Identification of a Ubiquitin Related Genes Signature for Predicting Prognosis of Prostate Cancer

Abstract

Background: Ubiquitin and ubiquitin-like (UB/UBL) conjugations are one of the most important post-translational modifications and involve in the occurrence of cancers. However, the biological function and clinical significance of ubiquitin related genes (URGs) in prostate cancer (PCa) are still unclear. Methods: The transcriptome data and clinicopathological data were downloaded from The Cancer Genome Atlas (TCGA), which was served as training cohort. The GSE21034 dataset was used to validate. The two datasets were removed batch effects and normalized using the "sva" R package. Univariate Cox, LASSO Cox, and multivariate Cox regression were performed to identify a URGs prognostic signature. Then Kaplan-Meier curve and receiver operating characteristic (ROC) curve analyses were used to evaluate the performance of the URGs signature. Thereafter, a nomogram was constructed and evaluated. Results: A six-URGs signature was established to predict biochemical recurrence (BCR) of PCa, which included ARIH2, FBXO6, GNB4, HECW2, LZTR1 and RNF185. Kaplan-Meier curve and ROC curve analyses revealed good performance of the prognostic signature in both training cohort and validation cohort. Univariate and multivariate Cox analyses showed the signature was an independent prognostic factor for BCR of PCa in training cohort. Then a nomogram based on the URGs signature and clinicopathological factors was established and showed an accurate prediction for prognosis in PCa. Conclusion: Our study established a URGs prognostic signature and constructed a nomogram to predict the BCR of PCa. This study could help with individualized treatment and identify PCa patients with high BCR risks.

Keywords: bioinformatics; prognosis; prognostic signature; prostate cancer; ubiquitin.

FIGURE 1

The flowchart of the study procedures. PCa, prostate cancer; TCGA, the Cancer Genome Atlas; URGs, ubiquitin related genes; LASSO, the least absolute shrinkage and selection operator; ROC, receiver operating characteristic.

FIGURE 2

Selection of prognostic URGs by LASSO regression. (A) LASSO coefficient profiles of the prognostic genes. (B) Parameter selection in LASSO model. URGs, ubiquitin related genes; LASSO, the least absolute shrinkage and selection operator.

FIGURE 3

Multivariate Cox regression analysis to identify prognostic genes.

FIGURE 4

Evaluation and validation of the predictive value of URGs signature in the TCGA dataset and GSE21034 dataset. (A) Kaplan-Meier curve analysis between high-risk and low-risk subgroups in the TCGA dataset. The subgroups were stratified by the optimal cut-off value for the risk scores. (B) Kaplan-Meier curve analysis between high-risk and low-risk subgroups in the GSE21034 dataset. The subgroups were stratified by the optimal cut-off value for the risk scores. (C) The AUCs under ROC for first, second, third, fourth and fifth year BCR predictions based on URGs signature in the TCGA dataset. (D) The AUCs under ROC for first, second, third, fourth and fifth year BCR predictions based on URGs signature in GSE21034 dataset. (E, G, I) The distribution of survival status and risk score, and heat map of prognostic genes expression in the TCGA dataset. (F, H, J) The distribution of survival status and risk score, and heat map of prognostic genes expression in the GSE21034 dataset. URGs, ubiquitin related genes; TCGA, the Cancer Genome Atlas; AUC, area under ROC curve; ROC, receiver operating characteristic; BCR, biochemical recurrence.

FIGURE 5

Kaplan-Meier curve analysis of PCa patients in different clinicopathological stratifications in TCGA dataset. (A) Age>65. (B) Age ≤ 65. (C) Gleason score>7. (D) Gleason score ≤ 7. (E) PSA value >10. (F) PSA value <=10. (G) Pathological T stage: T2. (H) Pathological T stage: T3-T4. (I) Negative SMS. (J) Positive SMS. PCa, prostate cancer; PSA, prostate specific antigen; SMS, surgical margin status.

FIGURE 6

The differential distribution of URGs signature risk scores between different clinicopathological variables and survival status in the TCGA dataset. (A) Age. (B) GGS. (C) PSA. (D) pT. (E) SMS. (F) BCR. URGs: ubiquitin related genes; GGS, Gleason grade score; PSA, prostate specific antigen; pT, pathological T stage; SMS, surgical margin status; BCR, biochemical recurrence.

FIGURE 7

Evaluation of the independent prognostic parameters based on clinicopathological variables and the signature, and construction of a nomogram in the TCGA dataset. (A) Univariate Cox analysis and (B) multivariate Cox analysis for evaluating independent prognostic factors. (C) Nomogram for predicting 1-, 3-, 5-year BCR-free survival of PCa patients. TCGA, the Cancer Genome Atlas; PSA, prostate specific antigen; SMS, surgical margin status; pT, pathological T stage; GGS, Gleason grade score; BCR, biochemical recurrence.

FIGURE 8

Validation of the nomogram. (A–C) the calibration curve of the nomogram for predicting 1-, 3-, 5-year BCR of the PCa patients in the TCGA dataset. (D–F) the calibration curve of the nomogram for predicting 1-, 3-, 5-year BCR of the PCa patients in GSE21034 dataset. (G) Kaplan-Meier curve analysis based the nomogram in the TCGA dataset. (H) Kaplan-Meier curve analysis based the nomogram in GSE21034 dataset. (I) ROC curve analysis for predicting 1-, 2-, 3-, 4-, and5-year BCR of the PCa patients in the TCGA dataset. (J) ROC curve analysis for predicting 1-, 2-, 3-, 4-, and 5-year BCR of the PCa patients in GSE21034 dataset. BCR, biochemical recurrence; PCa, prostate cancer; TCGA, the Cancer Genome Atlas; AUC, area under ROC curve; ROC, receiver operating characteristic.