Identification of Subpathway Signatures for Ovarian Cancer Prognosis by Integrated Analyses of High-Throughput miRNA and mRNA Expression

Abstract

Background/Aims: Ovarian cancer (OC) causes more death and serious conditions than any other female reproductive cancers, and many expression signatures have been identified for OC prognoses. However, no significant overlap is found among signatures from different studies, indicating the necessity of signature identifications at the functional level. Methods: We performed an integrated analyses of miRNA and gene expressions to identify OC prognostic subpathways (pathway regions). Using The Cancer Genome Atlas data set, we identified core prognostic subpathways, and calculated subpathway risk scores using both miRNA and gene components. Finally, we performed global risk impact analyses to optimize core subpathways using the random walk algorithm. Results: Subpathway-level analyses displayed more robust results than the gene- and miRNA-level analyses. Moreover, we verified the advantage of core subpathways over the entire pathway-based results and their prognostic performance in two independent validation data sets. Based on the global impact score, 13 subpathway signatures were selected and a combined subpathway-based risk score was further calculated for OC patient prognoses. Conclusions: Overall, it was possible to systematically perform integrated analyses of the expression levels of miRNAs and genes to identify prognostic subpathways and infer subpathway risk scores for use in OC clinical applications.