Identification and mechanistic analysis of shared biomarkers and pathogenesis in acute pancreatitis and sepsis based on differential gene expression and protein interaction networks

Abstract

Acute pancreatitis (AP) is a common gastrointestinal inflammatory disease that requires hospitalization, with 40–70% of patients in moderate to severe stages potentially developing sepsis, which is closely related to high mortality rates and poor prognosis. Therefore, early identification of AP patients at risk of developing sepsis is crucial for reducing mortality. This study aims to identify core genes associated with sepsis to provide new core genes for early warning and management of patients with acute pancreatitis. The study utilized the GSE54514, GSE57065, GSE95233, and GSE194331 datasets for analysis, employing weighted gene co-expression network analysis (WGCNA) and protein-protein interaction (PPI) network construction. Six core genes were identified using two machine learning methods and validated with the GSE3644 and GSE28750 datasets. The analysis revealed that the identified core genes (NDUFA1, COX7A2, COX7B, UQCRQ, SNRPG, and NDUFA4) are related to the oxidative phosphorylation (OxPhos) pathway, and significant differences were observed in the immune cell composition between AP and sepsis patients. SNRPG may play a role in the progression from AP to sepsis by regulating NDUFA4, linking it to cellular metabolism and redox balance. The newly identified core genes and their associated molecular mechanisms provide important clinical insights into the progression of acute pancreatitis to sepsis, potentially offering new research directions for future therapeutic strategies. Clinical trial number: This study was approved by the Ethics Committee of (Municipal Hospital affiliated to Taizhou University), in accordance with the Declaration of Helsinki. Approval number: LWSL202400220.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10142-025-01600-6.

Introduction

Acute pancreatitis (AP) is an inflammatory condition of the pancreas, predominantly induced by gallstones, excessive alcohol consumption, or elevated levels of triglycerides (Zhu et al. 2024). During disease progression, AP can lead to systemic inflammation and multiorgan dysfunction, and in advanced stages, sepsis-related complications may occur, which is a significant cause of mortality in these patients (Watts et al. 2025). Sepsis poses a significant global health challenge, characterized by substantial morbidity and mortality rates (Bhavani et al. 2024; Rong et al. 2024). It induces a significant rise in pro-inflammatory cytokines, reduction in lymphocyte count, and compromised immune function, which collectively contribute to life-threatening organ dysfunction.

Oxidative phosphorylation (OXPHOS) is a critical metabolic pathway in cells(Fig. 1), facilitating the electron transport chain and ATP synthesis. Reactive oxygen species (ROS), a byproduct of OXPHOS, can negatively impact cellular integrity. During the electron transport chain, electrons pass through a series of complexes, ultimately combining with oxygen to produce water. However, some electrons may “leak” during this process, particularly at complexes I and III, prematurely reacting with oxygen to form ROS, such as peroxides(Tang et al. 2020). When ROS production exceeds the antioxidant capacity of intracellular defense systems—such as superoxide dismutase, glutathione, and catalase—oxidative stress occurs. Under oxidative stress, cellular damage increases, impairing cell function and potentially triggering apoptosis or necrosis (Kremer and Rehling 2024; Kami Reddy et al. 2024).

Fig. 1.

Electron Transport Chain and Mitochondrial Dysfunction in Oxidative Phosphorylation. This diagram illustrates the electron transport chain (ETC) within the inner mitochondrial membrane, depicting the process of oxidative phosphorylation. Electrons (e⁻) are transferred through a series of complexes (I-IV), leading to the pumping of protons (H⁺) into the intermembrane space, creating a proton gradient. This gradient drives ATP synthesis via ATP synthase

Elucidating the mechanisms underlying the progression of AP to sepsis is a crucial area of both clinical and basic research. Although patients with severe pancreatitis often develop sepsis, the exact pathways involved are not yet fully understood. To shed light on these mechanisms, we conducted a weighted gene co-expression network analysis (WGCNA) on differentially expressed genes associated with both pancreatitis and sepsis. We then performed further analyses on intersecting genes, including core gene identification and immune infiltration studies, to explore the potential connection between AP and sepsis. These insights may contribute to better clinical management of high-risk patients, more efficient allocation of intensive care resources, prevention of septic shock, and ultimately, improved patient outcomes.

Materials and methods

Data acquisition and preprocessing

We comprehensively analyzed microarray datasets related to AP and sepsis, all derived from peripheral plasma samples and retrieved from the Gene Expression Omnibus (GEO) database (https://www.ncbi.nlm.nih.gov/geo/). For the sepsis cohort, we included three publicly available datasets: GSE54514, GSE57065, and GSE95233. These datasets were all based on peripheral blood gene expression profiling from intensive care unit (ICU) patients, with serum samples collected within 3 to 5 days of admission. The resulting integrated sepsis expression matrix consisted of 83 healthy controls and 221 sepsis patients, and was subsequently used for differential gene expression analysis and machine learning-based feature selection.

To ensure data comparability across different platforms and experimental batches, we implemented a two-step preprocessing pipeline. First, expression matrices from each dataset were independently normalized using the “normalizeBetweenArrays” function from the “limma” package to correct for intra-array technical variability. Next, we applied the “ComBat” algorithm from the “sva” package to remove batch effects introduced by differences in experimental conditions or platforms. This procedure effectively harmonized expression profiles across datasets while preserving biological variation.

