Chronic pancreatitis complication risk grading system based on exosome transcriptome data

Abstract

The invention relates to the technical field of medical biology, in particular to a chronic pancreatitis complication risk grading system based on exosome transcriptome data, and aims to solve the problems that existing chronic pancreatitis (CP) typing lacks molecular basis, complication risk prediction is weak and clinical transformation is poor. According to the system, plasma exosomes are extracted and sequenced, a functional characteristic gene set is constructed in combination with pancreas single cell data, CP is divided into three risk increasing subtypes by using a COCA algorithm, and finally 12 core miRNAs are screened to construct a BPNN diagnosis model. The invention proves that the plasma exosome can be used for staging classification of chronic pancreatitis for the first time, miRNA non-invasive accurate layering illness conditions can be detected through qPCR, the risk of fatty diarrhea and 3c type diabetes mellitus can be predicted, and the non-invasiveness, convenience, classification accuracy and result repeatability of the plasma exosome have clinical application and transformation advantages.

Description

Technical Field

[0001] This invention relates to the field of molecular diagnostics and disease stratification technology for chronic pancreatitis (CP), specifically a three-stage taxonomy for CP based on plasma exosome transcriptome characteristics. This system accurately predicts the risk of exocrine dysfunction (such as steatorrhea) and endocrine dysfunction (such as type 3c diabetes) in patients with CP by detecting the expression levels of specific microRNAs (miRNAs) in plasma exosomes. It can also differentiate between CP and non-pancreatic diseases, providing molecular evidence for clinical treatment planning, complication risk monitoring, and personalized intervention. Background Technology

[0002] Chronic pancreatitis (CP) is a progressively deteriorating fibrotic and inflammatory state affecting pancreatic tissue. Driven by a complex interplay of genetic and environmental factors, it is characterized by recurrent inflammatory episodes, leading to the replacement of pancreatic parenchyma with fibrous tissue. This fibrosis gradually declines pancreatic exocrine and endocrine functions, triggering a variety of serious complications. Among these, steatorrhea and type 3c diabetes mellitus (T3cDM) are particularly harmful: Steatorrhea, as a core manifestation of pancreatic exocrine insufficiency, leads to malabsorption of nutrients, vitamin deficiencies, and secondary risks such as cardiovascular disease and osteoporosis; T3cDM, as a marker of endocrine dysfunction, in addition to conventional diabetic complications, makes patients more susceptible to hypoglycemia-related adverse events and even death, significantly reducing their quality of life and life expectancy.

[0003] Given the significant impact of pancreatic cholangiopancreatography (CP), assessing disease severity, monitoring treatment effectiveness, and predicting complication risks are core clinical needs. Traditional CP classification systems (such as the Marseille-Rome classification and the Cambridge classification) are largely based on morphological characteristics (e.g., ductal changes observed during endoscopic retrograde cholangiopancreatography (ERCP)) or clinical symptoms (e.g., pain intensity and intervention requirements). These systems fail to accurately reflect the molecular pathological state of pancreatic tissue, easily overlooking early functional impairment and struggling to predict the risk of complications such as steatorrhea and diabetes. While recent molecular typing based on gene variations such as PRSS1 and CASR has revealed disease heterogeneity, it focuses on etiology and fails to integrate multiple genome biomarkers. Currently, all these methods suffer from insufficient classification accuracy and difficulty in accurately predicting complication risks.

[0004] Exosomes, as lipid vesicles carrying cell-specific RNAs (mRNA, miRNA, lncRNA, circRNA) and proteins, can be isolated from peripheral blood and are ideal liquid biopsy carriers. Studies have confirmed that exosomes derived from pancreatic parenchymal cells are key to pancreatic cell communication and participate in the fibrosis and inflammatory regulation of pancreatic cystic fibrosis (CP). Peripheral exosomes have great potential as biomarkers for disease diagnosis and prediction; however, currently there is no system for constructing a CP staging classification system using exosomal transcriptome characteristics, nor are there corresponding clinical translational tools.

[0005] Therefore, developing a classification system based on exosome molecular characteristics that can accurately stratify CP disease and predict the risk of complications is crucial for solving the current dilemma in CP diagnosis and treatment. Summary of the Invention

[0006] Purpose of the invention: Addressing the shortcomings of existing chronic pancreatitis classification techniques, such as a lack of molecular basis, weak predictive ability for complication risks, and poor clinical translation, the purpose of this invention is to: To establish a personalized risk typing system for CP complications based on multi-omics characteristics of plasma exosomes, so as to achieve accurate prediction of the risk of steatorrhea and T3cDM.

