Regression analysis of data with repeated measurements using the method of successive differences

A Novel Cluster-Based Computational Method to Identify miRNA Regulatory Modules

The identification of miRNA regulatory modules can help decipher miRNAs combinatorial regulation effects on the pathogenesis underlying complex diseases, especially in cancer. By integrating miRNA/mRNA expression profiles and sequence-based predicted target site information, we develop a novel cluster-based computational method named CoModule for identifying miRNA regulatory modules (MRMs). The ultimate goal of CoModule is to detect the MRMs, in which the miRNAs in each module are expected to present cooperative mechanisms in regulating their targets mRNAs. Here, the co-expression of miRNAs are believed to present cooperative regulatory relationship, therefore, the critical step of CoModule is first to partition the miRNAs with similar expression into a cluster by employing rough set clustering. After gaining credible miRNA clusters, the targets of regulator are naturally added into corresponding clusters to produce the final miRNA regulatory modules. We apply this present method to ovarian cancer datasets and make a comparison with the other two existing prominent approaches. The results indicate that the modules identified by CoModule perform better than the other two methods ranging from the topological aspects to the biological function. Survival analysis detects a number of prognostic modules with statistical significance, which can help reveal the potential diagnostic for ovarian cancer.