Bioinformatics Analysis of a Prognostic miRNA Signature and Potential Key Genes in Pancreatic Cancer

Abstract

Background: In this study, miRNAs and their critical target genes related to the prognosis of pancreatic cancer were screened based on bioinformatics analysis to provide targets for the prognosis and treatment of pancreatic cancer.

Methods: R software was used to screen differentially expressed miRNAs (DEMs) and genes (DEGs) downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases, respectively. A miRNA Cox proportional hazards regression model was constructed based on the miRNAs, and a miRNA prognostic model was generated. The target genes of the prognostic miRNAs were predicted using TargetScan and miRDB and then intersected with the DEGs to obtain common genes. The functions of the common genes were subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) analyses. A protein-protein interaction (PPI) network of the common genes was constructed with the STRING database and visualized with Cytoscape software. Key genes were also screened with the MCODE and cytoHubba plug-ins of Cytoscape. Finally, a prognostic model formed by the key gene was also established to help evaluate the reliability of this screening process.

Results: A prognostic model containing four downregulated miRNAs (hsa-mir-424, hsa-mir-3613, hsa-mir-4772 and hsa-mir-126) related to the prognosis of pancreatic cancer was constructed. A total of 118 common genes were enriched in two KEGG pathways and 33 GO functional annotations, including extracellular matrix (ECM)-receptor interaction and cell adhesion. Nine key genes related to pancreatic cancer were also obtained: MMP14, ITGA2, THBS2, COL1A1, COL3A1, COL11A1, COL6A3, COL12A1 and COL5A2. The prognostic model formed by nine key genes also possessed good prognostic ability.

Conclusions: The prognostic model consisting of four miRNAs can reliably predict the prognosis of patients with pancreatic cancer. In addition, the screened nine key genes, which can also form a reliable prognostic model, are significantly related to the occurrence and development of pancreatic cancer. Among them, one novel miRNA (hsa-mir-4772) and two novel genes (COL12A1 and COL5A2) associated with pancreatic cancer have great potential to be used as prognostic factors and therapeutic targets for this tumor.

Keywords: Gene Expression Omnibus; The Cancer Genome Atlas; biomarkers; miRNAs; pancreatic cancer; target genes.

Figure 1

Volcano diagrams of DEMs (A) and DEGs (B). Volcano diagrams show the P value and the fold change of differentially expressed miRNAs and genes. Green and red circles represent downregulated and upregulated miRNAs or genes, respectively.

Figure 2

Construction of prognostic model based on four prognostic miRNAs. (A) The risk score curve, survival status and heatmap are shown from top to bottom. (B) Survival curve. (C). ROC curve of the prognostic model.

Figure 3

Functional enrichment analysis of 118 common genes. The x-axis represents the P value, and the y-axis represents the pathways and annotations. The bubble size increases with the number of enriched genes.

Figure 4

PPI network diagram. (A) PPI network of 118 common genes. (B) Results of the cytoHubba topological analysis. Different node colors represent the logFC value of the DEGs. (C) MCODE network module diagram. The shade and shallowness of the red lines represent the combined score between proteins.

Figure 5

The miRNA-gene pathway and annotation networks represent the relationships among miRNAs, common genes, key genes and gene enrichment results. MiRNAs, common genes, key genes and enrichment terms are represented by red, blue, green and orange circles, respectively.

Figure 6

Construction of a prognostic model based on nine key genes. (A) The risk score curve, survival status and heatmap are shown from top to bottom. (B) Survival curve. (C) ROC curve of the prognostic model.