For AP, we selected GSE194331, which includes 32 healthy samples and 87 AP samples. Additionally, GSE3644 and GSE28750 were chosen as external cohorts for sepsis and AP, respectively, to evaluate expression levels and predictive capabilities of candidate markers. The detailed information of the datasets is presented in Table 1, and the specific workflow of this study is shown in Fig. 2.

Table 1.

The datasets used in the study specify their platforms, organisms, diseases, and the number of controls and cases

Fig. 2.

Integrative Analysis of Sepsis and AP Gene Expression Data Using Bioinformatics Approaches.This flowchart represents the workflow of a bioinformatics study integrating gene expression data from sepsis and AP cohorts. Data were obtained from the GEO database for three sepsis datasets (GSE54514, GSE95233, GSE57065) and one AP dataset (GSE194331)

Construction of WGCNA network and identification of gene modules

Co-expression networks were developed using the WGCNA package in R to identify gene co-expression modules. Specifically, (1) hierarchical clustering analysis was conducted to screen for any notable sample outliers; (2) the soft-thresholding power (β) was selected through an R-based algorithm to construct a biologically relevant scale-free network; and (3) network connectivity was established using the topological overlap matrix (TOM), with gene modules identified via the dynamic tree cut algorithm. Genes within each identified module were then carried forward for further analysis.

Detection of shared DEGs

Differential expression analysis was performed between healthy and sepsis groups as well as between healthy and AP groups using the limma package in R. The threshold for identifying DEGs was set to p-value < 0.05 and|lgFC| ≥ 0.6 to determine significantly differentially expressed genes. The results were visualized using volcano and heatmaps to elucidate differential expression patterns among the study groups. Finally, the genes from WGCNA modules most associated with sepsis and pancreatitis were extracted and cross-analyzed to obtain common DEGs.

Functional enrichment analysis

Functional enrichment analysis of the identified common differentially expressed genes (DEGs) was conducted using the “clusterProfiler” package in R, encompassing analyses of Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Disease Ontology (DO). In the GO analysis, DEGs were classified into three categories: biological processes (BP), cellular components (CC), and molecular functions (MF), thereby providing a comprehensive overview of their functional implications. KEGG analysis was utilized to identify potential pathways linked to the DEGs, offering insights into the biological mechanisms underlying the observed changes in expression. To ensure the reliability of the results, stringent criteria were employed, with significance thresholds for enrichment set at a p-value of less than 0.05.

Construction of PPI network and identification of key immune-related DEGs

The protein-protein interaction (PPI) network was constructed using the STRING database (version 11.5; http://string-db.org) and visualized with Cytoscape software (version 3.9.0; http://www.cytoscape.org). The core functional modules were subsequently identified using the MCODE plugin in Cytoscape, with selection criteria set as follows: k-core = 2, degree cutoff = 2, maximum depth = 100, node score cutoff = 0.2. Immune-related DEGs were then identified from the ImmPort database (https://www.immport.org) (Table S1). A Venn diagram was used to cross-compare sepsis and AP to obtain core immune-related DEGs. Finally, a co-expression network of core immune-related DEGs was constructed using the GeneMANIA database (http://www.genemania.org).

Machine learning for common core genes

To identify reliable diagnostic biomarkers from differentially expressed genes, we employed two supervised machine learning algorithms: Support Vector Machine–Recursive Feature Elimination (SVM-RFE) and Random Forest (RF), both implemented in R. These methods were used to rank gene features based on their importance in distinguishing sepsis patients from controls.

SVM-RFE was conducted using the “e1071” R package with a 10-fold cross-validation strategy to minimize overfitting and enhance model reliability. Genes were iteratively eliminated based on their weight contribution to the SVM classifier, and the optimal feature subset was determined by identifying the gene set associated with the lowest cross-validation classification error.

RF analysis was performed using the “randomForest” R package. The model was initially trained using 500 decision trees, and the optimal model was selected based on the minimum out-of-bag (OOB) error rate. Feature importance was quantified using the mean decrease in Gini impurity, and genes with importance scores greater than 2 were retained as candidate biomarkers. The top 30 most informative genes were visualized using lollipop plots.

To improve robustness, overlapping genes identified by both SVM-RFE and RF were considered as final candidate biomarkers. Their diagnostic performance was assessed using receiver operating characteristic (ROC) curve analysis, with evaluation metrics including the area under the curve (AUC), sensitivity, specificity, and accuracy.

Construction of the lncRNA-miRNA-mRNA network

Long non-coding RNAs (lncRNAs) can regulate miRNA activity by directly binding to miRNAs, which in turn suppress mRNA translation or trigger mRNA degradation by binding to the 3’ untranslated region (3’UTR) of target mRNAs. To demonstrate these complex regulatory relationships, we integrated regulatory information from lncRNA-miRNA and miRNA-mRNA interactions to construct a global lncRNA-miRNA-mRNA network. Significant correlated pairs were identified using the Pearson correlation coefficient to ensure that the nodes and edges in the network accurately reflect biological relationships. The network was visualized and analyzed for differential co-expression using Cytoscape software (version 3.5.1), enabling intuitive representation of regulatory relationships and further analysis of the network’s topological structure and functional annotation, ultimately revealing key regulatory molecules under disease or specific biological states.

Potential therapeutic drugs related to core genes

The Connectivity Map (CMap) is a powerful tool for drug screening that can predict potential molecular targeting drugs based on DEGs. Using the CMap database (CMap: https://clue.io/cmap), researchers can utilize the L1000 analysis platform to explore the relationships between drugs, genes, and disease states. The database contains cellular expression profiles for 164 drugs/small molecules and data on gene overexpression or knockdown, providing a valuable resource for identifying drugs that may reverse specific pathological conditions. As such, CMap shows great potential for application in precision medicine and drug discovery.

