A pyroptosis-related gene signature for the diagnosis of acute pancreatitis

Yuting Wang; Jun Li; Zhongsu Yu; Shuyuan Li; Yuxia Chen; Yun Pan; Liangping Cheng; Guangyuan Yu

doi:10.1371/journal.pone.0347204

. 2026 Apr 17;21(4):e0347204. doi: 10.1371/journal.pone.0347204

A pyroptosis-related gene signature for the diagnosis of acute pancreatitis

Yuting Wang ^1,^#, Jun Li ^1,^#, Zhongsu Yu ¹, Shuyuan Li ¹, Yuxia Chen ¹, Yun Pan ¹, Liangping Cheng ^1,^2,^*,^#, Guangyuan Yu ^1,^3,^*,^#

Editor: Alexis G Murillo Carrasco⁴

PMCID: PMC13089757 PMID: 41996313

Abstract

Acute pancreatitis (AP) is a severe inflammatory disorder in which pyroptosis—a pro-inflammatory form of programmed cell death—may contribute to pathogenesis. However, the complete transcriptional profile of pyroptosis-related genes (PRGs) in AP and their potential as diagnostic biomarkers remain underexplored. This study aimed to systematically characterize pyroptosis-associated transcriptional signatures and identify the reliable biomarkers for diagnostic purposes. Three transcriptomic datasets from murine AP models were integrated to identify pyroptosis-related differentially expressed genes (PRDEGs). Functional enrichment and immune cell infiltration analyses were conducted to elucidate the biological pathways and immune microenvironment alterations associated with these genes. mRNA-transcription factor (TF) and mRNA-microRNA (miRNA) regulatory networks were constructed to investigate underlying molecular interactions. Machine learning techniques, including support vector machine (SVM) and least absolute shrinkage and selection operator (LASSO), were applied for feature selection, leading to the identification of key diagnostic markers and the development of a logistic regression model. The regression model were then assessed using an independent cohort of human peripheral blood samples. Eleven PRDEGs were identified, with enrichment observed in processes such as cytoskeletal organization, cell-substrate adhesion, and critical inflammatory signaling pathways, including MAPK and NF-κB. Immune infiltration analysis revealed significant correlations between these PRDEGs and various immune cell subsets, particularly M1 macrophages, Treg cells, and monocytes. A four-gene diagnostic signature, comprising ANXA3, IQGAP1, RELA, and VTN, was established through SVM and LASSO analysis. In the independent human cohort, the fixed-coefficient four-gene model demonstrated reduced discrimination, which likely reflects interspecies and tissue-specific variations. However, after optimizing the model to exclude non-significant predictors, a refined two-gene signature (ANXA3 and IQGAP1) exhibited improved accuracy, with excellent calibration and clinical net benefit. This study offers a comprehensive transcriptomic analysis of the pyroptosis-mediated landscape and immune microenvironment in AP. An optimized two-gene signature, comprising ANXA3 and IQGAP1, was validated in a human cohort with superior accuracy, reflecting critical disruptions in inflammatory pathways and cytoskeletal organization. Notably, ANXA3 demonstrated potential for stratifying disease severity. Although these markers hold potential for molecular diagnosis, further prospective studies are essential to establish their clinical specificity and generalizability across diverse populations.

Introduction

AP is a sudden inflammation of the pancreas that can vary in severity from mild, self-limiting conditions to severe, life-threatening illnesses. It is a prevalent gastrointestinal cause of hospital admissions globally [1], and the mortality rate for severe acute pancreatitis can reach as high as 30% [2]. Despite advancements in medical care, the management of AP primarily consists of supportive measures, as there are currently no specific pharmacological treatments available [3]. This underscores the pressing need for innovative diagnostic markers and therapeutic targets to enhance the clinical outcomes of patients with AP.

Pyroptosis has hallmarks of cell swelling, lysis, and secretion of pro-inflammatory cytokines, contributing to the inflammatory response [4]. Recent studies have highlighted the significant role of pyroptosis in various inflammatory diseases such as sepsis, atherosclerosis, and neurodegenerative disorders [5–7]. Despite this, the involvement of pyroptosis in acute pancreatitis remains poorly understood. Exploring the mechanisms of pyroptosis in acute pancreatitis could may offer new insights into the disease’s pathophysiology and potentially identify novel therapeutic targets.

The objective of this study was to explore the involvement of pyroptosis in acute pancreatitis through the identification of PRDEGs. Utilizing bioinformatics methodologies, we conducted comprehensive analyses of gene expression datasets obtained from the Gene Expression Omnibus (GEO) database. Subsequently, we performed functional enrichment analyses and developed a four-gene diagnostic model using machine-learning approaches, which was then externally validated in an independent human RNA-seq cohort.

Methods

Data acquisition

The GEOquery package [8] from the R programming language was used to download data from the GEO database. Three datasets—GSE109227 [9], GSE65146, and GSE121038—focused on AP were obtained. These datasets, derived from Mus musculus, include pancreas tissue samples. A detailed overview of these datasets is provided in Table 1.

Table 1. GEO microarray chip information.

	GSE109227	GSE65146	GSE121038
Platform	GPL6246	GPL6246	GPL10787
Species	Mus musculus	Mus musculus	Mus musculus
Tissue	Pancreas	Pancreas	Pancreas
Samples in AP group	6	39	8
Samples in Control group	5	5	7
Reference	PMID: 30139658	\	\

Open in a new tab

In the GSE109227 dataset, the GPL6246 chip platform was used, which included 6 AP samples and 5 control samples. Similarly, the GSE65146 dataset utilized the GPL6246 chip platform and contained 39 AP samples, 5 control samples, and 27 mutant mouse samples. The GSE121038 dataset, on the other hand, employed the GPL10787 chip platform and consisted of 8 AP samples and 7 control samples. For the purpose of this study, only the AP and control groups were considered.

We utilized the GeneCards database [10] (https://www.genecards.org/) to identify PRGs. By searching for the terms “Pyroptosis” and “Protein Coding,” we selected genes with a relevance score > 1, identifying 405 PRGs. A subsequent search for “Pyroptosis” in the PubMed database (https://pubmed.ncbi.nlm.nih.gov/) retrieved 33 PRGs from the literature [11]. After removing duplicates, we combined the results, yielding a total of 413 human PRGs. These were cross-referenced with 334 mouse PRGs using the R package Homologene (Version 1.4.68.19.3.27).

The sva R package [12] (v3.50.0) was used to remove batch effects across the GSE109227, GSE65146, and GSE121038 datasets, generating a combined dataset comprising 53 AP cases and 17 controls. The integrated dataset was then normalized using the limma R package [13]. Probe annotations were harmonized and standardized across platforms. Finally, principal component analysis (PCA) was performed on the expression matrix to verify the effectiveness of batch-effect correction.

Differentially expressed genes related to pyroptosis associated with AP

Based on the sample annotations in the GEO database, samples were assigned to either the control or AP group. Differential expression analysis between groups was performed in R using appropriate Bioconductor packages. Differentially Expressed Gene (DEGs) were defined as those with an absolute log2 fold change (|log2FC|) > 1 and adjusted P value < 0.05. Genes with log2FC > 1 were considered upregulated, whereas those with log2FC <−1 were considered downregulated. DEGs were visualized using volcano plots.

To identify PRDEGs associated with AP, PRGs were intersected with the DEGs identified from combined GEO datasets, with |log2FC| > 1 and adjusted P value < 0.05. A Venn diagram was then constructed to visualize the overlapping genes. The PRDEGs were further represented using a heatmap.

GO and KEGG enrichment analysis

Gene Ontology (GO) analysis [14] is a widely used method for investigating functional enrichment across categories such as Biological Process (BP), Cellular Component (CC), and Molecular Function (MF). The Kyoto Encyclopedia of Genes and Genomes (KEGG) [15] is a well-established resource for understanding genomic data, biological pathways, diseases, and drugs. Enrichment analyses of PRGs were performed using the clusterProfiler package [16]. The criteria for significant differences were set at a false discovery rate (FDR, q-value) < 0.25 and adjusted P value < 0.05.

Establishment of diagnostic model for AP

To develop diagnostic models for AP using the combined GEO datasets, we performed logistic regression to identify PRDEGs that discriminate AP from control samples. This analysis evaluated the association between the candidate predictors (PRDEGs) and the binary outcome (AP vs control). PRDEGs with P < 0.05 were considered statistically significant and were included in the final logistic regression model; the resulting effect estimates were visualized using a forest plot.

Utilizing PRDEGs identified through logistic regression, SVM [17] was employed for establishing an optimized predictive model. The selection of the number of genes that provided the highest accuracy and lowest error rate was a key aspect of model optimization.

Subsequently, the study applied the R package glmnet [18] (Version 4.1–8) to perform LASSO regression analysis. The seed was set to 600, and family = “binomial” was used as the parameter. LASSO regression analysis included a term known as penalty (lambda × absolute coefficient of slope) in the calculation of linear regression, which helped to reduce overfitting and improve the model’s capacity for applicability. The results were visualized using trajectory plots and model diagnostic plots. Ultimately, an AP diagnostic model was constructed by integrating the PRDEGs selected by LASSO.

RiskScore was obtained with risk coefficients from the LASSO regression analysis using the following formula: [].

Validation of the diagnostic model for AP.

ROC curves and AUC values were generated using the pROC package [19] to assess the diagnostic performance of the AP model RiskScore. Subsequently, a nomogram was constructed based on the logistic regression model to visualize the contribution and combined effects of the variables on a two-dimensional coordinate scale [20].

The correctness and identification capabilities of an AP diagnostic model created using LASSO regression analysis will be assessed using calibration analysis. A Calibration Curve was established to visualize the consistence of expected likelihoods and actual results.

Furthermore, DCA [21] made by R package ggDCA (Version 1.1) was performed to produce DCA maps based on the model genes in the combined GEO Datasets. To evaluate the overall clinical utility of the diagnostic model in healthcare decision-making, DCA was performed.

Functional similarity (Friends) analysis

The quantitative assessment of gene and genome similarity through semantic comparison of GO annotations has become a foundational approach in bioinformatics analyses. To assess the functional similarity of the model genes, referred to as Friends analysis, the R package GOSemSim [22] was employed.

DEG verification and ROC curve

To analyze the differential expression of PRDEGs between control and AP samples, a group comparison map was constructed, enabling a visual evaluation of the DEGs’ expression levels. ROC curves for the PRDEGs were then plotted using the R package pROC, and AUC values were calculated. AP samples selected from the GEO Database were classified into HighRisk and LowRisk groups using the median RiskScore from the AP diagnostic models. To further evaluate the DEGs of model genes, a group comparison map was drawn to illustrate gene expression differences between the two groups. ROC curves for the model genes were plotted using the pROC package, and AUC values were calculated to assess their diagnostic value in AP progression.

Gene set enrichment analysis (GSEA)

GSEA [23] was performed to evaluate gene distribution trends within predefined gene sets, ranked by their correlation with the phenotype, to determine their contribution to AP. In this study, genes from the Combined GEO Datasets were ranked by their logFC values. GSEA was then performed using the R package clusterProfiler, with the following parameters: seed = 2020, number of computations = 1000, minimum number of genes in each gene set = 10, and maximum number of genes in each gene set = 500.

The gene sets for GSEA were from MSigDB [24](https://www.gsea-msigdb.org/gsea/msigdb) with screening criteria of adj.p < 0.05 and q < 0.25.

AP samples from the Combined GEO Datasets were categorized into HighRisk and LowRisk groups based on the median RiskScore of the diagnostic model. DEGs were analyzed using the limma R package, and volcano plots were generated to display genes with |logFC| > 1 and adj.p < 0.05 in the HighRisk and LowRisk groups. The R package pheatmap was employed to create heatmaps of these gene expression values.

Finally, GSEA was performed on all genes in the AP samples using clusterProfiler, with gene sets from the MSigDB c2.cp.all.v2022.1.Hs.symbols.gmt. The screening criteria for GSEA were adj.p < 0.05 and q < 0.25, with the minimum and maximum numbers of genes in each gene set set to 10 and 500, respectively.

Protein-Protein Interaction (PPI) Network

Through association with model genes in the post-transcriptional phase, TFs regulate gene expression. The ChIPBase database [25] (http://rna.sysu.edu.cn/chipbase/) was employed to identify TFs and examine their regulatory functions on the model genes. Cytoscape [21] was used to visualize the mRNA-TF regulatory network.

MiRNAs play a critical ecological role in cellular growth and progression by controlling multiple target genes and regulating the same gene through several miRNAs. MiRNAs associated with the model genes were obtained from the StarBase v3.0 database [26], allowing for the investigation of miRNA-gene associations. Cytoscape software was subsequently used to illustrate the mRNA-miRNA regulatory network.

Immune infiltration analysis

CIBERSORT [27] employs linear support vector regression to deconvolute transcriptomic expression matrices and estimate the composition and abundance of immune cells in mixed cell populations. By applying the CIBERSORT algorithm combined with immune cell gene matrix features and filtering for data with an immune cell enrichment score greater than zero, the immune cell infiltration results in the Combined Datasets were obtained. The proportion of immune cells was visualized using a bar chart.

