Construction Of Cancer-specific Gene Regulatory Networks

Cancer is a complex disease,and the process by which it arises is linked to a variety of factors.It is often accompanied by aberrant gene transcriptional regulation.Therefore,studying the transcriptional regulation of genes in cancer can contribute to the understanding of the complex mechanisms of cancer.Currently,more attention has been paid to the general gene regulation relationships in pan-cancer or a particular cancer type,while little research has been done on the cancer-specific gene regulation.Recent studies have shown that different cancer types reflect specific chromatin accessibility,and that open-state chromatin is often a prerequisite for transcriptional regulation to occur.This makes it possible to study cancer-specific gene regulation relationships.In this paper,we innovatively propose an idea for constructing cancer-specific gene regulatory networks and develop the computational framework.The computational framework mines cancer-specific open chromatin fragments and chromatin 3D interactions through integrated analysis of genomic,transcriptomic and 3D genomic data.On this basis,linear proximity gene regulatory relationships in the genome are inferred based on the cancer-specific open chromatin fragments,and remote gene regulatory relationships are inferred based on the cancer-specific chromatin 3D spatial interactions.By integrating proximity and remote gene regulation and combining gene expression information in regulatory relationships,a reliable cancer type-specific gene regulation network is constructed.In this paper,the computational framework was applied to construct cancer-specific gene regulatory networks for breast cancer,prostate cancer and colorectal cancer respectively.By analyzing all genes in the networks,392,88 and 86 genes in the three cancer networks were found to be differentially expressed in disease samples and normal samples.This result was shown to be significant(p-value<0.05)by using hypergeometric test,reflecting the value of the network for cancer risk gene prediction.Analysis of the regulatory edges and gene nodes in the network based on existing biological database validation revealed 4,1 and 2 regulatory relationships were validated,and 2,4 and 2 gene nodes were covered by cancer genes validated by biological experiments,respectively.This result was achieved by focusing only on the specificity case,indicating the high reliability of the cancer network in this paper.By topological analysis,9,12 and 13 important genes were screened in the three cancers,of which 5,6 and 8 were cancer-causing genes that have been reported in the literature,respectively.The remaining genes could be useful for guiding bio-validation experiments of cancer-related risk genes.GO function and Reactome pathway enrichment analysis of the network modules showed that the gene modules in the network have significant cancerrelated biological functions.This result has a potential functional predictive role for genes in the modules whose functions are currently unknown.In summary,the work in this paper indicates that the idea of constructing a cancer-specific transcriptional regulatory network is feasible.This network will help to identify risk factors associated with cancer and will have a positive effect on the understanding of cancer pathogenesis,and hence diagnosis and treatment.The computational framework proposed in this paper enables the fusion of chromatin state and structure information with gene expression and function information.Following the fundamental law that structure determines function,the accessibility of chromatin,linear and three-dimensional structure of chromatin and gene expression are analyzed through data integration to construct cancer-specific gene regulatory network innovatively.The analysis and mining on this specific network needs to be further developed,which is expected to provide new ideas and references for cancer research.