Introduction

Pancreatic cancer (PC) is the 8th leading cause of death from cancer worldwide and the 4th most common cancer in the USA, with a mortality rate of about 30,000 deaths each year (Vincent et al. 2011; Pourshams et al. 2018; Han et al. 2015). Unfortunately, no definitive therapy has been found, and most of the patients die because of this disease (Pourshams et al. 2018). Modern tumor diagnosis and therapy is now looked at the molecular level ranging of the biomarkers (AbdelGhafar et al. 2019; El-Guindy et al. 2019; Tobin et al. 2015, Habib et al. 2020). To the best of our knowledge K-ras gene with over 85% of prevalence is one of the genes most involved in pancreatic cancers (Li et al. 2004; Uehara et al. 1999, Rungjarernarrejitt et al. 2018). P16, a tumor suppressor gene, as well as p53, are deactivated in about 95% of pancreatic cancers (Schutte et al. 1997). Imaging and marker analyses don't really discriminate between pancreatic adenocarcinoma and non-neoplastic pancreatic diseases in patients which pancreatic cancer is suspected. For these purposes, to promote early screening and identification, much work has been made to recognize reliable markers of pancreatic cancer (Goggins 2007). While some conventional markers such as cancer antigen (CA) 15-3 and carcinoembryonic antigen (CEA) are used for early screening and diagnosis of pancreatic cancer, its sensitivity and specificity are low (Ghafar et al. 2020). Other genes include HER2, MYB, AKT2, BRCA2, FHIT, CDKN2A, PALB2, STK11, and PRSS1, which are associated with the activation or inactivation of proto-oncogenes and other related genes, such as tumor suppressor genes. Although the new diagnosis and therapeutic approaches have grown over recent years (Rezaie et al. 2020a; Baudino 2015); the underlying causes of this cancer have not yet been fully identified (Han et al. 2015; Li et al. 2004; Hidalgo 2010). The microRNAs (miRNAs) are essential regulators of gene expression in cancer cells. The miRNAs are short, non-coding, and single-stranded RNA molecules (17–25-nucleotide), that can control gene expression, by joining to 3-untranslated region (3-UTR) genes or mRNA degradation (Andorfer et al. 2011; Schetter et al. 2008). The microRNAs participate in the regulation of cellular pathways such as apoptosis, differentiation, proliferation, metastasis, and senescence (Wang et al. 2015; Khafaei et al. 2019). Thus, they can be applied as potential therapeutic or diagnostic biomarkers in cancer (Andorfer et al. 2011; Schetter et al. 2008; Mendell and Olson 2012; Shah et al. 2016). They have been extremely involved in the pathogenesis of human cancer and some investigations showed that the miRNAs specifically act in various cancers. For example in breast cancer, miR-200 showed upregulation as an oncogene (Korpal et al. 2011), while in the lung, kidney, and ovarian cancers downregulation of miR-200 significantly correlated with decreased angiogenesis and poor survival (Pecot et al. 2013). In pancreatic cancer the hsa-miR-217, hsa-miR-96, miR-216a, and miR-148a/b by targeting KRAS and other genes, are downregulated, and also the miR-221, miR-210, miR-155, and miR-21 are upregulated (Xue et al. 2013). Nevertheless, in some cancers, microRNAs have dual functions. For example, the miR-221, which has been shown to be upregulated in liver and colon cancers, acts as an oncogenic factor by inhibiting tumor suppressors PTEN and RECK, whereas it acts as a tumor suppressor by derepression of oncogenes such as c-KIT and ETV1 (Berindan‐Neagoe et al. 2014). Although many miRNA and genes have been reported for various cancers (Berindan‐Neagoe et al. 2014; Sohrabi et al. 2020) such as miR-200c for ovarian cancer, hsa-miR-142-5p and hsa-miR-375 for gastric cancer, miR-133 for breast cancer, and miR-21 for colon cancer, and lung cancer, to best of our knowledge, for pancreatic cancer diagnosis, suitable markers with adequate accuracy are needed (Ghafar et al. 2020; Berindan‐Neagoe et al. 2014). Molecular markers are increasingly being suggested for the screening and diagnosis of PC, but they are not feasible for routine clinical use (Fry et al. 2008; Hasan et al. 2019).