[0007] Screening for clinically applicable core molecular biomarkers (12-miRNA combination) to develop low-cost, high-accuracy detection tools.

[0008] This provides molecular evidence for stratified diagnosis and treatment of CP patients and early intervention for complications, filling the technological gap in "exosome liquid biopsy-complication risk typing".

[0009] To elucidate the mechanism of exosome-derived miRNAs in the progression of chronic pancreatitis and provide molecular theoretical support for the classification system.

[0010] Technical solution: A risk stratification system for complications of chronic pancreatitis based on exosome transcriptome data, including the following core steps: S1, Plasma exosome extraction and transcriptome data acquisition: Peripheral blood was collected from patients with chronic pancreatitis (CP) and healthy controls. After anticoagulation with EDTA, plasma was separated by centrifugation and frozen. Plasma exosomes were extracted using Exo-spin exosome separation reagent. Total RNA was extracted from exosomes, and the normalized expression matrices of mRNA, miRNA, lncRNA, and circRNA were obtained by whole transcriptome sequencing. S2, Construction of pancreatic functional characteristic gene sets: Single-cell sequencing data of pancreatic cells from CP patients were downloaded from public databases, and 13 pancreatic cell types were identified by UMAP clustering; the "FindAllMarkers" function was used to screen cells related to exocrine function to construct an exocrine functional characteristic gene set, and cells related to endocrine function were screened to construct an endocrine functional characteristic gene set; it was verified that the expression of two types of characteristic genes in exosomes of CP patients was significantly higher than that in healthy controls, and the association between pancreatic function and exosome molecules was established; S3, COCA three-stage subtyping model construction: The "Cluster of Cluster Assignments (COCA)" algorithm was used for integrated clustering. The optimal number of clusters was determined to be 3 by calculating Δ(K), and CP patients were divided into three subtypes: COCA1, COCA2, and COCA3. The risk of steatorrhea and type 3c diabetes in each subtype was verified by combining clinical data, and the increasing trend of complication risk in the COCA1-COCA3 subtypes was confirmed.

[0011] Preferably, the 12 core miRNAs are obtained by screening in the following manner: based on the exosomal miRNA expression data of COCA three-stage typing according to claim 1, candidate miRNAs are first screened by univariate logistic regression, and then miRNAs with non-zero coefficients are retained by LASSO regression dimensionality reduction. The model is constructed using a backpropagation neural network, and parameters such as the learning rate, activation function, and number of hidden layer nodes are optimized through grid search. The model is used in a derivation cohort. The expression levels of these 12 miRNAs in plasma exosomes can be detected by qPCR to classify COCA subtypes in CP patients and to differentiate between CP and non-CP patients. The 12 core miRNAs include miR-24-3p, miR-20a-5p, miR-486-3p, let-7b-3p, miR-6134, miR-652-3p, miR-6511a-5p, miR-4420, miR-7845-5p, miR-1285-3p, miR-1910-5p, and miR-640. Beneficial effects

[0012] 1. Accuracy: This invention is the first to integrate exosome multi-omics features to construct a three-stage classification system for CP, achieving a classification accuracy significantly higher than traditional morphological or clinical symptom-based classifications. In a derived cohort (89 CP cases + 22 healthy controls), the system achieved a classification accuracy of 94.6%, with a sensitivity of 87.6%, specificity of 77.8%, and accuracy of 86% in distinguishing CP from non-CP diseases.

[0013] 2. Non-invasive: Based on peripheral blood plasma exosomes, it eliminates the need for invasive procedures such as pancreatic tissue biopsy and ERCP, reducing patient trauma and making it suitable for early screening and long-term dynamic monitoring of CP.

[0014] 3. Clinical applicability: The accompanying 12-miRNA diagnostic model can be achieved through routine qPCR detection, which is simple to operate; the classification system facilitates clinicians to develop individualized treatment plans (such as strengthening pancreatic function protection and complication prevention for patients with COCA3 subtype).

[0015] 4. Significant risk prediction value: It can identify high-risk CP patients in advance, providing a basis for early intervention (such as extracorporeal shock wave lithotripsy and endoscopic retrograde cholangiopancreatography) and reducing the incidence of complications.

