Identification of cancer prognosis-associated functional modules using differential co-expression networks

// Wenshuai Yu 1 , Shengjie Zhao 1, 2 , Yongcui Wang 3 , Brian Nlong Zhao 4 , Weiling Zhao 5 and Xiaobo Zhou 6, 7, 8 1 Key Laboratory of Embedded System and Service Computing, College of Electronics and Information Engineering, The Ministry of Education, Tongji University, Shanghai, China 2 College of Software Engineering, Tongji University, Shanghai, China 3 Key Laboratory of Adaptation and Evolution of Plateau Biota, Northwest Institute of Plateau Biology, Chinese Academy of Sciences, Xining, China 4 Shanghai High School International Division, Shanghai, China 5 Department of Radiology and Comprehensive Cancer Center, Wake Forest University School of Medicine, Winston Salem, NC, USA 6 College of Electronics and Information Engineering, Tongji University, Shanghai, China 7 Center for Big Data Sciences and Network Security, Tongji University, Shanghai, China 8 Center for Bioinformatics and System Biology, Wake Forest University School of Medicine, Winston Salem, NC, USA Correspondence to: Shengjie Zhao, email: shengjiezhao@tongji.edu.cn Xiaobo Zhou, email: zhouxb2015@163.com Keywords: co-expression network; prognosis; HO-GSVD; gene module; cancer Abbreviations: DAVID: the Database for Annotation, Visualization, and Integrated Discovery; TCGA: The Cancer Genome Atlas; KEGG: Kyoto Encyclopedia of Genes and Genomes Received: July 12, 2017      Accepted: November 15, 2017      Published: December 04, 2017 ABSTRACT The rapid accumulation of cancer-related data owing to high-throughput technologies has provided unprecedented choices to understand the progression of cancer and discover functional networks in multiple cancers. Establishment of co-expression networks will help us to discover the systemic properties of carcinogenesis features and regulatory mechanisms of multiple cancers. Here, we proposed a computational workflow to identify differentially co-expressed gene modules across 8 cancer types by using combined gene differential expression analysis methods and a higher-order generalized singular value decomposition. Four co-expression modules were identified; and oncogenes and tumor suppressors were significantly enriched in these modules. Functional enrichment analysis demonstrated the significantly enriched pathways in these modules, including ECM-receptor interaction, focal adhesion and PI3K-Akt signaling pathway. The top-ranked miRNAs (mir-199, mir-29, mir-200) and transcription factors ( FOXO4 , E2A , NFAT , and MAZ ) were identified, which play an important role in deregulating cellular energetics; and regulating angiogenesis and cancer immune system. The clinical significance of the co-expressed gene clusters was assessed by evaluating their predictability of cancer patients’ survival. The predictive power of different clusters and subclusters was demonstrated. Our results will be valuable in cancer-related gene function annotation and for the evaluation of cancer patients’ prognosis.