Functional Clusters Analysis And Research Based On Differential Coexpression Networks

In the era of postgenomics,one of the most important purpose of bioinformatics is to help people understand relationship between molecules in biological cells,and reveal the intermolecular interaction between molecules and the internal mechanism of controlling cell function.After nearly twenty years of development,genetic studies has transformed to gene network research from single gene.Right now,differential coexpression analysis has gradually become an important approach to improve the conventional method of analyzing differentially expressed genes.With this approach,it is possible to discover disease mechanisms and underlying regulatory dynamics which remain obscure in differential expression analysis.The detection of differential coexpression links and functional clusters between different disease states is a demanding task,in order to effectively study this dynamic regulation mechanism,many differential coexpression network research methods have been studied and applied,nevertheless,there is no gold standard for detecting differential coexpression links and functional clusters,consequently,we developed a novel fusion algorithm FDvDe(Fusion of differential vertex and differential edge)to detect differential coexpression links by aggregating the set of “differential vertex”(the gene pairs whose correlation coefficient are different between the study group and control group)and “differential edge”(the genes whose topology property are different between the study group and control group).Then,in order to study the hidden value information,we constructed differential coexpression networks between normal and tumour states by integrating the differential coexpression links.With this approach,we identified 1823 genes and 29370 links.Then,we developed the algorithms GTHC(GO term hierarchical clusters)to identify functional modules.The distance matrix used in the hierarchical process was formed by the GO semantic similarity.Furthermore,we aggregated the densities among clusters describing the connectivity among clusters and topological property analysis to discover the hub genes and hub pathways which play an important role in disease mechanism.In this paper,we showed that our approach worked well on a data set of breast cancer samples(68 tumour samples)and normal samples(73 normal samples),and revealed that the research of the coexpression network is meaningful to understand the disease mechanisms,and we obtain the modules having biological significance and hub genes found in this approach to support the above conclusion.