Development of a Diagnostic Model for Barrett's Esophagus and Esophageal Adenocarcinoma Based on Machine Learning and Immune Infiltration

Development of a Diagnostic Model for Barrett's Esophagus and Esophageal Adenocarcinoma Based on Machine Learning and Immune Infiltration | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Development of a Diagnostic Model for Barrett's Esophagus and Esophageal Adenocarcinoma Based on Machine Learning and Immune Infiltration Hui Feng, Junyao Liang, Tao Zhou, Jie Xiao, Dexu Zhang, Man Zhou, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8004374/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Esophageal adenocarcinoma (EAC) is a highly lethal cancer, with Barrett's esophagus (BE) as its only known precursor. Early diagnosis is challenging, and the key biomarkers and mechanisms driving the BE-to-EAC progression are not fully understood. Methods We integrated transcriptomic data from BE (from the Gene Expression Omnibus (GEO)) and EAC (from The Cancer Genome Atlas (TCGA)) to identify shared differentially expressed genes (SDEGs). We then performed functional enrichment, protein-protein interaction, and immune infiltration analyses. Machine learning algorithms, including Least Absolute Shrinkage and Selection Operator (LASSO) and Random Forest, were applied to screen for diagnostic biomarkers. Results We identified 101 SDEGs. Subsequent analysis highlighted immune-related regulatory genes. Machine learning pinpointed MUC13 and PBLD as robust diagnostic biomarkers. The diagnostic model exhibited excellent performance, with area under the curve (AUC) values of 0.94 for BE and 0.98 for EAC, supported by nomograms and calibration curves (p > 0.9). Mechanistically, MUC13 promotes tumor progression via NF-κB, Wnt/β-catenin, MAPK/ERK, HIF-1α/VEGF, and epithelial-mesenchymal transition (EMT) pathways, driving inflammation-immune crosstalk and metastasis. In contrast, PBLD appears to suppress these processes. Conclusions MUC13 and PBLD are identified as potential biomarkers for the progression from BE to EAC. They play opposing roles in regulating key oncogenic pathways, immune response, and metastasis, offering significant potential for improving early diagnosis and developing targeted therapies. Barrett’s Esophagus Esophageal Adenocarcinoma biomarker diagnostic model MUC13 PBLD transcriptomics immune infiltration machine learning Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Esophageal cancer (EC) ranks as the seventh most prevalent malignant tumor worldwide, exhibiting an exceptionally poor prognosis with a five-year overall survival rate of less than 30%[ 1 ] and ranks as the sixth leading cause of cancer-related mortality[ 2 ]. The clinical manifestations of EC include dysphagia, chest pain or retrosternal burning sensation, dyspepsia or heartburn, vomiting, and hematemesis. Definitive diagnosis requires endoscopic visualization and pathological biopsy confirmation. The two principal histopathological subtypes of EC are esophageal adenocarcinoma (EAC) and esophageal squamous cell carcinoma (ESCC). While ESCC remains the predominant variant globally, the incidence of EAC has increased markedly and steadily over the past four decades[ 3 ], particularly in Western populations. Notably, more than 40% of EAC patients present with metastatic disease at diagnosis, and their five-year survival rate is below 20%[ 4 ]. EAC originates from metaplastic columnar epithelium (such as Barrett's esophagus[BE]). Chronic exposure to risk factors such as obesity, alcohol consumption, tobacco use, and especially sustained gastroesophageal reflux disease (GERD) induces the replacement of native stratified squamous epithelium with intestinal-type columnar epithelium (intestinal metaplasia, IM) in the distal esophagus, which may then progress to EAC. The treatment of EAC includes surgery (transthoracic esophagectomy with lymph node dissection, endoscopic submucosal dissection [ESD], endoscopic mucosal resection [EMR]), chemoradiotherapy, targeted therapy, and immunotherapy[ 5 ]. Barrett's esophagus (BE) is a metaplastic transformation of esophageal squamous mucosa to intestinal-type columnar epithelium induced by chronic acid exposure, representing the only established precursor lesion and a major risk factor for EAC[ 6 ]. The diagnosis of BE principally relies on endoscopic examination and histopathological confirmation due to the lack of specific clinical symptoms. Epidemiological data show a significant increase in BE incidence in developed nations in recent decades[ 7 ]. Risk factors for BE include advancing age, chronic GERD, male sex, tobacco use, Caucasian ethnicity, familial predisposition, and central obesity[ 8 ]. The progression from BE to EAC is a stepwise process, starting from non-dysplastic Barrett's esophagus (NDBE), advancing to low-grade dysplasia (LGD), then to high-grade dysplasia (HGD), and ultimately to invasive EAC. Patients with BE have the highest risk of progressing to HGD or EAC within the first 1–3 months after diagnosis, with a 30- to 125-fold higher incidence of EAC compared to the general population, and individuals with HGD have a significantly higher rate of malignant transformations[ 9 ]. The main treatments for BE include endoscopic eradication therapy (EET), esophagectomy for advanced cases, and proton pump inhibitor (PPI)--based anti-reflux therapy, along with regular endoscopic surveillance[ 10 ]. EAC often presents with insidious early-stage symptoms that are easily overlooked, posing significant challenges for timely diagnosis. Although BE is a well-established risk factor for EAC, the molecular mechanisms driving this progression remain incompletely understood. Therefore, this study aims to identify shared genetic signatures between BE and EAC using integrative bioinformatics approaches and construct a robust genetic diagnostic model. This will help elucidate the molecular mechanisms of disease evolution and provide new insights for early detection and targeted therapy of Barrett's esophagus-derived esophageal adenocarcinoma. Material and methods Study Design The workflow of this study is schematically presented in Fig. 1 . Data acquisition and preprocessing The raw matrix datasets for BE were acquired from the Gene Expression Omnibus (GEO) database ( https://www.ncbi.nlm.nih.gov/geo/ ). Dataset selection was based on the following specific criteria: (1)data from human subjects, (2)samples from both the BE group and healthy control groups, while excluding EAC group, (3)genome-wide expression profiles. Based on these criteria, four datasets (GSE13083, GSE1420, GSE39491, and GSE36223) were obtained. After annotation and normalization, these datasets were merged to form a combined dataset comprising 78 BE samples and 78 healthy