Assessment of immune infiltration

We conducted an analysis of immune cell distribution across 22 immune cell types utilizing the CIBERSORT algorithm (https://cibersortx.stanford.edu), following the provided R script and performing 1,000 permutations to estimate the relative abundance of infiltrating immune cells. The v-SVR function was executed using the “e1071” R package. Samples yielding p-values less than 0.05 in CIBERSORT were considered to exhibit statistically significant results. Subsequently, the relationship between DECuGs and the infiltrating immune cells was analyzed using R software. Finally, the findings were visualized with the “reshape2” and “ggpubr” R packages.

Quantitative real-time PCR (qRT-PCR)

Peripheral whole blood samples were collected from three patients with AP and three healthy individuals at Municipal Hospital affiliated to Taizhou University(Zhejiang, China). Total RNA was extracted and subjected to quantitative real-time polymerase chain reaction (qRT-PCR). None of the enrolled subjects had a history of autoimmune disease, malignancy, or oral immunosuppressive medication use. This study was approved by the institutional ethics committee (Approval No. LWSL202400220).

Western blotting

After protein quantification using a BCA Protein Assay Kit (Thermo), equal amounts of protein (20 µg) from each sample were separated by 10% SDS-PAGE and transferred onto PVDF membranes (Millipore, Billerica, MA, USA). Membranes were blocked with 5% non-fat milk at room temperature for 2 h and incubated overnight at 4 °C with the following primary antibodies: anti-UQCRQ (Abcam, ab267244, 1:1000), anti-COX7A2 (Abcam, ab131143, 1:1000), anti-SNRPG (Abcam, ab204569, 1:1000), anti-NDUFA4 (Abcam, ab129752, 1:1000), and anti-ALBUMIN (Abcam, ab207327, 1:1000). After washing, membranes were incubated with the appropriate secondary antibodies overnight. Protein bands were visualized using the ChemiDoc imaging system (Bio-Rad).

Statistical analysis

Data and statistical analyses were performed using R software (version 4.1.3). The Wilcoxon test was utilized to assess significant differences between the two groups, and Spearman correlation analysis was conducted to examine the relationship between core genes expression levels and circulating immune cells. A p-value of less than 0.05 was regarded as statistically significant.

Results

Identification of common genes

We first applied Weighted Gene Co-expression Network Analysis (WGCNA) to identify gene modules significantly associated with disease phenotypes. In the AP dataset, a soft-thresholding power of β = 6 was selected to construct a scale-free network, achieving a scale-free topology fit index (R²) of 0.85 with high mean connectivity. Based on the resulting gene correlation-based hierarchical clustering dendrogram (Figure S1A, B), the “brown” module (Cor = 0.56, p = 3 × 10⁻¹¹) and “pink” module (Cor = − 0.51, p = 3 × 10⁻⁹) were identified as the most significantly correlated with AP status (Fig. 3A).

Fig. 3.

Identification of trait-related gene modules and differentially expressed genes in AP and sepsis. (A–B) Heatmaps showing module–trait relationships based on weighted gene co-expression network analysis (WGCNA) in acute pancreatitis (A) and sepsis (B). Each row represents a co-expression module labeled by color, and each column represents a clinical trait (Control or Treated). The values within each cell represent the Pearson correlation coefficient between the module eigengene and the trait, with the corresponding p-value shown in parentheses. Modules with strong positive or negative correlations are highlighted in red and blue, respectively, indicating potential trait relevance. (C–D) Volcano plots illustrating differentially expressed genes (DEGs) in AP (C) and sepsis (D) datasets. Each point represents a gene, with red and blue indicating significantly upregulated and downregulated genes, respectively, based on thresholds of|lgFC| ≥ 0.6 and adjusted p-value < 0.05. Grey dots represent non-significant genes

For the sepsis dataset, batch effects among GSE54514, GSE57065, and GSE95233 were corrected using the ComBat algorithm. Principal component analysis (PCA) revealed distinct separation between datasets prior to correction (Figure S2A), while post-correction samples clustered tightly together, indicating effective batch adjustment (Figure S2B). A soft threshold of β = 4 was then chosen, achieving a scale-free topology (R² = 0.85) with satisfactory mean connectivity (Figure S2C, D).

Hierarchical clustering identified several key modules associated with sepsis traits, including the “green” module (Cor = 0.11, p = 0.02), “royalblue” module (Cor = 0.12, p = 0.02), and “cyan” module (Cor = − 0.14, p = 0.004) (Fig. 3B), which were selected for further analysis based on their relevance.Differential expression analysis of the AP and sepsis datasets was performed using the limma package in R, identifying 4,459(Fig. 3C, FigS2E) and 667 DEGs (Fig. 3D, FigS1C), respectively.

Functional characterization analysis

We performed a cross-analysis of module characteristic genes with DEGs, resulting in 37 common significantly expressed genes shared by sepsis and AP, as depicted in the Venn diagram (Fig. 4A). These differentially expressed genes were further subjected to Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses. The GO enrichment analysis revealed that biological processes (BP), cellular components (CC), and molecular functions (MF) were all associated with oxidative phosphorylation. Specifically, BP terms included aerobic electron transport chain, ATP synthesis coupled electron transport, respiratory electron transport chain, and oxidative phosphorylation; CC terms included respiratory chain complex and mitochondrial respirasome; and MF terms included oxidoreduction-driven active transmembrane transporter activity (Fig. 4B, C). The KEGG pathway analysis demonstrated that these significantly enriched DEGs were notably involved in the oxidative phosphorylation pathway (Fig. 4D).

