The workflow of our study was depicted in Figure 1. Briefly, eight GEO datasets were divided into three groups, including 110 PC samples and 67 normal samples in GPL570, 120 PC samples, and 112 normal samples in GPL6244, 118 PC tissues, and 13 normal tissues in GPL13667. TCGA and GTEx database of PC were enrolled in 178 PC tissues and 171 normal tissues. The detail of each group can be found in Supplementary Table 1 (Table S1). Next, DEGs in each group were identified and displayed in the volcano plot with thresholds of |log2FC| >1 and adjust P value < 0.05 (Figure 2A–D). As depicted in Venn plots (Supplementary Figure 1A–B), we integrated commonly expressed genes that were intersected in each group, and successfully identified 170 up-regulated and 99 down-regulated DEGs. The top 10 up and down regulated DEGs identified by integrated analysis of all groups are shown in Figure 2E. Also, We provided the result of DEGs among four groups and common DEGs in supplementary Table 4 (Table S4), and DEGs in each group were chosen for the following enrichment analysis.
Enrichment analysis for DEGs
To detect the potential biological functions among DEGs in each group, GO enrichment analysis and KEGG pathway analysis was conducted. The result of the GO enrichment analysis was shown in Supplementary Figure 2A–D. It is well acknowledged that cancer-related pathways play key roles in the progression of PC, and genes with similar functions cluster together to form a regulatory pathway. So we try to figure out the similarity of the KEGG pathway among DEGs. As depicted in Figure 3A–D, the significant KEGG pathway with adjust P value < 0.05 were commonly enriched in the pathway in cancer, PI3K-Akt signaling pathway and Focal adhesion pathway, which indicated that those pathways were closely related to the progression of PC. There are 51 significant pathways which commonly enriched in every group (Figure 3E). As depicted in Supplementary Figure 3A–D, most of the DEGs were enriched in the pathway of pathway in cancer, which suggested that pathway in cancer played key roles in the progression of pancreatic cancer. Additionally, we focus on DEGs that enriched in the pathway of pathway in cancer and chosen for subsequent analysis.
Screening and validation of Hub genes
There are altogether 224 genes which enriched in the pathway of pathway in cancer. To understand the mutual interaction between DEGs and pathway in cancer, a PPI network was constructed. Also, we calculated the node degree of the PPI network using cytoHubba tools from Cytoscape software and identified the top 20 hub genes in the pathway of pathway in cancer (Figure 3F). Subsequently, GEPIA and KM plot database was performed to assess the expression and prognosis roles of pathway-related hub genes. For pathway-related hub genes, 19 genes were up-regulated in PC and one that was down-regulated in PC. Only EGF expressed at a low level but didn't associate with a good prognosis in PC (Supplementary Figure 4F). Notably, CCND1, FN1, and MET were up-regulated in all groups as depicted in Supplementary Figure 1D-F. There are eleven genes (CCND1, FN1, CTNNB1, CASP3, RHOA, FGF2, CXCL8, STAT1, MMP9, NRAS, and MET) that were not only significantly up-regulated in PC but also obviously related to poor prognosis of PC (Figure 4A-F and Supplementary Figure 4A–E), and selected for candidate hub genes.
Identification and validation of upstream miRNA
Based on the result from candidate hub genes, we further identified the upstream miRNA of those genes through the miRTarBase database that included the experimentally validated microRNA-target interactions. Only these miRNAs that proved by strong evidence, including reporter assay, western blot, and qPCR methods were considered as candidate miRNAs. There are a total of 146 miRNAs that were predicted to regulate 11 hub genes (Table S2). Next, miRNA-Seq data were obtained from TCGA databases to evaluate the expression role of candidate miRNAs, and the KM plot database was selected for verifying the prognostic value of candidate miRNAs. As shown in Supplementary Figure 5A–D, we confirmed four miRNAs related to pathway-related hub genes (hsa-miR-20b, hsa-miR-139, hsa-miR-451a, and hsa-miR-144), which were not only expressed down-regulated in PC but also linked with poor prognosis in PC. Additionally, we validated the survival outcome of four miRNAs in the OncoLnc database to increase the reliability of our results. The results indicated that four miRNAs were associated with good survival, P < 0.05 (Supplementary Figure 6A–D). All the qualified miRNAs were selected for further tests.
Prediction and validation of upstream lncRNA
To predict the upstream lncRNA of candidate miRNAs, the miRNet database was employed for screening miRNA-linked lncRNAs. According to prediction in Table S3, we have found out 171 lncRNAs for four down-regulated miRNAs. Next, GEPIA and KM plot database was conducted to evaluate the expression role and prognostic value of predicted lncRNAs. According to the ceRNA hypothesis motioned above, we screened out two eligible lncRNAs (PVT1 and LINC01578) associated with down-regulated miRNAs that were both significantly up-regulated in PC and indicated dismal survival (Figure 5A–B). Finally, we identified the eligible lncRNAs, miRNAs, and mRNAs that not only satisfied the standards of expression and prognosis but also complied with the hypothesis of ceRNA network (Figure 5C).
Construction and verification of ceRNA network
After previous prediction and validation, we constructed a pathway-related lncRNA-miRNA-mRNA ceRNA network. There were three ceRNA networks, including PVT1/miR-20b/CCND1, LINC01578/miR-139/MET, and LINC01578/miR-144/MET, which was constructed by our prediction. It's widely accepted that the eligible miRNA has an opposite interaction with mRNA and lncRNA, whereas lncRNA has a positive coexpression relationship with mRNA. Furthermore, the co-expression analysis was performed to validate the inter-relationships among lncRNA-miRNA, lncRNA-miRNA, and miRNA-mRNA. Ultimately, we successfully established the PVT1/miR-20b/CCND1 cancer-related ceRNA network, which was not only significantly associated with the prognosis of PC patients but also played key roles in the progression of PC (Figure 6A-C Supplementary Figure 7A–E). Finally, the PVT1/miR-20b/CCND1 pathway-related ceRNA network and its potential roles in the progression of PC was vividly displayed in schematic representations (Figure 6D).