Group comparison plots were generated using the R package ggplot2 to visualize differences in immune cell composition between control and AP samples. Spearman correlation analysis was performed to assess relationships, and the results were visualized in a correlation heatmap. Spearman correlation was also applied to evaluate the association between immune cells and model genes, which was then visualized using a bubble plot.

Using the same procedures described above, immune cell infiltration was further compared between the LowRisk and HighRisk AP subgroups.

Validation of the diagnostic model for AP in a human RNA-seq cohort (GSE194331)

We downloaded the raw count matrix for the independent human RNA-seq cohort GSE194331, consisting of adult patients (AP = 87, Control = 32), from the GEO database. The counts were normalized using DESeq2 size factors, and the normalized expression values were log2-transformed for downstream analyses. The samples were grouped into the AP and control cohorts based on the phenotype annotations in the GEO metadata. To visualize predefined genes of interest, faceted boxplots were generated for the 11 PRDEGs, and group differences (AP vs Control) were assessed using the Mann-Whitney U test.

To assess the cross-domain transportability, the four-gene diagnostic model, previously derived from the murine discovery cohort, was applied to the human blood cohort using fixed coefficients (without re-estimation). Diagnostic performance was evaluated using ROC analysis and AUC. Due to the reduced discrimination observed with fixed coefficients, an exploratory optimization of the model was performed. The Akaike Information Criterion (AIC) was first used to compare the fit and parsimony between the full four-gene model and the reduced two-gene model (ANXA3, IQGAP1). Subsequently, the coefficients for the optimal gene signature were re-estimated using logistic regression. Diagnostic discrimination and clinical utility were assessed via ROC analysis and DCA, respectively, following the procedures described in the ‘Validation of the diagnostic model for AP’ section. Calibration was quantified by the Brier score and calibration slope, alongside visual inspection of calibration plots (Bootstrap B = 1000) and the Hosmer-Lemeshow test. To prevent overfitting and ensure the reliability of the model recalibration, a rigorous split-sample validation strategy was implemented. The external human cohort (N = 119) was randomly partitioned into a training set (50%, n = 60) and an independent testing set (50%, n = 59). The coefficients for the optimal two-gene signature were re-estimated using logistic regression exclusively within the training set. Diagnostic discrimination was then verified on the unseen testing set using ROC analysis and AUC to confirm that the high diagnostic accuracy was not an artifact of overfitting.

For severity-related analyses within AP, samples were stratified into Mild AP, Moderate AP, and Severe AP based on phenotype descriptors. The expression of the four model genes across severity categories was visualized using faceted boxplots, with overall differences assessed by using Mann-Whitney U test. ANXA3 was additionally evaluated as a single-gene severity marker using ROC analysis.

Statistical analysis

All analyses were conducted using R (v4.3.0). Continuous variables were summarized as mean ± SD or median (IQR), based on the Shapiro–Wilk normality test, and were compared using the independent Student’s t-test (after testing for homoscedasticity) or the Mann–Whitney U test as appropriate. For comparisons across more than two groups, one-way ANOVA or the Kruskal–Wallis test with Dunn’s post-hoc comparisons was used. Gene–gene associations were evaluated using Spearman’s rank correlation. For prespecified candidate genes, p values were adjusted for multiple testing using the Benjamini–Hochberg (BH) method and reported as adjusted p (adj.p), with significance set at adj.p < 0.05 (or two-sided p < 0.05). And all R scripts have been made publicly available via our GitHub repository: https://github.com/guangyuanyu90-ai/AP-diagnostic-model.

Ethics statement

All data analyzed in this study were obtained from publicly available, de-identified datasets in the GEO database. Because this study involved only the secondary analysis of anonymous public data and did not include direct human participant involvement or animal experiments, additional institutional ethical approval and informed consent were not required.

Results

Technology roadmap

Fig 1 outlines the comprehensive pipeline used to identify AP-related pyroptosis signatures and develop a diagnostic model. Three discovery datasets were merged and normalized to form the training cohort. DEGs and PRGs were integrated to identify PRDEGs, followed by GO and KEGG enrichment analyses. A four-gene diagnostic model was derived and subsequently used for downstream analyses in the discovery cohort, including risk grouping, immune infiltration by CIBERSORT, regulatory network construction, and GSEA. An independent human RNA-seq cohort was utilized for external validation. The expression levels of PRDEGs were checked, and the four-gene model was re-evaluated through ROC, Calibration and DCA.

Merging of AP datasets

Batch effects in the GSE109227, GSE65146, and GSE121038 datasets were eliminated using the “sva” R package, resulting in combined GEO datasets. A distribution boxplot was generated to analyze dataset levels, and a PCA plot was created to examine data distribution. Both the boxplot and PCA plot demonstrated effective batch effect removal in the AP dataset samples (Fig 2).

Fig 2 — (A) Boxplot before batch effect correction. (B) Boxplot after batch effect correction.(C) PCA plot before batch effect correction. (D) PCA plot after batch effect correction.GSE109227 is shown in green, GSE65146 in blue, and GSE121038 in brown.

Pyroptosis associated DEGs in AP

The combined GEO datasets, including AP and control groups, were analyzed for gene level differences. Differential analysis identified 213 DEGs, meeting the criteria of |logFC| > 1 and adj.p < 0.05. Among these, 177 genes were up-regulated (logFC > 1 and adj.p < 0.05) and 36 were down-regulated. A volcano plot visualized these findings.

To identify PRDEGs, DEGs with |logFC| > 1 and adj.p < 0.05 were intersected with PRGs. This intersection yielded 11 PRDEGs: Rela, Iqgap1, Actn4, Flna, Anxa3, Vtn, Lcn2, Mpeg1, Pah, Car9, and Pyhin1. A Venn diagram was used to illustrate the overlap. The expression levels of these PRDEGs were assessed, and heat maps are presented in Fig 3C.

Fig 3 — (A)Volcano plot of DEGs between the AP and control groups. (B) Venn diagram showing the overlap between DEGs and PRGs. (C) Heatmap of the 11 PRDEGs. Brown indicates the AP group and green indicates the control group; red indicates relatively high expression and blue indicates relatively low expression.

Verification of PRDEGs and ROC curve

Variable PRDEGs were evaluated in the combined GEO datasets, with a group comparison displaying the differential expression of PRDEGs between AP and control samples. The analysis revealed statistically significant expression levels of the 11 PRDEGs (p < 0.01), which included Rela, Iqgap1, Actn4, Flna, Anxa3, Vtn, Lcn2, Mpeg1, Pah, Car9, and Pyhin1 (Fig 4).

Fig 4 — (A) (A) Gene Expression by Group. Boxplots show the expression levels of 11 genes (Actn4, Anxa3, Car9, Flna, Iqgap1, Lcn2, Mpeg1, Pah, Pyhin1, Rela, and Vtn) in Control and AP groups. Statistical significance was assessed using the Kruskal-Wallis test, with p-values indicated by asterisks (***p < 0.001, **p < 0.01). (B-G) ROC Curves for Gene Markers. ROC curves show the diagnostic performance of individual genes in distinguishing AP from Control groups. The area AUC is reported for each gene, with 95% confidence intervals indicated where applicable.

ROC curve analysis was performed for PRDEGs, with results showing high diagnostic accuracy (AUC > 0.9) for five PRDEGs—Rela, Iqgap1, Actn4, Flna, and Anxa3. Vtn, Lcn2, Mpeg1, Pah, Car9, and Pyhin1 demonstrated moderate accuracy (AUC 0.7–0.9) in distinguishing AP from control samples.

Enrichment analysis

GO and KEGG enrichment analyses were conducted to investigate the relationships between the 11 PRDEGs and AP. As shown in Table 2, the 11 PRDEGs were primarily enriched in the following categories:

Table 2. Enrichment analysis for PRDEGs.

ONTOLOGY	ID	Description	GeneRatio	BgRatio	pvalue	p.adjust	qvalue
BP	GO:0010810	regulation of cell-substrate adhesion	4/10	230/28564	8.28 e-07	4.99 e-04	1.86 e-04
BP	GO:0031589	cell-substrate adhesion	4/10	369/28564	5.41 e-06	1.63 e-03	6.07 e-04
BP	GO:0001667	ameboidal-type cell migration	4/10	470/28564	1.40 e-05	2.12 e-03	7.89 e-04
BP	GO:0010811	positive regulation of cell-substrate adhesion	3/10	142/28564	1.41 e-05	2.12 e-03	7.89 e-04
BP	GO:0150105	protein localization to cell-cell junction	2/10	24/28564	3.03 e-05	3.66 e-03	1.36 e-03
CC	GO:0030863	cortical cytoskeleton	3/9	119/28585	5.80 e-06	3.60 e-04	1.10 e-04
CC	GO:0005938	cell cortex	3/9	300/28585	9.17 e-05	2.84 e-03	8.69 e-04
CC	GO:0030864	cortical actin cytoskeleton	2/9	84/28585	3.03 e-04	5.12 e-03	1.57 e-03
CC	GO:0032432	actin filament bundle	2/9	97/28585	4.04 e-04	5.12 e-03	1.57 e-03
CC	GO:0015629	actin cytoskeleton	3/9	500/28585	4.13 e-04	5.12 e-03	1.57 e-03
MF	GO:0051015	actin filament binding	3/9	221/28171	3.86 e-05	3.25 e-03	1.26 e-03
MF	GO:0003779	actin binding	3/9	447/28171	3.10 e-04	1.30 e-02	5.07 e-03
MF	GO:0031490	chromatin DNA binding	2/9	115/28171	5.84 e-04	1.63 e-02	6.35 e-03
MF	GO:0044325	transmembrane transporter binding	2/9	152/28171	1.02 e-03	2.13 e-02	8.27 e-03
MF	GO:0005506	iron ion binding	2/9	170/28171	1.27 e-03	2.13 e-02	8.27 e-03
KEGG	mmu04510	Focal adhesion – Mus musculus (house mouse)	3/8	202/9710	4.60 e-04	2.01 e-02	1.67 e-02
KEGG	mmu05205	Proteoglycans in cancer – Mus musculus (house mouse)	3/8	204/9710	4.73 e-04	2.01 e-02	1.67 e-02

Open in a new tab

BP: Regulation of cell-substrate adhesion; Cell-substrate adhesion; Ameboidal-type cell migration; Positive regulation of cell-substrate adhesion; Protein localization to cell-cell junction.
CC: Cortical cytoskeleton; Cell cortex; Cortical actin cytoskeleton; Actin filament bundle; Actin cytoskeleton.
MF: Actin filament binding; Actin binding; Chromatin DNA binding; Transmembrane receptor protein tyrosine kinase activity; Iron ion binding.
KEGG: Focal adhesion (Mus musculus); Proteoglycans in cancer (Mus musculus)

The results of the enrichment analyses were visualized through histograms (Fig 5) and network diagrams for BP, CC, MF, and KEGG. Larger nodes represented terms containing additional molecules, while lines depicted relevant molecular interactions. These analyses highlighted the significant involvement of genes in regulating cell-substrate adhesion and ameboid-type cell migration.

GSEA for AP

To evaluate the impact of the identified genes on AP in the Combined GEO Datasets, GSEA was conducted to examine the levels of all genes and the BP, CC, and MF involved. The results are summarized in Table 3 and visualized in Fig 6A.

Table 3. GSEA for combined datasets.

