Transcriptomic profiling of severe acute pancreatitis reveals neutrophil-driven inflammatory signatures and a five-gene diagnostic panel

Severe acute pancreatitis (SAP) is a life-threatening condition associated with systemic inflammatory response syndrome, multi-organ failure, and mortality rates of 20–40%. Early and accurate stratification of SAP from milder forms of acute pancreatitis (AP) remains a critical clinical challenge. Blood-based transcriptomic signatures offer a promising avenue for non-invasive biomarker discovery. We analyzed the publicly available dataset GSE194331, comprising human whole blood bulk RNA-seq from 119 samples: healthy controls (HC, n = 32), SAP ( n = 10), moderately severe AP (MSAP, n = 20), and mild AP (MildAP, n = 57). Differential expression analysis was performed using DESeq2, followed by gene set enrichment analysis (GSEA) with MSigDB Hallmark and KEGG gene sets. Single-sample immune cell deconvolution for 12 peripheral blood immune cell types was performed using ssGSEA-like rank-weighted enrichment with curated marker panels. A diagnostic panel was developed using elastic net/LASSO with leave-one-out cross-validation (LOO-CV) and bootstrap stability selection. Cross-species validation was conducted using mouse pancreatic single-cell RNA-seq data (GSE188819; 25,870 cells, 10x Genomics). DESeq2 identified 472 differentially expressed genes (DEGs; 367 upregulated, 105 downregulated) at padj < 0.01 and |log2FC|≥2, dominated by neutrophil granule proteins and lymphocyte depletion markers. GSEA revealed 67 significant pathways, with top activated pathways including Oxidative Phosphorylation (NES = 2.71) and Inflammatory Response (NES = 2.13), and top suppressed pathways including MYC Targets V2 (NES = − 2.19). Single-sample immune cell deconvolution revealed a continuous, severity-graded innate–adaptive imbalance, with M2 macrophages (Spearman r = + 0.66 with severity), neutrophils (r = + 0.51), and monocytes (r = + 0.46) progressively expanding, and CD4 + T cells ( r = − 0.61), B cells ( r = − 0.47), and NK cells ( r = − 0.40) progressively contracting from HC through MildAP and MSAP to SAP. A five-gene diagnostic panel (CAMP, LIPN, RAB13, SERPINB10, LILRA4) achieved LOO-CV AUC = 1.000 in the HC vs. SAP discovery comparison. The panel score showed a strong monotonic relationship with AP severity (Spearman r = 0.672, P = 5.8 × 10 − 17 ) and discriminated SAP from milder AP forms within the AP cohort itself (SAP vs. Mild+MSAP AUC = 0.890, 95% CI 0.756–0.982). Diagnostic performance in the discovery comparison was comparable to a previously published S100A8/S100A9/S100A12 panel (AUC = 0.997), while the dual-axis composition of our panel (innate-myeloid activation plus adaptive-compartment integrity) provided complementary biological information not captured by single-axis innate-immune markers. In mouse scRNA-seq, massive macrophage expansion (546 WT → 8,976 AP cells) and T/NK cell depletion were observed. Key human SAP genes—Anxa3, S100a9, and Rab13—were consistently upregulated across multiple mouse cell types in AP. In vitro, siRNA-mediated knockdown of CAMP in ATRA-differentiated HL-60 cells markedly attenuated LPS-induced neutrophil extracellular trap (NET) formation, providing direct functional evidence that the panel’s most stably selected gene is mechanistically required for innate-immune effector responses. [Revision note: The LOO-CV AUC of 1.000 in the SAP vs. HC discovery comparison ( n = 42) should be interpreted as a discovery-cohort upper bound. Under 200-repeat stratified 5-fold cross-validation on the same cohort the mean AUC remained 1.000 [95% percentile CI 1.000–1.000], confirming that the perfect value is not an artefact of the LOO scheme; the clinically more informative primary read-out is the SAP vs. Mild+MSAP AUC of 0.906 [95% CI 0.815–0.971]. External validation in an independent prospective cohort remains mandatory before any clinical use.] SAP is characterized by a neutrophil-driven blood transcriptomic signature with concurrent metabolic reprogramming and adaptive immune suppression. The innate–adaptive immune imbalance unfolds as a continuous, severity-graded process rather than a binary switch. The five-gene panel encodes a dual-axis molecular fingerprint—innate-myeloid activation (CAMP/LIPN/RAB13/SERPINB10) and adaptive-compartment integrity (LILRA4)—translating into superior diagnostic performance across the full AP severity spectrum. Cross-species conservation of key inflammatory genes supports their mechanistic and translational relevance.