Bioinformatics and system biology tools are remarkably used in the different fields of medical biology (Gasparini-Junior et al. 2019; Lu et al. 2018; Mohammadi et al. 2020; Rezaie et al. 2020b, 2018; Keshtvarz et al. 2017). Analyzing and understanding the genes-miRNAs network for cancer enables us to find genes and miRNAs that have significant expression level differences over cancer states. In addition, we can find genes significantly affected by miRNAs, identify different target genes for miRNAs, identify the mechanism of the disease, and finally introduce potential new and reliable biomarkers.

Aimed at gene expression analysis, the microarray provides a high-throughput platform that in recent years, has been extensively conducted as a powerful tool for assessing general genetic alteration during carcinogenesis (Kim 2004; Zhang et al. 2017). The microarray technology, which can calculate the expression changes of the different genes in the healthy and tumor groups, enables the researcher to study the expression of several thousands of genes simultaneously (Zhang et al. 2017). For example, in Yi lu study the ITGA2, MMP7, ITGB4, ITGA4, ITC3, VCAN, and PLAU genes have been identified after the bioinformatics analysis of the microarray data from patients with pancreatic ductal adenocarcinoma (Berindan‐Neagoe et al. 2014). Several studies also revealed that metabolic pathways and signaling interactions were associated with the shortening of life expectancy in this disease (Lu et al. 2018; Li et al. 2018). The MMP7 and ITGA2 genes have also been correlated with the prognosis and diagnosis of pancreatic cancer (Zhang et al. 2017; Li et al. 2018; Long et al. 2016). It is necessary to explore new molecular biomarkers facilitating detection or increasing the screening and treatment rate of pancreatic cancer. As a result, in the present work, we tried to study the genes-miRNAs-drugs network and understand their signaling pathways in pancreatic cancer to introduce the new diagnosis and therapeutic targets.

Methods

Pancreatic Cancer Biomarkers

We applied the medical subject headings (Rezaie et al. 2020a) based keywords (pancreatic cancer AND (biomarker* OR current marker*) AND discovery AND human) in Google Scholar, Pubmed from 2015 to 2020, and selected the most common biomarkers of pancreatic cancer.

Microarray Data

We searched expression profiles of miRNAs and mRNAs in the GEO database (https://www.ncbi.nlm.nih.gov/geo/) using the following keywords: pancreatic neoplasms AND Homo sapiens. The inclusion criteria were as follows: (1) samples diagnosed with pancreatic cancer (PC) and normal tissue samples, (2) profiling of gene expression for mRNA and miRNA, and (3) enough data to plan the study.

Data Processing

GEO2R online webserver was used to identify differentially expressed genes and miRNAs between PC and normal tissue samples. As the criteria, a P value < 0.05 and a |logFC|> 1 were set. To increase the specificity, further analyses with common miRNAs in both datasets (P value < 0.05) were designed and submitted to the MirDIP database to obtain the miRNA’s target genes. Finally, we identified common genes between targets of miRNAs and differential expression genes (DEGs) using the Venn diagram.

Functional and Pathway Enrichment Analysis

Kyoto encyclopedia of genes and genomes (KEGG) and GO analysis, including biological process (BP), molecular function (Sandhu et al. 2016), and cellular component (Tahira et al. 2011), were used for functional study of single genes or large-scale genome. Both analyses were done in the EnrichR database (Chen et al. 2013; Kuleshov et al. 2016). The cut-off criterion was set to P value < 0.01.

Protein–Protein Interaction (PPI) Network Construction and Analysis of Modules

The DEGs were mapped to STRING (recovery tool for interacting genes) database (http://string-db.org/) to evaluate the protein–protein interaction integration and then visualized with Cytoscape. The molecular complex detection (MCODE) plug-in (http://apps.cytoscape.org/apps/mcode) was used to screen PPI network hub gene modules with cut-off degree = 10, haircut = on, node score cut-off = 0.2, k-core = 2, and max. Depth = 100. Clusters with more than 5 scores were selected for further analysis (Fig. 1).