Fig. 4.

GO and KEGG Enrichment Analysis of Common Core Genes in Acute Pancreatitis and Sepsis. (A) Venn diagram of common differentially expressed genes between acute pancreatitis and sepsis identified through WGCNA network analysis and DEG overlap; (B, C) Circular and dot plots showing GO enrichment analysis of common genes; (D) KEGG pathway enrichment analysis of common genes. An adjusted p-value < 0.05 was considered statistically significant. The ordinate represents the enriched terms, and the abscissa represents the proportion of genes involved in each term. The size of the dots indicates the number of genes, while the color of the dots reflects the p-value

Construction and analysis of the protein-protein interaction (PPI) network for core genes

Using the STRING database, we generated a PPI network comprising 26 nodes and 134 edges, reflecting a dense and highly interconnected structure among proteins encoded by the selected core genes. The network was visualized with Cytoscape (Fig. 5A), where node color intensity indicates connectivity, with darker nodes representing higher centrality and potential biological significance.

Fig. 5.

PPI Network and Functional Enrichment Analysis of Common Genes in Acute Pancreatitis and Sepsis. (A) PPI network of common genes identified in acute pancreatitis and sepsis, visualized using the CytoHubba plugin in Cytoscape. Nodes represent genes, and the edges represent interactions between them. The color intensity of the nodes corresponds to the degree of centrality, with darker shades indicating higher degrees of connectivity. (B) Bar plot showing the gene expression levels of hub genes in the oxidative phosphorylation pathway. The horizontal axis shows the hub genes, and the vertical axis represents gene expression

To further interpret the biological functions of these genes, we carried out KEGG pathway enrichment analysis via the ClueGO and CluePedia plugins (Figure S3). The results highlighted significant enrichment in the OxPhos pathway, implicating these genes in mitochondrial metabolism, ROS production, and inflammatory responses. Moreover, their involvement in pathways regulating immune function and energy balance suggests a potential mechanistic link between AP and sepsis progression.

For pinpointing central genes within the interaction network, we utilized four complementary topological analysis methods—MCC, DMNC, MNC, and Degree—available through the CytoHubba plugin. Cross-method consensus revealed seven hub genes that were consistently top-ranked across all algorithms (Fig. 5B), reinforcing their importance as regulatory nodes within the co-expression network.

To identify reliable diagnostic markers, we applied two machine learning algorithms—SVM-RFE and Random Forest (RF)—using 10-fold cross-validation to reduce overfitting. SVM-RFE selected the optimal gene set based on the lowest classification error (Fig. 6A, C), while RF, built with 500 trees, ranked genes by mean decrease in Gini impurity, retaining those with scores above 2 (Fig. 6B, D). The top 30 genes were visualized using a lollipop plot. By intersecting results from both methods across AP and sepsis cohorts, we identified six core hub genes—NDUFA1, COX7A2, COX7B, UQCRQ, SNRPG, and NDUFA4—consistently ranked as top features (Fig. 6E).

Fig. 6.

Identification and Network Analysis of Common Key Genes in AP and Sepsis. (A, B) Importance plots of hub genes identified in AP using machine learning methods (SVM-RFE and RF); (C, D) Importance plots of hub genes identified in sepsis using machine learning methods (SVM-RFE and RF);(E) Venn diagram showing the overlap of key genes identified by RF and SVM-RFE in AP and sepsis. Six common hub genes (NDUFA1, COX7A2, COX7B, UQCRQ, SNRPG, and NDUFA4) were identified across both diseases; (F) GeneMANIA analysis of core related DEGs and their co-expressed genes in the network; (G) Box plot depicting the differential expression of the six shared hub genes between AP patients and control groups; (H) Box plot comparing the expression levels of the six common hub genes in sepsis patients and controls

To validate the biological relevance of these candidates, we performed GeneMANIA network analysis, which confirmed that five of the six genes (NDUFA1, COX7A2, COX7B, UQCRQ, and NDUFA4) were tightly linked to the OxPhos pathway (Fig. 7F). Additionally, differential expression analysis via t-tests showed that all six hub genes were significantly upregulated in both AP and sepsis samples relative to controls, supporting their potential utility as diagnostic biomarkers (Fig. 6G, H).

Fig. 7.

ROC Curves for Key Hub Genes in Training and Validation Sets for AP and Sepsis. (A) ROC curves for individual hub genes (NDUFA1, UQCRQ, COX7A2, NDUFA4, SNRPG, COX7B) in the training set of AP patients, with their corresponding AUC values. The overall model achieved an AUC of 0.969, with a 95% confidence interval (CI) of 0.933–0.995; (B) ROC curves for individual hub genes in sepsis patients, with corresponding AUC values. The overall model performance is shown on the right, achieving an AUC of 0.869, with a 95% CI of 0.826–0.908; (C) ROC curves for hub genes in the GSE3644 dataset, showing high predictive accuracy with AUC values ranging from 0.889 to 1.000. The model‘s performance achieved an AUC of 1.000 (95% CI: 1.000–1.000); (D) ROC curves for hub genes in the GSE28750 dataset, showing varying levels of predictive accuracy, with AUC values ranging from 0.698 to 0.975. The model’s performance achieved an AUC of 1.000 (95% CI: 1.000–1.000)

