Enhancer-gene networks for the identification of cancer driver genes affected by enhancer mutations

Motivation: Reversible epigenetic modifications that happen on the DNA’s histones, namely histone modifications, play an important role in gene regulation by controlling the accessibility of different functional genomic regions. Such modifications have been measured and primarily studied on genes or their promoters, but a currently interesting and less studied category of functional elements are enhancers. Enhancer regions can be found virtually anywhere along our non-coding DNA and through looping towards the promoters of target genes they contribute to their normal ex-pression patterns. Abnormal epigenetic signatures and somatic mutations in those regions can interfere with the en-hancer’s looping procedure and result in irregular gene expression patterns, a crucial contributor in cancer develop-ment. Results: In this paper we propose a method which utilizes epigenetic information across multiple cell types, to form reliable enhancer – gene pairs. The formation of each pair is based on the multiple correlation between epigenetic mark enrichment of an enhancer set and a gene’s expression. The pairing procedure deals with the increased computational requirements caused by the multi-dimensional nature of the data and constructs pairs which follow the literature notion of frequent linearly proximal enhancers. The distribution of distances of each pair’s elements, showed different results in regulation proximity between our method’s pairs and a number of randomized sets of pairs. These pairs are assem-bled into multiple enhancer-gene (EG) networks which include multiple connected subnetworks of different sizes. On each EG network we overlaid non-coding somatic mutations and found that enhancers of cancer census genes have a higher percentage of mutated enhancers as well as more regulating enhancers, than non-cancer genes. The mutation percentages and the within-pair distances showed different behaviour among chromosomes as well as epigenetic marks. Finally, we inserted these into network databases so that they can accept queries and be easily extendable. Conclusion: The core of this paper deals with the subject of enhancer cis-regulation and we have performed analyses in order to link regulating enhancers to their target genes, a recent and increasingly important task since the availability of high quality data. The subsequent EG networks depict the circuitry of regulation within a cell’s nucleus as well as the mutational landscape and are excellent candidates for further exploration of either the role of specific enhancers in gene misregulation during cancer or more generic characteristics of cancer related enhancers that affect cancer driver genes.