Genome-Wide Network-Based Analysis of Colorectal Cancer Identifies Novel Prognostic Factors and an Integrative Prognostic Index

• Main Authors: Xiaolin Hou, Xuelai He, Kang Wang, Nengyi Hou, Junwen Fu, Guiqing Jia, Xiaofei Zuo, Haibo Xiong, Minghui Pang
• Format: Article
• Language: English
• Published: Cell Physiol Biochem Press GmbH & Co KG 2018-09-01
• Series: Cellular Physiology and Biochemistry
• Subjects: Colorectal cancer; Prognosis; Prognostic factor; Genome-wide; Network analysis
• Online Access: https://www.karger.com/Article/FullText/493614
• ISSN: 1015-8987; 1421-9778

Background/Aims: Colorectal cancer (CRC) is one of leading cancers in both incidence and mortality rate. The 5-year survival rate varies considerably depending on the pathological stage of the tumor. Although prominent progress has been made through screening for survival-associated factors from a certain type of genetic or epigenetic modifications, few attempts have been made to apply a network-based approach in prognostic factor identification, which could prove valuable for a complex, multi-faceted disease such as CRC. Methods: In this study, a TCGA dataset of 379 CRC patients was subjected to a network-based analysis strategy consisting of multivariate regression, co-expression network and gene regulatory network analyses, and survival analyses. Both genetic and epigenetic aberrations, including those in gene expression and DNA methylation at specific sites, were screened for significant association with patient survival. A prognostic index (PI) integrating all potential prognostic factors was subsequently validated for its prognostic value. Results: A collection of six miRNAs, eleven mRNAs, and nine DNA methylation sites were identified as potential prognostic factors. The low- and high-risk patient groups assigned based on PI level showed significant difference in overall survival (hazard ratio = 1.32, 95% confidence interval 1.29-1.36, p < 0.0001). Patients in the low- and high-risk groups can be further divided into a total of four subgroups, based on pathological staging. In the two high-risk subgroups (PI > 0), there was significant different (Cox p < 0.0001) in OS between the earlier (stages I/II) and later stages (stages III/IV). However, in the two low-risk subgroups (PI < 0), earlier (stages I/II) and later stages (stages III/IV) showed no significant difference in OS (Cox p = 0.185). On the other hand, there were significant differences in OS between the low- and high-risk subgroups when both subgroups were of earlier stages (Cox p < 0.001) or of later stages (Cox p < 0.0001). Conclusion: The novel network-based, integrative analysis adopted in this study was efficient in screening for prognostic predictors. Along with PI, the set of 6 miRNAs, 11 mRNAs, and 9 DNA methylation sites could serve as the basis for improved prognosis estimation for CRC patients in future clinical practice.