Diagnostic value and validation of core biomarkers

We conducted a thorough evaluation of the diagnostic potential of these six core hub genes in samples from sepsis and AP. Initially, receiver operating characteristic (ROC) analysis was performed on the training set. The analysis of AP samples indicated a strong correlation among the six genes, with a combined ROC curve AUC value of 0.969 (95% CI: 0.933–0.995). Individually, the AUC values for each gene were all above 0.75: NDUFA1: 0.960, COX7A2: 0.863, COX7B: 0.824, UQCRQ: 0.814, SNRPG: 0.965, and NDUFA4: 0.782 (Fig. 7A). In sepsis samples, all six genes had an AUC value greater than 0.70: NDUFA1: 0.806, COX7A2: 0.772, COX7B: 0.730, UQCRQ: 0.813, SNRPG: 0.749, and NDUFA4: 0.789. The combined ROC analysis of the six genes yielded an AUC value of 0.869 (95% CI: 0.826–0.908) (Fig. 7B).

To validate the effectiveness and reliability of our model, we selected GSE3644 and GSE28750 datasets as validation sets. Given the high genetic, signaling pathway, immune response, and inflammatory regulation similarities between mice and humans, mouse models serve as valuable tools for studying human diseases. In particular, they effectively simulate key pathological processes such as immune cell activation, cytokine regulation, and metabolic stress. Additionally, due to the limited availability of comprehensive human AP datasets, we utilized mouse models for validation in this study.Analysis of the AP dataset revealed that five key genes demonstrated strong diagnostic performance, with ROC curve AUC values all exceeding 0.8, as follows: NDUFA1: 1.000, COX7A2: 0.944, UQCRQ: 0.944, SNRPG: 0.972, NDUFA4: 0.889. Furthermore, when combining all six genes for joint ROC analysis, the AUC reached 1.000, with all genes displaying statistically significant differential expression (p < 0.05) (Fig. 7C). Similarly, sepsis dataset analysis indicated a strong association among these six genes, with a combined ROC AUC of 1.000. Individually, these genes also exhibited high AUC values: NDUFA1: 0.897, COX7A2: 0.698, COX7B: 0.935, UQCRQ: 0.765, SNRPG: 0.840, NDUFA4: 0.975 (Fig. 7D). These genes also showed significant differential expression in the validation sets.

In conclusion, our findings demonstrate that these genes possess high diagnostic value for both AP and sepsis, highlighting their potential as reliable biomarkers for early disease detection.

Analysis of transcriptional regulation and prediction of potential drugs

We constructed a lncRNA-miRNA-mRNA regulatory network for the six hub genes (Fig. 8), revealing the complex regulatory relationships among lncRNAs, miRNAs, and mRNAs. The network comprises two sub-networks centered around SNRPG and NDUFA4. SNRPG is mainly regulated by multiple lncRNAs, such as LINC00265 and SSTR5-AS1, through hsa-miR-93-3p and hsa-miR-185-5p. These lncRNAs may bind to miRNAs via the competing endogenous RNA (ceRNA) mechanism, thereby reducing the direct inhibition of SNRPG by miRNAs. Similarly, NDUFA4 is indirectly regulated by multiple lncRNAs (e.g., TP73-AS1 and CDR1-AS) through hsa-miR-7-5p and hsa-miR-877-5p. This figure illustrates the intricate interactions between lncRNAs, miRNAs, and mRNAs, suggesting that multiple lncRNAs may modulate the expression levels of key genes SNRPG and NDUFA4 by competing with miRNAs. This regulatory network is crucial for understanding disease mechanisms and identifying molecular targets.

Fig. 8.

mRNA-miRNA-lncRNA Co-expression Network for Hub Genes SNRPG and NDUFA4 in Acute Pancreatitis and Sepsis. Red oval nodes represent mRNA, green triangular nodes indicate miRNA, and blue diamond-shaped nodes represent lncRNA

We downloaded drug data closely related to the six hub genes from the CMap. The results indicated that selenium, metformin hydrochloride, hydralazine, retinoic acid, pingyangmycin, and terpineol could alleviate or even reverse the disease state (Fig. 9A). The molecular structures of some of the predicted drugs are presented in Fig. 9B.

Fig. 9.

Correlation Analysis of Various Compounds with Hub Genes and Chemical Structures. (A) displays the statistical correlation data for several compounds (selenium, metformin hydrochloride, hydralazine, vitinoin, pingyangmycin, and pinosylvin), including significance (P-value), odds ratio, combined score, and associated genes. (B-F) show the chemical structures of each compound

Immune infiltration analysis of shared diagnostic genes

Our analysis evaluated the immune cell composition in AP and sepsis patients, revealing significant differences in immune cell profiles between patient and control groups (Fig. 10A, B). In both AP and sepsis patients, compared to healthy controls, significant differences were observed in T cells CD4 memory resting, T cells CD4 memory activated, NK cells resting, neutrophils, and mast cells resting (Fig. 10C, D). Specifically, T cells CD4 memory activated and mast cells resting were significantly increased in AP and sepsis patients, whereas T cells CD4 memory resting were decreased. Additionally, NK cells resting and neutrophils showed significant changes in the patient groups, reflecting distinct effects of these diseases on innate and adaptive immunity. Regarding B cell subsets, B cells naive were abnormally expressed in AP patients, while B cells memory showed notable changes in sepsis patients, suggesting differences in B cell differentiation and function regulation under these disease states. These findings highlight specific immune system alterations in AP and sepsis, providing critical insights into immune dysregulation mechanisms in these pathological states.

