Screening differential genes and prognostic analysis of gastric cancer based on TCGA database / 国际肿瘤学杂志

ABSTRACT

Objective:To extract the genes associated with prognosis from the differential expressed genes in gastric cancer tissues by using a large number of gastric cancer genome data in the cancer genome atlas (TCGA) database.Methods:Gene expression data of gastric adenocarcinoma were downloaded from TCGA database. After R language data preprocessing, edgeR was used to analyze the gene differential expression, and R language was used to identify the significant gene ontology (GO) terms and KEGG pathways in gene differential expression. Multivariate Cox stepwise regression analysis was used to predict the genes that affected survival. Genes obtained above were used for survival analysis online in Kaplan-Meier Plotter website (http //Kaplan-Meier Plotter.com).Results:A total of 305 gastric cancer and 30 normal gastric tissues were retrieved in TCGA database, and 3 231 differential genes were screened out, including 2 005 up-regulated genes and 1 226 down-regulated genes. These genes were enriched in GO terms including antigen binding, serine hydrolase activity, receptor ligands activity, serine peptidase activity, serine type endopeptidase activity, glycosaminoglycans binding, cytokine activity, hormone activity, peptidase inhibitor activity, metallopeptidase activity and so on. The genes in KEGG pathway analysis were enriched in chemical carcinogen, neuractive receptor-ligand interaction, cytokine-cytokine receptor interaction, metabolism of xenobiotics by cytochrome P450, protein digestion and absorption, staphylococcus aureus infection, retinol metabolism, drug metabolism P450, steroid hormone metabolism, pancreatic secretion and so on. Cox analysis showed that GPX3 and SERPINE1 had significant effect on the survival of gastric cancer patients. Receiver operating characteristic curve analysis showed that the expressions of GPX3 and SERPINE1 had a certain predictive value for the survival time of gastric cancer patients, when the critical values of GPX3 and SERPINE1 were 0.46 and 0.68 respectively, the sensitivity was 60.35%, the specificity was 82.06%, and the area under the curve was 0.763 (95% CI 0.828-0.936). Kaplan-Meier analysis showed that the high expressions of GPX3 ( P<0.001) and SERPINE1 ( P=0.001) were significantly related to the poor prognosis of gastric adenocarcinoma. Conclusion:The higher expression of SERPINE1 and GPX3 genes, the shorter survival time of gastric cancer patients. They may be the targets for predicting the prognosis of gastric cancer.