Identifying diagnostic markers and constructing a prognostic model for pancreatic cancer based on microarray and bioinformatic analysis

Pancreatic cancer (PC) is one of the leading causes of cancer-related death worldwide. The lack of effective diagnostic biomarkers and therapeutic targets makes PC difficult to screen and treat. The aim of this study was to develop a diagnostic and survival-related gene signature for PC to construct a prognostic model. An Arraystar RNA microarray was used to identify differentially expressed genes (DEGs) in clinical plasma samples between the PC group and the control group. We performed weighted gene co-expression network analysis (WGCNA) to identify significant modules of DEGs in the Gene Expression Omnibus (GEO) cohort and to obtain potential diagnostic hub genes by intersecting the significant module genes with microarray-derived messenger RNA (mRNA). In addition, survival analysis and univariate and multivariate Cox regression analyses were performed on the hub genes to construct a prognostic model. Our microarray data revealed 228 significantly upregulated mRNA in the PC group compared with the control group. Moreover, we identified 5 feature mRNA (FERMT1, S100A14, KCNN4, PKM, and ITGA3) with good diagnostic performance. According to survival analysis based on The Cancer Genome Atlas (TCGA) dataset, higher expression of the hub genes was related to a poorer survival rate in patients with PC. Univariate and multivariate Cox proportional hazard analyses revealed that the expression of FERMT1, S100A14, and ITGA3 was anindependent risk factor for poor prognosis. Our results revealed the potential biomarkers for the prediction of PC prognosis in addition to clinicopathological factors. Moreover, this study provides new insights into the molecular mechanisms of PC.