Fig. 10.

Immune Cell Infiltration Analysis for AP and Sepsis Datasets. (A, B) Correlation heatmaps of immune cell infiltration in AP and sepsis, showing the relationships between various immune cell types. Positive correlations are represented in red, while negative correlations are shown in blue, with the color intensity indicating the strength of the correlation. (C, D) Violin plots illustrating the proportions of various immune cell types in the control (blue) and disease (red) groups in the AP and sepsis datasets, respectively. Significant differences in cell type abundance are observed between the groups. The p-values indicate the statistical significance of these differences between groups

Additionally, we examined the correlation between core immune-related DEGs and immune cell components in patients with sepsis and AP (Fig. 11A). The results indicated that in AP patients, COX7A2 and NDUFA4 exhibited negative correlations with resting CD4 memory T cells, whereas COX7B and NDUFA1 showed positive correlations with activated CD4 memory T cells. Furthermore, COX7B and SNRPG were negatively correlated with resting NK cells. COX7A2 and NDUFA4 also displayed negative correlations with neutrophils, while UQCRQ was positively correlated with resting mast cells. In sepsis patients(Fig. 11B), COX7A2 and COX7B were negatively correlated with resting CD4 memory T cells, while NDUFA4 and SNRPG exhibited positive correlations with activated CD4 memory T cells. Additionally, COX7B was negatively correlated with resting NK cells, and neutrophils showed positive correlations with COX7A2 and NDUFA1. Resting mast cells were negatively correlated with UQCRQ. These findings highlight specific associations between certain DEGs and immune cell components across different pathological states, suggesting a crucial role for these genes in immune regulation in both AP and sepsis. This discovery offers new insights into the mechanisms of immune dysregulation in these conditions and establishes a foundation for future therapeutic strategies.

Fig. 11.

Correlation Analysis of Hub Genes with Immune Cell Infiltration in AP and Sepsis. (A) Correlation analysis of six hub genes (COX7A2, COX7B, NDUFA1, NDUFA4, SNRPG, UQCRQ) with various immune cell types in acute pancreatitis (AP). The correlation coefficients are represented along with the p-values, indicating the strength and significance of the correlation between gene expression and immune cell infiltration. The size of the circles represents the absolute correlation (abs(cor)), and the color scale represents the p-value, with darker colors showing more significant correlations; (B) Correlation analysis of the same six hub genes with immune cell infiltration in sepsis

Predictive value of core genes

In order to explore the potential of these hub genes as diagnostic markers for acute pancreatitis (AP), we measured the protein and transcript levels of UQCRQ, COX7A2, SNRPG, and NDUFA4 in serum specimens collected from AP patients (n = 3) and healthy individuals (n = 3). All four genes exhibited a clear upregulation trend in the AP group. Western blot analysis revealed markedly elevated protein levels of these genes in AP patients relative to controls (Fig. 12A). Consistently, qRT-PCR results confirmed a significant increase in their mRNA expression levels (Fig. 12B–E). These differences were statistically significant across three independent biological replicates (P < 0.05 to P < 0.0001).These findings suggest that mitochondrial dysfunction may contribute to the pathophysiological progression of acute pancreatitis, with UQCRQ, COX7A2, SNRPG, and NDUFA4 potentially acting as central regulators in this process. Notably, the consistent overexpression of these hub genes highlights their promise as candidate diagnostic biomarkers for identifying and monitoring severe cases of acute pancreatitis.

Fig. 12.

Expression analysis of UQCRQ, COX7A2, SNRPG, and NDUFA4 in control and AP groups. (A) Representative Western blot images showing the protein expression levels of UQCRQ, COX7A2, SNRPG, and NDUFA4 in control samples (CON-1, CON-2, CON-3) and acute pancreatitis samples (AP-1, AP-2, AP-3). ALBUMIN was used as a loading control.(B–E) Quantitative qRT-PCR results showing mRNA expression levels of UQCRQ (B), COX7A2 (C), SNRPG (D), and NDUFA4 (E) in three biological replicates from the control group (blue) and AP group (red). Data are presented as fold change relative to the control group. Statistical analysis was performed using unpaired t-tests (*P < 0.05, **P < 0.01, ***P < 0.001, ***P < 0.0001)

Discussion

AP is one of the most common gastrointestinal inflammatory diseases requiring hospitalization and a leading cause of patient mortality (Czapári et al. 2023). During the progression of AP, 40–70% of patients in the middle and late stages develop sepsis due to factors such as pancreatic or peripancreatic tissue necrosis and gut microbiota translocation (Wang et al. 2023a, b; Susak et al. 2021). This complication is closely associated with increased mortality and poor prognosis (Xia et al. 2024). Therefore, early identification of AP patients at risk of developing sepsis is crucial for reducing mortality and alleviating disease burden.

In this study, we selected the GSE54514, GSE57065, and GSE95233 datasets as training sets for sepsis analysis, while the GSE194331 dataset was used as the training set for AP. Using WGCNA, PPI network construction, and two machine learning methods, we identified six core genes: NDUFA1, COX7A2, COX7B, UQCRQ, SNRPG, and NDUFA4. Immune cell composition analysis revealed significant differences between disease and control groups. We further validated these genes using the GSE3644 and GSE28750 datasets as validation sets for sepsis and AP, respectively, and confirmed their accuracy through Mendelian randomization.