ID	Set Size	Enrichment Score	NES	p value	p.adjust	q value
REACTOME_TOLL_LIKE_RECEPTOR_CASCADES	138	6.29 e-01	2.54 e + 00	1.32 e-03	9.73 e-03	5.49 e-03
REACTOME_INTERFERON_SIGNALING	146	6.17 e-01	2.52 e + 00	1.32 e-03	9.73 e-03	5.49 e-03
KEGG_PATHOGENIC_ESCHERICHIA_COLI_INFECTION	41	7.61 e-01	2.51 e + 00	1.57 e-03	9.73 e-03	5.49 e-03
WP_PATHOGENIC_ESCHERICHIA_COLI_INFECTION	41	7.61 e-01	2.51 e + 00	1.57 e-03	9.73 e-03	5.49 e-03
REACTOME_MYD88_INDEPENDENT_TLR4_CASCADE	93	6.56 e-01	2.50 e + 00	1.40 e-03	9.73 e-03	5.49 e-03
PID_PDGFRB_PATHWAY	115	6.35 e-01	2.50 e + 00	1.37 e-03	9.73 e-03	5.49 e-03
WP_VEGFAVEGFR2_SIGNALING_PATHWAY	373	5.50 e-01	2.48 e + 00	1.16 e-03	9.73 e-03	5.49 e-03
REACTOME_RHO_GTPASES_ACTIVATE_FORMINS	104	6.37 e-01	2.46 e + 00	1.39 e-03	9.73 e-03	5.49 e-03
WP_TOLLLIKE_RECEPTOR_SIGNALING_PATHWAY	88	6.42 e-01	2.43 e + 00	1.39 e-03	9.73 e-03	5.49 e-03
REACTOME_TOLL_LIKE_RECEPTOR_9_TLR9_CASCADE	92	6.36 e-01	2.43 e + 00	1.38 e-03	9.73 e-03	5.49 e-03
REACTOME_TOLL_LIKE_RECEPTOR_TLR1_TLR2_CASCADE	102	6.30 e-01	2.43 e + 00	1.39 e-03	9.73 e-03	5.49 e-03
WP_EBOLA_VIRUS_INFECTION_IN_HOST	113	6.19 e-01	2.43 e + 00	1.38 e-03	9.73 e-03	5.49 e-03
WP_INTEGRINMEDIATED_CELL_ADHESION	87	6.43 e-01	2.42 e + 00	1.40 e-03	9.73 e-03	5.49 e-03
KEGG_TOLL_LIKE_RECEPTOR_SIGNALING_PATHWAY	87	6.41 e-01	2.41 e + 00	1.40 e-03	9.73 e-03	5.49 e-03
KEGG_LEISHMANIA_INFECTION	59	6.81 e-01	2.41 e + 00	1.47 e-03	9.73 e-03	5.49 e-03
REACTOME_SIGNALING_BY_INTERLEUKINS	373	5.35 e-01	2.41 e + 00	1.16 e-03	9.73 e-03	5.49 e-03
BIOCARTA_MAPK_PATHWAY	78	5.97 e-01	2.20 e + 00	1.43 e-03	9.73 e-03	5.49 e-03
BIOCARTA_NFKB_PATHWAY	19	7.25 e-01	2.02 e + 00	1.69 e-03	9.73 e-03	5.49 e-03
BIOCARTA_TGFB_PATHWAY	19	7.13 e-01	1.99 e + 00	1.69 e-03	9.73 e-03	5.49 e-03
WP_PI3KAKT_SIGNALING_PATHWAY	285	3.99 e-01	1.76 e + 00	1.19 e-03	9.73 e-03	5.49 e-03

Open in a new tab

Fig 6 — (A) Bubble plot summarizing the significantly enriched pathways in the combined GEO datasets. (B–E) Enrichment plots for the MAPK pathway (B), NF-κB pathway (C), TGF-β pathway (D), and PI3K-Akt signaling pathway (E). In the bubble plot, bubble size represents gene count and color represents enrichment significance.

GSEA revealed significant enrichment of genes in several key pathways and functions, which including: MAPK Pathway, NF-κB Pathway, TGF-β Pathway, PI3K-Akt Signaling Pathway, providing novel insights into the molecular mechanisms underlying AP and highlighting the BP, CC, and MF influenced by gene expression changes in AP.

Establishment of a diagnostic model for AP

To assess the diagnostic potential of the 11 PRDEGs in AP, logistic regression analysis was performed and visualized using a Forest Plot. All 11 PRDEGs were significantly associated with AP (p < 0.05), including Rela, Iqgap1, Actn4, Flna, Anxa3, Vtn, Lcn2, Mpeg1, Pah, Car9, and Pyhin1.

An SVM model was then constructed using the 11 PRDEGs to determine the number of genes that maximized accuracy and minimized error. The analysis revealed that the highest accuracy occurred with the inclusion of ten genes in the SVM model.

Subsequently, an AP diagnostic model was developed using LASSO regression analysis incorporating the 11 PRDEGs identified through SVM. Visualizations of the LASSO regression model (Fig 7D) and the LASSO variable trajectory (Fig 7E) were generated. The model identified four key PRDEGs—Vtn, Anxa3, Rela, and Iqgap1—as the most relevant genes for inclusion in the diagnostic model.

Fig 7 — (A) Forest plot of the 11 PRDEGs in the AP diagnostic model. (B, C) SVM model performance showing the number of genes associated with the lowest error rate (B) and highest accuracy (C). (D, E) LASSO regression model plot (D) and coefficient trajectory plot (E).

The RiskScore for the LASSO model was calculated based on the risk coefficients from the LASSO regression analysis using the following formula: RiskScore = Vtn × (−0.175) + Anxa3 × 0.21 + Rela × 2.754 + Iqgap1 × 0.319.

Validation and Friends analysis of the AP diagnostic model

ROC curves were plotted using the RiskScore with the R package pROC. The ROC curves (Fig 8A) indicated that the RiskScore had high diagnostic accuracy, with an AUC > 0.9.

Nomograms were generated to visualize the interrelationships of the model genes in the diagnostic model (Fig 8B). The analysis showed that Rela had the most significant contribution to the AP diagnostic model, while Anxa3 showed a comparatively lower effect.

Calibration analysis was conducted to assess the model’s precision and specificity. The calibration curve (Fig 8C) demonstrated that the observed and expected probabilities were closely aligned, indicating robust predictive power.

DCA was performed to evaluate the clinical utility of the model (Fig 8D). The DCA results showed that the model’s net benefit remained higher than the “All Positive” and “All Negative” lines, indicating excellent clinical performance.

Functional similarity (Friends) analysis scores were used to identify key genes involved in the biological processes of AP. Fig 8E highlighted Iqgap1 as a pivotal gene in AP, with a score closest to the critical value (cut-off = 0.55).

Verification and ROC curve of model genes

To examine the levels of the model genes in AP samples, a group comparison (Fig 9A) was performed, showing significant differential expression of the four model genes (Vtn, Anxa3, Rela, and Iqgap1) between the HighRisk and LowRisk groups. The expression levels of these genes were significantly different (p < 0.001), further supporting their relevance in the diagnostic model.

ROC curves (Fig 9B-9E) demonstrated that two model genes, Iqgap1 and Rela, exhibited high diagnostic accuracy (AUC > 0.9). In contrast, the expression levels of Anxa3 and Vtn showed moderate accuracy (0.7 < AUC < 0.9) in distinguishing between High Risk and Low Risk groups.

High and low risk GSEA

Differential analysis, conducted using the R package limma, identified 113 DEGs between the HighRisk and LowRisk groups. Of these, 82 genes were up-regulated (logFC > 1 and adj.p < 0.05), and 31 were down-regulated (logFC < −1 and adj.p < 0.05).

A volcano plot was created to visualize these results (Fig 10A), and a heatmap displaying the top 10 up-regulated and down-regulated DEGs based on logFC was generated (Fig 10B).

GSEA was performed to explore the involvement of all expressed genes in biological processes, cellular components, and molecular functions. The results were visualized using a bubble chart (Fig 10C) and are summarized in Table 4.

Table 4. GSEA for risk groups.

ID	Set Size	Enrichment Score	NES	p value	p.adjust	q value
REACTOME_INTERLEUKIN_4_AND_INTERLEUKIN_13_SIGNALING	92	6.69 e-01	2.44 e + 00	1.54 e-03	1.32 e-02	7.96 e-03
REACTOME_INTERLEUKIN_10_SIGNALING	34	7.92 e-01	2.41 e + 00	1.71 e-03	1.32 e-02	7.96 e-03
KEGG_PATHOGENIC_ESCHERICHIA_COLI_INFECTION	41	7.61 e-01	2.40 e + 00	1.71 e-03	1.32 e-02	7.96 e-03
WP_PATHOGENIC_ESCHERICHIA_COLI_INFECTION	41	7.61 e-01	2.40 e + 00	1.71 e-03	1.32 e-02	7.96 e-03
REACTOME_TOLL_LIKE_RECEPTOR_CASCADES	138	6.22 e-01	2.40 e + 00	1.50 e-03	1.32 e-02	7.96 e-03
REACTOME_TOLL_LIKE_RECEPTOR_TLR1_TLR2_CASCADE	102	6.50 e-01	2.39 e + 00	1.55 e-03	1.32 e-02	7.96 e-03
REACTOME_FORMATION_OF_THE_CORNIFIED_ENVELOPE	66	6.92 e-01	2.39 e + 00	1.60 e-03	1.32 e-02	7.96 e-03
WP_INTEGRINMEDIATED_CELL_ADHESION	87	6.59 e-01	2.38 e + 00	1.54 e-03	1.32 e-02	7.96 e-03
REACTOME_SIGNALING_BY_INTERLEUKINS	373	5.46 e-01	2.36 e + 00	1.31 e-03	1.32 e-02	7.96 e-03
PID_PDGFRB_PATHWAY	115	6.21 e-01	2.34 e + 00	1.52 e-03	1.32 e-02	7.96 e-03
WP_NETWORK_MAP_OF_SARSCOV2_SIGNALING_PATHWAY	173	5.84 e-01	2.33 e + 00	1.46 e-03	1.32 e-02	7.96 e-03
REACTOME_RHO_GTPASES_ACTIVATE_FORMINS	104	6.24 e-01	2.31 e + 00	1.55 e-03	1.32 e-02	7.96 e-03
WP_FIBRIN_COMPLEMENT_RECEPTOR_3_SIGNALING_PATHWAY	39	7.38 e-01	2.30 e + 00	1.70 e-03	1.32 e-02	7.96 e-03
WP_TGFBETA_SIGNALING_PATHWAY	122	5.99 e-01	2.29 e + 00	1.49 e-03	1.32 e-02	7.96 e-03
WP_SPINAL_CORD_INJURY	98	6.24 e-01	2.28 e + 00	1.56 e-03	1.32 e-02	7.96 e-03
KEGG_LEUKOCYTE_TRANSENDOTHELIAL_MIGRATION	99	6.21 e-01	2.28 e + 00	1.55 e-03	1.32 e-02	7.96 e-03
PID_IL6_7_PATHWAY	44	6.74 e-01	2.14 e + 00	1.72 e-03	1.32 e-02	7.96 e-03
BIOCARTA_TGFB_PATHWAY	19	6.96 e-01	1.87 e + 00	3.55 e-03	1.89 e-02	1.14 e-02
REACTOME_SIGNALING_BY_HIPPO	20	6.40 e-01	1.74 e + 00	1.05 e-02	3.88 e-02	2.35 e-02
PID_HEDGEHOG_GLI_PATHWAY	44	5.36 e-01	1.70 e + 00	3.44 e-03	1.89 e-02	1.14 e-02

Open in a new tab

GSEA indicated significant enrichment of genes in key pathways and functions (IL-6/JAK/STAT3 Signaling Pathway; TGF-β Pathway; Signaling by Hippo; Hedgehog/Gli Pathway), essential for understanding the molecular mechanisms of AP, and highlighted BP, CC, and MF significantly affected by gene expression changes in AP.

Construction of control network

Next, using the genetic ChIPBase database and TF data for model genes, an mRNA-TF regulatory network was constructed, involving three model genes and 20 TFs (Fig 11A). Detailed information is provided in S1 Table.

Fig 11 — (A) mRNA–TF regulatory network of the model genes.(B) mRNA–miRNA regulatory network of the model genes.

Additionally, miRNAs related to the model genes were obtained from the StarBase database, and an mRNA-miRNA regulatory network was constructed and visualized using Cytoscape software (Fig 11B). This network includes four model genes and 25 miRNAs, with specific details listed in S2 Table.

Immune infiltration analysis of AP (CIBERSORT)

The abundance of 22 immune cell types was estimated from the integrated GEO datasets. Immune infiltration analysis revealed the proportion of immune cells in the combined GEO datasets, with Fig 12A showing the immune cell composition. A group comparison chart (Fig 12B) highlighted five immune cell types with significant differences (p < 0.05): M1 macrophages, Treg cells, CD4 memory T cells, Th17 cells, and monocytes.

A correlation heatmap (Fig 12C) was generated to show immune cell infiltration in AP samples. Most immune cells exhibited strong correlations, with M1 macrophages and monocytes showing the highest positive relevance.

Finally, a bubble chart (Fig 12D) was created to illustrate the relationship between model genes and immune cell infiltration abundance. This chart revealed that Iqgap1 was strongly correlated with most immune cells, with Treg cells exhibiting the strongest negative correlation.

Immune infiltration analysis of High and Low Risk Groups (CIBERSORT)

The immune infiltration abundance of 22 immune cell types was calculated using the CIBERSORT algorithm, applied to AP samples from the combined GEO datasets. Based on the immune infiltration analysis, a group comparison chart was generated to highlight differences in immune cell infiltration between various groups. The analysis revealed significant differences in the abundance of eight immune cell types: CD8 memory T cells, M1 macrophages, M2 macrophages, Treg cells, Th1 cells, monocytes, gamma delta T cells, and activated NK cells (Fig 13A).

Correlation heatmaps were generated to further investigate immune infiltration in AP samples. Fig 13B-13C revealed that in LowRisk samples, most immune cells exhibited strong correlations, with Treg cells and monocytes showing the strongest negative correlation. In HighRisk samples, strong correlations were also observed, with M1 macrophages and gamma delta T cells demonstrating the most significant negative correlations.

Bubble charts were created to illustrate the relationship between model genes and immune cell infiltration abundance. The bubble charts indicated that in LowRisk samples, most immune cells were strongly correlated with Iqgap1, with Treg cells showing the strongest negative correlation. In HighRisk samples, the majority of immune cells were strongly correlated with ANXA3, with monocytes displaying the strongest positive correlation.