controls. The EAC dataset was obtained from The Cancer Genome Atlas (TCGA) database ( https://portal.gdc.cancer.gov/analysis_page?app=CohortBuilder&tab=general ), which include 81 EAC samples and 13 healthy controls. After annotation and normalization, the data were processed for bioinformatics analysis.Batch effects across samples were corrected using the “ComBat” algorithm from the R package “sva”. And variance-stabilizing transformation (VST) was performed using the “DESeq2” package. For subsequent analysis, both the BE and EAC datasets were randomly divided into a 70% training set and a 30% validation set. Principal component analysis (PCA) was then conducted using the “tinyarray” and “ggplot2” packages, and the results were visualized as PCA plots (Fig. 2 a and b). Differentially expressed genes analysis All feature selection procedures, including differential gene expression analysis, protein-protein interaction network analysis, immune infiltration assessment, and machine learning-based selection, were performed exclusively on the training sets (70% of the normalized BE and EAC cohorts). The hold-out sets (30%) were used solely for final performance evaluation. To identify differentially expressed genes (DEGs) between disease groups (BE and EAC) and healthy controls, we performed differential expression analysis using both the “limma” and “DESeq2” packages on the processed expression data. Significant DEGs were defined with a threshold of adjusted p-value 1.5 to ensure the inclusion of biologically relevant genes. Volcano plots were generated to visualize the identified DEGs. Subsequently, shared differentially expressed genes (SDEGs) between BE and EAC were identified by taking the intersection of the DEG sets using Venn diagrams implemented in TBtools. Gene ontology and pathway enrichment analyses Functional and pathway enrichment analyses of the SDEGs were performed using R with a significance threshold of p-value < 0.05. Within the Gene Ontology (GO) framework, genes were systematically classified and functionally annotated three primary categories: biological processes (BP), cellular components (CC), and molecular functions (MF)[ 11 ]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) database enables systematic analysis of gene functions and the exploration of potential pathways through integrated genomic information[ 12 ][ 13 ]. Using the R package "clusterProfiler", we performed GO and KEGG enrichment analyses on these SDEGs. Significantly enriched pathways were selected based on thresholds of p-value < 0.05 and q-value < 0.05. The top 10 most significantly enriched pathways were visualized using bar plots, bubble plots, and circle plots generated by the "ggplot2" package in R. Protein-protein interaction (PPI) network analysis Based on the identified SDEGs, we constructed a PPI network for BE and EAC using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING v12.0) database ( https://www.string-db.org/ ), with an interaction confidence score threshold set at ≥ 0.7 (the default medium confidence level in STRING) for filtering interactions(p-value < 0.05). Employing the Maximal Clique Centrality (MCC) algorithm, we ranked genes by their centrality within the network. Protein interactions were visualized with “Cytoscape v3.10.2” software, and its “cytoHubba” plugin was utilized to identify the top 10 hub genes in the PPI network. These hub genes were subsequently integrated with genes obtained from functional enrichment analysis. Immune infiltration and correlation analysis To elucidate immune cell infiltration patterns in BE, EAC, and healthy control samples, we employed the computational tool “CIBERSORT” to quantify the relative abundance of 22 human immune cell subtypes (p-value < 0.05). Using the LM22 signature matrix, we applied the CIBERSORT algorithm to the intersection of genome-wide profiles between BE and EAC cohorts, comprising 13435 shared genes. This gene set fully encompasses all 547 immune-related genes present in the LM22 signature matrix. With selection criteria of false discovery rate (FDR) 0.05 between tumor and normal samples, thereby we identified shared immune-related SDEGs. Subsequent correlation analysis was performed using the R package "pheatmap" to examine associations between these genes and the 22 immune cell types. The results were visualized as a correlation heatmap generated with the "ggplot2" package in R. Finally, we integrated the shared immune-related genes from BE and EAC with the previously identified hub genes and genes derived from functional enrichment analysis. Machine learning and feature genes selection To optimize feature genes selection, we implemented two machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO) regression and Random Forest (RF). LASSO employs L1 regularization to compress the variable space, making it suitable for sparse feature selection in high-dimensional data, whereas RF effectively captures non-linear relationships and interactions among variables through the integration of multiple decision trees. Their combination can thereby complement and optimize the biomarker identification pipeline. Using the six candidate genes previously identified through functional enrichment, PPI, and immune infiltration analysis as inputs, LASSO and RF analyses were performed separately. The RF model was implemented using the R package “randomForest” [ 14 ]. The number of trees (ntree) was set to 500, and the number of features considered at each split (mtry) was set to √p, where p represents the total number of features minus one. Bootstrap sampling was enabled with out-of-bag (OOB) samples retained for every tree. Model stability was evaluated by plotting the OOB error against the number of trees. Feature importance was ranked based on the MeanDecreaseGini (also known as IncNodePurity) metric [ 15 ]. To determine the feature selection threshold, a stepwise elimination strategy was applied: features with lower importance were removed sequentially, and the corresponding OOB error was computed after each removal. The relationship between the number of retained features and the OOB error was analyzed to identify a critical subset of features. Additionally, LASSO regression was also employed for feature selection. The LASSO regression was carried out using the R package “glmnet”[ 16 ] with 10-fold cross-validation to determine the optimal penalty coefficient λ (lambda). Genes with non-zero coefficients at the selected λ value were ultimately identified as potential feature genes. This procedure represents the standard approach for selecting the λ value, aimed at minimizing the cross-validation error of the model[ 17 ]. Ultimately, the overlapping genes identified by both algorithms were selected as a core set of feature genes