Fig. 1
figure 1

Flow chart of data analysis. Firstly, we identified differentially expressed mRNAs and miRNAs in pancreatic cancer through integrating expression microarray data then we constructed network of miRNA–mRNA based on Microarray data. Also, we enrichment genes and analyzed protein–protein network and finally survival of genes were investigated

Full size image

Survival Analysis

Survival analysis was performed using five studies with terms of Pancreatic Adenocarcinoma (TCGA, PanCancer Atlas), Pancreatic Adenocarcinoma (TCGA, Provisional), Pancreatic Cancer (UTSW, Nat Commun 2015), Pancreatic Adenocarcinoma (QCMG, Nature 2016) and Pancreatic Adenocarcinoma (ICGC, Nature 2012) from the cBioPortal database (Gao et al. 2013), (survival analysis refers to the overall survival Kaplan–Meier estimate). The datasets used in this study were composed of 1034 pancreatic cancer/adenocarcinoma samples/patients. Based on a relationship network built by MCODE (clusters 1 and 2 with score > 5), gene names were submitted in cBioPortal, and survival analysis was accomplished, of which genes with Log rank test of P < 0.05 were presented.

Construction of Regulatory miRNA-Drug-mRNA Network

Herein, we identified the mRNAs targeted by miRNA through mirDIP 4.1 (Tokar et al. 2017; Shirdel et al. 2011). Then, we identified the mRNAs that correlated with miRNA expression in pancreatic cancer through a Venn diagram. The target genes of the common miRNA from GSE71533 and GSE2479 were predicted with mirDIP, an online MicroRNA target-predicting database, that comprise of 28 database (BCmicrO,BiTargeting, CoMeTa, Cupid, DIANA,ElMMo3,GenMir++, microrna.org, mirbase, mirCoX, miRcode, miRDB, miRTar2GO, MAMI, MBStar, MirAncesTar, MirMAP, MirSNP, MirTar, Mirza-G, MultiMiTar, PACCMIT, PicTar, PITA, RepTar, RNA22, RNAhybrid, TargetRank, TargetScan and TargetSpy). The network was constructed between 43 selected miRNAs and 17 selected genes, including cluster one and two MCODE analysis (score > 5) in cytoscape version 3.6.0. Genes correlated with different miRNAs also entered into the DGIdb website (http://www.dgidb.org/) to find target gene drugs. A general network of genes, miRNAs, and their drugs was constructed with cytoscape software.

Results

Previous Biomarkers for Pancreatic Cancer

Some genes and miRNAs, which have been mentioned in previous studies as pancreatic cancer biomarkers, were identified. These genes and miRNAs are summarized in Table 1.

Table 1 Advantages and disadvantages of previous biomarkers for pancreatic cancer
Full size table

Many of these genes and miRNAs were excluded from our study due to insignificant P-Value and/or sharing between datasets (Table 2; Fig. 1). So far, many genes or miRNAs in pancreatic cancer had relatively good conditions for biomarkers, but often for various reasons, their importance has been diminished. Some of them have same expression profile in different cancers, others are not detectable in the early stages of cancer or at different stages. Others need to be studied with a larger sample size, some of which have only been tested in a pancreatic cancer subtype or under the influence of a particular drug, and some of them are not able to distinguish cancer from inflammatory conditions. These factors reduce their specificity and sensitivity.

Table 2 Reasons for rejecting often previous biomarkers
Full size table

Differential Expression Profiles of miRNAs and Genes

In this study after analysis of the 341 cases and 108 controls (Table 3), we tried to identify the miRNA and genes associated with pancreatic cancer and their interactions. Eventually, we could introduce the new potential biomarkers for this cancer. Efforts to understand the relationships between miRNAs and genes in different diseases already have been performed by the analysis of microarray data. After a search in the GEO database with the related keywords, five datasets (GSE15471, GSE16515, GSE30134, GSE46234, and GSE107610) were collected for gene expression and two datasets (GSE71533 and GSE24279) for miRNA expression analysis (Table 3).