External validation of the diagnostic model in the human RNA-seq cohort

To evaluate the potential of the four – gene model for the diagnosis of AP, we performed external validation using the independent human RNA-seq dataset GSE194331. An initial analysis of the 11 PRDEGs revealed that, despite the inherent interspecies transcriptional variability, most of the genes remained significantly dysregulated in human blood samples from AP patients (S1 Fig).

To assess the cross-species applicability of the discovery-phase signature, we first applied the four-gene model (ANXA3, IQGAP1, RELA, VTN) to an independent human blood cohort (GSE194331), using fixed coefficients derived from the murine dataset. However, this direct application resulted in poor discrimination, with an AUC of approximately 0.5. This finding underscores the limitations of direct model transferability and indicates considerable biological differences in gene expression profiles between species.

To address this limitation, we conducted an exploratory optimization. Differential expression analysis of the human cohort revealed that while ANXA3 and IQGAP1 consistently provided diagnostic signals (P < 0.001; S1 Fig), RELA and VTN showed inconsistent trends and lacked statistical significance. Model comparison using the AIC further demonstrated that the two-gene model (ANXA3 + IQGAP1) offered a superior fit with lower complexity (AIC: 61.12) compared to the four-gene model (AIC: 63.01). Consequently, we recalibrated the diagnostic model by re-estimating the coefficients for ANXA3 and IQGAP1 using logistic regression in the human dataset.

Following recalibration, the optimized two-gene model exhibited a substantial restoration in performance, with an AUC of 0.957 (95% CI: 0.926–0.989) (Fig 14A; S5 Table). This notable improvement highlights the importance of incorporating species-specific expression weights to enhance diagnostic accuracy. Calibration analysis further confirmed the reliability of the recalibrated model. The calibration curve demonstrated strong alignment between predicted probabilities and observed outcomes (Fig 14C, 14D), as evidenced by a low Brier score of 0.0774 and a perfect calibration slope of 1.00 (S4 Table). Furthermore, DCA validated that the two-gene model delivers substantial net clinical benefit across a wide range of threshold probabilities, surpassing the performance of default strategies (Fig 14B).

Fig 14 — (A) ROC curve showing the diagnostic performance of the recalibrated two-gene signature (ANXA3 + IQGAP1) in the human peripheral blood cohort. The model achieved an AUC of 0.957 (95% CI: 0.926–0.989). (B) DCA showing the clinical net benefit of the two-gene model across different threshold probabilities, compared with the treat-all and treat-none strategies. (C) Bootstrap calibration curve based on 1,000 resamples, demonstrating the agreement between predicted and observed probabilities. (D) Calibration plot showing the relationship between predicted and observed probabilities, with the Hosmer–Lemeshow test result indicated.

In a secondary exploratory analysis of disease severity (Mild vs. Severe AP), the four-gene model demonstrated moderate discriminative power, with an AUC of 0.779 (S2B Fig). Notably, ANXA3 alone exhibited a stronger ability to differentiate between disease severity levels, as shown by the boxplots for different severity categories (S2A Fig). It also displayed comparable performance (AUC = 0.721) (S2C Fig), suggesting that while the primary utility of the signature lies in diagnosis, it may also provide modest insight into disease progression.

Discussion

AP is an acute inflammatory condition with outcomes ranging from self-limited illness to life-threatening organ failure, contributing to significant morbidity and healthcare burdens, particularly in severe cases [28,29]. Clarifying the molecular mechanisms and developing reproducible diagnostic markers remain critical clinical priorities.

Principal findings: Three murine transcriptomic datasets were integrated to map the pyroptosis-related transcriptional landscape, identifying 11 PRDEGs associated with AP. Functional enrichment analysis revealed that these genes are involved in cytoskeletal organization and key inflammatory pathways, including MAPK and NF-κB. Immune infiltration analysis further highlighted a distinct immune microenvironment in AP, characterized by an increased presence of M1 macrophages and monocytes, while Iqgap1 was strongly negatively correlated with Treg cell abundance. Based on these findings, a diagnostic model comprising four candidate genes (ANXA3, IQGAP1, RELA, and VTN) was initially constructed in the murine discovery cohort. Risk stratification based on this model effectively classified patients into high- and low-risk groups. GSEA revealed significant enrichment of IL-6/JAK/STAT3 and TGF-β signaling pathways in high-risk samples, indicating a more severe inflammatory state. However, considering the complexity of cross-species translation, we performed rigorous model optimization in the human blood cohort. Although RELA and VTN were significant in murine pancreas tissue, they exhibited inconsistent expression in human blood. Consequently, using the AIC, the panel was refined into a robust two-gene diagnostic model (ANXA3 and IQGAP1). In the independent human cohort, this recalibrated signature maintained superior accuracy with an AUC of 0.957. Its robustness was further validated by excellent calibration (Brier score: 0.0774) and DCA, while secondary analyses highlighted the potential utility of ANXA3 alone in stratifying disease severity. These findings establish the two-gene signature as a precise, clinically interpretable, and highly translatable tool for AP diagnosis. Although the transition from murine pancreatic tissue to human peripheral blood represents a shift in biological domains, our results suggest that circulating leukocyte transcriptomes reliably reflect the local inflammatory state [30]. This indicates that the immune signature initiated in the pancreas is effectively captured in the systemic circulation, providing a non-invasive method to monitor organ-specific injury.

Biological plausibility and relation to prior work: Pyroptosis has been implicated in various inflammatory conditions [31]. The prominence of RELA, IQGAP1, ACTN4, FLNA, and ANXA3 aligns with their known roles in cytokine induction, cytoskeletal dynamics, and injury responses [32–36]. Specifically, ANXA3 is recognized as a distinct marker of neutrophil granules. Its substantial upregulation in peripheral blood suggests that the model effectively captures the extensive mobilization and activation of neutrophils, a hallmark of the early systemic inflammatory response in AP [37]. Meanwhile, VTN is involved in adhesion and complement regulation [38]. Other PRDEGs are linked to complementary processes: LCN2 and MPEG1 to innate immunity, PAH to amino acid metabolism, CA9 to acid–base homeostasis, and PYHIN1 to inflammatory signaling [39–43]. Consistent with these gene functions, enrichment analysis emphasized cell–substrate adhesion and amoeboid-type migration, processes critical to leukocyte trafficking and tissue remodeling in AP [44–46]. The immune deconvolution findings corroborate previous observations of macrophages, neutrophils, and T-cell subsets in AP [47–51], providing mechanistic insight into the PRDEG panel [52].

Clinical Implications: This study provides a molecular framework for incorporating pyroptosis into the diagnostic and mechanistic evaluation of AP. By identifying specific pyroptosis-related signatures, it offers an alternative approach for early diagnosis, potentially serving as a complementary tool when conventional biochemical markers are inconclusive. The finding that ANXA3 alone demonstrates significant discriminative power in differentiating severe AP suggests its dual potential as both a diagnostic marker and an early indicator of disease severity. This could aid clinicians in rapid risk stratification and triage. Furthermore, the correlation between IQGAP1 and immune tolerance, specifically Treg abundance, provides valuable mechanistic insights into the immune-inflammatory state of patients, offering a foundation for future investigations into immunomodulatory therapies targeting pyroptosis pathways.

Strengths and Limitations: This study has several notable strengths. First, multiple independent transcriptomic datasets were integrated, and batch effects were systematically controlled, enhancing the robustness of the findings. Second, by focusing on pyroptosis-related pathways, a clear and interpretable gene diagnostic model was developed. Third, comprehensive external validation was performed in an independent human cohort, where the model’s discriminative performance, calibration accuracy, and clinical net benefit were assessed, with full transparency regarding model coefficients and per-sample predictions. Several limitations should be considered. The most significant is the biological domain shift from murine pancreatic tissue in the discovery phase to human peripheral blood in the validation phase. This transition means that, while the murine model focused on localized pancreatic injury, the human validation phase emphasized systemic inflammatory signals. Our analysis identified tissue-specific markers, such as RELA and VTN, that did not translate well to blood-based diagnostics, requiring recalibration of the model. Although this optimization reduced species-specific expression differences, biological biases due to sample-type variation could not be fully eliminated. Moreover, while the two-gene model demonstrated excellent diagnostic accuracy, its ability to stratify disease severity was primarily driven by ANXA3, suggesting that different components of the signature may have distinct clinical implications. Furthermore, as a retrospective study with a moderate sample size, residual confounding factors—such as clinical comorbidities or unmeasured covariates—may have influenced the results. While immune cell deconvolution provided valuable insights, it relied on algorithmic assumptions and lacked direct flow cytometric validation. Finally, the absence of prospective multi-center validation and targeted functional experiments (such as in vitro gene knockdown) represents a significant gap. Future research should focus on investigating the mechanistic roles of ANXA3 and IQGAP1 in AP-driven pyroptosis and validating the signature in larger, prospective clinical cohorts.

This study defines a reproducible pyroptosis-related transcriptional signature for AP and establishes a streamlined two-gene diagnostic model (ANXA3 and IQGAP1) with demonstrated cross-species translatability. Initially derived from murine pancreatic tissue, the recalibrated version of this signature maintained exceptional diagnostic performance in human peripheral blood, highlighting its potential as a reliable biomarker. Despite its high accuracy, the transition from localized tissue signals to systemic circulation requires careful biological interpretation. Future research will focus on: (i) biological validation through protein-level confirmation and assessment of tissue–blood expression concordance; (ii) specificity testing against other acute inflammatory conditions (such as sepsis or bacteremia) to verify the pancreas-specific diagnostic relevance; (iii) prospective multi-center clinical evaluation to further validate its utility in early severity stratification, particularly the independent predictive value of ANXA3, and its incremental benefit beyond conventional markers; and (iv) the development of standardized clinical assays, such as RT-qPCR, to facilitate robust and reproducible translation into emergency clinical settings.

Supporting information

S1 Fig. Boxplot of 11 PRDEGs expression across groups.

Boxplots illustrating the relative expression levels of 11 pyroptosis-related differentially expressed genes (PRDEGs)—RELA, IQGAP1, ACTN4, FLNA, ANXA3, VTN, LCN2, MPEG1, PAH, CA9 (homologous to murine Car9), and PYHIN1—were generated within the independent human peripheral blood dataset. This analysis assesses the cross-species translatability of the initial murine-derived signature by comparing the differential expression patterns of these candidate genes between human patients with AP and healthy controls.

(TIF)

pone.0347204.s001.tif^{(4MB, tif)}

S2 Fig. Association of signature genes with disease severity and their predictive performance in the human blood cohort.

The figure depicts the relationship between gene expression and the severity of acute pancreatitis (Mild, Moderate, and Severe) using a human peripheral blood dataset. (A) Boxplots show the relative expression levels of the signature genes across the three severity groups, emphasizing significant differential expression as the disease progresses. (B-C) ROC curves assess the discriminatory ability (AUC) of the multi-gene panel compared to ANXA3 alone in predicting disease severity.

(TIFF)

pone.0347204.s002.tiff^{(7MB, tiff)}

S3 Fig. Diagnostic performance comparison between the initial four-gene and optimized two-gene models.

The figure illustrates the diagnostic performance, calibration, and clinical utility of the initial four-gene signature (ANXA3, IQGAP1, RELA, and VTN) compared to the AIC-optimized two-gene signature (ANXA3 and IQGAP1) in the human validation cohort. Performance is evaluated through ROC curves to assess discriminatory accuracy (AUC) (A), calibration plots (B) to evaluate the goodness-of-fit between predicted and observed outcomes, and (DCA) (C) to determine clinical net benefit across various threshold probabilities.

(TIF)

pone.0347204.s003.tif^{(4.7MB, tif)}

S4 Fig. Split-sample validation and diagnostic performance of the recalibrated two-gene model.

The ROC curve was used to assess the discriminatory accuracy of the optimized two-gene signature (ANXA3 and IQGAP1) in an independent human testing set. To mitigate the risk of overfitting, the external human cohort (GSE194331, N = 119) was randomly divided into a training set (50%, n = 60) for coefficient re-estimation and an independent testing set (50%, n = 59). The curve illustrates the model’s performance when applied exclusively to the testing set, demonstrating that the high diagnostic accuracy is robust and not a result of overfitting.

(TIFF)

pone.0347204.s004.tiff^{(7.5MB, tiff)}

S1 Table. Predicted TF regulatory network for the four model genes.

This table lists the potential TF associated with the identified four model genes.

(CSV)

pone.0347204.s005.csv^{(352B, csv)}

S2 Table. Predicted miRNA–mRNA regulatory network of the four model genes.

This table details the miRNAs targeted to the four model genes.

(CSV)

pone.0347204.s006.csv^{(592B, csv)}

S3 Table. Detailed logistic regression parameters and odds ratios for the optimized two-gene diagnostic model.

This table presents the statistical summary of the refined diagnostic model in the human validation cohort. It includes the estimated coefficients, standard errors, z-values, and p-values. Odds ratios (OR) with their corresponding 95% CI are provided to quantify the association between the expression levels of individual genes and the likelihood of AP.

(CSV)

pone.0347204.s007.csv^{(244B, csv)}

S4 Table. Recalibration metrics and Brier score for the two-gene model.