for subsequent validation of their diagnostic potential. All key parameters, filtering thresholds, and dataset partitioning information related to the machine learning analyses in this study are summarized in Supplementary Table S1 . Logistic regression validation To assess the clinical diagnostic value of these potential feature genes identified by LASSO regression and RF algorithms, we partitioned the BE and EAC datasets into a 70% training set and a 30% validation set. Subsequently, we employed logistic regression modeling based on the selected feature genes and evaluated its performance using the validation set. The model was visualized through diagnostic plots. Using the validation set, we generated the receiver operating characteristic (ROC) curves with the R package “pROC” and calculated the area under the curve (AUC) to evaluate the model’s predictive performance. The contribution of the model to disease risk was visualized using nomograms generated by the R package “nomogramFormula”. To assess model calibration, we performed the Hosmer-Lemeshow goodness-of-fit test (H-L test) and generated calibration curves using the R package “ResourceSelection”, which validated the model's discrimination and calibration. Detailed performance metrics are summarized in Supplementary Table S2 and S3. Results Identification of shared differentially expressed genes Differential expression analysis of BE and EAC was performed using thresholds of adjusted p-value 1.5. The identified DEGs were visualized in volcano plots (Fig. 2 c and d), while heatmaps illustrated expression patterns across multiple samples and genes (Figs. 2 e and f).The analysis revealed 609 DEGs in BE (352 upregulated, 257 downregulated) and 4,042 DEGs in EAC (1,618 upregulated, 2,424 downregulated). Intersection analysis of both disease groups identified 101 shared differentially expressed genes (SDEGs), visualized in a Venn diagram (Fig. 2 g). Complete gene lists are summarized in Supplementary Table S4. GO and KEGG enrichment analyses results The results of GO and KEGG enrichment analyses were visualized using bar plots and bubble plots to represent the significantly enriched terms and pathways. The GO enrichment analysis of the 101 SDEGs revealed significant associations with multiple processes (Fig. 3 a and b). In the biological process (BP) category, these genes were predominantly enriched in digestive functions (13 genes), gastrointestinal epithelium maintenance, epithelial structure maintenance, non-protein amino acid biosynthesis, cellular aldehyde metabolism, and sodium/potassium ion homeostasis; For cellular components (CC), the SDEGs were primarily localized to the apical plasma membrane, cell apical part, Golgi lumen, collagen-containing extracellular matrix, granzyme, cornified envelope; molecular function (MF) analysis demonstrated these genes were significantly involved in branched-chain amino acid transmembrane transporter activity, glycosaminoglycan binding, ATPase-coupled transmembrane transporter activity, extracellular matrix structural constituency. These findings collectively indicate that the identified SDEGs are functionally associated with: nutrient digestion and absorption, maintenance of epithelial barriers, transmembrane transport-driven ion, and molecular homeostasis.KEGG pathway analysis revealed significant enrichment patterns among the SDEGs (Fig. 3 c and d). Five genes were notably enriched in the digestive system-related pathway "gastric acid secretion." Additionally, these SDEGs demonstrated prominent enrichment in several key pathways including "Collecting duct acid secretion", "Nitrogen metabolism", "Amino acid biosynthesis", "IL-17 signaling pathway", "Rheumatoid arthritis", etc. The core commonalities among these pathways suggest their involvement in the regulation of inflammatory responses, immune functions, and metabolic processes. By integrating the results from both GO and KEGG analyses, we identified 13 pathway-associated SDEGs that showed consistent enrichment across these functional categories (with the threshold of p - value < 0.05). PPI network analysis and Hub genes selection To identify shared molecular signatures between BE and EAC, we selected shared SDEGs based on stringent criteria (adjusted p - value 1.5). This approach ensured a focus on genes with significant differential expression, which are more likely to play pivotal roles in BE and EAC pathogenesis. After that, we delved into the exploration of their intricate PPI network. To ensure the reliability and relevance of the PPI network analysis, we removed nodes that were either unconnected or weakly connected to the main network, which might represent spurious interactions or genes with limited functional relevance in the context of BE and EAC. Subsequently, we used the “CytoHubba” plugin, a well-established tool in network analysis, to identify critical nodes and subnetworks within the refined PPI network. Through this in-depth analysis, we were able to pinpoint the top 10 hub genes, namely VIL1 , CDH17 , MUC2 , OLFM4 , MUC13 , ATP4B , GKN1 , LIPF , ATP4A , and PGC (Fig. 3 e and f). These hub genes are likely to serve as key regulators in the molecular pathways underlying BE and EAC, given their central positions in the PPI network. Ultimately, by the combination of the 10 hub genes with 13 genes from functional enrichment analyses, we identified a set of 18 core regulatory SDEGs strongly implicated in the development and progression of BE and EAC, including MUC4 , MDK , GCNT3 , ADRA2A , MUC2 , VIL1 , SST , CAPN9 , MUC13 , PBLD , SI , PGC , GKN1 , CDH17 , OLFM4 , ATP4B , LIPF , and ATP4A . Immune infiltration and correlation analysis Given the significant enrichment of SDEGs in immune-related pathways identified in functional analyses, we further investigated these genes from an immune perspective. Box plots demonstrated significantly elevated immune cell infiltration in both BE and EAC groups compared to healthy controls (Fig. 4 a and b). Analysis of 22 immune cell subtypes revealed distinct patterns: in BE, increased plasma cells and mast cells but decreased T lymphocytes and dendritic cells compared to controls; in EAC, increased T lymphocytes, macrophages, and dendritic cells but reduced B lymphocytes, plasma cells, and mast cells relative to healthy controls (Fig. 4 c and d). Applying stringent thresholds (FDR 0.05), we identified 100 immune-related SDEGs in BE and 40 immune-related SDEGs in EAC (p - value < 0.05). The intersection of the two yielded 40 immune-associated DEGs shared by BE and EAC (Fig. 4 e). Correlation heatmaps illustrated associations between these 40 immune-related SDEGs and various immune cell types (Fig. 5 a and b). By integrating these with the previously identified 18 core regulatory genes, we