Introduction

Current diagnostic technologies for pancreatic cancer are limited, making it one of the most aggressive solid malignant tumors and posing a significant threat to human health1. Pancreatic ductal adenocarcinoma (PDAC) constitutes over 90% of pancreatic tumors cases2 and has a higher mortality rate than other tumors3. Despite advancements in pancreatic tumor screening and treatment, including endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic ultrasound-guided fine-needle aspiration (EUS-FNA), 1- and 5-year survival rates for PDAC remain at 24% and 9%, respective4. Surgical resection is the only curative option for localized PDAC. However, early diagnosis is challenging because of the tumor’s insidious nature, leading to metastatic disease at surgery and high recurrence rates5. As traditional methods have not significantly improved postoperative survival rates, identifying specific biomarkers is crucial for early PDAC diagnosis and treatment, potentially enhancing patient survival rates.

With high-throughput technologies, chip technology has been increasingly applied across various domains, including methylation modification and sequencing. These applications are invaluable for tumor classification based on histological characteristics and clinical records and for clustering and identifying tumor-related genes and signaling pathways6. Bioinformatics has recently become an essential interdisciplinary field that integrates biology and mathematics to analyze gene epigenetic information from large-scale microarray data, filtering pertinent information. This approach significantly aids in screening disease-related information and exploring the underlying molecular mechanisms7.

In this study, we selected normal and PDAC tissues from the comprehensive gene expression database (GEO), and compared the differentially expressed genes (DEGs) between the groups. We screened four genes closely related to the survival and prognosis of pancreatic cancer patients and conducted expression verification.

Materials and methods

Data collection

The GEO dataset stores gene expression, sequencing, and other data for various disease models, enabling researchers to freely access, download, and analyze these original data. We selected three datasets, GSE28375, GSE41372 and GSE62452, all from the GPL6244 platform. These datasets included 45 pairs of pancreatic tumors and adjacent tissues, 6 pairs of pancreas and normal tissues, and 69 pairs of pancreatic tumors and 61 pairs of adjacent tissues. All tumor tissues were diagnosed as PDAC by pathology8,9 (Table 1).

Table 1 Information of datasets in the analysis of PDAC tissues vs. normal or adjacent tissues.
Full size table

Identification of the DEGs

We conducted online analysis using GEO2R for comparison and screening to obtain DEGs. Screening criteria included |logFC|≥1 and the P value < 0.05, considered statistically significant. Genes with logFC ≥ 1 were considered up-regulated, and genes with logFC≤-1 were considered down-regulated. Venn software was used to display common DEGs across the three datasets.

Functional enrichment analysis of the DEGs

We used Database Annotation, Visualization, and Integrated Discovery (DAVID 2021 update, http://david.ncifcrf.gov)10 to annotate, visualize, and integrate DEGs for enrichment analysis. Results are displayed as biological process (BP), cell composition (CC), molecular function (MF) and KEGG pathway. A P-value of <0.05 indicated statistically significance.

Establishment of the PPI network

We utilized the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING)11 to construct the PPI network and visualized it using Cytoscape (version 3.8.2)12. We used MCODE, a Cytoscape plugin, to decode PPI network relationships and identify gene clusters. Screening criteria: degree cutoff = 2; node score cutoff = 0.2; k‑core = 2 and Max Depth = 100. These operations authenticate the relationship between proteins with different expression abundance and clarify the relationship between genes and diseases.

Survival rate and validation analysis of hub genes in the TGCA dataset

We used the Gene Expression Profiling Interactive Analysis tool (GEPIA 2)13 to further analyze the Cancer Genome Atlas (TCGA) and Genotype-Tissue Expression data by comparing gene expression and interaction in normal and cancer tissues. Then we used the Kaplan-Meier Plotter database (https://kmplot.com/analysis/)14 and the university of Alabama at Birmingham cancer data analysis Portal (UALCAN, https://ualcan.path.uab.edu/index.html)15,16 to further analyze the Cancer Genome Atlas (TCGA) by comparing pathological staging and prognosis in normal and cancer tissues. We verified protein expression levels of the central genes by the Human Protein Atlas (HPA) (http://www.proteinatlas.org/).

Cell culture

We acquired human embryonic kidney cells (293T), the normal human pancreatic ductal cell line (HPDE6-C7) and human pancreatic cancer cells (SW1990, PANC-1, BxPC-3) were purchased from the Chinese Academy of Sciences and cultured in Dulbecco’s modified Eagle’s medium (DMEM) with 10% fetal bovine serum (FBS) under conditions of 37 ℃ and 5% CO2.