Table 3 Five gene expression profiles and two miRNA expression profiles and their features
Full size table

Genes and miRNAs showing significant increase (P value < 0.05 and logFC > 1) and decrease (P value < 0.05 and logFC < − 1) were identified as DEGs and differential expression miRNAs (DEMs) (Fig. 1). Finally, as shown in Figs. 2 and 3, we reached the 309 DEGs (59 down and 250 upregulated distributed in 3 databases) and 6 DEMs (hsa-miR-210 upregulated and hsa-miR-217, hsa-miR-216a, hsa-miR-216b, hsa-miR-375, and hsa-miR-634 downregulated) (Fig. 3a, b). To increase the specificity and identify more genes, we performed further analysis with common miRNAs (43 numbers with P value < 0.05) in both datasets (Fig. 3c). They were submitted to the MirDIP database and 6006 genes were predicted for these miRNAs. The common genes between 309 (DEGs) and 6006 miRNA target genes were identified, which are shown in the Venn diagram. We ultimately obtained 109 mRNAs, of which 22 and 87 were down and upregulated respectively (Fig. 4). Unlike some previous studies and as it is shown in Fig. 1, there are some important criteria in our study such as presence in at least 3 datasets, over- or under-expression with P value < 0.05 and |logFC|> 1, the similarity with miRNAs targets, MCODE and survival analysis. These criteria have been caused to exclude many previously introduced biomarkers like CA19-9, TFF1, COL6A3 and CEACAM1 from our further analysis (Table 2).

Fig. 2
figure 2

Venn diagrams represent the commonly down and upregulated genes (309 DEGs), among the five identified databases. a Downregulated DEGs in pancreatic cancer among the five datasets were recognized via Venn diagrams, 59 genes were reported to be downregulated at least in three datasets. b Upregulated DEGs in pancreatic cancer among the five datasets, 250 genes were reported to be upregulated at least in three datasets

Full size image
Fig. 3
figure 3

Venn diagrams represent common miRNAs with different filters identified in the two datasets. a DEMs with P value less than 0.05 and logFC < − 1 were identified in both datasets. b DEM with P value less than 0.05 and logFC > 1 were identified in both datasets. c Common miRNAs with P value less than 0.05 were identified in both datasets that were recognized via Venn diagrams

Full size image
Fig. 4
figure 4

Venn diagrams shows common genes between 309 (DEGs) and 6006 (miRNA targets) genes were identified. Commonality between the miRNAs predicted targets (6006 genes) and common DEGs in at least 3 databases for down and upregulation genes (59 and 250 genes) which 22 and 87 genes were associated with down and upregulated genes respectively. Finally, enrichment for these genes (22 down and 87 up) was carried out with used EnrichR and their protein networks were constructed String database. Then, these protein networks were analyzed with MCODE in Cytoscape and the important genes were introduced

Full size image

Functional and Pathway Enrichment Analyses

To further understand the functions and mechanisms of the identified DEGs, functional and pathway enrichment studies have been performed, including GO and KEGG, with the EnrichR database. The GO term enrichment analysis showed that the upregulated genes were massively enriched in the biological process-associated group like extracellular matrix organization, collagen fibril organization, skeletal system development, regulation of cell migration, regulation of cell proliferation, regulation of the apoptotic process, cellular response to cytokine stimulus, neutrophil degranulation and regulation of transcription, while the downregulated genes mainly involved in peptide metabolic process, cellular protein localization, cellular response to insulin stimulus and organonitrogen compound catabolic process (Table S1).

