Identification of a Novel Protein-Based Prognostic Model in Gastric Cancers

Abstract

Gastric cancer (GC) is the third leading cause of cancer-related deaths worldwide. However, there are still no reliable biomarkers for the prognosis of this disease. This study aims to construct a robust protein-based prognostic prediction model for GC patients. The protein expression data and clinical information of GC patients were downloaded from the TCPA and TCGA databases, and the expressions of 218 proteins in 352 GC patients were analyzed using bioinformatics methods. Additionally, Kaplan-Meier (KM) survival analysis and univariate and multivariate Cox regression analysis were applied to screen the prognosis-related proteins for establishing the prognostic prediction risk model. Finally, five proteins, including NDRG1_pT346, SYK, P90RSK, TIGAR, and XBP1, were related to the risk prognosis of gastric cancer and were selected for model construction. Furthermore, a significant trend toward worse survival was found in the high-risk group (p = 1.495 × 10-7). The time-dependent ROC analysis indicated that the model had better specificity and sensitivity compared to the clinical features at 1, 2, and 3 years (AUC = 0.685, 0.673, and 0.665, respectively). Notably, the independent prognostic analysis results revealed that the model was an independent prognostic factor for GC patients. In conclusion, the robust protein-based model based on five proteins was established, and its potential benefits in the prognostic prediction of GC patients were demonstrated.

Keywords: TCPA; TIGAR; gastric cancer; overall survival.

Figure 1

Survival analysis for each model protein. Low SYK expression (A), NDRG1-PT346 (B), P90RSK (C), and TIGAR (D) and high XBP1 expression (E) were positively associated with poorer overall survival of STAD patients.

Figure 2

Construction of a protein risk prognostic model in STAD. (A) The patients were divided into high-risk and low-risk groups based on the risk scores. The risk scores for all patients were ranked in ascending order and were divided by the threshold (vertical dotted line). The green dots in the left and red dots in the right indicated the low- and high-risk groups respectively. (B) The scatter diagram showed the survival time and survival status of each patient in high- and low-risk groups. The red and green dots represented death and alive respectively. The high-risk group had more deaths compared to low-risk group. (C) The risk heatmap showed the expression levels of five prognostic-risk model proteins between high-risk and low-risk groups. The dark red and green represented higher expression and lower expression respectively.

Figure 3

The protein risk prognostic model had good prognostic predictive value and was an independent prognostic factor for STAD patients. (A) The survival analysis for STAD patients between high-risk and low-risk groups. (B) Time-dependent ROC of the risk model; the AUC was assessed at 1, 2, and 3 years (AUC = 0.685, 0.673, and 0.665). (C–E) Time-dependent ROC curve analysis revealed that the predictive model has higher accuracy and sensitivity compared to the clinical features. (F,G) The univariate and multivariate Cox analysis revealed that this risk model was an independent prognostic factor for STAD patients.

Figure 4

Correlation analyses between model proteins and other proteins based on the TCPA database.