To assess the concordance between predicted risks and observed outcomes in the human cohort, this table provides key calibration metrics, including the calibration intercept (calibration-in-the-large), calibration slope, and the Brier score (a measure of overall prediction accuracy). These metrics indicate the goodness-of-fit and the reliability of the model’s probability estimates.

(CSV)

pone.0347204.s008.csv^{(90B, csv)}

S5 Table. AUC for the two-gene signature.

The table summarizes the diagnostic discrimination performance of the combined ANXA3 and IQGAP1 signature. The AUC value is provided along with its 95% CI to illustrate the model’s ability to accurately distinguish AP patients from healthy controls.

(CSV)

pone.0347204.s009.csv^{(71B, csv)}

S6 Table. Performance metrics from 5-fold repeated cross-validation (5x5 CV).

This table summarizes the stability and robustness of the two-gene model through repeated cross-validation (5 folds, 5 repeats). It includes the mean values and standard deviations (SD) for the area under the ROC curve (ROC), sensitivity (Sens), and specificity (Spec), demonstrating that the model’s performance is consistent and not biased by a specific data split.

(CSV)

pone.0347204.s010.csv^{(166B, csv)}

S7 Table. Internal validation and optimism correction of the initial four-gene diagnostic model.

This table details the rigorous internal validation metrics for the initial four-gene signature (Anxa3, Iqgap1, Rela, and Vtn) based on 500 bootstrap iterations. It compares the “Apparent” performance (measured on the original dataset) with the “Bootstrap-corrected” performance to estimate “Optimism.” Metrics include the AUC, Brier score, calibration intercept, and calibration slope. The Optimism-corrected values provide a more realistic and conservative estimate of the model’s predictive performance by accounting for potential over-fitting during the initial model construction phase.

(CSV)

pone.0347204.s011.csv^{(500B, csv)}

S8 Table. Bootstrap-based selection frequencies of the 11 PRDEGs in the diagnostic model.

To evaluate the stability of the feature selection process, this table presents the frequency with which each of the 11 initial PRDEGs was selected across 500 bootstrap iterations. Genes with higher selection frequencies (Such as Anxa3 at 0.91) represent the most robust and reliable predictors for the diagnostic signature.

(CSV)

pone.0347204.s012.csv^{(188B, csv)}

Data Availability

All relevant data are publicly available without restriction. Raw microarray and RNA-seq datasets were obtained from the NCBI Gene Expression Omnibus (GEO) repository under accession numbers GSE109227, GSE65146, GSE121038 and GSE194331. Other supporting data and analysis details are provided in the Supporting Information files accompanying the paper.

Funding Statement

This work was supported by the National Natural Science Foundation of China (https://www.nsfc.gov.cn, Grant No. 82200011 to YGY). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

1.Banks PA, Freeman ML, Practice parameters committee of the American college of gastroenterology. Practice guidelines in acute pancreatitis. Am J Gastroenterol. 2006;101:2379–400. doi: 10.1111/j.1572-0241.2006.00856.x [DOI] [PubMed] [Google Scholar]
2.Forsmark CE, Baillie J, AGA Institute Clinical Practice and Economics Committee, AGA Institute Governing Board. AGA institute technical review on acute pancreatitis. Gastroenterology. 2007;132:2022–44. doi: 10.1053/j.gastro.2007.03.065 [DOI] [PubMed] [Google Scholar]
3.Tenner S, Baillie J, DeWitt J, Vege SS, American College of Gastroenterology. American college of gastroenterology guideline: management of acute pancreatitis. Am J Gastroenterol. 2013;108(9):1400–15; 1416. doi: 10.1038/ajg.2013.218 [DOI] [PubMed] [Google Scholar]
4.Jorgensen I, Miao EA. Pyroptotic cell death defends against intracellular pathogens. Immunol Rev. 2015;265(1):130–42. doi: 10.1111/imr.12287 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.McKenzie BA, Dixit VM, Power C. Fiery cell death: pyroptosis in the central nervous system. Trends Neurosci. 2020;43(1):55–73. doi: 10.1016/j.tins.2019.11.005 [DOI] [PubMed] [Google Scholar]
6.Zhaolin Z, Guohua L, Shiyuan W, Zuo W. Role of pyroptosis in cardiovascular disease. Cell Prolif. 2019;52(2):e12563. doi: 10.1111/cpr.12563 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Shi J, Gao W, Shao F. Pyroptosis: gasdermin-mediated programmed necrotic cell death. Trends Biochem Sci. 2017;42(4):245–54. doi: 10.1016/j.tibs.2016.10.004 [DOI] [PubMed] [Google Scholar]
8.Davis S, Meltzer PS. GEOquery: a bridge between the Gene Expression Omnibus (GEO) and BioConductor. Bioinformatics. 2007;23(14):1846–7. doi: 10.1093/bioinformatics/btm254 [DOI] [PubMed] [Google Scholar]
9.Norberg KJ, Nania S, Li X, Gao H, Szatmary P, Segersvärd R, et al. RCAN1 is a marker of oxidative stress, induced in acute pancreatitis. Pancreatology. 2018;18(7):734–41. doi: 10.1016/j.pan.2018.08.005 [DOI] [PubMed] [Google Scholar]
10.Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishilevich S, et al. The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics. 2016;54:1.30.1-1.30.33. doi: 10.1002/cpbi.5 [DOI] [PubMed] [Google Scholar]
11.Wu J, Zhu Y, Luo M, Li L. Comprehensive analysis of pyroptosis-related genes and tumor microenvironment infiltration characterization in breast cancer. Front Immunol. 2021;12:748221. doi: 10.3389/fimmu.2021.748221 [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics. 2012;28(6):882–3. doi: 10.1093/bioinformatics/bts034 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. doi: 10.1093/nar/gkv007 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Gene Ontology Consortium, Aleksander SA, Balhoff J, Carbon S, Cherry JM, Drabkin HJ, et al. The Gene Ontology knowledgebase in 2023. Genetics. 2023;224:iyad031. doi: 10.1093/genetics/iyad031 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28(1):27–30. doi: 10.1093/nar/28.1.27 [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Yu G, Wang L-G, Han Y, He Q-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16(5):284–7. doi: 10.1089/omi.2011.0118 [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Sanz H, Valim C, Vegas E, Oller JM, Reverter F. SVM-RFE: selection and visualization of the most relevant features through non-linear kernels. BMC Bioinformatics. 2018;19(1):432. doi: 10.1186/s12859-018-2451-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Engebretsen S, Bohlin J. Statistical predictions with glmnet. Clin Epigenetics. 2019;11(1):123. doi: 10.1186/s13148-019-0730-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez J-C, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinform. 2011;12:77. doi: 10.1186/1471-2105-12-77 [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Wu J, Zhang H, Li L, Hu M, Chen L, Xu B, et al. A nomogram for predicting overall survival in patients with low-grade endometrial stromal sarcoma: a population-based analysis. Cancer Commun (Lond). 2020;40(7):301–12. doi: 10.1002/cac2.12067 [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Van Calster B, Wynants L, Verbeek JFM, Verbakel JY, Christodoulou E, Vickers AJ, et al. Reporting and interpreting decision curve analysis: a guide for investigators. Eur Urol. 2018;74(6):796–804. doi: 10.1016/j.eururo.2018.08.038 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Yu G, Li F, Qin Y, Bo X, Wu Y, Wang S. GOSemSim: an R package for measuring semantic similarity among GO terms and gene products. Bioinformatics. 2010;26(7):976–8. doi: 10.1093/bioinformatics/btq064 [DOI] [PubMed] [Google Scholar]
23.Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545–50. doi: 10.1073/pnas.0506580102 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Liberzon A, Subramanian A, Pinchback R, Thorvaldsdóttir H, Tamayo P, Mesirov JP. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011;27(12):1739–40. doi: 10.1093/bioinformatics/btr260 [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Zhou K-R, Liu S, Sun W-J, Zheng L-L, Zhou H, Yang J-H, et al. ChIPBase v2.0: decoding transcriptional regulatory networks of non-coding RNAs and protein-coding genes from ChIP-seq data. Nucleic Acids Res. 2017;45(D1):D43–50. doi: 10.1093/nar/gkw965 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Li J-H, Liu S, Zhou H, Qu L-H, Yang J-H. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res. 2014;42(Database issue):D92-7. doi: 10.1093/nar/gkt1248 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453–7. doi: 10.1038/nmeth.3337 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Yadav D, Lowenfels AB. The epidemiology of pancreatitis and pancreatic cancer. Gastroenterology. 2013;144(6):1252–61. doi: 10.1053/j.gastro.2013.01.068 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Banks PA, Bollen TL, Dervenis C, Gooszen HG, Johnson CD, Sarr MG, et al. Classification of acute pancreatitis--2012: revision of the Atlanta classification and definitions by international consensus. Gut. 2013;62(1):102–11. doi: 10.1136/gutjnl-2012-302779 [DOI] [PubMed] [Google Scholar]
30.Zhao Y, Hosking C, Cunningham D, Langhorne J, Lin J-W. Transcriptome analysis of blood and spleen in virulent and avirulent mouse malaria infection. Sci Data. 2020;7(1):253. doi: 10.1038/s41597-020-00592-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
31.Bergsbaken T, Fink SL, Cookson BT. Pyroptosis: host cell death and inflammation. Nat Rev Microbiol. 2009;7(2):99–109. doi: 10.1038/nrmicro2070 [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Gerke V, Creutz CE, Moss SE. Annexins: linking Ca2+ signalling to membrane dynamics. Nat Rev Mol Cell Biol. 2005;6(6):449–61. doi: 10.1038/nrm1661 [DOI] [PubMed] [Google Scholar]
33.Jakkampudi A, Jangala R, Reddy BR, Mitnala S, Nageshwar Reddy D, Talukdar R. NF-κB in acute pancreatitis: mechanisms and therapeutic potential. Pancreatology. 2016;16(4):477–88. doi: 10.1016/j.pan.2016.05.001 [DOI] [PubMed] [Google Scholar]
34.Wei T, Lambert PF. Role of IQGAP1 in carcinogenesis. Cancers (Basel). 2021;13(16):3940. doi: 10.3390/cancers13163940 [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Lappalainen P, Kotila T, Jégou A, Romet-Lemonne G. Biochemical and mechanical regulation of actin dynamics. Nat Rev Mol Cell Biol. 2022;23(12):836–52. doi: 10.1038/s41580-022-00508-4 [DOI] [PubMed] [Google Scholar]
36.Zhou J, Kang X, An H, Lv Y, Liu X. The function and pathogenic mechanism of filamin A. Gene. 2021;784:145575. doi: 10.1016/j.gene.2021.145575 [DOI] [PubMed] [Google Scholar]
37.Li M, Liu D, Jing F, Liu R, Yi Q. The role of Annexin A3 in coronary arterial lesions in children with Kawasaki disease. Front Pediatr. 2023;11:1111788. doi: 10.3389/fped.2023.1111788 [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Preissner KT. Structure and biological role of vitronectin. Annu Rev Cell Biol. 1991;7:275–310. doi: 10.1146/annurev.cb.07.110191.001423 [DOI] [PubMed] [Google Scholar]
39.Bertin J, DiStefano PS. The PYRIN domain: a novel motif found in apoptosis and inflammation proteins. Cell Death Differ. 2000;7(12):1273–4. doi: 10.1038/sj.cdd.4400774 [DOI] [PubMed] [Google Scholar]
40.Flo TH, Smith KD, Sato S, Rodriguez DJ, Holmes MA, Strong RK, et al. Lipocalin 2 mediates an innate immune response to bacterial infection by sequestrating iron. Nature. 2004;432(7019):917–21. doi: 10.1038/nature03104 [DOI] [PubMed] [Google Scholar]
41.Spilsbury K, O’Mara MA, Wu WM, Rowe PB, Symonds G, Takayama Y. Isolation of a novel macrophage-specific gene by differential cDNA analysis. Blood. 1995;85(6):1620–9. doi: 10.1182/blood.v85.6.1620.bloodjournal8561620 [DOI] [PubMed] [Google Scholar]
42.Salam MT, Lin P-C, Avol EL, Gauderman WJ, Gilliland FD. Microsomal epoxide hydrolase, glutathione S-transferase P1, traffic and childhood asthma. Thorax. 2007;62(12):1050–7. doi: 10.1136/thx.2007.080127 [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Kalinin S, Malkova A, Sharonova T, Sharoyko V, Bunev A, Supuran CT, et al. Carbonic anhydrase IX inhibitors as candidates for combination therapy of solid tumors. Int J Mol Sci. 2021;22(24):13405. doi: 10.3390/ijms222413405 [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Wang X, Yang Y, Wang Y, Lu C, Hu X, Kawazoe N, et al. Focal adhesion and actin orientation regulated by cellular geometry determine stem cell differentiation via mechanotransduction. Acta Biomater. 2024;182:81–92. doi: 10.1016/j.actbio.2024.05.017 [DOI] [PubMed] [Google Scholar]
45.Harjunpää H, Llort Asens M, Guenther C, Fagerholm SC. Cell adhesion molecules and their roles and regulation in the immune and tumor microenvironment. Front Immunol. 2019;10:1078. doi: 10.3389/fimmu.2019.01078 [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Barros-Becker F, Lam P-Y, Fisher R, Huttenlocher A. Live imaging reveals distinct modes of neutrophil and macrophage migration within interstitial tissues. J Cell Sci. 2017;130(22):3801–8. doi: 10.1242/jcs.206128 [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Hu F, Lou N, Jiao J, Guo F, Xiang H, Shang D. Macrophages in pancreatitis: mechanisms and therapeutic potential. Biomed Pharmacother. 2020;131:110693. doi: 10.1016/j.biopha.2020.110693 [DOI] [PubMed] [Google Scholar]
48.Hajishengallis G, Chavakis T. Mechanisms and therapeutic modulation of neutrophil-mediated inflammation. J Dent Res. 2022;101(13):1563–71. doi: 10.1177/00220345221107602 [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Habtezion A. Inflammation in acute and chronic pancreatitis. Curr Opin Gastroenterol. 2015;31(5):395–9. doi: 10.1097/MOG.0000000000000195 [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Baer JM, Zuo C, Kang L-I, de la Lastra AA, Borcherding NC, Knolhoff BL, et al. Fibrosis induced by resident macrophages has divergent roles in pancreas inflammatory injury and PDAC. Nat Immunol. 2023;24(9):1443–57. doi: 10.1038/s41590-023-01579-x [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Sturgeon R, Goel P, Singh RK. Tumor-associated neutrophils in pancreatic cancer progression and metastasis. Am J Cancer Res. 2023;13(12):6176–89. [PMC free article] [PubMed] [Google Scholar]
52.Parmar T, Parmar VM, Perusek L, Georges A, Takahashi M, Crabb JW, et al. Lipocalin 2 plays an important role in regulating inflammation in retinal degeneration. J Immunol. 2018;200(9):3128–41. doi: 10.4049/jimmunol.1701573 [DOI] [PMC free article] [PubMed] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0347204.r001