Packaging of the virus and construction of stable cell lines

We designed the knockout primers and inserted them into sgRNA-Cas9 vector. The sequences are shown as following: forward (5’-GCGAACTGTGACTGTCTAAA-3’) and reverse (5’-GCGAACTGTGACTGTCTAAA-3’). 293FT cells were seeded in a 6-well plate one day in advance. The constructed vector, together with PMD and SPA plasmids, was transfected into 293FT cells. Supernatant was collected 72 h post-transfection, virus concentrate added, and centrifuge at 4℃ for a high-titer lentivirus solution. SW1990 and PANC-1 cells were seeded one day in advance, and 48 h after viral infection, puromycin was added at a final concentration of 2 µg/ml. After 1 week, monoclonal cell colonies were selected for limited dilution screening. Continue screening used 1 µg/ml puromycin. Post-complete cell growth, culture was expanded. NC (negative control) represents negative control cells, while KO (knockout) represents gene knockout cells.

Real-time quantitative polymerase chain reaction (qPCR) for gene expression measurement at the mRNA level

We extracted total RNA from cells using the Easy Pure RNA Purification Kit (Transgen, China). We confirmed RNA purity with A260/280 between 1.8 and 2.0 and concentration between 800 and 1000 ng/µl. We designed primers using Primer 7.0 software. A quantitative polymerase chain reaction (qPCR) was performed using SYBR Green (MCE, USA) on the ABI 7500 fast real-Time PCR Systems (Thermo Fisher Scientific, Sunnyvale, CA). Table 2 shows the primer sequences; GAPDH was the internal control.

Table 2 Primers used for real-time PCR analysis.
Full size table

Western blotting analysis

Total protein was extracted from the cells using RIPA lysis buffer (R0010, Solarbio, Beijing, China). After exlectrophoresis separation using 10% SDS-PAGE, the protein was transferred to 0.22 μm PVDF membrane. The membrane was incubated with primary antibodies (PLAU, HA500429, HUABIO, Hangzhou, China; GAPDH, AC002, ABclonal, Wuhan, China) overnight at 4℃ after being sealed with 5% skimmed milk. We then incubated the membrane at room temperature for 1 h using HRP-labeled secondary antibodies (1:5000, AS014, AS003, Abclonal, Wuhan, China), after being washed with TBST. We used electrochemiluminescence western blot reagents (complex2000, COMPLEX, Nanjing, China) to visualize the protein blots.

Cell proliferation analysis

We added cells to 96-well plates 1 day in advance. Cell viability at 24, 48, 72, 96, and 120 h was detected using Cell Counting Kit-8 (CCK-8, CA1210, Solarbio, Beijing, China). Absorbance was red at a wavelength of 450 nm. All the experiments were repeated three times.

Clone formation assay

We added 5 × 102 cells to 12-well plates and cultured them for 12 days. Mediums was changed every 2–3 days during this period. We then discarded the culture medium and rinsed the cells once with cold PBS. After fixing cells with 4% paraformaldehyde at room temperature for 20 min, they were rinsed once with PBS. We added 0.1% crystal violet staining solution, stained at room temperature for 20 min, and gently rinsed with water. After the plates dried, we took photos and counted the number of clones.

Cell apoptosis assay

After digestion with trypsin, cells in each group were resuspended and centrifuged at 1000 g for 5 min. The supernatant was discarded, and 195 µl of Annexin V-FITC binding solution was added to gently resuspend the cells. We added 5 µl Annexin V-FITC and mixed gently. We then added 10 µl of propidium iodide (PE) staining solution and mixed gently. The solution was incubated at room temperature in the dark for 15 min, then placed in an ice bath. Detection was performed on a flow cytometer.

Statistical analysis

We completed all experiments three times independently. We used SPSS 25.0 for statistical analysis. P-values of less than 0.05 (*) were considered statistically significant.

Results

Identification of DEGs in pancreatic carcinoma

The three GEO datasets we selected included 120 PDAC tissues and 112 normal pancreatic tissues. According to the cut-off criteria |logFC|≥1 and P<0.05, we obtained 267, 1391, 207 upregulated genes and 163, 544, and 111 downregulated genes in GSE28735, 41,372 and 62,452, respectively. Using Venn diagram software, we screened 260 common DEGs in the PDAC specimen from the three GEO datasets, including 95 downregulated genes and 165 upregulated genes (Fig. 1; Table 3).