identified 6 candidate feature genes: GCNT3 , ADRA2A , MUC13 , PBLD , SI , and OLFM4 (Fig. 5 c). Their distinct correlation patterns with different immune cell populations are shown in Fig. 5 d and e. Machine learning modeling to screen for signature genes The LASSO and RF algorithms were employed to screen the final characteristic genes from the six candidate key genes. For the LASSO regression, genes with non-zero coefficients were selected at the optimal λ value (λ =0.019 for BE, λ = 0.012 for EAC)(Fig. 6 a and b). This resulted in the identification of two genes ( MUC13 and PBLD) for BE (Fig. 6 c) and four genes ( ADRA2A , MUC13 , PBLD , and OLFM4) for EAC (Fig. 6 d). In the RF model, the OOB error decreased progressively with an increasing number of trees and stabilized between 400 and 500 trees (Fig. 7 c and g). Feature importance was then ranked by the mean decrease Gini index (Fig. 7 a, b, e and f). Using a stepwise feature elimination strategy, we was employed evaluated how different feature subsets affected the OOB error. The results showed a clear inflection point in the OOB error as features were sequentially removed (Fig. 7 d and h). Based on this criterion, two genes ( PBLD and MUC13) were selected as the most important features for BE, and three genes ( MUC13 , PBLD , and GCNT3) for EAC. Finally, by integrating the results from both algorithms, we identified two shared feature genes: MUC13 and PBLD (Fig. 7 i). Differential expression analysis indicated that both genes were up-regulated in BE, while in EAC, MUC13 was up-regulated and PBLD was down-regulated. Logistic regression validation To preliminarily evaluate the diagnostic performance of the feature genes MUC13 and PBLD , we performed logistic regression analysis on the internal validation set, which comprised 30% of samples randomly selected from the normalized and merged dataset. The discriminatory ability of the model was assessed by plotting ROC curves and calculating the AUC values, which range from 0.5 (random guessing) to 1 (perfect classification). The internal validation results showed AUC values of 0.94 (95% CI: 0.858–1.000) for BE and 0.98 (95% CI: 0.957–1.000) for EAC (Fig. 8 a and b), indicating that the diagnostic model based on these two feature genes exhibits high sensitivity and specificity. We also constructed nomograms to visualize the model's ability to predict individual disease risk (Fig. 8 c and d). The substantial contributions of both genes to disease risk, as indicated by their positions on the horizontal axes, confirm the model's strong predictive power. To evaluate the calibration between predicted probabilities and observed outcomes, we performed the Hosmer-Lemeshow (H-L) goodness-of-fit test (detailed metrics are provided in Supplementary Table S2) and plotted calibration curves (Fig. 8 e and f). The H-L test showed no significant deviation (BE: p > 0.9; EAC: p > 0.9), and the calibration curves closely followed the diagonal reference line, indicating good model fit and high calibration accuracy, with predicted risks aligning well with actual observations. These validation results demonstrate that the diagnostic model based on MUC13 and PBLD has favorable diagnostic value and potential for distinguishing EAC arising from BE. This suggests that both genes may play critical roles in BE-to-EAC progression and could serve as biomarkers for early diagnosis and targeted intervention. Discussion Esophageal adenocarcinoma (EAC) is a highly lethal malignancy, with studies projecting a continued global increase in its incidence in the coming years[ 18 ]. EAC typically presents with nonspecific early symptoms, and most patients are diagnosed at advanced stages. Although Barrett's esophagus (BE) is the only known precursor lesion of EAC, reliable risk stratification tools and clinical monitoring strategies remain limited. Therefore, the development of novel early diagnostic biomarkers is critical to advancing precision prevention and early therapeutic intervention for esophageal cancer. In this study, we employed multi-modal bioinformatics approaches to systematically investigate shared molecular pathways in BE and EAC pathogenesis. Differential expression analysis of training datasets identified 101 SDEGs common to both conditions. Subsequent functional annotation through GO and KEGG enrichment analyses, combined with PPI network analysis, revealed 10 hub genes. Integration of these results yielded a core set of 18 key regulatory genes. Furthermore, immune infiltration analysis identified 40 shared immune-related SDEGs. Intersection analysis of these gene sets ultimately pinpointed six high-priority candidate feature genes: GCNT3 , ADRA2A , MUC13 , PBLD , SI , and OLFM4 . Through machine learning approaches employing LASSO regression and RF algorithms, we identified two feature genes: MUC13 and PBLD . Subsequent validation using the hold-out sets (30% of normalized sets) demonstrated strong diagnostic performance, as evidenced by: ROC curve analysis showing high discriminative power (AUC = 0.94 for BE, and AUC = 0.98 for EAC), nomogram analysis confirming accurate disease risk prediction, and Hosmer-Lemeshow goodness-of-fit test indicating good calibration (close alignment with ideal curves). These comprehensive validation results confirm the accuracy and clinical diagnostic potential of this genetic model. GO and KEGG enrichment analyses of SDEGs in BE and EAC, combined with PPI network analysis, revealed significant associations with inflammatory responses, immune regulation, and epithelial barrier function. Both BE and EAC develop from reflux-induced esophageal mucosal inflammation, with established inflammation-to-cancer transformation mechanisms. Cytokine storms and proinflammatory microenvironments play pivotal roles in this process, where interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF- α) promote immune cell chemotaxis through nuclear factor kappa B (NF-κB) signaling pathway activation[ 19 , 20 ]. The interplay between the prostaglandin/epithelial growth factor/peroxisome proliferator-activated receptor-γ (PG/EGF/PPARγ) system and proinflammatory cytokines, TNFα and IL-1β, drives the progression from chronic inflammation to BE and subsequent EAC[ 21 ]. The reciprocal activation between the interleukin-6/signal transducer and activator of transcription 3 (IL-6/STAT3) and NF-κB signaling pathways fosters an immunosuppressive microenvironment and accelerates the malignant progression of BE[ 22 ]. Reflux-induced dysbiosis promotes EAC progression by amplifying procarcinogenic signals through the toll-like receptor (TLR) /NF-κB/tumor protein p53 (TP53) signaling cascade while suppressing antitumor immune responses[ 23 ]. Studies have demonstrated aberrant activation of the Wingless/Integrated (Wnt)/β-catenin pathway during BE carcinogenesis[ 24 ]. Hyperactive Wnt/β-catenin signaling induces immune exclusion and impairs T cell-mediated antitumor immunity, while enhancing resistance to immunotherapy[ 25 ]. Recent evidence indicates that poly ADP-ribose glycohydrolase (PARG) activates the Wnt/β-catenin pathway to promote epithelial-mesenchymal transition (EMT), thereby facilitating EC cell growth, metastasis, and invasion[ 26 ]. Research indicates that chronic inflammatory exposure in the esophagus promotes EMT process through AP endonuclease 1 (APE1) redox-dependent E-cadherin cleavage, establishing a persistent positive feedback loop between inflammatory microenvironment and EMT that drives EAC initiation and metastasis[ 27 ]. Additionally, acid exposure activates the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) pathway, enhancing proliferation and survival while suppressing apoptosis in human Barrett's adenocarcinoma cell line (SEG-1), thereby accelerating BE malignant progression[ 28 ]. These established mechanisms of BE-to-EAC progression are consistent with the functional enrichment results observed in our SDEG analysis. The progression from BE to EAC represents a classic example of inflammation–cancer transformation. Interestingly, key pathways including Wnt/β-catenin, NF-κB, and EMT do not operate in isolation, but rather form a highly interconnected and mutually reinforcing signaling network. This robust tripartite alliance collaboratively drives malignant progression. (i) The inflammatory microenvironment induced by chronic reflux serves as the central driver of this process. Refluxates such as bile acids and gastric acid continuously stimulate the lower esophageal mucosa, activating the NF-κB signaling pathway and upregulating a series of pro-inflammatory cytokines (e.g., TNF-α, IL-1β, IL-6, IL-8) and cyclooxygenase-2 (COX-2). This creates a microenvironment rich in growth factors and genotoxic reactive oxygen species (ROS), leading to DNA damage, promoting epithelial cell proliferation, inhibiting apoptosis, and facilitating the activation of other oncogenic pathways. These events represent the initial steps of inflammation–cancer transformation[ 25 ][ 29 , 30 ]. (ii) The extensive bidirectional crosstalk between NF-κB and Wnt signaling represents a critical mechanism in the malignant transformation of BE. Inflammatory cytokines such as TNF-α upregulate the expression of Wnt ligands (e.g., Wnt 1, Wnt 2) while suppressing Wnt pathway antagonists (e.g., DKK1) in an NF-κB-dependent manner. This leads to nuclear accumulation and transcriptional activation of β-catenin. Activated β-catenin physically interacts with key components of the NF-κB pathway, such as p65/RelA, thereby enhancing NF-κB transcriptional activity. This interaction establishes a positive feedback loop that perpetuates pro-inflammatory and pro-proliferative signaling[ 31 ]. (iii) EMT is a critical step in the malignant progression of BE, conferring invasive and metastatic capabilities. Both Wnt/β-catenin and NF-κB signaling pathways are centrally involved in the core regulation of EMT and act synergistically to drive this process. The NF-κB complex and β-catenin/TCF complex can cooperatively bind to the promoter regions of EMT-related transcription factors such as SNAIL, SLUG, and ZEB1, thereby upregulating their expression[ 32 , 33 ]. These transcription factors subsequently suppress epithelial markers (e.g., E-cadherin) and promote mesenchymal markers (e.g., Vimentin, N-cadherin), initiating the EMT program. Concurrently, under conditions of oxidative stress, APE1 cleaves E-cadherin, compromising epithelial integrity, while factors such as PARG further enhance β-catenin activity. Additionally, reflux-related acid and bile acid stimuli induce EMT activation through NF-κB and MAPK/ERK pathways[ 26 – 28 ], collectively accelerating the EMT process and promoting tumor cell invasion. Given the pivotal involvement of inflammation-immune regulation and tumor metastasis pathways in BE- to- EAC progression, we further investigated the functional mechanisms underlying the two potential biomarkers, MUC13 and PBLD . Mucin 13 ( MUC13 ), a member of the transmembrane mucin family, is typically expressed at low levels across epithelial surfaces of the gastrointestinal tract, respiratory system, and reproductive tract, where it contributes to barrier function maintenance. Substantial evidence indicates that MUC13 overexpression disrupts normal cellular polarity, promoting malignant transformation through enhanced tumorigenesis, metastasis, invasive capacity, and colony formation[ 34 ]. Studies have demonstrated aberrant overexpression of MUC13 across multiple cancer types, where it significantly correlates with adenocarcinoma initiation, progression, and poor clinical outcomes[ 35 ]. The mechanistic investigations are listed as follows: In digestive system cancers, in esophageal cancer (EC), immunohistochemical (IHC) analysis revealed significant overexpression of MUC13 in tumor tissues compared with adjacent non-tumor tissues. Silencing MUC13 expression suppressed cell proliferation, reduced colony formation, induced cell cycle arrest, and promoted apoptosis. It also markedly decreased tumor volume and tumor weight. These findings suggest that MUC13 acts as a key regulator of O-glycosylation processes and contributes to esophageal cancer progression[ 36 ]. MUC13 promotes cancer progression by activating the Wnt/β-catenin pathway in colorectal cancer (CRC)[ 37 , 38 ] and hepatocellular carcinoma (HCC)[ 34 ], enhancing cell proliferation, EMT, immune evasion, and anti-tumor immune responses; additionally, MUC13 suppresses apoptosis and facilitates metastasis in colon cancer cells by augmenting TNF-induced NF-κB activation[ 39 ]. In pancreatic cancer, MUC13 activates pro-oncogenic signaling by modulating human epidermal growth factor receptor-2 (HER2) kinase activity, inducing NF-κB Inhibitory Protein (IκB ) phosphorylation and NF-κB p65 nuclear translocation[ 40 ]. In gastric cancer (GC), both miR-361-3p and miR-132-3p can suppress tumor progression by targeting and inhibiting MUC13 expression[ 41 , 42 ]; furthermore, Takahiro Shimamura et al investigated MUC13 expression in gastric tissues using immunohistochemistry and RT-PCR, demonstrating its progressive upregulation across different tissue types (absent in normal gastric mucosa, detectable in intestinal metaplasia, and highly expressed in 64.9% of GC specimens); and, notably, MUC13 showed particularly prominent overexpression in the intestinal-type GC[ 43 ]. In lung cancer, MUC13 promotes tumor progression by activating the MAPK/ERK signaling pathway[ 44 ]. In esophageal squamous cell carcinoma (ESCC), Wang H et al performed immunohistochemical