Construction and validation of a prognostic risk model for uterine corpus endometrial carcinoma based on alternative splicing events*

Abstract

Abstract Objective To establish a prognostic risk model for uterine corpus endometrial carcinoma (UCEC) based on alternative splicing (AS) event data from The Cancer Genome Atlas (TCGA) and assess the accuracy of the model. Methods TCGA and SpliceSeq databases were used to acquire a summary of AS events and clinical data related to UCEC. Bioinformatic analysis was performed to identify differentially expressed AS events in UCEC. Least absolute shrinkage and selection operator (LASSO) regression and multivariate Cox regression analyses were used for constructing a prognostic risk model. Next, using the receiver operating characteristic (ROC) curve, Kaplan-Meier survival analysis, and independent prognostic analysis, we assessed the accuracy of the model. In addition, a splicing network was established based on the association between potential splicing factors and AS events. Results We downloaded clinical data and AS events of 527 UCEC cases from TCGA and SpliceSeq databases, respectively. We obtained 18,779 survival-associated AS events in UCEC using univariate Cox regression analysis and 487 AS events using LASSO regression analysis. Multivariate Cox regression analysis established a prognostic risk model for UCEC based on the percentage splicing value of 13 AS events. Independent prognostic effect on UCEC risk was then assessed using multivariate and univariate Cox regression analyses (P < 0.001). The area under the curve was 0.827. The pathological stage and risk score were independent prognostic factors for UCEC. Herein, we established a regulatory network between alternative endometrial cancer-related splicing events and splicing factors. Conclusion We constructed a prognostic model of UCEC based on 13 AS events by analyzing datasets from TCGA and SpliceSeq databases with medium accuracy. The pathological stage and risk score were independent prognostic factors in the prognostic risk model.