was <0.05. Differentially expressed genes that matched methylated genes were then studied.

In addition, we conducted unsupervised clustering analysis, which demonstrated the separation of NOA and OA samples with evidence of gene clusters that were differentially expressed or methylated. Hierarchical clustering was performed using Ward linkage with Euclidean distance for samples. A clustering method available in the Charm package in R was applied to the differentially methylated CpGs³⁵.