analysis of 186 ESCC specimens from patients receiving neoadjuvant chemotherapy and demonstrated that the MUC13/MUC20 combination may serve as a potential prognostic biomarker for ESCC patients following neoadjuvant chemoradiotherapy[ 45 ]. In ovarian cancer, studies have demonstrated elevated MUC13 expression in tumor tissues; MUC13 -transfected cell lines exhibited malignant phenotypes and xenograft models showed upregulated HER2, p21-Activated Kinase 1 (PAK1), and Tumor Protein p38 Mitogen-Activated Protein Kinase (p38) signaling pathways coupled with enhanced tumor growth in response to MUC13 expression[ 46 ]. These findings collectively establish MUC13 's oncogenic role and promoting cancer cell proliferation, invasion/ metastasis. The phenazine biosynthesis-like domain-containing protein ( PBLD ), the sole representative of the phenazine biosynthesis-like protein family, has been shown to exhibit reduced expression in multiple cancers. Accumulating evidence indicates that PBLD functions as a tumor suppressor across various cancer types, ameliorating inflammatory responses, enhancing immune responses, and suppressing tumorigenesis and invasion. In the inflammation-cancer transformation of colorectal tissues, studies have revealed that vitexin modulates macrophage polarization through the Vitamin D Receptor (VDR)/ PBLD pathway, thereby attenuating malignant progression in chronic colitis[ 47 ]. Chen S et al have demonstrated reduced mRNA and protein levels of PBLD in ulcerative colitis (UC) tissues, and PBLD deficiency has been found to exacerbate immune cell infiltration in intestinal epithelial cells, while mechanistic studies have revealed that PBLD ameliorates intestinal inflammation and enhances barrier function through suppression of NF-κB signaling transduction[ 48 ]; further studies have demonstrated that rapamycin promotes intestinal barrier repair in UC through activation of the Mechanistic Target of Rapamycin (mTOR)/ PBLD /Angiomotin (AMOT) signaling pathway[ 49 ]. In gastric cancer (GC), studies have confirmed that PBLD suppresses tumor progression and invasion by inhibiting the Transforming Growth Factor β (TGF-β) signaling-mediated differentiation process and EMT program[ 50 ]. In hepatocellular carcinoma (HCC), PBLD has been identified as a critical regulator of tumor angiogenic microenvironment remodeling, exerting its inhibitory effects on HCC angiogenesis through the Extracellular Signal-Regulated Kinase (ERK)/Hypoxia-Inducible Factor 1α (HIF-1α)/Vascular Endothelial Growth Factor (VEGF) signaling axis[ 51 ]; furthermore, studies have demonstrated that the PBLD activator, Cedrelone, suppresses EMT phenotypes and HCC progression via PBLD upregulation, while PBLD overexpression promotes apoptosis through activation of Ras and Ras-proximate-1 (Rap1) signaling pathways[ 52 ]. In breast cancer, studies have revealed that Circular RNA Lysine Demethylase 4C (circKDM4C) suppresses tumor progression and attenuates doxorubicin resistance by modulating the miR-548p/ PBLD axis[ 53 ]. Furthermore, the clinical translational value of MUC13 and PBLD as potential biomarkers for BE-EAC progression lies primarily in their feasibility across two commonly used diagnostic modalities: tissue biopsy and liquid biopsy. The most straightforward strategy, compatible with current BE surveillance practices, involves detecting their protein expression via immunohistochemical (IHC) analysis in endoscopic biopsy samples. As a transmembrane mucin, MUC13 is technically suitable for IHC detection. Although direct evidence in EAC remains to be further substantiated, existing studies have demonstrated its consistent detection in FFPE tissue specimens of various gastrointestinal adenocarcinomas—including esophageal[ 36 ] and colorectal cancer[ 39 ]—indicating that the associated antibody-based detection system is reasonably reliable. Similarly, PBLD protein has also been validated in FFPE tissues from other cancer types such as hepatocellular carcinoma[ 52 ][ 54 ] and breast cancer[ 53 ]. These studies from other malignancies provide a methodological reference for preliminary assay development. Therefore, future exploration of a dual-marker IHC assay based on MUC13 and PBLD for esophageal biopsy specimens may represent a translatable hypothesis worthy of further validation in prospective clinical cohorts. In summary, current evidence suggests that MUC13 is frequently highly expressed across multiple malignant tumors, including esophageal cancer and other adenocarcinoma subtypes, whereas PBLD often shows reduced expression. In other cancer types, they have been implicated in potential oncogenic and tumor-suppressive functions, respectively. However, their specific roles in Barrett’s esophagus and esophageal adenocarcinoma remain limited. It is noteworthy that several signaling pathways associated with these genes—such as Wnt/β-catenin, NF-κB, MAPK/ERK, and EMT—have been widely reported to play critical roles in the malignant progression from BE to EAC. These include NF-κB-mediated regulation of the inflammatory-immune microenvironment[ 19 – 23 ], aberrant activation of the Wnt/β-catenin pathway[ 24 – 26 ], the EMT process[ 27 ], MAPK/ERK-driven proliferation and anti-apoptotic mechanisms[ 28 ], as well as crosstalk among these pathways[ 29 – 33 ]. Combined with the differential expression results from this study—which showed significant upregulation of MUC13 in both BE and EAC tissues, while PBLD was mildly elevated in non-dysplastic BE but markedly downregulated in EAC—we propose that MUC13 and PBLD may act as key regulatory molecules during BE-EAC progression, potentially through interactive mechanisms involving the aforementioned pathways. This hypothesis warrants further experimental validation. Such efforts could provide new research directions for early diagnosis, mechanistic insight, and targeted intervention in BE malignant transformation. Study Limitations This study identified a two-gene diagnostic signature through integrated bioinformatic analyses and demonstrated good performance in internal validation (AUC: 0.94–0.98). However, two key limitations must be emphasized. First, all reported performance metrics are derived from internal validation only. The limited number of normal samples in the EAC cohort increases the risk of overfitting and may have led to optimistic performance estimates. Second, the study lacks experimental validation. The biological and diagnostic relevance of MUC13 and PBLD, while supported by bioinformatic evidence and previous literature, requires confirmation through experimental methods such as immunohistochemistry, Western blot, or functional assays, as well as validation in independent external cohorts. Therefore, our findings should be regarded as preliminary yet