Fig. 1
figure 1

Selection of 260 common DEGs from three datasets (GSE28375, GSE41372 and GSE62452) (A–C). Volcano plot of DEGs from the three datasets; (D). 165 DEGs are upregulated (logFC≥1); (E). 95 DEGs are downregulated (logFC<1).

Full size image
Table 3 260 commonly DEGs were screened from three profile datasets, including 165 upregulated and 95 downregulated genes in PDAC tissues compared to normal tissues.
Full size table

Functional enrichment analysis of the DEGs

We conducted gene functional analysis via DAVID to understand the functional characteristics of DEGs, and identified 110 significant enriched entries, including BP (61), CC (20), MF (29). Figure 2 showed that the DEGs were mostly concentrated in extracellular matrix disassembly and organization, cell adhesion, and reactive oxygen species metabolic process in terms of BP (Fig. 2A). In the CC group, DEGSs were particularly accumulated in the extracellular matrix, integrin complex, basement membrane, and basolateral plasma membrane (Fig. 2B). For MF, DEGs were strongly enriched in serine-type endopeptidase activity, calcium ion binding, and transmembrane receptor protein tyrosine kinase activator activity (Fig. 2C). Pathway analysis revealed that DEGs were concentrated in 11 pathways including pancreatic secretion, PI3K/Akt signaling pathway, pathways in cancer, ECM-receptor interaction, and glycerolipid metabolism (Fig. 2D).

Fig. 2
figure 2

The top 20 GO and significantly enriched KEGG pathways. (A) BP; (B) CC; (C) MF; (D) KEGG pathways. The Y-axis represents remarkably enriched items. Conversely, the X-axis indicates the degree of enrichment; P-value is represented by the color of the dots, and the size of the dot indicates the number enriched genes.

Full size image

PPI network establishment and analysis

The PPI network of DEGs visualized by Cytoscape software consisted of 246 nodes and 835 edges, including 67 downregulated and 137 upregulated genes. Cluster analysis identified 36 central nodes containing 16 downregulated and 20 upregulated genes (Fig. 3; Table 4).

Fig. 3
figure 3

PPI network display and module analysis using STRING and Cytoscape. (A) Nodes indicate protein; edges indicate protein interactions; red circles show up-regulated DEGs, while blue ones represent down-regulated DEGs; (B) Module 1; (C) Module 2. Red circles indicate upregulated DEGs, while blue ones represent downregulated DEGs.

Full size image
Table 4 36 central genes were selected by STRING and cytoscape software from PPI network including 20 upregulated and 16 downregulated genes.
Full size table

Sreening of hub genes and detection of the expression levels

The box plot from GEPIA 2 indicated that these genes expression levels were obviously increased in pancreatic cancer samples than in normal samples (Fig. 4A). A total of 177 cases of clinical data from TCGA database confirmed that the mRNA expression of COL12A1, ITGA2, MMP14 and PLAU were obviously elevated in pancreatic cancer tissues than normal tissues (Fig. 4B). Based on Kaplan-Meier Plotter database, the expression levels of 4 hub genes and the relationship with the OS and DFS ratio of pancreatic cancer patients were analyzed. The data indicated that four highly expressed genes, namely COL12A1 (HR = 1.23, P = 0.028), ITGA2 (HR = 1.49, P = 1.33 × 10− 5), MMP14 (HR = 1.39, P = 7.56 × 10− 4) and PLAU (HR = 1.52, P = 1.92 × 10− 6), were considered to be remarkably related with shorter overall survival (OS) time in patients of pancreatic cancer (Fig. 5A–D). Moreover, the four genes, COL12A1 (HR = 2.31, P = 2.41 × 10− 5), ITGA2 (HR = 2.27, P = 1.23 × 10− 5), MMP14 (HR = 0.49, P = 2.54 × 10− 4) and PLAU (HR = 1.8, P = 0.00196), were considered to be remarkably related with shorter disease free survival (DFS) time in patients of pancreatic cancer (Fig. 5E–H). Furthermore, qPCR assay suggested that the expression of COL12A1, ITGA2, MMP14 and PLAU mRNAs were remarkably increased in the SW1990, PANC-1 and BxPC-3 cells compared with the HPDE6-C7 cell (Fig. 6). Furthermore, inmmunohisitochemical results indicated that the expression levels of COL12A1, ITGA2, MMP14 and PLAU were higher in PDAC as compared with normal pancreatic tissues, based on extracting clinical data from the HPA database (Fig. 7).