PPI analysis and disease progression

PPI analysis identified core proteins involved in AP and sepsis, revealing their roles in inflammation, immune regulation, and cellular stress(Liu et al. 2018). These proteins are primarily associated with OxPhos, cytokine signaling, and immune modulation pathways, which are critical for disease progression. Dysregulation of these proteins may lead to immune cell dysfunction, oxidative stress-induced damage, and metabolic imbalances(Ozger 2023). Notably, proteins with high centrality in the PPI network represent potential therapeutic targets for AP and sepsis. To further validate these findings, we plan to integrate additional datasets and conduct experimental studies to enhance the reliability and clinical applicability of our research.

Core genes and oxidative phosphorylation (OxPhos) pathway

Through GO and KEGG enrichment analyses, combined with PPI network analysis, we identified a strong association between the core genes (NDUFA1, COX7A2, COX7B, UQCRQ, and NDUFA4) and the OxPhos pathway. PPI analysis, a powerful tool for elucidating disease mechanisms, helped us uncover key protein interactions involved in AP progression to sepsis.

These six core genes play critical roles in different stages of the OxPhos process. Specifically, NDUFA1 (a subunit of Complex I), COX7A2 and COX7B (subunits of Complex IV), and UQCRQ (a subunit of Complex III) contribute to mitochondrial dysfunction when dysregulated. This leads to impaired OxPhos, excessive ROS production, oxidative stress, and inflammatory responses (Uehara et al. 2014; Yan et al. 2023; Carty et al. 2024; Liu et al. 2024; Xiong et al. 2024). Furthermore, SNRPG may indirectly influence OxPhos by regulating NDUFA4 expression and promoting immune cell dysfunction, thereby exacerbating sepsis progression.

Additionally, the PPI network highlights the crucial role of immune cell dysregulation in sepsis. COX7B and NDUFA4 may enhance T-cell activation, contributing to systemic inflammation, while SNRPG downregulation may impair NK cell function, facilitating local infection spread. These insights suggest that restoring mitochondrial function, regulating OxPhos, and modulating immune responses could serve as novel therapeutic strategies for AP-associated sepsis.

The role of oxidative phosphorylation (OxPhos) and ROS in AP and sepsis

OxPhos is the core process of cellular energy metabolism, responsible for generating over 90% of cellular ATP (Hüttemann et al. 2012). In this process, the electron transport chain (ETC) transfers electrons from reduced coenzymes (e.g., NADH and FADH₂) to oxygen, ultimately producing water. However, electron leakage, particularly at Complex I and III, leads to ROS production, including superoxide anions (O₂⁻), hydrogen peroxide (H₂O₂), and hydroxyl radicals (·OH). These ROS trigger oxidative stress, damaging cellular membranes, proteins, and DNA, ultimately initiating inflammation and cell death pathways (Singer and London 2017; Cho and Kim 2024).

ROS plays a pivotal role in AP progression. Studies have shown that AP is often accompanied by excessive ROS production, which directly damages pancreatic cells and activates inflammatory mediators (e.g., cytokines, chemokines, and enzymes), exacerbating tissue injury(Liu et al. 2022; Li et al. 2023). For example, elevated VNN1, a molecule that induces oxidative stress responses, promotes ROS generation, thereby intensifying AP-related inflammation and disease progression (Martin et al. 2004; Kang et al. 2016). Moreover, ROS activates multiple signaling pathways, such as NF-κB, MAPK, and JNK, amplifying local inflammatory responses, leading to sustained immune cell activation and further pancreatic tissue damage (Das et al. 2024; Reiter et al. 2024).

Similarly, oxidative stress and ROS dysregulation play a crucial role in sepsis progression. Mitochondrial dysfunction, particularly disruptions in ATP synthesis and the ETC, is a hallmark of sepsis. Excessive ROS leads to lipid peroxidation, DNA damage, and protein denaturation, which impair mitochondrial function, decrease membrane potential, and induce apoptosis and necrosis, ultimately contributing to organ failure (Patoli et al. 2020; Wang et al. 2023a, b). In addition to driving cell death and tissue injury, ROS also plays a dual role in immune regulation. While moderate ROS levels activate immune cells (e.g., neutrophils and macrophages) for immune defense, excessive ROS disrupts immune homeostasis, leading to immune dysfunction and systemic immune imbalance (Pei et al. 2024). Targeting ROS overproduction or restoring mitochondrial function may provide a viable strategy to mitigate AP progression to sepsis.

Comparison with previous studies and immune cell dysregulation

Unlike our findings, Si-Jiu Yang et al. identified ARG1 and HP as common core genes in AP and sepsis using GEO database analysis(Yang et al. 2024). These genes reflect immunosuppressive and inflammatory states, respectively, playing complementary roles in disease progression. However, our study consistently highlights the critical role of immune cells in AP and sepsis. Immune cells play a key role in coordinating immune responses and maintaining immune homeostasis. Understanding immune cell alterations, including their quantity, functional state, and interactions, is essential for elucidating sepsis mechanisms, predicting treatment efficacy, and developing novel therapeutic strategies.

Our immune cell infiltration analysis revealed significant alterations in immune cell composition in AP and sepsis patients, particularly in T cells, NK cells, neutrophils, mast cells, and B-cell subsets. This suggests distinct responses between the innate and adaptive immune systems, potentially contributing to immune dysfunction.