Also cell component analysis showed that the upregulated genes were enriched in the endoplasmic reticulum lumen, membrane rafts and tertiary granule, integral component of the plasma membrane, focal adhesion, and lysosome. The downregulation of the genes was mainly found in the integral component of the plasma membrane and endoplasmic reticulum-Golgi intermediate compartment (Table S2). Moreover, in terms of molecular function, the upregulated genes were enriched in platelet-derived growth factor binding, transcriptional activator activity, RNA polymerase II core promoter proximal region sequence-specific binding, collagen binding, kinase binding, peptidase activity, acting on l-amino acid peptides, and transcription regulatory region sequence-specific DNA binding. The downregulated genes were enriched in zinc ion binding, metalloaminopeptidase activity, transition metal ion binding, protein homodimerization activity, and aminopeptidase activity (Table S3).

Furthermore, the KEGG pathway analysis showed upregulated genes involved in ECM-receptor interaction, amoebiasis, Axon guidance, PI3K-Akt signaling pathway, Focal adhesion Protein digestion and absorption, adherens junction, Platelet activation, and human papillomavirus infection. Downregulated genes involved in riboflavin metabolism (Table 4).

Table 4 Enrichment up and downregulated genes in Kegg database
Full size table

Analysis of PPI Network Modules

Up- and downregulated DEGs in pancreatic cancer were mapped via the STRING database with a PPI score > 0.4. Twenty-two nodes and one edge for downregulated genes (Fig. 5a) and 87 nodes and 134 edges for upregulated genes were constructed (Fig. 5b). MCODE analysis cannot be used for downregulated genes due to the lack of network, but such analysis was constructed for three networks (Fig. 6a–c) with different scores, which are detailed in Table 5. Seventeen genes had scores > 5 after the MCODE analysis.

Fig. 5
figure 5

PPI networks of DEGs identified in pancreatic cancer status versus normal status. a Downregulation of DEGs identified in at least three datasets and common with DEMs target gene, were used to construct the PPI network. The lines between nodes represent the interactions between genes. b Upregulation of DEGs identified in at least three datasets and common with DEMs target gene. The color of each gene is determined based on the logFC. Guide for logFC are listed at the bottom right of the page

Full size image
Fig. 6
figure 6

Three PPI modules were extracted from the PPI network using MCODE in Cytoscape. PPI protein–protein interaction, DEGs differentially expressed genes, blue nodes indicate upregulated genes. The color of each gene is determined based on the logFC. Guide for logFC are listed at the bottom right of the page

Full size image
Table 5 Three networks constructed with MCODE analysis for DEG identified in at least three datasets and common with DEMs target gene
Full size table

Survival Analysis

After MCODE analysis, seventeen genes have been assessed in cluster 1 and 2 for pancreatic cancer survival (derived from 5 studies by cBioPortal database). As shown in Table 6, among all genes, seven upregulated genes (CDH11, COL5A1, DTL, ITGA2, KIF14, SMC4, and VCAN) significantly shortened the life expectancy (P < 0.05) (Fig. 7).

Table 6 Median months survival in patients with altered 7 genes expression
Full size table
Fig. 7
figure 7

Overall Survival Kaplan–Meier estimate of 7 genes in 1034 samples from TCGA datasets with term of Pancreatic Adenocarcinoma (TCGA, PanCancer Atlas), Pancreatic Adenocarcinoma (TCGA, Provisional), Pancreatic Cancer (UTSW, Nat Commun 2015), Pancreatic Adenocarcinoma (QCMG, Nature 2016) and Pancreatic Adenocarcinoma (ICGC, Nature 2012). Red line represents cases with alterations. Blue line represents cases without alterations (VCAN, P = 1.815e−3), (SMC4, P = 0.0146), (KIF14, P = 0.0161), (ITGA2, P = 6.090e−3), (DTL, P = 5.988e−5), (COL5A1, P = 0.0435), (CDH11, P = 0.0475)

Full size image

This significant difference can be due to five datasets related to pancreatic cancer, the sharing of genes in at least three datasets, and having other applied criteria in this study (Fig. 1). We analyzed the survival of five previously identified genes (TFF1, COL6A3, NUSAP1, POSTN, and ARHGEF15), which were removed in our latest filter. Nevertheless, they did not seem to have significant effects on patients' survival length (log rank test P value > 0.05) (Fig. 8).