Decision Letter 0

Alexis Carrasco

7 Jan 2026

Dear Dr. yu,

Please revise carefully and reply to reviewers' comments. Additionally, I suggest proposing a more rigorous validation before clinical translation. It is necessary to compare and properly discuss the models used (i.e., mice to humans), as well as their limitations. How feasible is it to translate these interpretations? What is the potential participation of involved genes?

Please submit your revised manuscript by Feb 20 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org . When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols ..

We look forward to receiving your revised manuscript.

Kind regards,

Alexis G. Murillo Carrasco

Academic Editor

PLOS One

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. Thank you for stating the following in the Competing Interests section:

The authors have no conflicts to disclose.

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests). If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

4. Please include a copy of Table 4 which you refer to in your text on page 26.

5. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

6. We are unable to open your Supporting Information file [S1_Data_ExternalValidation_Predictions.tsv]. Please kindly revise as necessary and re-upload.

7. If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Partly

Reviewer #4: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously? -->?>

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

Reviewer #4: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available??>

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English??>

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: Yes

Reviewer #4: Yes

**********

Reviewer #1: Even though it is a rectrospective and cross species study, still it holds promise for finding newer avenues for better understanding for what is happening at cellular level in this disease subgroup. Also it might pave way for finding a treatment which can alter the natural course of disease.

Reviewer #2: The statistical analysis is appropriate in method and implementation, and the findings are data-supported, but the degree of rigor is moderate rather than high. The results are credible for exploratory bioinformatics, yet insufficiently validated for clinical translation at this stage.

Recommendations for improvement before publication:

Include internal k-fold or bootstrap cross-validation in the discovery cohort.

Assess gene selection stability via resampling.

Report calibration slope/intercept and Brier score.

Consider inclusion of covariates or stratified analyses to address confounding.

The manuscript is scientifically intelligible but linguistically below publication quality in its current form. The English is serviceable for review but would not meet PLOS ONE’s final publication standard without professional copyediting.

Recommendation:

Comprehensive language revision by a native English-speaking scientific editor.

Eliminate translation artifacts and redundant phrasing.

Standardize capitalization, terminology, and figure caption style.

Reviewer #3: The manuscript "Pyroptosis-related transcriptional signature and diagnostic modeling in acute pancreatitis with cross-species external validation" presents an interesting approach to identifying diagnostic markers for Acute Pancreatitis (AP). However, the study design involves a significant biological domain shift—training on mouse pancreatic tissue and validating on human peripheral blood. While the statistical recalibration is impressed, the biological plausibility is not sufficiently addressed.

- Could the authors provide evidence (from literature or additional analysis) that the expression levels of the four model genes (ANXA3, IQGAP1, RELA, VTN) in the pancreas are concordant with their expression in peripheral blood during AP?

- Since the "recalibration" method mathematically adjusts for prevalence and distribution shifts, how can we be certain that the model's performance in the human cohort is not merely reflecting non-specific systemic inflammation (e.g., sepsis or general infection) rather than pancreas-specific pathology?

- What is the specific clinical value added by this 4-gene signature? Is it intended for cases with indeterminate enzyme levels, or for predicting severity? Can the authors compare the AUC of their model (0.932) against standard biochemical markers (like Lipase or CRP) within the validation cohort (if such metadata is available) or discuss this comparison critically? Without this, the practical "net benefit" shown in the DCA is theoretical rather than practical.

- Please clarify if the "recalibration" parameters were derived from a subset of the external cohort (e.g., a 50% split) and tested on the remainder, or if they were fitted on the whole external dataset. Fitting on the whole dataset and then reporting AUC on the same dataset can artificially inflate performance (overfitting to the validation set).

Reviewer #4: This work is very interesting and has potential significance in the field of precision medicine for acute pancreatitis. The analysis of target gene expression in acute pancreatitis may help predict disease occurrence and severity in the future, particularly in patients with recurrent acute pancreatitis, and may contribute to the prevention of recurrent episodes and progression to chronic pancreatitis.

The study provides valuable information, including comprehensive gene datasets and a variety of statistical analyses. I would like to suggest that the difference in sample size between the acute pancreatitis group and the control group may affect the robustness of the results. If feasible, using comparable population sizes in both groups could improve the consistency and reliability of the findings.

Additionally, could the authors please clarify whether the acute pancreatitis data are derived from a pediatric population or the general population?

It would also be helpful to know whether patients were stratified according to disease severity (mild versus severe acute pancreatitis) and whether there is an association between PRDEG expression levels and disease severity.

Furthermore, regarding the association between acute pancreatitis and PRDEGs/DEGs, could the authors elaborate on whether these findings have potential implications for clinical management or therapeutic decision-making in current clinical practice?

Finally, it would be important to consider whether the observed gene expression changes are specific to acute pancreatitis or if similar expression patterns are seen in other acute inflammatory conditions, such as sepsis, bacteremia, or systemic lupus erythematosus (SLE). Validation in these settings may help determine the specificity and clinical applicability of the findings.

**********

what does this mean? ). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy .-->

Reviewer #1: No

Reviewer #2: No

Reviewer #3: No

Reviewer #4: Yes: Tanawat PattarapuntakulTanawat Pattarapuntakul

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures

You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation.

NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications.

PLoS One. 2026 Apr 17;21(4):e0347204. doi: 10.1371/journal.pone.0347204.r002

Author response to Decision Letter 1

19 Feb 2026

1、Response to Reviewer #1

Comment: Even though it is a rectrospective and cross species study, still it holds promise for finding newer avenues for better understanding for what is happening at cellular level in this disease subgroup. Also it might pave way for finding a treatment which can alter the natural course of disease.

Response:

We sincerely thank the reviewer for this encouraging and positive evaluation of our work. We appreciate your recognition of the translational potential and clinical relevance of our study, despite the inherent limitations of its retrospective, cross-species design.

We fully agree with your perspective. A central aim of this study was to bridge transcriptomic signatures and cellular-level pathophysiology in acute pancreatitis (AP). By delineating the pyroptosis-related transcriptional landscape—particularly the putative involvement of ANXA3 in neutrophil mobilization and IQGAP1 in cytoskeletal dynamics and immune regulation (including potential effects on Treg cells)—we sought to provide a clearer view of the immune microenvironment during AP progression.

As you noted, these cellular-level insights extend beyond diagnostic utility. They also offer a mechanistic rationale for future studies of targeted immunomodulatory strategies, for example by interrogating the pyroptosis-cytoskeleton axis. We share your view that precision medicine approaches may ultimately help modify the trajectory of severe AP and reduce recurrence risk.

We thank you again for your constructive input and for affirming the clinical and scientific direction of our work.

2、Response to Reviewer #2

General Comment: The statistical analysis is appropriate in method and implementation, and the findings are data-supported, but the degree of rigor is moderate rather than high. The results are credible for exploratory bioinformatics, yet insufficiently validated for clinical translation at this stage.

Response:

We greatly appreciate the reviewer’s positive feedback on the methodology and data support of our study. We agree that rigorous validation is crucial for demonstrating translational potential. In this revision, we have significantly enhanced the statistical rigor by implementing internal resampling validation methods, including Bootstrap and Repeated Cross-Validation, to assess gene selection stability and conduct detailed calibration analyses. Additionally, we have employed a professional English editing service to ensure that the manuscript adheres to the highest publication standards.

Comment 1: Include internal k-fold or bootstrap cross-validation in the discovery cohort.

Response:

Thank you for your valuable suggestion. To enhance the robustness of our discovery phase findings, we have conducted the following additional internal validations:

Repeated 10-Fold Cross-Validation: Using the caret package, we performed 10-fold cross-validation, repeated 10 times. The model exhibited excellent stability, with a mean AUC of 0.983 across all folds.

Bootstrap Validation: We further validated the model through 500 bootstrap resamples. The results demonstrated high predictive accuracy, yielding an optimism-corrected AUC of 0.958.

These rigorous validation results have been detailed in S7 Table of the supplementary materials.

Comment 2: Assess gene selection stability via resampling.

Response:

We conducted the requested stability analysis by performing 500 iterations of bootstrap resampling with LASSO regression to calculate the selection frequency for each candidate gene, distinguishing robust markers from random noise.

High Stability: ANXA3 (91.0%), VTN (90.0%), and RELA (88.8%) were consistently selected in the majority of iterations, demonstrating their robustness as core diagnostic markers.

Moderate Stability: IQGAP1 exhibited moderate selection frequency (28.8%) in the LASSO bootstrap analysis. However, it was retained in the final model due to two key factors: (1) its robust identification by the SVM-RFE algorithm (the other component of our dual-strategy feature selection), and (2) its critical validation as one of the two most significant predictors in the independent human cohort.

The gene selection frequency data is provided in S8 Table of the supplementary materials.

Comment 3: Report calibration slope/intercept and Brier score.

Response:

As requested, we have calculated and reported the following calibration metrics. The analysis, based on 500 bootstrap resamples, demonstrates that our model is both well-calibrated and stable: Brier Score: The optimism-corrected Brier score was 0.108, indicating high probabilistic precision. Calibration Slope: The optimism-corrected slope was 0.95, which is very close to the ideal value of 1.0, suggesting excellent alignment between predicted probabilities and observed outcomes. Calibration Intercept: The intercept was -0.05, reflecting minimal calibration bias.These metrics are provided in S9 Table .

Comment 4: The manuscript is scientifically intelligible but linguistically below publication quality in its current form. The English is serviceable for review but would not meet PLOS ONE’s final publication standard without professional copyediting.

Response:

We apologize for the linguistic shortcomings in the previous version. The revised manuscript has been thoroughly edited by a professional, native English-speaking scientific editor. We have corrected grammatical errors, enhanced sentence flow, and ensured that all terminology complies with standard academic usage. We believe the language quality now meets the publication standards of PLOS ONE.

3、Response to Reviewer #3

General Comment: The manuscript "Pyroptosis-related transcriptional signature and diagnostic modeling in acute pancreatitis with cross-species external validation" presents an interesting approach to identifying diagnostic markers for Acute Pancreatitis (AP). However, the study design involves a significant biological domain shift—training on mouse pancreatic tissue and validating on human peripheral blood. While the statistical recalibration is impressed, the biological plausibility is not sufficiently addressed.

Response:

We sincerely thank the reviewer for their critical and constructive insights. We agree that bridging the "tissue-to-blood" and "mouse-to-human" gaps is a significant challenge. In the revised manuscript, we have addressed these concerns by providing evidence for the "tissue-blood axis," selecting the appropriate diagnostic gene model based on AIC, and implementing a rigorous split-sample validation to avoid overfitting.

Comment 1: Could the authors provide evidence (from literature or additional analysis) that the expression levels of the four model genes (ANXA3, IQGAP1, RELA, VTN) in the pancreas are concordant with their expression in peripheral blood during AP?

