A four-gene model for prognostic prediction in bladder urothelial carcinoma

Abstract

Background

Bladder urothelial carcinoma (BLCA) is one of the most common malignant tumors in urinary system worldwide. High possibility of recurrence and progression leads to poor prognosis, revealing the significant role of long-term postoperative monitoring to the patients. However, effective noninvasive diagnosis is currently limited.

Materials and methods

Differentially expressed genes (DEGs) were analyzed using R-packages. Functional enrichment analyses were performed on Metascape. The prognostic model was established by multi-step cox regression and evaluated by survival plots and receiver operating characteristic (ROC) curves. The nomogram was then constructed based on three identified prognostic factors. C-index and calibration curves were calculated to testify the capacity for predicting survival possibility of BLCA patients. The transcription levels of model genes were further verified in a gemcitabine-resistant bladder cancer cell line T24 (TGR) by quantitative real-time PCR (qRT-PCR).

Results

360 genes were differentially expressed between BLCA and normal bladder mucosae simultaneously in three GEO datasets, of which 59 were up-regulated and 301 were down-regulated. 159 prognostic genes were obtained from DEGs. Lasso and multivariate cox regression were conducted in sequence and the prognostic model was eventually optimized to four genes (EHBP1, RHOJ, FASN, STXBP6). Survival analyses demonstrated that the overall survival (OS) of patients in high-risk group was significantly shorter than that in low-risk group. The area under the curve (AUC) values of 3–5 years survival were basically above 0.7. Moreover, cox regression analyses showed that age, T stage and risk score were independent indicators for BLCA prognosis. For further clinical application, a nomogram was then constructed by integrating these factors. The C-index (0.72, CI 95 %, 0.669–0.775) and calibration curves demonstrated the good performance of nomogram. Importantly, the mRNA level of model genes was significantly up-regulated in TGR compared to T24, indicating a better prediction for chemotherapy-resistant BLCA patients.

Conclusion

Collectively, our findings suggest a novel four-gene predictive model for BLCA prognosis. It is expected to provide a valuable reference for prognostic evaluation and treatment in BLCA patients.