Fig. 8
figure 8

Overall survival Kaplan–Meier estimate of five genes that were previously identified as pancreatic cancer biomarkers and they removed in our latest filter in 1034 samples from TCGA datasets with term of Pancreatic Adenocarcinoma (TCGA, PanCancer Atlas), Pancreatic Adenocarcinoma (TCGA, Provisional), Pancreatic Cancer (UTSW, Nat Commun 2015), Pancreatic Adenocarcinoma (QCMG, Nature 2016) and Pancreatic Adenocarcinoma (ICGC, Nature 2012). Red line represents cases with alterations. Blue line represents cases without alterations (COL6A3, P = 0.725), (POSTN, P = 0.209), (ARHGEF15, P = 0.368). TFF1 and NUSAP1 did not have enough information in this database

Full size image

miRNA-drug-mRNA Network Construction

Our goal in constructing this network is to identify important genes that are affected by different drugs and miRNAs that may be used as new agents in future studies of pancreatic cancer.

As shown in Fig. 9, initially, interaction analysis showed that multiple drugs were introduced for 12 genes (RACGAP1, ITGA2, COL5A1, COL3A1, COL1A2, COL1A1, CDH11, CCNA2, VCAN, TGFBR1, SDC1, and RRM2). Then the 23 miRNAs associated with these genes were identified, among those, 12 miRNAs (hsa-mir-107, hsa-mir-532-3p, hsa-mir-320a/b, hsa-mir-320c/d, hsa-mir-302e, hsa-mir-342-3p, hsa-mir-375, hsa-mir-217, hsa-mir-429 and hsa-mir-125a-5p) affects at least two gene.

Fig. 9
figure 9

Interaction network between the 17 gene from MCODE analysis (cluster one and two) and 43 reported miRNAs. Circle represent genes, rectangle represent miRNAs and round rectangle show drugs. Each miRNA and gene is recognized by a specific color and miRNAs with yellow rectangles have targets of more than 2 genes. 23 of 43 miRNA and 147 related drugs were associated with 17 genes, then these drugs were inserted into the network and the final form was constructed

Full size image

Based on the figure, we found that Hsa-mir-107 and hsa-mir-125a-5p can regulate 6 genes in this network, that their common genes are COL5A1, CDH11, and TGFBR1. These two miRNA and three genes can be introduced as major miRNAs and genes of this network.

Discussion

The microarray data analysis has been used in many studies on the different types of tumors to find the important gene, miRNA, and metabolic pathways (Tu et al. 2019). In general, the role of miRNAs and genes in carcinogenesis and cancer progression is well known, but their expression, function, and association with pancreatic cancer development are not yet fully investigated (Sun et al. 2017). Early detection of small cancers before the onset of metastasis is currently the only tools to significantly improving resection, post-resection prognosis, and ultimately survival. However, not all available diagnostic tools and biomarkers for PDAC are able to detect early or preventive cancer and suffer from low specificity and sensitivity (Sharma et al. 2020).

In this study, we attempted to explore the interactions between genes, miRNAs, and important pathways in pancreatic cancer with the lowest filtration limitation and relatively high sample size through an accepted analysis method such as GEO2R. We identified the DTL, CDH11, COL5A1, ITGA2, KIF14, SMC4, and VCAN genes and hsa-mir-210, hsa-mir-217, hsa-mir-216a, hsa-mir-216b, hsa-mir-375, and hsa-mir-634 miRNAs associated with pancreatic cancer. Consequently, we could introduce the new potential biomarkers for this cancer.

One important gene is VCAN (belonging to the sulfate proteoglycans family), which has been expressed in many malignancies, such as pancreas, gastric, and prostate cancers. This gene often associated with poor prognosis (Barry et al. 2013; Jiang et al. 2019). In an integrated bioinformatics analysis for pancreatic ductal adenocarcinoma, some genes like MMP7, MMP1, COLA1, CDH11, COL3A1, ITGA2, and VCAN were introduced as hub genes. These may be potential targets for diagnosis, treatment, and prognostic prediction (Lu et al. 2018). Two studies conducted in 2014 and 2019 have shown that S100A6 was significantly upregulated in pancreatic carcinoma compared to normal tissues. They also showed that ITGA2 overexpression and hypomethylation of MET and ITGA2 were significantly associated with overall survival (Liu et al. 2019; Nones et al. 2014).