[0016] 5. Mechanism elucidation: It was also clarified that exosomal miR-24-3p regulates pancreatic stellate cell fibrosis, macrophage inflammation, and β-cell apoptosis by targeting the S100A8 gene of neutrophils, providing a theoretical basis for targeted therapy of CP (such as the S100A8 inhibitor paquimod).

[0017] 6. Potential for cross-disease applications: The correlation paradigm between exosomes and disease subtyping and prognosis established in this invention can be extended to the molecular subtyping of other chronic organ diseases such as cirrhosis and chronic kidney disease, and has broad technical radiation value. Attached Figure Description

[0018] Figure 1 Main workflow and accuracy validation of the three-stage classification system for chronic pancreatitis. A. Overview of the research workflow; BE. Association of complication risks with COCA subtypes; FG. Validation of the accuracy of the 12-miRNA model.

[0019] Figure 2 Plasma exosome extraction and whole transcriptome sequencing workflow and validation results. A. Schematic diagram of sample processing and exosome extraction workflow; B. Exosome TEM validation image; C. Exosome biomarker Western Blot validation image; D. Principal component analysis (PCA) of mRNA expression differences between the healthy donor (NC) group and the CP group; E. Principal component analysis (PCA) of miRNA expression differences between the healthy donor (NC) group and the CP group; F. Principal component analysis (PCA) of lncRNA expression differences between the healthy donor (NC) group and the CP group; G. Principal component analysis (PCA) of circRNA expression differences between the healthy donor (NC) group and the CP group; H. Pathways of enrichment of differentially expressed miRNAs in target genes.

[0020] Figure 3Construction and clinical validation of the COCA three-stage classification system. A. UMAP clustering analysis results of human CP single-cell sequencing data; B. UMAP map of exocrine and endocrine signal expression; C. Four-quadrant diagram showing the expression of exocrine or endocrine signals in each sample (divided into CP and non-CP patients); D. Conceptual diagram of the COCA typing method; E. Four-quadrant diagram showing the expression of exocrine or endocrine signals in each sample (COCA typing); F. Enrichment pathways of different COCA subtypes.

[0021] Figure 4 : Construction and Validation of a 12-miRNA Diagnostic Model. A. LASSO Regression Feature Selection Path Diagram; B. BPNN Model Structure Diagram; C. Expression Heatmap of 12 miRNAs in Various Sample Types; D. Expression Heatmap of 12 miRNAs in 53 qPCR Validation Samples; E. Confusion Matrix of the 12-miRNA Model Distinguishing Between CP and Non-CP Patients.

[0022] Figure 5 Validation of the regulatory mechanism of exosome miR-24-3p / S100A8. A. Effect of miR-24-3p mimics on S100A8 protein expression in HL-60 cells; B. Design of miR-24-3p / S100A8 dual-luciferase reporter gene; C. Results of luciferase reporter assays using empty plasmid (NC), wild-type (WT), and 3UTR mutant (MUT) plasmids; D. Effect of exogenous S100A8 on the expression of col-1 and FN proteins in pancreatic stellate cells; E. Exogenous S100A8 and CD36, TLR4, RAGE, or NC. The expression of col-1 and FN proteins in pancreatic stellate cells after siRNA transfection; F. Effects of exogenous S100A8 on the expression of IL-1B, IL-6 and tumor necrosis factor genes in RAW264.7 cells with and without TLR4; G. Relative protein levels of bcl2 and bax in MIN6 cells after culture with RAW264.7 cells, TLR4 knockout RAW264.7 cells, and cells cultured with exogenous S100A8 supplementation; H. Schematic diagram of the mir-24-3p / S100A8 regulatory pathway. Detailed Implementation

[0023] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] This invention aims to establish a personalized risk classification system for pancreatic complication (CP), and its core research approach includes non-invasive acquisition of pancreatic molecular information, multi-dimensional integrated classification, clinical translational validation, and mechanism analysis. (See attached document.) Figure 1 As shown in A, this invention is mainly carried out in four stages: The first stage is the basic data construction stage: through plasma exosome extraction and whole transcriptome sequencing, the expression characteristics of various molecules such as mRNA and miRNA in exosomes of CP patients and healthy individuals are obtained. At the same time, combined with pancreatic single-cell sequencing data, a characteristic gene set reflecting exocrine and endocrine functions is constructed to provide molecular basis for classification.

