A seven-gene prognosis model to predict biochemical recurrence for prostate cancer based on the TCGA database.

Abstract

The incidence rate of prostate cancer is increasing rapidly. This study aims to explore the gene-associated mechanism of prostate cancer biochemical recurrence (BCR) after radical prostatectomy and to construct a biochemical recurrence of prostate cancer prognostic model. The DEseq2 R package was used for the differential expression of mRNA. The ClusterProfiler R package was used to analyze the functional enrichment of Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) to explore related mechanisms. The Survival, Survminer, and My.stepwise R packages were used to construct the prognostic model to predict the biochemical recurrence-free probability. The RMS R package was used to draw the nomogram. For evaluating the prognostic model, the timeROC R package was used to draw the time-dependent ROC curve (receiver operating characteristic curve). To investigate the association between mRNA and prostate cancer, we performed differential expression analysis on the TCGA (The Cancer Genome Atlas) database. Seven protein-coding genes (VWA5B2, ARC, SOX11, MGAM, FOXN4, PRAME, and MMP26) were picked as independent prognostic genes by regression analysis. Based on their Cox coefficient, a risk score formula was proposed. According to the risk scores, patients were divided into high- and low-risk groups based on the median score. Kaplan-Meier plot curves showed that the low-risk group had a better biochemical recurrence-free probability compared to the high-risk group. The 1-year, 3-year, and 5-year AUCs (areas under the ROC curve) of the model were 77%, 81%, and 86%, respectively. In addition, we built a nomogram based on the result of multivariate Cox regression analysis. Furthermore, we select the GSE46602 dataset as our external validation. The 1-year, 3-year, and 5-year AUCs of BCR-free probability were 83%, 82%, and 80%, respectively. Finally, the levels of seven genes showed a difference between PRAD tissues and adjacent non-tumorous tissues. This study shows that establishing a biochemical recurrence prediction prognostic model comprising seven protein-coding genes is an effective and precise method for predicting the progression of prostate cancer.