Screening and Validation of Independent Predictors of Poor Survival in Pancreatic Cancer.

Pancreatic cancer is a digestive system malignant tumor with high mortality and poor prognosis, but the mechanisms of progression remain unclear in pancreatic cancer. It's necessary to identify the hub genes in pancreatic cancer and explore the novel potential predictors in the prognosis of pancreatic cancer. We downloaded two mRNA expression profiles from Gene Expression Omnibus and The Cancer Genome Atlas Pancreatic Cancer (TCGA-PAAD) datasets to screen the commonly differentially expressed genes in pancreatic cancer by limma package in R. Subsequently, measurement of the functional similarity among the 38 DEGs in common was performed to identify the hub genes using GOSemSim package. Then, survival analysis and Cox regression were applied to explore prognosis-related hub genes using the survival package. Statistics analysis by two-tailed Student's t-test or one-way based on TCGA-PAAD datasets and qPCR detection in clinical samples were performed to explore the correlations between expression of hub genes in pancreatic cancer tissues and clinical parameters. Based on integrated analysis of TCGA and GEO datasets, we screened 38 DEGs in common, which were all up-regulated. The functional similarity results showed that 10 DEGs including TSPAN1, MSLN, C1orf116, PKP3, CEACAM6, BAIAP2L1, PPL, RAB25, ERBB3, and AP1M2 in the DEGs in common, which had the higher average functional similarity, were considered as the hub genes. Survival analysis results and Cox regression analysis showed that TSPAN1, CEACAM6, as well as ERBB3 were all associated with poor overall survival of PC. qPCR results showed that the expression levels of TSPAN1 and ERBB3 were significantly upregulated in the PC tissues. The statistical analysis results revealed that TSPAN1 expression correlated significantly with histologic grade, T stage, clinical stage, and vital status by two-tailed Student's t-test or one-way ANOVA; ERBB3 expression correlated significantly with T stage, clinical stage, and vital status by two-tailed Student's t-test or one-way ANOVA. We found that TSPAN1 and ERBB3 could be independent predictors of poor survival in pancreatic cancer.