[0025] The second phase involved developing a classification model: The COCA algorithm was used to integrate multi-omics data, identifying three disease stages (COCA1-3). The effectiveness of the classification was validated through clinical feature association (complication incidence rate) and survival analysis, ensuring that the classification reflected disease severity and complication risk. (See attached image) Figure 1 As shown in the BE study, we followed up patients who did not have steatorrhea or diabetes at the time of sample collection and observed that the incidence of steatorrhea and diabetes was significantly higher in patients with COCA3 or COCA2 subtypes than in patients with COCA1 subtype.

[0026] The third phase involves the development of clinical translational tools: The most stable miRNAs in circulation are selected as research subjects. Core biomarkers are screened from differentially expressed miRNAs. A 12-miRNA diagnostic model is constructed using LASSO dimensionality reduction and BPNN modeling. This model is then validated using qPCR on 53 randomly selected samples to ensure its rapid and accurate application in clinical testing. Figure 1 As demonstrated by FG, the model achieved accuracies of 96.7%, 85%, and 94.6% on the training set, test set, and all patient cohorts, respectively.

[0027] The fourth stage is mechanism verification: focusing on the interaction between exosome miR-24-3p and neutrophil S100A8, the mechanism by which it regulates pancreatic fibrosis and β-cell apoptosis is verified through cell experiments and animal models. This not only elucidates the molecular basis of the classification system, but also provides potential targets for CP-targeted therapy.

[0028] Experimental Materials and Methods Sample source and processing methods Sample sources: 140 CP patients and 27 healthy controls from Shanghai Changhai Hospital between October 2017 and October 2019; 39 CP patients and 5 healthy controls from the First Affiliated Hospital of Zhengzhou University between October 2018 and October 2020.

[0029] Follow-up protocol: All CP patients were followed up for 48 months, and new occurrences of steatorrhea (fecal fat quantification >7g / 24h) and type 3c diabetes mellitus (T3cDM, CP history + abnormal blood glucose, i.e. fasting blood glucose ≥7.0mmol / L or glycated hemoglobin ≥6.5% and excluding type 2 diabetes) were recorded.

[0030] All samples excluded patients experiencing acute exacerbations of chronic pancreatitis. Diagnosis and clinical staging of chronic pancreatitis followed the 54th edition of the Chinese Guidelines for the Diagnosis and Treatment of Chronic Pancreatitis (2018 edition). Informed consent was obtained from all patients, and the procedures were approved by the hospital ethics committees (Changhai Hospital CHEC2017-232, First Affiliated Hospital of Zhengzhou University 2018-KY-0324). The follow-up period for all patients was approximately 48 months.

[0031] Sample processing procedure: Venous blood is collected in EDTA anticoagulant tubes, centrifuged at 2890×g for 10 minutes at 4℃ to separate plasma, aliquoted and frozen at -80℃.

[0032] Methods for Exosome Extraction and Identification Exosome extraction: Exosomes were separated using a commercially available kit. The specific steps were performed according to the instructions for the Exo-spin exosome separation reagent (CGS Exo-spin) provided by Cell Guidance, Inc. (USA). Exosomes were obtained from plasma samples.

[0033] Exosome identification: Morphological verification: Exosomes were resuspended in 50µL PBS, and the resuspension was dropped onto filter paper. After covering with a copper grid, the paper was incubated at room temperature for 2 minutes. Then, 1% uranium acetate staining solution was added for 30 seconds. The morphology of exosomes was then observed using a Hitachi H7650 transmission electron microscope. Biomarker detection: Total protein was extracted from exosomes, quantified by BCA, and then subjected to SDS-PAGE electrophoresis. The protein was transferred to a PVDF membrane, blocked, and then incubated overnight at 4°C with primary antibodies (CD63, TSG101, and Calnexin, all purchased from Cell Signaling Technology). The membrane was then incubated with HRP secondary antibody at room temperature for 1 hour and then developed by ECL luminescence.

[0034] Exosome whole transcriptome sequencing and data analysis methods RNA extraction and sequencing: Total RNA was extracted from exosomal tissues using the TRIzol method, and RNA quality and purity were detected using an Agilent 2100 bioanalyzer; Sequencing library construction: Sequencing libraries were constructed using the NEXTflex® Small RNA-Seq Kit and the Ovation Human FFPERNA-Seq Library System, and the expression of the sequencing libraries was described using an Illumina HiSeq 2500.