highly promising. The clinical translational potential of these two gene biomarkers warrants further confirmation through well-designed multi-center prospective studies and external validation efforts. Conclusion In conclusion, our integrated bioinformatic study identifies MUC13 and PBLD as potential diagnostic biomarker for Barrett's esophagus (BE) and esophageal adenocarcinoma (EAC). Their strong discriminatory power throughout the BE-to-EAC sequence, coupled with their involvement in regulating the inflammatory-immune microenvironment and tumor invasion, suggests their potential involvement in malignant progression. These findings provide a foundation for developing early diagnostic, risk stratification and management strategies for BE patients, as well as for elucidating the molecular mechanisms of EAC progression. Further validation of these biomarkers in independent external cohorts and through functional assays is needed to confirm their clinical utility and translational value for improving patient outcomes. Declarations Acknowledgements We are grateful to public databases for providing valuable data, and we sincerely appreciate the contributions and efforts of R software and R package developers for their convenience. Funding statement This research was supported by the Hunan Provincial Natural Science Foundation Science and Health Joint Project, grant number 2023JJ60044; Hunan Province Traditional Chinese Medicine Research Program Project, grant number B2023079; Hunan Province Clinical Medical Technology Innovation Guidance Project, grant number 2021SK51413; Excellent Youth Project of the Research Program of Hunan Provincial Department of Education, grant number 21B0389. The APC was funded by the Project for Advancing Clinical Evidence-Based Capacity of Traditional Chinese Medicine in Treating Advantageous Diseases, grant number czxm-kyb-2025001. Author contributions FH conceived and designed the study. TCS and ZM participated in data collection. LJY and XJ performed data analysis and visualization. ZDX and ZT conducted methodological validation. FH and LJY contributed to manuscript drafting and writing. XY provided overall research guidance and secured funding. All authors critically reviewed the manuscript and approved the final version, taking full responsibility for the content. Competing interests The author(s) declare no competing interests. Ethics approval Not applicable. Availability of data and materials The datasets analysed during the current study are available in Gene Expression Omnibus (GSE13083, GSE1420, GSE39491, and GSE36223) and The Cancer Genome Atlas (TCGA-EAC). These datasets are accessible at https://www.ncbi.nlm.nih.gov/geo/ and https://portal.gdc.cancer.gov/analysis_page?app=CohortBuilder&tab=general. The data used to support the findings of this study are available from the corresponding author upon request. Consent to participate/Consent to publish Not applicable. References Xia C, et al. Cancer statistics in China and United States, 2022: profiles, trends, and determinants. Chin Med J (Engl). 2022;135:584–90. Santucci C, et al. Global trends in esophageal cancer mortality with predictions to 2025, and in incidence by histotype. Cancer Epidemiol. 2023;87:102486. Bray F, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68:394–424. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70:7–30. Obermannová R, et al. Oesophageal cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol. 2022;33:992–1004. Curtius K, Rubenstein JH, Chak A, Inadomi JM. Computational modelling suggests that barrett’s oesophagus may be the precursor of all oesophageal adenocarcinomas. Gut. 2020;70:1435–40. Thrift AP. Global burden and epidemiology of barrett oesophagus and oesophageal cancer. Nat Rev Gastroenterol Hepatol. 2021;18:432–43. Beydoun AS, Stabenau KA, Altman KW, Johnston N. Cancer risk in barrett’s esophagus: a clinical review. Int J Mol Sci. 2023;24:6018. Holmberg D, Ness-Jensen E, Mattsson F, El-Serag HB, Lagergren J. Risk of oesophageal adenocarcinoma in individuals with barrett’s oesophagus. Eur J Cancer. 2017;75:41–6. Weusten BLAM, et al. Diagnosis and management of barrett esophagus: european society of gastrointestinal endoscopy (ESGE) guideline. Endoscopy. 2023;55:1124–46. The Gene Ontology Consortium. The gene ontology resource: 20 years and still GOing strong. Nucleic Acids Res. 2019;47:D330–8. Kanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016;44:D457–462. Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27–30. Ellis K, et al. A random forest classifier for the prediction of energy expenditure and type of physical activity from wrist and hip accelerometers. Physiol Meas. 2014;35:2191–203. Cutler A, Stevens JR. Random forests for microarrays. Methods Enzymol, 422–32 (2006). Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33:1–22. Tibshirani R. The lasso method for variable selection in the cox model. Stat. Med . 16, 385–395 (1997).Liu, C.-Q. Epidemiology of esophageal cancer in 2020 and projections to 2030 and 2040. Thorac. Cancer 14, 3–11 (2023). Liu C-Q, et al. Epidemiology of esophageal cancer in 2020 and projections to 2030 and 2040. Thorac Cancer. 2023;14:3–11. O’Riordan JM, et al. Proinflammatory cytokine and nuclear factor kappa-B expression along the inflammation-metaplasia-dysplasia-adenocarcinoma sequence in the esophagus. Am J Gastroenterol. 2005;100:1257–64. Dunbar KB, et al. Association of acute gastroesophageal reflux disease with esophageal histologic changes. JAMA. 2016;315:2104–12. Majka J, et al. Interaction of epidermal growth factor with COX-2 products and peroxisome proliferator-activated receptor-γ system in experimental rat barrett’s esophagus. Am J Physiol Gastrointest Liver Physiol. 2020;318:G375–89. Dvorak K, Dvorak B. Role of interleukin-6 in barrett’s esophagus pathogenesis. World J Gastroenterol. 2013;19:2307–12. Weh KM, et al. Prebiotic proanthocyanidins inhibit bile reflux-induced esophageal adenocarcinoma through reshaping the gut microbiome and esophageal metabolome. JCI Insight. 2024;9:e168112. Lyros O, et al. Dickkopf-1 (DKK1) promotes tumor growth via akt-phosphorylation and independently of wnt-axis in barrett’s associated esophageal adenocarcinoma. Am J Cancer Res. 2019;9:330–46. Zhou Y, et al. Wnt signaling pathway in cancer immunotherapy. Cancer Lett. 2022;525:84–96. Yan J, Zhou Y, Wang Y, Liu Y. PARG promotes esophagus cancer cell metastasis by activation of the wnt/β-catenin pathway. Biochem Genet. 2024;62:761–74. Lu H, et al. Reflux conditions induce E-cadherin cleavage and EMT via APE1 redox function in oesophageal adenocarcinoma. Gut. 2023;73:47–62. Srinivas BK, Shivamadhu MC, Jayarama S. Musa acuminata lectin exerts anti-cancer effects on HeLa and EAC cells via activation of caspase and inhibitions of akt, erk, and jnk pathway expression and suppresses the neoangiogenesis in in-vivo models. Int J Biol Macromol. 