Fig. 4
figure 4

Expression analysis of hub genes. (A) Comparison of the transcript levels of four genes in tumor (n = 179) and normal (n = 171) tissues via GEPIA 2; (B) Analysis of four genes expression levels from the TCGA database.

Full size image
Fig. 5
figure 5

Correlation between the OS and DFS of patients and the expression of four hub genes. (A–D). The OS of patients; (E–H). The DFS of patients.

Full size image
Fig. 6
figure 6

Relative mRNA expression of (A) COL12A1, (B) ITGA2, (C) MMP14 and (D) PLAU in SW1990, PANC-1 and BxPC-3 cells compared with that in HPDE6-C7 cells. * indicates P<0.05.

Full size image
Fig. 7
figure 7

IHC validation of hub gene expression from HPA. Staining showed the protein expressions of pivotal genes in the pancreatic cancer tissue samples than in the normal pancreatic tissue samples. To ensure homogeneity and reliability, we use the same antibody immunohistochemical pictures for display. The antibody information is as follows, COL12A1 (HPA009143), ITGA2 (HPA063556), MMP14 (HPA051432), PLAU (HPA070796).

Full size image

Knockdown of PLAU repressed the proliferation of pancreatic carcinoma cells

To investigate PLAU’s function and mechanism in pancreatic carcinoma, we constructed stable PLAU knockdown transfected SW1990 and PANC-1 cell lines (KO) and negative control cell line (NC). Knockdown efficiency was verified via western blot analysis (Fig. 8A,B). Subsequently, we investigated cell biological behaviors using CCK-8, colony formation and apoptosis assays. PLAU knockdown significantly suppressed SW1990 and PANC-1 cell viabilities (Fig. 8C,D). Moreover, colony formation was notably decreased in KO cells than that in NC cells (Fig. 8E–H). Additionally, the apoptotic rates of SW1990 and PANC-1 cells increased upon PLAU knockdown (Fig. 8I–L). These findings suggested that PLAU knockdown could inhibit the proliferation and viability of pancreatic carcinoma cells while inducing apoptosis.

Fig. 8
figure 8

The effect of PLAU knockdown on pancreatic cancer cell phenotype. (A, B). Western blotting verification of PLAU knockdown effect in SW1990 and PANC-1 cells. (C, D). Effects of PLAU knockdown on proliferation of SW1990 and PANC-1 cells. (E–H). Effects of PLAU knockdown on colony formation of SW1990 and PANC-1 cells. (I–L). Effects of PLAU knockdown on apoptosis of SW1990 and PANC-1 cells.

Full size image

Discussion

PDAC is known for its invasiveness, invisibility and lethality17. The average lifespan of untreated of PDAC patients is reported to be only four months. Due to late detection, survival time has not significantly prolonged even with surgery and other treatment18. Therefore, identifying specific markers for early and accurate diagnosis and effective targeted treatment of pancreatic cancer is crucial.

In this study, we screened gene expression profiles and identified 260 DEGs via GEO2R, including 165 upregulated and 95 downregulated genes. The three chosen datasets are from the platform GPL6244 to ensure data reliability and uniformity. GO enrichment and KEGG pathway analysis showed that DEGs were significantly enriched in 11 pathways, including pancreatic secretion, PI3K-Akt signaling, and cancer pathways. Activation of PI3K signaling is strongly associated with poor prognosis of PDAC patients19. PI3K and its downstream signaling elements are involved in tumor-promoting effects of mutant K-Ras in diseases, including PDAC. They also transmit signals from growth factor stimulation and cytokines through receptor tyrosine kinases, regulating various carcinogenic function20.

Thereafter, we built the PPI network via STRING and identified 36 candidate genes using MCODE plus in Cytoscape software. Among these, four genes—COL12A1, ITGA2, MMP14, and PLAU—were highly expressed in pancreatic cancer tissues and related to the survival and prognosis, as determined through GEPIA and Kaplan-Meier Plotter analysis. We then examined the mRNA expression of these genes and found higher expression levels of COL12A1, ITGA2, MMP14, and PLAU in pancreatic cancer cells compare to pancreatic ductal epithelial cell. Further validation from HPA database showed significant differences in the translational levels of COL12A1, ITGA2, MMP14, and PLAU in PDAC and normal tissues, suggesting their potential involvement in PDAC tumorigenesis and progression.