[0035] Data analysis workflow: Raw data filtering: Cutadapt removes adapters and low-quality reads; Tophat is used for mapping to the human reference genome; Bowtie2 is used to build a reference gene index. CircRNAs and their host genes are identified using FIND_CIRC59 and CIRCExplorer60; transcripts are assembled using StringTie software; and known lncRNAs are screened using Perl scripts. Quantitative analysis of lncRNAs and mRNAs is performed using the BALLOWOND R package.

[0036] Differential RNA screening: The expression of exon mRNA, lncRNA, miRNA, and circRNA in the two groups of patients was analyzed using the LIMMA software package. Fold change (Fc) was used as an indicator of differential expression between the CP group and the control group. The t-test was used to evaluate the statistical significance of the differences.

[0037] Functional enrichment analysis: The clusterProfiler package was used to perform GO / KEGG enrichment analysis to identify the rich biological processes, molecular functions, cellular components and pathways associated with these mRNAs, with a focus on pathways related to fibrosis, inflammation and glucose metabolism.

[0038] Methods for constructing a gene set of pancreatic functional characteristics Single-cell data acquisition and preprocessing: Single-cell data of chronic pancreatic tissue from http: / / singlecell.charite.de / cellbrowser / pancreas / ?ds=Chronic_Pancreatitis26 were downloaded using the SEURAT software package. After data filtering, logarithmic normalization and scaling were performed, and intercellular differences were regressed based on the percentage of UMI counts and mitochondrial readings. Cell clustering was performed at a resolution of 0.1 using the "FindCluster" function. Dimensionality reduction was performed using the "RunUMAP" function, and visualization was performed using UMAP. For subpopulation cell clustering, different cell types were extracted separately, and clustering was performed at different resolutions based on visual inspection of their respective top 30 principal components.

[0039] Feature gene screening: The "FindAllMarkers" function in Seurat was used to identify specific genes in each cell subpopulation. Exosome feature gene sets (Top 40 genes from acinar cells + activated stellate cells) and endocrine feature gene sets (Top 40 genes from α cells + β cells) were constructed. The median expression level of feature genes in the exosome sequencing data was calculated, and heatmaps were generated.

[0040] COCA Three-Stage Classification Model Construction and Validation Methods COCA clustering analysis: PAM clustering was performed on the expression matrices of mRNA, miRNA, lncRNA, and circRNA. Specifically, 1000 iterations of 80% resampling were performed using the PAM method based on the Pearson correlation distance metric, resulting in clusters ranging from 2 to 10. The required number of clusters was determined by increasing Δ(K) according to the proportion of the area under the CDF. The value of Δ(K) determined the optimal number of clusters to be 3 (COCA1-3).

[0041]

[0042] Classification validity verification: Molecular verification: WGCNA identifies the core modules of each subtype and analyzes the expression trends of characteristic genes; Clinical verification: Patients without steatorrhea or diabetes at the time of sample collection were followed up, and survival analysis was performed using the Kaplan-Meier method.

[0043] Methods for constructing and validating 12-miRNA diagnostic models Biomarker screening: Univariate logistic regression (P<0.01) was used to initially screen candidate miRNAs; LASSO regression (10-fold cross-validation) was used to reduce dimensionality and select 12 miRNAs with non-zero coefficients.

[0044] BPNN Model Construction and Validation: The training and test sets are split in an 8:2 ratio, and the model is built using the "neuralnet" package. Parameters are optimized through grid search—including the learning rate, loss function, activation function, number of hidden layers, and number of nodes per layer.

[0045] Verification method of miR-24-3p / S100A8 regulation mechanism: Identifying key action pairs: First, 12-miRNA-related mRNAs with potential binding ability were screened. Then, based on the whole transcriptome sequencing results, 31 miRNA-mRNA pairs were identified in the control groups COCA1, COCA2, and COCA3. Next, miRNAs with consistent trends in the COCA1-3 groups were screened. Finally, effector molecules that interact with the screened miRNAs with consistent trends were identified.

[0046] Molecular binding verification: Dual-luciferase reporter assay verified the direct binding relationship between miR-24-3P and S100A8.

[0047] Cellular function validation: Neutrophils, pancreatic stellate cells, macrophages, and pancreatic β cells were co-cultured in Transwell to validate the effects of S100A8 on fibrosis and β-cell apoptosis.

