Gene signatures predict biochemical recurrence-free survival in primary prostate cancer patients after radical therapy

Abstract

Background: This study evaluated the predictive value of gene signatures for biochemical recurrence (BCR) in primary prostate cancer (PCa) patients.

Methods: Clinical features and gene expression profiles of PCa patients were attained from Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA) datasets, which were further classified into a training set (n = 419), a validation set (n = 403). The least absolute shrinkage and selection operator Cox (LASSO-Cox) method was used to select discriminative gene signatures in training set for biochemical recurrence-free survival (BCRFS). Selected gene signatures established a risk score system. Univariate and multivariate analyses of prognostic factors about BCRFS were performed using the Cox proportional hazards regression models. A nomogram based on multivariate analysis was plotted to facilitate clinical application. Kyoto Encyclopedia of Gene and Genomes (KEGG) and Gene Ontology (GO) analyses were then executed for differentially expressed genes (DEGs).

Results: Notably, the risk score could significantly identify BCRFS by time-dependent receiver operating characteristic (t-ROC) curves in the training set (3-year area under the curve (AUC) = 0.820, 5-year AUC = 0.809) and the validation set (3-year AUC = 0.723, 5-year AUC = 0.733).

Conclusions: Clinically, the nomogram model, which incorporates Gleason score and the risk score, could effectively predict BCRFS and potentially be utilized as a useful tool for the screening of BCRFS in PCa.

Keywords: LASSO-Cox regression; biochemical recurrence-free survival; gene signature; primary prostate cancer; radical therapy.

FIGURE 1

Selection strategy for gene signatures. (A) “Leave‐one‐out” cross‐validation for parameter selection in LASSO‐COX regression models, and the optimal λ value of 0.11517381 with log(λ) = −2.1613129 was selected; (B) six BCRFS‐associated gene signatures were selected by LASSO‐COX models. LASSO, least absolute shrinkage and selection operator method; BCRFS, biochemical recurrence‐free survival

FIGURE 2

Risk score distribution, BCRFS status, and expression pattern of BCRFS‐associated gene signatures in both cohorts. (A) The scattergram of the risk score in the training set; (B) the scattergram of the risk score in the validation set; (C) BCRFS time/BCR status in the training set; (D) BCRFS time/BCR status in the validation set; (E) the expression pattern of six BCRFS‐associated gene signatures in the training set; (F) the expression pattern of six BCRFS‐associated gene signatures in the validation set. BCRFS, biochemical recurrence‐free survival

FIGURE 3

Gene signatures can predict BCRFS in both cohorts. (A) K–M survival curves for the training set indicated that better BCRFS was associated with significantly lower risk score; (B) K–M survival curves for the validation set indicated that better BCRFS was associated with significantly lower risk score; (C) Time‐dependent ROC revealed that the risk score was an excellent predictor for BCRFS in the training set; (D) Time‐dependent ROC revealed that the risk score was an excellent predictor for BCRFS in the validation set. BCRFS, biochemical recurrence‐free survival; K–M, Kaplan–Meier; ROC, receiver operating characteristic

FIGURE 4

Nomogram prediction of BCRFS probability. The risk score and Gleason score were used to establish the nomogram for predicting 3 and 5‐year BCRFS in the training set. The dominant factor was the risk score. BCRFS, biochemical recurrence‐free survival

FIGURE 5

Calibration plots of the nomogram. (A) Three‐year calibration plot of nomogram in the training set; (B) 5‐year calibration plot of nomogram in the training set; (C) 3‐year calibration plot of nomogram in the validation set; (D) 5‐year calibration plot of nomogram in the validation set; the nomogram's performance was excellent for predicting the 3‐year BCRFS and 5‐year BCRFS in both cohorts. BCRFS, biochemical recurrence‐free survival

FIGURE 6

Bioinformatical analysis of three DEGs. (A) Three major categories were included in the bubble plots of GO analysis; (B) two enriched terms of KEGG pathway shown in bubble plot; (C) a chord plot was used to visualize the top three GO terms of BP, CC, and MF, respectively. DEGs, differentially expressed genes; GO, Gene Ontology; BP, biological process; CC, cellular component; MF, molecular function; KEGG, Kyoto Encyclopedia of Gene and Genomes