Response:

We appreciate this important question and acknowledge that local tissue injury and systemic immune responses represent distinct biological compartments. However, our findings support a functional concordance between these compartments:

Systemic Reflection of Local Injury: As detailed in the updated Results and Fig S1, despite inherent interspecies variability, the core genes (ANXA3 and IQGAP1) were consistently dysregulated in human blood samples. This supports the concept of "Liquid Biopsy," where circulating leukocytes reflect the systemic inflammatory activation (e.g., pyroptosis and cytoskeletal remodeling) triggered by pancreatic necrosis.

Mechanistic Link: ANXA3, a known marker of neutrophil granules, is elevated in blood, reflecting the massive mobilization of neutrophils—a hallmark of AP pathology. We have expanded the Discussion to reference literature confirming that transcriptomic changes in blood serve as a reliable proxy for the immune microenvironment during AP progression.

Comment 2: How can we be certain that the model reflects pancreas-specific pathology rather than non-specific systemic inflammation (e.g., sepsis)?

Response:

To ensure specificity and minimize non-specific "noise," we implemented the following measures:

Optimization via AIC: We compared models using the Akaike Information Criterion (AIC). The two-gene model (ANXA3 + IQGAP1) exhibited a significantly better fit (AIC: 61.12) compared to the four-gene model (AIC: 63.01), effectively filtering out less specific inflammatory mediators, such as RELA.

Severity Correlation: Our secondary analysis (Results section and S2A Fig) demonstrates a significant correlation between ANXA3 expression and disease severity (Mild vs. Severe AP), suggesting that the signature is sensitive to the specific pathological intensity of the pancreatic injury, rather than merely reflecting a generic binary "inflammation" signal.

Comment 3: What is the specific clinical value compared to standard biochemical markers (Lipase/CRP)?

Response:

We have highlighted the added value of the optimized two-gene signature (originally derived from a four-gene model):

Clinical Net Benefit: We performed Decision Curve Analysis (DCA, Fig. 14C), which confirmed that our recalibrated model offers substantial net clinical benefit across a wide range of threshold probabilities, outperforming both the "treat all" and "treat none" strategies.

Severity Prediction: Standard biomarkers, such as Lipase, often fail to predict disease severity. Our model, particularly ANXA3, demonstrated an AUC of 0.772 in differentiating disease severity (S2C Fig), providing prognostic insights that are not captured by simple enzymatic tests.

Limitations: Unfortunately, the public GEO datasets used for validation do not provide individual-level metadata for Serum Lipase or CRP, preventing a direct statistical comparison. Future studies will directly compare the differences between CRP or Lipase and gene-based diagnostics using clinical samples.

Comment 4: Please clarify if the "recalibration" parameters were derived from a subset of the external cohort (e.g., a 50% split) and tested on the remainder, or if they were fitted on the whole external dataset.

Response:

We appreciate the reviewer’s concern regarding the risk of overfitting. To ensure the reliability of the recalibrated model, we strictly avoided fitting and testing on the same data:

Split-Sample Validation: The external human cohort (N=119) was randomly partitioned into a Training Set (50%, n=60) for coefficient re-estimation (recalibration) and an independent Testing Set (50%, n=59) for performance evaluation.

Performance Verification: When applied to the unseen testing set, the model achieved an AUC of 0.952 (S4 Fig), which is highly consistent with the full-cohort AUC of 0.957 ( S5 Table).

Robustness: This consistency, along with a Brier score of 0.0774 and a calibration slope of 1.00 (S4 Table), confirms that our recalibration process is robust and the high diagnostic accuracy is not an artifact of overfitting.

We have updated the Statistical Analysis section to explicitly describe this split-validation procedure.

4、Response to Reviewer #4

General Comment: This work is very interesting and has potential significance in the field of precision medicine for acute pancreatitis. The analysis of target gene expression in acute pancreatitis may help predict disease occurrence and severity in the future, particularly in patients with recurrent acute pancreatitis, and may contribute to the prevention of recurrent episodes and progression to chronic pancreatitis.

Response:

We sincerely thank the reviewer for their positive evaluation of our work and for recognizing its translational potential for precision medicine in AP. The reviewer’s constructive comments regarding sample size, population demographics, severity stratification, and disease specificity were invaluable in strengthening the manuscript and clarifying its clinical implications.

Comment 1: The study provides valuable information... I would like to suggest that the difference in sample size between the acute pancreatitis group and the control group may affect the robustness of the results. If feasible, using comparable population sizes in both groups could improve the consistency and reliability of the findings.

Response:

We fully agree with the reviewer that balanced datasets are ideal for statistical modeling. Because our study relied on retrospective public cohorts from the GEO database, we were constrained by the sample sizes provided in the original datasets. However, this imbalance naturally reflects the clinical reality of hospital-based cohorts, where patients with AP frequently outnumber healthy controls.

To mitigate potential statistical bias arising from this sample size imbalance, we implemented the following measures:

We employed robust non-parametric tests (like the Mann-Whitney U test), which do not assume equal variances or sample sizes.

During model recalibration, we utilized rigorous split-sample validation and bootstrap resampling (B = 500 iterations). The resulting optimism-corrected metrics (like an AUC of 0.958 and a Brier score of 0.077) confirm that our model’s high performance is robust and unaffected by majority-class skew.

We explicitly acknowledged the limitations associated with sample size imbalance in the revised Limitations section of the manuscript.

Comment 2: Additionally, could the authors please clarify whether the acute pancreatitis data are derived from a pediatric population or the general population?

Response:

We apologize for omitting this demographic information. The acute pancreatitis data evaluated in our validation cohort (GSE194331) were derived from an adult population, rather than a pediatric cohort. To ensure the target demographic is unambiguous, we have explicitly clarified this detail in the Materials and Methods section.

Comment 3: It would also be helpful to know whether patients were stratified according to disease severity (mild versus severe acute pancreatitis) and whether there is an association between PRDEG expression levels and disease severity.

Response:

We appreciate this insightful suggestion. In the revised manuscript, we conducted a secondary exploratory analysis to address this point. Briefly, we stratified patients with AP into mild, moderately severe, and severe groups using the available phenotype metadata. We observed a significant association between the proposed signature and disease severity. Specifically, ANXA3 discriminated severity categories with an AUC of 0.721, and the combined model achieved an AUC of 0.779. The boxplots (now included as S2A Fig) further show a stepwise increase in ANXA3 expression with greater disease severity. Collectively, these results have been added to the Results section and suggest that, beyond its primary diagnostic utility, the signature may also capture clinically relevant information related to disease progression.

Comment 4: Furthermore, regarding the association between acute pancreatitis and PRDEGs/DEGs, could the authors elaborate on whether these findings have potential implications for clinical management or therapeutic decision-making in current clinical practice?

Response:

We have expanded the Clinical Implications subsection within the Discussion to further elucidate the translational value of our findings:

Early Triage and Severity Prediction: Because ANXA3 expression correlates with disease severity, its elevation in early peripheral blood samples could serve as a critical prognostic indicator. This allows clinicians to triage high-risk patients to the intensive care unit (ICU) before overt organ failure manifests, which is particularly valuable given that traditional biochemical markers, such as serum lipase, do not reliably correlate with disease severity.

Diagnostic Adjunct: The identified transcriptomic signature can function as a complementary diagnostic tool for patients with indeterminate biochemical markers or delayed clinical presentations, particularly after transient enzymatic markers have normalized.

Therapeutic Targets: Our immune infiltration analysis revealed a strong inverse correlation between IQGAP1 expression and Treg abundance. This finding provides a mechanistic rationale for investigating immunomodulatory therapies that target the pyroptosis-cytoskeleton axis to res

Attachment

Submitted filename: response to reviewers.docx

pone.0347204.s014.docx^{(24.4KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0347204.r003

Decision Letter 1

Alexis Carrasco

30 Mar 2026

A pyroptosis-related gene signature for the diagnosis of acute pancreatitis

PONE-D-25-44646R1

Dear Dr. yu,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support ..

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Alexis G. Murillo Carrasco

Academic Editor

PLOS One

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

Reviewer #1: All comments have been addressed

Reviewer #3: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions??>

Reviewer #1: Yes

Reviewer #3: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously? -->?>

Reviewer #1: Yes

Reviewer #3: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available??>

Reviewer #1: Yes

Reviewer #3: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English??>

Reviewer #1: Yes

Reviewer #3: Yes

**********

Reviewer #1: I have to repeat same things as i wrote last time. This study is clinically relevant and has the potential to answer to a very challenging disease condition.Even though it is a rectrospective and cross species study, still it holds promise for finding newer avenues for better understanding for what is happening at cellular level in this disease subgroup. Also it might pave way for finding a treatment which can alter the natural course of disease.

Reviewer #3: The manuscript is now methodologically sound, clinically relevant, and meets the high academic standards of the journal. I strongly recommend it for publication in its current form.

**********

what does this mean? ). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Reviewer #1: Yes: Dr Vinod KumarDr Vinod Kumar

Reviewer #3: No

**********

PLoS One. doi: 10.1371/journal.pone.0347204.r004

Acceptance letter

Alexis Carrasco

PONE-D-25-44646R1

PLOS One

Dear Dr. Yu,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Alexis G. Murillo Carrasco

Academic Editor

PLOS One

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Fig. Boxplot of 11 PRDEGs expression across groups.

(TIF)

pone.0347204.s001.tif^{(4MB, tif)}

S2 Fig. Association of signature genes with disease severity and their predictive performance in the human blood cohort.

(TIFF)

pone.0347204.s002.tiff^{(7MB, tiff)}

S3 Fig. Diagnostic performance comparison between the initial four-gene and optimized two-gene models.

(TIF)

pone.0347204.s003.tif^{(4.7MB, tif)}

S4 Fig. Split-sample validation and diagnostic performance of the recalibrated two-gene model.

(TIFF)

pone.0347204.s004.tiff^{(7.5MB, tiff)}

S1 Table. Predicted TF regulatory network for the four model genes.

This table lists the potential TF associated with the identified four model genes.

(CSV)

pone.0347204.s005.csv^{(352B, csv)}

S2 Table. Predicted miRNA–mRNA regulatory network of the four model genes.

This table details the miRNAs targeted to the four model genes.

(CSV)

pone.0347204.s006.csv^{(592B, csv)}

S3 Table. Detailed logistic regression parameters and odds ratios for the optimized two-gene diagnostic model.

(CSV)

pone.0347204.s007.csv^{(244B, csv)}

S4 Table. Recalibration metrics and Brier score for the two-gene model.

(CSV)

pone.0347204.s008.csv^{(90B, csv)}

S5 Table. AUC for the two-gene signature.

(CSV)

pone.0347204.s009.csv^{(71B, csv)}

S6 Table. Performance metrics from 5-fold repeated cross-validation (5x5 CV).

(CSV)

pone.0347204.s010.csv^{(166B, csv)}

S7 Table. Internal validation and optimism correction of the initial four-gene diagnostic model.

(CSV)

pone.0347204.s011.csv^{(500B, csv)}

S8 Table. Bootstrap-based selection frequencies of the 11 PRDEGs in the diagnostic model.

(CSV)