[0048] Animal model validation: In the CP mouse model (induced by taurine), neutrophil depletion and intervention with the S100A8 inhibitor (paquimod) were used to validate the source and function of exosomal S100A8.

[0049] Cell and animal models Cell lines: HL-60 cells (RIKEN RCB3683): ​​cultured in RPMI 1640 medium containing 10% FBS, induced to differentiate into neutrophil-like cells in 1.25% DMSO for 3 days.

[0050] Human pancreatic stellate cells (HPSCs): donated by a professor from the MD Anderson Cancer Center in the United States, cultured in DMEM medium containing 10% FBS.

[0051] RAW264.7 macrophages, HEK293T cells, and MIN6 pancreatic β cells (all purchased from ATCC) were cultured in DMEM medium containing 10% FBS and 1% penicillin / streptomycin, respectively.

[0052] All cells were cultured at 37°C in a 5% carbon dioxide environment, with the culture medium changed every 2 days.

[0053] Animal models: CP model: C57BL / 6 mice (6-8 weeks old, male) were injected intraperitoneally with cynomorium (50 μg / kg, once every hour for a total of 7 times), 3 times a week for 6 weeks.

[0054] Neutrophil exhaustion model: At week 6 and week 6+4, anti-Ly6g antibody (12.5 mg / kg) was injected intraperitoneally, while the control group was injected with isotype IgG.

[0055] S100A8 inhibitor intervention model: Starting from week 3 of modeling, paquimod (5 mg / kg) was administered by gavage daily, while the control group was administered the same amount of DMSO by gavage.

[0056] Statistical analysis Data processing and plotting were performed using R, Python, and GraphPad Prism. Intergroup comparisons were conducted using t-tests or ANOVA. The risk of complications was assessed using Kaplan-Meier curves and the Log-rank test. A p-value < 0.05 was considered statistically significant. Example

[0057] Example 1. Plasma exosome extraction verification and transcriptome differential analysis (corresponding to...) Figure 2 ) Exosome extraction: First, such as Figure 2As shown in Figure A, plasma was collected from CP patients and healthy donors (NCs) to extract exosomes.

[0058] Exosome morphology: Transmission electron microscopy revealed that, for example... Figure 2 As shown in Figure B, the extracted exosomes are typical "tea saucer-shaped" lipid bilayer vesicles with a particle size concentrated in the range of 60-90 nm.

[0059] Symbolic representation: Figure 2 Western blot results showed that CD63 and TSG101 were positively expressed in exosomes of both the CP and NC groups, while Calnexin was expressed only in cell lysis buffer (negative control), demonstrating that the extracted exosomes were free of endoplasmic reticulum contamination and of high quality. RNA content sequencing: RNA content in exosomes was sequenced, and principal component analysis was performed on mRNAs, miRNAs, lncRNAs, and circRNAs between patients and healthy groups. Results are as follows: Figure 2 As shown in the DG, the CP group and the NC group samples are significantly separated, and the samples within each group have a high degree of clustering.

[0060] Differential gene enrichment analysis: Pathway enrichment analysis was performed on differentially expressed mRNAs, miRNAs, lncRNAs, and circRNAs targeting genes. Figure 2 H shows the main enrichment pathways of differentially expressed miRNA genes, namely the Foxo pathway, the MAPK pathway, and the insulin signaling pathway, which are associated with pancreatic fibrosis and abnormal glucose metabolism.

[0061] In summary, there are significant differences in the plasma exosome transcriptome between CP patients and healthy individuals. The differentially expressed molecules are involved in key pathways of CP pathological progression, providing a molecular basis for the construction of classification models.

[0062] Example 2. COCA typing construction (corresponding to) Figure 1 , Figure 3 ) Feature gene set construction ( Figure 3 A, 3B): Single-cell sequencing data of pancreatic tissue from CP patients were downloaded from a public database. Annotations identified 13 cell types in the CP group. The results are presented in... Figure 3 A. Then as follows Figure 3 As shown in Figure B, we identified and visualized the expression levels of corresponding gene tags in the endocrine and exocrine systems. Exocrine tags consist of the first 40 marker genes from REG+ cells of acinar cells and activated stellate cells, while endocrine tags consist of the first 40 marker genes from α cells and β cells. Next, we defined the median expression level of each tag gene in CP patients as the expression level of that tag, and displayed this as a dot plot. Figure 3As can be seen in C, the expression of these tag genes in the CP group was significantly higher than that in the NC group.

[0063] COCA Clustering: The Cluster Assignment (COCA) algorithm was used to analyze each of the four molecular types (mRNA, miRNA, lncRNA, and CircRNA), and the optimal number of clusters was determined to be 3 (COCA1, COCA2, COCA3). The results are shown in... Figure 3 D. Then, as in... Figure 3 E. We continue to use dot plots to show the expression of endocrine and exocrine tag genes in different COCA subtypes. We can see that these tag genes show an upward trend in COCA1-3 subtypes.

[0064] Clinical validation: To determine whether the COCA classification system could be used to predict the risk of diabetes and seborrheic arrhythmia in patients with CP, we followed up patients who did not have steatorrhea or diabetes at the time of sample collection. The results are shown in... Figure 1 BE (Bioequivalence) revealed that patients with COCA3 or COCA2 subtypes had a significantly higher incidence of steatorrhea and diabetes compared to those with COCA1 subtypes. This indicates that our developed COCA system can predict the risk of early diabetes and seborrheic arrhythmia in CP (prolapse and cerebral embolism) patients.

[0065] Molecular-level validation: To identify the key pathways activated in each COCA subtype, we built upon the above findings and used weighted gene co-expression network analysis (WGCNA) to identify the core modules of each subtype. We discovered that COCA1 is associated with stress response, COCA2 with metabolic changes, and COCA3 with inflammation. The results are presented in… Figure 3 F in.

[0066] In summary, the three-stage classification of COCA can accurately reflect the differences in pancreatic exocrine and endocrine function and the risk of complications in CP patients. Each subtype has unique molecular characteristics and pathological associations, and has clinical predictive value.

[0067] Example 3. Construction and Validation of 12-miRNA Diagnostic Model (corresponding) Figure 4 ) While the aforementioned COCA genotyping system can correctly classify CP patients, it still has some limitations. For example, it requires complete sequencing data and cannot distinguish between CP patients and healthy individuals. To improve the model's clinical translation potential and reduce testing costs, we further optimized the model based on the above research. Previous studies have shown that miRNAs are relatively stable in vitro and are the best diagnostic indicators; therefore, we focused our research on miRNAs.

[0068] LASSO Screening: First, to address the inherent challenges of high dimensionality and multicollinearity in exosome miRNA sequencing data, we used LASSO regression for dimensionality reduction, ultimately identifying 12 core miRNAs. The flowchart is shown below. Figure 4 A.

[0069] BPNN Modeling and Validation: Next, we construct the BNPP model. The structure of the BNPP model is shown in... Figure 2 B. The derived queue (89 CP cases + 22 NC cases) was divided into a training set (71 CP cases + 18 NC cases) and a test set (18 CP cases + 4 NC cases) at an 8:2 ratio. After continuous parameter adjustments, the final model achieved accuracies of 96.7%, 85%, and 94.6% on the training set, test set, and full dataset, respectively. Figure 1 F). Next, we randomly selected 53 samples for qPCR detection of 12 miRNAs, and the results were input into a classifier for validation, achieving an accuracy of 83% (F). Figure 1 G). Finally, validation was performed in an external queue. Notably, all normal samples in the external validation queue were accurately identified.

[0070] Model Expansion and Validation: We downloaded miRNA data representing 15 osteoarthritis patients and 3 rheumatoid arthritis patients from public databases (GSE263996, GSE218599) and combined them with miRNA data from our local cohort of 89 CP patients. The model was applied to predict CP status, and the resulting confusion matrix is ​​shown below. Figure 4 As shown in Figure E, among 107 subjects, the COCA algorithm correctly identified predicted CP patients with a sensitivity of 87.6% and a specificity of 77.8%. These results preliminarily indicate that our algorithm not only has predictive power for CP-related complications (such as steatorrhea and diabetes) but also has potential diagnostic value in distinguishing between CP and non-CP conditions.

[0071] In summary, the 12-miRNA diagnostic model has high accuracy and stability, can be rapidly detected by qPCR, and is feasible for clinical translation.

[0072] Example 4. Verification of the regulatory mechanism (corresponding to) Figure 5 ) By estimating miRNA-mRNA interactions and correlation analysis, as well as the expression trends in COCA1-3 samples, we chose to focus on whether the interaction between miR-24-3p and S100A8 affects the occurrence of CP. Through a series of sequencing and in vitro and in vivo experiments, we determined that miR-24-3p and S100A8 are key interaction pairs in the exonucleosomes of CP patients, and neutrophils also play a key role in this interaction.

