Screening Protein Prognostic Biomarkers for Stomach Adenocarcinoma Based on The Cancer Proteome Atlas

Abstract

The objective was to construct a prognostic risk model of stomach adenocarcinoma (STAD) based on The Cancer Proteome Atlas (TCPA) to search for prognostic biomarkers. Protein data and clinical data on STAD were downloaded from the TCGA database, and differential expressions of proteins between carcinoma and para-carcinoma tissues were screened using the R package. The STAD data were randomly divided into a training set and a testing set in a 1:1 ratio. Subsequently, a linear prognostic risk model of proteins was constructed using Cox regression analysis based on training set data. Based on the scores of the prognostic model, sampled patients were categorized into two groups: a high-risk group and a low-risk group. Using the testing set and the full sample, ROC curves and K-M survival analysis were conducted to measure the predictive power of the prognostic model. The target genes of proteins in the prognostic model were predicted and their biological functions were analyzed. A total of 34 differentially expressed proteins were screened (19 up-regulated, 15 down-regulated). Based on 176 cases in the training set, a prognostic model consisting of three proteins (COLLAGEN VI, CD20, TIGAR) was constructed, with moderate prediction accuracy (AUC=0.719). As shown by the Kaplan-Meier and survival status charts, the overall survival rate of the low-risk group was better than that of the high-risk group. Moreover, a total of 48 target proteins were identified to have predictive power, and the level of proteins in hsa05200 (Pathways in cancer) was the highest. According to the results of the Univariate and multivariate COX analysis, tri-protein was identified as an independent prognostic factor. Therefore, the tri-protein prognostic risk model can be used to predict the likelihood of STAD and guide clinical treatment.