The four genes identified in our study had been confirmed to be involved in tumor development. COL12A1, a member of the fibril-associated collagen family, encodes the α chain of type XII collagen21, a homotrimer associated with type I collagen that regulates interactions between type I collagen fibrils and the surrounding matrix22. Bioinformaitics analysis have shown that COL12A1 expression is linked to poor prognosis in pancreatic cancer, especially PDAC patients23,24. In addition, experimental studies have confirmed COL12A1’s role in PDAC biology. Overexpression of COL12A1 could counteract the decreased migration and invasion ability of pancreatic cancer cells caused by PABPC1 silence25. In studying the relationship between cancer-associated fibroblasts (CAFs) and tumor microenvironment (TME) and progression, Song et al. found that COL12A1 was highly expressed in PDAC tissues. Knockout of COL12A1 reduces the proliferation and migration of CAFs, which indicates that COL12A1 may be a new prognostic biomarker for TME targeted treatment of pancreatic cancer. Interestingly, they found that COL12A1 was mainly expressed in CAFs but not in pancreatic cancer cells26. Therefore, we did not detect the expression level of COL12A1 in pancreatic cancer cells. In addition, we used Kaplan Meier Plotter analysis to find a negative correlation between high COL12A1 expression and low survival ratio in PDAC patients. But using UALCAN, we found no significant correlation between them, which might be related to uneven sample distribution between different groups (Supplementary material). Our research suggests that COL12A1 is closely related to the PDAC process, and COL12A1 may be a potential therapeutic target of PDAC.

Receptor integrins are transmembrane receptors composed of α subunit and β subunit. The α subunit, encoded by ITGA2, heterodimerizes with β subunit to form collagen and laminin surface receptors that play a role in intercellular adhesion and promote the formation of newly synthesized extracellular matrix27. Bioinformatics analyses have found that ITGA2 is highly expressed in PDAC and closely associated with poor prognosis. ITGA2 is linked to INF-α response, EMT, and immune regulation in PDAC28. Single-cell sequencing analysis showed ITGA2 expression by ductal cells in PDAC29. Through comparative analysis of PDAC cells and normal cells, it was found that ITGA2 is highly expressed during the proliferation, apoptosis, and migration of PDAC cells in vitro, and the knockout of ITGA2 can effectively inhibit the EMT process of BxPC-3 and MIA PaCa2 cells28. ITGA2 silencing significantly reduced the proliferation and migration of gemcitabine-resistance PDAC cell clones, indicating ITGA2 might be a new potential therapeutic target to counteract gemcitabine resistance in PDAC30. Hiroki found high ITGA2/ITGB1 expression in PDAC clinical specimens, and ITGA2 silencing attenuated malignant phenotypes of PDAC cells31. Wang et al. confirmed that ITGA2 is associated with low OS and short DFS in PDAC patients29. We used Kaplan Meier Plotter analysis to find a negative correlation between high ITGA2 expression and low survival ratio in PDAC patients. But using UALCAN, we found no significant correlation between them, which might be related to uneven sample distribution between different groups (Supplementary material). Therefore, more samples are needed to verify the relationship between ITGA2 and the survival of PDAC patients. These results indicate that ITGA2 plays an important role in the PDAC process.

MMPs are a family of zinc-dependent endopeptidases that degrade extracellular collagen, laminin, fibronectin, and proteoglycans. They play roles in physiological processes, including tissue repair and remodeling, cell proliferation, differentiation, migration, apoptosis, angiogenesis, and immune response32. MMPs include membrane-anchored and secreted types. MMP14 belongs to the latter and plays a role in tumor metastasis by degrading ECM and activating MMP2 and MMP933. Liang et al. predicted through bioinformatics that MMP14 is highly expressed in PAAD and found that MMP14 knockout in PANC1 cells significantly inhibited cell migration and tumorigenesis in nude mice34. CCL9, as a downstream target of KrasG12D, can upregulate MMP14 expression and promote acinar to ductal metaplasia (ADM). PDAC originates from ADM, and pancreatic acini carrying carcinogenic Kras mutations can transform into ductal like phenotypes, leading to the development of pancreatic intraepithelial neoplasia (PanIN) lesions35. MMP14 is the main differentially secreted protein of gemcitabine resistant PDAC cells, which can be packaged into exosomes and spread from chemotherapy resistant cells to sensitive cells. MMP14 transferred through extracellular vesicles can enhance the drug resistance and migration ability of sensitive cells, promote sphere formation, and stabilize the tumor stem cell marker CD4436. Luan et al. found through analysis of the GEO database that MMP14 is highly expressed in PDAC and is closely related to adverse patient outcomes37. We used Kaplan Meier Plotter analysis to find a negative correlation between high MMP14 expression and low survival ratio in PDAC patients. But using UALCAN, we found no significant correlation between them, which might be related to uneven sample distribution between different groups (Supplementary material). It can be confirmed that MMP14 is also an important regulatory factor in the development process of PDAC.