In compliance with our results, a study conducted by Mingge Shang et al. (2019 has shown an association between increased expression of six genes (KRT7, KRT19, SEMA3C, ITGA2, MYOF, and ANXA1) and their poor survival. PI3K-Akt, Focal adhesion and ECM-receptor interaction were also recognized as important pathways (Shang et al. 2019). It is noteworthy that these results were based on other databases and smaller samples than our investigation. Ivane Abiatari et al. (2009 showed upregulation of kinesin family member 14 (KIF14) and Rho-GDP dissociation inhibitor beta (ARHGDIbeta) mRNA levels in patients with pancreatic cancer (Abiatari et al. 2009).

Song Ye in 2014 analyzed the GEO-derived pancreatic cancer microarray data with 60 cancer samples and 21 controls. Through investigating the association between genes and miRNAs, they identified different miRNAs and genes as important elements in this cancer (Ye et al. 2014). In comparison to their study, the present research includes a higher number of samples and a new data analysis method. miRNA-210 had a significant upregulation in both studies, while miR-217 and miR-216a had a significant downregulation.

In agreement with our results, another study in 2014 on pancreatic cancer showed a significant increase in miR-210 and a significant decrease in miR-217, miR-216, and miR-375. Furthermore, our results showed a significant decrease in miR-216a and miR-634, but their expression has not been mentioned in that study (Ye et al. 2014). This difference can be due to the number of samples and the type of data analysis methods.

High expressions of miR-196a and miR-210 were associated with poor survival (P-Value = 0.001 and P value = 0.003) (Yu et al. 2017); however, miR-196a was excluded from our study due to lack of sharing between two miRNA datasets. miR-216a may potentially serve as a novel tumor marker and prognostic factor for pancreatic cancer and miR-217 functions as a prognosis predictor and inhibits pancreatic cancer cell proliferation, invasion and promotes apoptosis via targeting E2F3.

The next miRNA showing a significant downregulation in our study was miR-375. Previously, downregulation of this miRNA was demonstrated by in vitro methods such as real-time PCR in pancreatic, glioma, and gastric cancers. The hsa-miR-375 is associated with inhibition of migration and proliferation of pancreatic and glioma cancer cells via downregulating of RWDD3 gene (Ji et al. 2020). Our mRNA–miRNA interaction network also demonstrated that the hsa‑miR‑375 can regulate the RACGAP1, TGFBR1, SMC4, VCAN genes. The RACGAP1 gene is one of the most important predictors of poor prognosis in the pancreas adenocarcinoma (Khalid et al. 2019). TGFBR1 has been involved in invasion, migration, angiogenesis, and tumorigenesis of pancreatic cancer cells and it is also active in cellular pathways such as TGF-β/SMAD (Gasparini-Junior et al. 2019; Zhou et al. 2018).

Ultimately the last miRNA that was examined in the present study was miR-634, with a significant downregulation. This event has also been reported in different cancers (Gao et al. 2016; Fischer et al. 2001). Previous studies have already shown its anti-tumor role and promotive action in the apoptosis of cancer cells. The mRNA–miRNA interaction network showed that hsa‑miR-634 could regulate COL3A1. Upregulation of the COL3A1 gene is directly associated with gliomas, renal cell carcinoma (RCC), and bladder malignancies growth, metastasis, progression, and prognosis (Su et al. 2014; Shi and Tian 2019). It may be referred to as a diagnostic or therapeutic biomarker (Gao et al. 2016). In general, the inconsistency between the genes and miRNAs reported in the different system biology studies maybe depend on many factors such as the cancer type, the sample size, filtration approaches, and the procedures of data analysis.