[0073] Molecular interaction verification: First, such as Figure 5 As shown in Figure A, we confirmed that the upregulation of miR-24-3p in neutrophils inhibited S100A8 expression at both the mRNA and protein levels. Next, as... Figure 5 B. We predicted the 3'-untranslated region (UTR) binding site between miR-24-3p and S100A8 using Miranda software and constructed a luciferase reporter vector. Experiments showed ( Figure 5 C), miR-24-3p mimics reduced the luciferase activity of the S100A8 WT 3'UTR vector, while the Mut vector showed no significant change, demonstrating that miR-24-3p directly targets S100A8.

[0074] Cellular function and animal validation: Neutrophil-astrocytocyte co-culture showed that S100A8 increased the expression of fibrosis markers in a dose-dependent manner. Figure 5 D); and then, by knocking out three potential receptors for S100A8, the downstream effector molecules were identified as RAGE and TLR4 (D). Figure 5 E). Finally, by stimulating macrophages with and without TLR4 with S100A8, it was confirmed that S100A8 promotes the inflammatory response of macrophages through TLR4. Figure 5 F), and then, through co-culture of pancreatic β cells and macrophages, it was determined that S100A8 can promote pancreatic β cell apoptosis by activating TLR4 in macrophages. Figure 5 G).

[0075] In summary, the molecular mechanism of this classification system has been elucidated. Figure 5 E), downregulation of exosomal miR-24-3p can lead to upregulation of neutrophil S100A8, which can alleviate CP pathological damage by activating stellate cells through TLR4, inducing macrophage inflammation, promoting β-cell apoptosis, and promoting CP progression.

[0076] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A risk stratification system for complications of chronic pancreatitis based on exosome transcriptome data, characterized in that, The core steps include the following: S1, Plasma exosome extraction and transcriptome data acquisition: Peripheral blood was collected from patients with chronic pancreatitis (CP) and healthy controls. After anticoagulation with EDTA, plasma was separated by centrifugation and frozen. Plasma exosomes were extracted using Exo-spin exosome separation reagent. Total RNA was extracted from exosomes, and the normalized expression matrices of mRNA, miRNA, lncRNA, and circRNA were obtained by whole transcriptome sequencing. S2, Construction of pancreatic functional characteristic gene sets: Single-cell sequencing data of pancreatic cells from CP patients were downloaded from public databases, and 13 pancreatic cell types were identified by UMAP clustering; the "FindAllMarkers" function was used to screen cells related to exocrine function to construct an exocrine functional characteristic gene set, and cells related to endocrine function were screened to construct an endocrine functional characteristic gene set; it was verified that the expression of two types of characteristic genes in exosomes of CP patients was significantly higher than that in healthy controls, and the association between pancreatic function and exosome molecules was established; S3, COCA three-stage subtyping model construction: The "Cluster of Cluster Assignments (COCA)" algorithm was used for integrated clustering. The optimal number of clusters was determined to be 3 by calculating Δ(K), and CP patients were divided into three subtypes: COCA1, COCA2, and COCA3. The risk of steatorrhea and type 3c diabetes in each subtype was verified by combining clinical data, and the increasing trend of complication risk in the COCA1-COCA3 subtypes was confirmed.

2. The risk stratification system for complications of chronic pancreatitis based on exosome transcriptome data according to claim 1, characterized in that: The 12 core miRNAs were obtained through screening in the following manner: based on the exosomal miRNA expression data of the COCA three-stage typing in claim 1, candidate miRNAs were first screened by univariate logistic regression, and then miRNAs with non-zero coefficients were retained by LASSO regression dimensionality reduction. The model is constructed using a backpropagation neural network, and parameters such as the learning rate, activation function, and number of hidden layer nodes are optimized through grid search. The model is used in a derivation cohort. The expression levels of these 12 miRNAs in plasma exosomes can be detected by qPCR to classify COCA subtypes in CP patients and to differentiate between CP and non-CP patients. The 12 core miRNAs include miR-24-3p, miR-20a-5p, miR-486-3p, let-7b-3p, miR-6134, miR-652-3p, miR-6511a-5p, miR-4420, miR-7845-5p, miR-1285-3p, miR-1910-5p, and miR-640.

Images