DNA co-methylation analysis suggests novel functional associations between gene pairs in breast cancer samples

Abstract

Localized promoter hypermethylation and overall DNA hypomethylation have been associated with the presence of tumor in humans. Yet, despite the large amount of recently produced epigenetic data, there is still a lack of understanding on how several genes behave in tumor cells with respect to their epigenetic alterations such as DNA methylation. Here we performed a novel type of analysis that measures the correlation of DNA methylation levels between two genes across many samples. We linked this so-called co-methylation to the genomic distance of these genes, their functional similarity and their expression levels. Co-methylation analysis of more than 300 breast cancer samples from the TCGA portal yielded 187 pairs of genes showing Pearson correlation coefficients |r| ≥ 0.75. These pairs were formed by 133 genes. Less than half of these pairs are located on the same chromosome. For these, we found that the level of co-methylation is weakly anti-correlated with genomic distance (r = -0.29). Linking co-methylation with the functional similarity of genes showed that genes with r ≥ 0.8 tend to have similar molecular function and to be involved in the same biological process as described in the Gene Ontology project. Clustering of highly co-methylated genes identified four enriched KEGG pathways. Hence we have introduced co-methylation as a new indicator to discover functional associations between gene pairs in breast cancer and furthermore to discover new candidate genes that should be inspected more closely in the context of the studied disease.