There have been studies exploring the roles of the three genes mentioned above in PDAC cells through genes knockdown. However, the direct regulatory effect of PLAU on PDAC cells has not been experimentally confirmed, promoting us to conduct in vitro experiments targeting the PLAU gene.

The PLAU encodes a secreted serine protease, mainly expressed in the lumen and viscera of the endoplasmic reticulum, playing a role in regulating cell and cellular component motility38. PLAU activates plasminogen into plasmin, which degrades the ECM, regulates cell migration and adhesion during tissue regeneration and intracellular signaling, and initiates and develops various tumors39. Some studies confirmed the upregulation of PLAU in various tumor tissues and that silencing of PLAU inhibits tumor cell process, such as glioma40, head and neck squamous cell carcinoma41, and gastric adenocarcinoma42. Mansouri analyzed GEO databases, PDAC, and found that PLAU is highly expressed in PDAC43. Xu et al. analyzed PDAC and normal pancreatic tissue in the TCGA database, and found the high expression of PLAU in PDAC was associated with its differentiation and lymph node metastasis status, positively correlated with cachexia symptoms, and negatively correlated with the survival rate of PDAC patients44. These findings are consistent with our bioinformatics predictions. Liu et al. found that PLAU expression was upregulated in senescent cancer associated fibroblasts (SCAFs). PDAC cells treated with PLAU knockdown SCAFs-conditioned medium (SCAFs-CM) showed weakened proliferation, migration, and invasion ability. Moreover, PDAC cells treated with CM from PLAU overexpression SACFs showed increased proliferation, migration, and invasion ability. These suggested that SCAFs derived PLAU promoted tumor proliferation and immunosuppressive TME formation45. PLAU regulation is associated with invasive phenotypes related to PDAC growth and metastasis, including epithelial mesenchymal transition, stemness, ECM degradation, and immune suppression of the tumor microenvironment46. Traditional Chinese medicines such as Thunder God Vine F (TwHF)47 and Dendrobium nobile44 inhibit the biological processes of PDAC. Both treatments inhibited the expression of PLAU, indicating that PLAU may be a target for these drugs to treat PDAC. These findings provide strong evidence for PLAU’s involvement in the biological processes of PDAC. In order to explore the effect of PLAU on pancreatic cancer cells, we constructed the PLAU knockout pancreatic cancer cell lines. The results showed that PLAU knockout led to decreased cell proliferation and increased apoptosis, suggesting that PLAU can promote the proliferation of pancreatic cancer cells and inhibit their apoptosis. This conclusion is consistent with other scholar’s research.

However, our research has limitations needing further exploration. First, we only selected three datasets for analysis, future research should verify more samples. Second, the study lacked more clinical cohort validation. Third, we should select more cell lines and perform in vivo and in vitro assays, such as targeted gene knockdown and rescue assays, and tumor formation experiments in nude mic to explore PLAU’s molecular role in PDAC progression.

Conclusion

We used a comprehensive bioinformatics approach to screen 260 DEGs and identified four central genes. The function of the PLAU gene was validated through cell proliferation and apoptosis experiments, and it was found that knocking out PLAU could inhibit PDAC cell proliferation and promote cell apoptosis. Based on the results of bioinformatics analysis and cell function experiment, we have reason to believe that PLAU plays an important role in the development of PDAC. The next step of research will establish stable cell and animal models for gene knockout and overexpression, and investigate the specific molecular role of PLAU in PDAC process through in vivo and in vitro experiments. These will provide more powerful evidence for identifying PLAU as a potential molecular marker for PDAC, and provide new research plans and strategies for the diagnosis and treatment of this disease.