pone.0347204.s012.csv^{(188B, csv)}

Attachment

Submitted filename: response to reviewers.docx

pone.0347204.s014.docx^{(24.4KB, docx)}

Data Availability Statement

[pone.0347204.ref001] 1.Banks PA, Freeman ML, Practice parameters committee of the American college of gastroenterology. Practice guidelines in acute pancreatitis. Am J Gastroenterol. 2006;101:2379–400. doi: 10.1111/j.1572-0241.2006.00856.x [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref002] 2.Forsmark CE, Baillie J, AGA Institute Clinical Practice and Economics Committee, AGA Institute Governing Board. AGA institute technical review on acute pancreatitis. Gastroenterology. 2007;132:2022–44. doi: 10.1053/j.gastro.2007.03.065 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref003] 3.Tenner S, Baillie J, DeWitt J, Vege SS, American College of Gastroenterology. American college of gastroenterology guideline: management of acute pancreatitis. Am J Gastroenterol. 2013;108(9):1400–15; 1416. doi: 10.1038/ajg.2013.218 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref004] 4.Jorgensen I, Miao EA. Pyroptotic cell death defends against intracellular pathogens. Immunol Rev. 2015;265(1):130–42. doi: 10.1111/imr.12287 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref005] 5.McKenzie BA, Dixit VM, Power C. Fiery cell death: pyroptosis in the central nervous system. Trends Neurosci. 2020;43(1):55–73. doi: 10.1016/j.tins.2019.11.005 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref006] 6.Zhaolin Z, Guohua L, Shiyuan W, Zuo W. Role of pyroptosis in cardiovascular disease. Cell Prolif. 2019;52(2):e12563. doi: 10.1111/cpr.12563 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref007] 7.Shi J, Gao W, Shao F. Pyroptosis: gasdermin-mediated programmed necrotic cell death. Trends Biochem Sci. 2017;42(4):245–54. doi: 10.1016/j.tibs.2016.10.004 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref008] 8.Davis S, Meltzer PS. GEOquery: a bridge between the Gene Expression Omnibus (GEO) and BioConductor. Bioinformatics. 2007;23(14):1846–7. doi: 10.1093/bioinformatics/btm254 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref009] 9.Norberg KJ, Nania S, Li X, Gao H, Szatmary P, Segersvärd R, et al. RCAN1 is a marker of oxidative stress, induced in acute pancreatitis. Pancreatology. 2018;18(7):734–41. doi: 10.1016/j.pan.2018.08.005 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref010] 10.Stelzer G, Rosen N, Plaschkes I, Zimmerman S, Twik M, Fishilevich S, et al. The GeneCards suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics. 2016;54:1.30.1-1.30.33. doi: 10.1002/cpbi.5 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref011] 11.Wu J, Zhu Y, Luo M, Li L. Comprehensive analysis of pyroptosis-related genes and tumor microenvironment infiltration characterization in breast cancer. Front Immunol. 2021;12:748221. doi: 10.3389/fimmu.2021.748221 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref012] 12.Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics. 2012;28(6):882–3. doi: 10.1093/bioinformatics/bts034 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref013] 13.Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. doi: 10.1093/nar/gkv007 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref014] 14.Gene Ontology Consortium, Aleksander SA, Balhoff J, Carbon S, Cherry JM, Drabkin HJ, et al. The Gene Ontology knowledgebase in 2023. Genetics. 2023;224:iyad031. doi: 10.1093/genetics/iyad031 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref015] 15.Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28(1):27–30. doi: 10.1093/nar/28.1.27 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref016] 16.Yu G, Wang L-G, Han Y, He Q-Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS. 2012;16(5):284–7. doi: 10.1089/omi.2011.0118 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref017] 17.Sanz H, Valim C, Vegas E, Oller JM, Reverter F. SVM-RFE: selection and visualization of the most relevant features through non-linear kernels. BMC Bioinformatics. 2018;19(1):432. doi: 10.1186/s12859-018-2451-4 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref018] 18.Engebretsen S, Bohlin J. Statistical predictions with glmnet. Clin Epigenetics. 2019;11(1):123. doi: 10.1186/s13148-019-0730-1 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref019] 19.Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez J-C, et al. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinform. 2011;12:77. doi: 10.1186/1471-2105-12-77 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref020] 20.Wu J, Zhang H, Li L, Hu M, Chen L, Xu B, et al. A nomogram for predicting overall survival in patients with low-grade endometrial stromal sarcoma: a population-based analysis. Cancer Commun (Lond). 2020;40(7):301–12. doi: 10.1002/cac2.12067 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref021] 21.Van Calster B, Wynants L, Verbeek JFM, Verbakel JY, Christodoulou E, Vickers AJ, et al. Reporting and interpreting decision curve analysis: a guide for investigators. Eur Urol. 2018;74(6):796–804. doi: 10.1016/j.eururo.2018.08.038 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref022] 22.Yu G, Li F, Qin Y, Bo X, Wu Y, Wang S. GOSemSim: an R package for measuring semantic similarity among GO terms and gene products. Bioinformatics. 2010;26(7):976–8. doi: 10.1093/bioinformatics/btq064 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref023] 23.Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102(43):15545–50. doi: 10.1073/pnas.0506580102 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref024] 24.Liberzon A, Subramanian A, Pinchback R, Thorvaldsdóttir H, Tamayo P, Mesirov JP. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011;27(12):1739–40. doi: 10.1093/bioinformatics/btr260 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref025] 25.Zhou K-R, Liu S, Sun W-J, Zheng L-L, Zhou H, Yang J-H, et al. ChIPBase v2.0: decoding transcriptional regulatory networks of non-coding RNAs and protein-coding genes from ChIP-seq data. Nucleic Acids Res. 2017;45(D1):D43–50. doi: 10.1093/nar/gkw965 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref026] 26.Li J-H, Liu S, Zhou H, Qu L-H, Yang J-H. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res. 2014;42(Database issue):D92-7. doi: 10.1093/nar/gkt1248 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref027] 27.Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453–7. doi: 10.1038/nmeth.3337 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref028] 28.Yadav D, Lowenfels AB. The epidemiology of pancreatitis and pancreatic cancer. Gastroenterology. 2013;144(6):1252–61. doi: 10.1053/j.gastro.2013.01.068 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref029] 29.Banks PA, Bollen TL, Dervenis C, Gooszen HG, Johnson CD, Sarr MG, et al. Classification of acute pancreatitis--2012: revision of the Atlanta classification and definitions by international consensus. Gut. 2013;62(1):102–11. doi: 10.1136/gutjnl-2012-302779 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref030] 30.Zhao Y, Hosking C, Cunningham D, Langhorne J, Lin J-W. Transcriptome analysis of blood and spleen in virulent and avirulent mouse malaria infection. Sci Data. 2020;7(1):253. doi: 10.1038/s41597-020-00592-1 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref031] 31.Bergsbaken T, Fink SL, Cookson BT. Pyroptosis: host cell death and inflammation. Nat Rev Microbiol. 2009;7(2):99–109. doi: 10.1038/nrmicro2070 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref032] 32.Gerke V, Creutz CE, Moss SE. Annexins: linking Ca2+ signalling to membrane dynamics. Nat Rev Mol Cell Biol. 2005;6(6):449–61. doi: 10.1038/nrm1661 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref033] 33.Jakkampudi A, Jangala R, Reddy BR, Mitnala S, Nageshwar Reddy D, Talukdar R. NF-κB in acute pancreatitis: mechanisms and therapeutic potential. Pancreatology. 2016;16(4):477–88. doi: 10.1016/j.pan.2016.05.001 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref034] 34.Wei T, Lambert PF. Role of IQGAP1 in carcinogenesis. Cancers (Basel). 2021;13(16):3940. doi: 10.3390/cancers13163940 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref035] 35.Lappalainen P, Kotila T, Jégou A, Romet-Lemonne G. Biochemical and mechanical regulation of actin dynamics. Nat Rev Mol Cell Biol. 2022;23(12):836–52. doi: 10.1038/s41580-022-00508-4 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref036] 36.Zhou J, Kang X, An H, Lv Y, Liu X. The function and pathogenic mechanism of filamin A. Gene. 2021;784:145575. doi: 10.1016/j.gene.2021.145575 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref037] 37.Li M, Liu D, Jing F, Liu R, Yi Q. The role of Annexin A3 in coronary arterial lesions in children with Kawasaki disease. Front Pediatr. 2023;11:1111788. doi: 10.3389/fped.2023.1111788 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref038] 38.Preissner KT. Structure and biological role of vitronectin. Annu Rev Cell Biol. 1991;7:275–310. doi: 10.1146/annurev.cb.07.110191.001423 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref039] 39.Bertin J, DiStefano PS. The PYRIN domain: a novel motif found in apoptosis and inflammation proteins. Cell Death Differ. 2000;7(12):1273–4. doi: 10.1038/sj.cdd.4400774 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref040] 40.Flo TH, Smith KD, Sato S, Rodriguez DJ, Holmes MA, Strong RK, et al. Lipocalin 2 mediates an innate immune response to bacterial infection by sequestrating iron. Nature. 2004;432(7019):917–21. doi: 10.1038/nature03104 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref041] 41.Spilsbury K, O’Mara MA, Wu WM, Rowe PB, Symonds G, Takayama Y. Isolation of a novel macrophage-specific gene by differential cDNA analysis. Blood. 1995;85(6):1620–9. doi: 10.1182/blood.v85.6.1620.bloodjournal8561620 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref042] 42.Salam MT, Lin P-C, Avol EL, Gauderman WJ, Gilliland FD. Microsomal epoxide hydrolase, glutathione S-transferase P1, traffic and childhood asthma. Thorax. 2007;62(12):1050–7. doi: 10.1136/thx.2007.080127 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref043] 43.Kalinin S, Malkova A, Sharonova T, Sharoyko V, Bunev A, Supuran CT, et al. Carbonic anhydrase IX inhibitors as candidates for combination therapy of solid tumors. Int J Mol Sci. 2021;22(24):13405. doi: 10.3390/ijms222413405 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref044] 44.Wang X, Yang Y, Wang Y, Lu C, Hu X, Kawazoe N, et al. Focal adhesion and actin orientation regulated by cellular geometry determine stem cell differentiation via mechanotransduction. Acta Biomater. 2024;182:81–92. doi: 10.1016/j.actbio.2024.05.017 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref045] 45.Harjunpää H, Llort Asens M, Guenther C, Fagerholm SC. Cell adhesion molecules and their roles and regulation in the immune and tumor microenvironment. Front Immunol. 2019;10:1078. doi: 10.3389/fimmu.2019.01078 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref046] 46.Barros-Becker F, Lam P-Y, Fisher R, Huttenlocher A. Live imaging reveals distinct modes of neutrophil and macrophage migration within interstitial tissues. J Cell Sci. 2017;130(22):3801–8. doi: 10.1242/jcs.206128 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref047] 47.Hu F, Lou N, Jiao J, Guo F, Xiang H, Shang D. Macrophages in pancreatitis: mechanisms and therapeutic potential. Biomed Pharmacother. 2020;131:110693. doi: 10.1016/j.biopha.2020.110693 [DOI] [PubMed] [Google Scholar]

[pone.0347204.ref048] 48.Hajishengallis G, Chavakis T. Mechanisms and therapeutic modulation of neutrophil-mediated inflammation. J Dent Res. 2022;101(13):1563–71. doi: 10.1177/00220345221107602 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref049] 49.Habtezion A. Inflammation in acute and chronic pancreatitis. Curr Opin Gastroenterol. 2015;31(5):395–9. doi: 10.1097/MOG.0000000000000195 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref050] 50.Baer JM, Zuo C, Kang L-I, de la Lastra AA, Borcherding NC, Knolhoff BL, et al. Fibrosis induced by resident macrophages has divergent roles in pancreas inflammatory injury and PDAC. Nat Immunol. 2023;24(9):1443–57. doi: 10.1038/s41590-023-01579-x [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref051] 51.Sturgeon R, Goel P, Singh RK. Tumor-associated neutrophils in pancreatic cancer progression and metastasis. Am J Cancer Res. 2023;13(12):6176–89. [PMC free article] [PubMed] [Google Scholar]

[pone.0347204.ref052] 52.Parmar T, Parmar VM, Perusek L, Georges A, Takahashi M, Crabb JW, et al. Lipocalin 2 plays an important role in regulating inflammation in retinal degeneration. J Immunol. 2018;200(9):3128–41. doi: 10.4049/jimmunol.1701573 [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

A pyroptosis-related gene signature for the diagnosis of acute pancreatitis

Yuting Wang

Jun Li

Zhongsu Yu

Shuyuan Li

Yuxia Chen

Yun Pan

Liangping Cheng

Guangyuan Yu

Roles

Abstract

Introduction

Methods

Data acquisition

Table 1. GEO microarray chip information.

Differentially expressed genes related to pyroptosis associated with AP

GO and KEGG enrichment analysis

Establishment of diagnostic model for AP

Validation of the diagnostic model for AP.

Functional similarity (Friends) analysis

DEG verification and ROC curve

Gene set enrichment analysis (GSEA)

Protein-Protein Interaction (PPI) Network

Immune infiltration analysis

Validation of the diagnostic model for AP in a human RNA-seq cohort (GSE194331)

Statistical analysis

Ethics statement

Results

Technology roadmap

Fig 1. Flowchart of the study design and analytical workflow.

Merging of AP datasets

Fig 2. Batch effects removal of GSE109227, GSE65146 and GSE121038.

Pyroptosis associated DEGs in AP

Fig 3. DGE analysis.

Verification of PRDEGs and ROC curve

Fig 4. Differential expression validation and ROC analysis.

Enrichment analysis

Table 2. Enrichment analysis for PRDEGs.

Fig 5. GO and KEGG enrichment analysis for PRDEGs.

GSEA for AP

Table 3. GSEA for combined datasets.

Fig 6. GSEA for combined datasets.

Establishment of a diagnostic model for AP

Fig 7. Construction of the diagnostic model for AP.

Validation and Friends analysis of the AP diagnostic model

Fig 8. Validation of the diagnostic model for AP and Friends analysis.

Verification and ROC curve of model genes

Fig 9. Differential expression and ROC analysis of model genes in high-risk and low-risk AP groups.

High and low risk GSEA

Fig 10. Differential expression analysis and GSEA of high-risk and low-risk AP groups.

Table 4. GSEA for risk groups.

Construction of control network

Fig 11. Regulatory networks of the model genes.

Immune infiltration analysis of AP (CIBERSORT)

Fig 12. Combined datasets immune infiltration analysis by CIBERSORT algorithm.

Immune infiltration analysis of High and Low Risk Groups (CIBERSORT)

Fig 13. Risk group immune infiltration analysis by CIBERSORT algorithm.

External validation of the diagnostic model in the human RNA-seq cohort

Fig 14. Validation and performance assessment of the recalibrated two-gene diagnostic model in the independent human cohort.

Discussion

Supporting information

Data Availability

Funding Statement

References

Decision Letter 0

Alexis Carrasco

Roles

Author response to Decision Letter 1

Decision Letter 1

Alexis Carrasco

Roles

Acceptance letter

Alexis Carrasco

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK