Whole Genome Analysis and Prognostic Model Construction Based on Alternative Splicing Events in Endometrial Cancer

• Main Authors: Caixia Wang, Mingjun Zheng, Shuang Wang, Xin Nie, Qian Guo, Lingling Gao, Xiao Li, Yue Qi, Juanjuan Liu, Bei Lin
• Format: Article
• Language: English
• Published: Hindawi Limited 2019-01-01
• Series: BioMed Research International
• Online Access: http://dx.doi.org/10.1155/2019/2686875

Objectives. A growing body of evidence has shown that aberrant alternative splicing (AS) is closely related to the occurrence and development of cancer. However, prior studies mainly have concentrated on a few genes that exhibit aberrant AS. This study aimed to determine AS events through whole genome analysis and construct a prognostic model of endometrial cancer (EC). Methods. We downloaded gene expression RNAseq data from UCSC Xena, and seven types of AS events from TCGA SpliceSeq. Univariate Cox regression was employed to analyze the prognostic-related alternative splicing events (PASEs) and splicing factors; multivariate Cox regression was conducted to analyze the effect of risk score (All) and clinicopathological parameters on EC prognosis. An underlying interaction network of PASEs of EC was constructed by Cytoscape Reactome FI, GO, and KEGG pathway enrichment was performed by DAVID. ROC curves and Kaplan-Meir analysis were used to assess the diagnostic value of prognostic model. The correlation between PASEs and splicing factors was analyzed by GraphPad Prism; then a network was constructed using Cytoscape. Results. In total, 28,281 AS events in EC were identified, which consisted of 1166 PASEs. RNPS1, NEK2, and CTNNB1 were the hub genes in the network of the top 600 PASEs. The area under the curve (AUC) of risk score (All) reached 0.819. Risk score (All) together with FIGO stage, cancer status, and primary therapy outcome success was risk factors of the prognosis of EC patients. Splicing factors YBX1, HNRNPDL, and HNRNPA1 were significantly related to the overall survival (OS). The splicing network indicated that the expression of splicing factors was significantly correlated with percent-splice-in (PSI) value of PASEs. Conclusion. We constructed a model for predicting the prognosis of EC patients based on PASEs using whole genome analysis of AS events and thereby provided a reliable theoretical basis for EC clinical prognosis evaluation.