These genes and miRNAs were found to be related to drug response. Inactivating mutations in both TGFBR1 and TGFBR2 have been observed in pancreatic carcinoma and constitutively decreased TGFBR1 expression may have a decreased risk of pancreatic cancer by regulating factors such as TGF-β-mediated growth inhibition (Adrian et al. 2009). In agreement with our results, in human pancreatic cancer (PC) samples and 12 PC cell lines, an increase in TGFBR1 expression has been observed (Adrian et al. 2009; Fan et al. 2020). Yue Fan and et al. (2020) showed that TGFBR1, VTCN1, and LGALS9 were found to be associated with the worse outcomes of patients with PC. Also, they demonstrated, upregulation of TGFBR1 was closely pertained to poor overall survival and significantly associated with the prognosis in PC (Fan et al. 2020).

Collagen type V (COL5) presents in most connective tissue matrix and plays a functional role in different cancers such as breast cancer (Barsky et al. 1982), colon cancer (Fischer et al. 2001), and pancreatic ductal adenocarcinoma (Berchtold et al. 2015). Bioinformatic identification showed that COL5A1 may be a major factor in many types of cancers like breast cancer, gastric cancer, papillary thyroid carcinoma, ovarian cancer, oral squamous cell carcinoma, and lung adenocarcinoma. Also, overexpression of COL5A1 significantly correlates with the overall survival of patients with clear cell renal cell carcinoma (Feng et al. 2019).

CDH11 signaling may play an essential role in mediating the function of activated stellate cells, defining the fibrotic and inflammatory microenvironment in both pancreatitis and cancer. In agreement with our results, Cadherin-11 (Cad-11, also known as OB cadherin or CDH11) has shown upregulation in pancreatic cancer and involved in pancreatic stellate cells (PSCs) activation and pancreatic cancer metastasis (Birtolo et al. 2017).

The miR-107 is abnormally expressed in several metastatic tumors like colorectal, breast, gastric, liver, lung, bladder, and cervical cancer, and its upregulation contributes to cancer progression and metastasis. Also, upregulation of miR-107 is associated with poor clinicopathological parameters and prognosis in pancreatic ductal adenocarcinoma (PDAC) patients (Xiong et al. 2017). miR‐107 had been considered as an oncogene miRNA in gastric and liver cancer (Gong et al. 2019).

miR-125a-5p has been shown to play critical functions in human malignancies. miR-125a-5p acted as a tumor suppressor in various human malignancies, such as hepatocellular carcinoma, breast cancer, lung cancer, gastric cancer, and glioblastoma (Gao et al. 2014). YANG and ZENG by integrating transcriptome analysis showed the upregulation of hsa-mir-125a-5p in miRNA–mRNA crosstalk in pancreatic cancer (Yang and Zeng 2015). Lichao Pan and et al. (2018) showed that the miR-125a by targeting mitofusin 2 (Mfn2), promotes apoptosis, metabolism disorder, and migration impairment in pancreatic cancer cells (Pan et al. 2018).

Conclusion

The important genes and miRNAs introduced in this study, due to the significant change in expression in tumor and normal tissue and the effect on patient survival, can be more specific and sensitive than previous tests for pancreatic cancer and may be used alone or in combination with other tests for diagnosis or prognosis. However, it is necessary for these identified genes and miRNAs to pass the experimental approvals later.

To the best of our knowledge, we attempted to identify DEGs using comprehensive bioinformatics evaluations and come up with potential biological markers to predict disease progression. After analysis, a total of 309 DEGs and 6 DEMs were screened with promising targets for pancreatic cancer treatment, management, and prognosis, among which the DTL, CDH11, COL5A1, ITGA2, KIF14, SMC4, VCAN, hsa-mir-210, hsa-mir-217, hsa-mir-216a, hsa-mir-216b, hsa-mir-375 and hsa-mir-634 had pivotal functions. Also, the wide relation of the hsa-mir-107 and hsa-mir-125a-5p and COL5A1, CDH11 and TGFBR1 genes to drug response was represented in the miRNA-drug-mRNA network. These outcomes call for more organized researches to model exact aspects of gene networks in pancreatic cancer.