The role of the OxPhos pathway in both AP and sepsis indicates a shared immunoregulatory mechanism. During AP progression to sepsis, significant changes in immune cell function and gene expression drive the transition from local inflammation to systemic immune dysregulation. Specifically, COX7B and NDUFA4 enhance T-cell memory activation, increasing immune responses and systemic inflammation, thereby elevating sepsis risk. SNRPG negatively regulates NK cell function, potentially impairing immune surveillance and facilitating local infection spread. UQCRQ promotes mast cell activation, enhancing immune regulation; however, excessive mast cell activation may contribute to systemic inflammatory responses, worsening sepsis symptoms. COX7A2 and NDUFA1 activate neutrophils, which aid in local infection response but may cause widespread tissue damage if excessively activated, ultimately leading to immune dysregulation and sepsis development.

Clinical significance assessment

Our lncRNA–miRNA–mRNA regulatory network analysis suggests that the interaction between NDUFA4 and SNRPG may promote the development of AP-associated sepsis by modulating the OxPhos pathway. SNRPG (Small Nuclear Ribonucleoprotein G) is a critical nuclear RNA-binding protein involved in RNA splicing, gene expression regulation, and RNA metabolism(Pinheiro et al. 2024; Mabonga et al. 2021). Emerging evidence indicates a bidirectional regulatory relationship between SNRPG and specific miRNAs, which may influence inflammatory responses and systemic homeostasis.

Based on these findings, we hypothesize that SNRPG may regulate the expression of hsa-miR-7-5p and hsa-miR-93-3p, thereby enhancing NDUFA4-mediated OxPhos activity and accelerating the progression of AP. By systematically investigating the expression profile of SNRPG in AP and evaluating its role in OxPhos regulation, we aim to elucidate the potential molecular mechanisms by which SNRPG modulates NDUFA4 via miRNA signaling, contributing to ROS production and inflammation.This study is of significant importance for assessing the therapeutic potential of SNRPG and its downstream pathways in AP. Our findings are expected to provide novel insights into the molecular pathogenesis of acute pancreatitis and establish a foundation for the development of early diagnostic biomarkers and targeted therapeutic strategies, ultimately advancing precision medicine approaches for this life-threatening condition.

This study reveals the critical role of OxPhos pathway dysfunction and immune cell dysregulation in AP progression to sepsis and therapeutic targets. Early identification of these core genes and immune cell changes could serve as biomarkers for predicting sepsis in AP patients and provide new insights for immunotherapy strategies. Future research should further explore the diagnostic and therapeutic potential of these genes in clinical applications.

Future perspectives

In future studies, we plan to further validate the roles of these six core genes in AP. First, we will establish both cellular models and mouse models of pancreatitis to investigate the specific functions of these genes in the onset and progression of AP. Additionally, we will actively collect serum samples from AP patients and integrate genomic analyses to examine the expression patterns of these genes and their correlations with clinical symptoms, disease severity, and prognosis.

Furthermore, based on findings from cellular and animal models, we will conduct pharmacological studies to explore whether modulating the expression of these core genes through therapeutic interventions could mitigate AP symptoms and improve patient outcomes. This approach not only enhances our understanding of the molecular mechanisms underlying AP but also identifies potential therapeutic targets for future clinical applications.

Limitations

This study identified disease-associated biomarkers through bioinformatics analysis and validated their expression via qRT-PCR and Western blot in patient serum samples. However, the lack of in vivo validation limits the understanding of their biological function under physiological conditions.

The limited sample size, particularly in the external validation cohort, may reduce the generalizability of the findings, highlighting the need for validation in larger, independent cohorts.

Additionally, data heterogeneity—such as differences in experimental design, platforms, and sample types across GEO datasets—may affect the consistency of the results.

Finally, the findings are influenced by the choice of computational methods, which may introduce bias. Further studies using complementary algorithms and experimental approaches are warranted to confirm these results.

Conclusion

In summary, our study identified core genes between acute pancreatitis and sepsis, including NDUFA1, COX7A2, COX7B, UQCRQ, SNRPG, and NDUFA4. Additionally, the study showed that acute pancreatitis may promote the oxidative phosphorylation pathway through SNRPG, which in turn contributes to the development of sepsis. Moreover, there are associations between these core genes and dysregulated immune cells. Overall, these dysregulated core genes provide potential research directions for acute pancreatitis and sepsis.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Weina Lu: Conceptualization, methodology, data analysis, writing– original draft preparation.Yifeng Mao: Methodology, data collection, writing– review and editing.Qingqing Chen: Data curation, formal analysis, visualization.Shangwen Cai: Methodology, investigation, supervision.Panpan Xu: Writing– review and editing, project administration.Chenghua Xu: Conceptualization, resources, funding acquisition.Cheng Zheng: Data analysis, writing– review and editing.Jian Lan: Conceptualization, supervision, writing– review and editing.All authors reviewed the manuscript.

Funding

This work was supported in part by grants from the Medical Health Science and Technology Project of Zhejiang Provincial Health Commission [No. 2024KY1824, Qingqing Chen; No. 2024KY1811, Panpan Xu; 2023KY401, Chenghua Xu; No. 2025KY1871, Cheng Zheng]; The Science and Technology Project of Taizhou [No. 23ywb70, Qingqing Chen; No.24ywa44, Cheng Zheng].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of(Municipal Hospital affiliated to Taizhou University), and was conducted in accordance with the Declaration of Helsinki. The ethics approval number for this study is LWSL202400220.

Consent for publication

The authors of this paper all consent to its publication.

Competing interests

The authors declare no competing interests.