2021;166:1173–87. Ahmad S, et al. Long non-coding RNAs regulated NF-κB signaling in cancer metastasis: micromanaging by not so small non-coding RNAs. Semin Cancer Biol. 2022;85:155–63. Xia Y, Shen S, Verma IM. NF-κB, an active player in human cancers. Cancer Immunol Res. 2014;2:823–30. Ma B, Hottiger MO. Crosstalk between wnt/β-catenin and NF-κB signaling pathway during inflammation. Front Immunol. 2016;7:378. Oh A, et al. NF-κB signaling in neoplastic transition from epithelial to mesenchymal phenotype. Cell Commun Signal : CCS. 2023;21:291. Solanas G, et al. E-cadherin controls beta-catenin and NF-kappaB transcriptional activity in mesenchymal gene expression. J Cell Sci. 2008;121:2224–34. Dai Y, et al. Overexpression of MUC13, a poor prognostic predictor, promotes cell growth by activating wnt signaling in hepatocellular carcinoma. Am J Pathol. 2018;188:378–91. Jonckheere N, Skrypek N, Van Seuningen I. Mucins and tumor resistance to chemotherapeutic drugs. Biochim Biophys Acta. 2014;1846:142–51. Han Y, et al. MUC13 promotes the development of esophageal cancer by upregulating the expression of O-glycan process-related molecules. Discov Oncol. 2023;14:123. Chen X, et al. RUNX1-MUC13 interaction activates wnt/β-catenin signaling implications for colorectal cancer metastasis. Int J Biol Sci. 2024;20:4999–5026. Doxtater K, et al. MUC13 drives cancer aggressiveness and metastasis through the YAP1-dependent pathway. Life Sci Alliance. 2023;6:e202301975. Sheng YH, et al. MUC13 protects colorectal cancer cells from death by activating the NF-κB pathway and is a potential therapeutic target. Oncogene. 2017;36:700–13. Kumari S, et al. MUC13 contributes to rewiring of glucose metabolism in pancreatic cancer. Oncogenesis. 2018;7:19. Cai T, Peng B, Hu J, He Y. Long noncoding RNA BBOX1-AS1 promotes the progression of gastric cancer by regulating the miR-361-3p/mucin 13 signaling axis. Bioengineered. 2022;13:13407–21. He L, Qu L, Wei L, Chen Y, Suo J. Reduction of miR-132-3p contributes to gastric cancer proliferation by targeting MUC13. Mol Med Rep. 2017;15:3055–61. Shimamura T, et al. Overexpression of MUC13 is associated with intestinal-type gastric cancer. Cancer Sci. 2005;96:265–73. Pang Y, et al. MUC13 promotes lung cancer development and progression by activating ERK signaling. Oncol Lett. 2022;23:37. Wang H, et al. The expression and prognostic significance of mucin 13 and mucin 20 in esophageal squamous cell carcinoma. J Cancer Res Ther. 2015;11(Suppl 1):C74–79. Malik S, et al. Emerging role of mucins in antibody drug conjugates for ovarian cancer therapy. J Ovarian Res. 2024;17:161. Chen Y, et al. Discovery of vitexin as a novel VDR agonist that mitigates the transition from chronic intestinal inflammation to colorectal cancer. Mol Cancer. 2024;23:196. Chen S, et al. Epithelial PBLD attenuates intestinal inflammatory response and improves intestinal barrier function by inhibiting NF-κB signaling. Cell Death Dis. 2021;12:563. Xu Y, et al. Rapamycin promotes the intestinal barrier repair in ulcerative colitis via the mTOR/PBLD/AMOT signaling pathway. Biochim Biophys Acta Mol Basis Dis. 2024;1870:167287. Li D, et al. Mitogen-activated protein kinase activator with WD40 repeats (MAWD) and MAWD-binding protein induce cell differentiation in gastric cancer. BMC Cancer. 2015;15:637. Han L, et al. PBLD inhibits angiogenesis via impeding VEGF/VEGFR2-mediated microenvironmental cross-talk between HCC cells and endothelial cells. Oncogene. 2022;41:1851–65. Wu J, Niu Q, Yuan J, Xu X, Cao L. Novel compound cedrelone inhibits hepatocellular carcinoma progression via PBLD and ras/Rap1. Exp Ther Med. 2019;18:4209–20. Correction. circKDM4C suppresses tumor progression and attenuates doxorubicin resistance by regulating miR-548p/PBLD axis in breast cancer - PubMed. https://pubmed.ncbi.nlm.nih.gov/33767441/ Li A, et al. Decreased expression of PBLD correlates with poor prognosis and functions as a tumor suppressor in human hepatocellular carcinoma. Oncotarget. 2016;7:524–37. Additional Declarations No competing interests reported. Supplementary Files SupplementaryTables.pdf Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8004374","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":553746426,"identity":"a741a0e4-3c4e-4a94-991d-10dc0cdd2eb5","order_by":0,"name":"Hui Feng","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Hui","middleName":"","lastName":"Feng","suffix":""},{"id":553746427,"identity":"65159552-aae6-4a84-8778-f111a33c2eb1","order_by":1,"name":"Junyao Liang","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Junyao","middleName":"","lastName":"Liang","suffix":""},{"id":553746428,"identity":"f4a8a043-ab13-4f1f-b708-6cad0042c196","order_by":2,"name":"Tao Zhou","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Tao","middleName":"","lastName":"Zhou","suffix":""},{"id":553746429,"identity":"28e822dd-584c-4b1a-901f-35d71d545b5c","order_by":3,"name":"Jie Xiao","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jie","middleName":"","lastName":"Xiao","suffix":""},{"id":553746430,"identity":"7e61826a-6615-4f7b-b40b-b955852f8cc6","order_by":4,"name":"Dexu Zhang","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Dexu","middleName":"","lastName":"Zhang","suffix":""},{"id":553746431,"identity":"e4f88f9a-0dfa-4807-be04-c6b3a8801bda","order_by":5,"name":"Man Zhou","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Man","middleName":"","lastName":"Zhou","suffix":""},{"id":553746432,"identity":"dec459bc-9a0a-4af7-8213-cd7aa358dc09","order_by":6,"name":"Chusen Tang","email":"","orcid":"","institution":"Hunan University of Traditional Chinese Medicine","correspondingAuthor":false,"prefix":"","firstName":"Chusen","middleName":"","lastName":"Tang","suffix":""},{"id":553746433,"identity":"0f86627f-67a1-403d-ba35-fc7196b089e0","order_by":7,"name":"Yin Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYBACeWbGhgMJFTYJUD4zYS2G7cyNBx6cSSNBC8N59uaDD9sOk6CFsRnksDPn88wlkp9uYKiwTmxgP3sArxZ2iF9uF1vOSDO7wXAmPbGBJy8BrxaoLbcTN9zIYbvB2HY4sUGCxwC/yw4DtSS2nYNq+Ue8lgNQLQ1EaDGEOCw5ccOZZ2Y3Eo6lG7fx5ODXIs9//PHHHxV2iRuOJz+78aHGWraf/QwBh6GABCBmI0H9KBgFo2AUjAIcAAA/S1Fblb3Z1wAAAABJRU5ErkJggg==","orcid":"","institution":"The First Hospital of Hunan University of Chinese Medicine","correspondingAuthor":true,"prefix":"","firstName":"Yin","middleName":"","lastName":"Xu","suffix":""}],"badges":[],"createdAt":"2025-11-01 09:08:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8004374/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8004374/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97371133,"identity":"bc51cd51-dc31-4386-bb42-c58bfb10097d","added_by":"auto","created_at":"2025-12-03 16:28:25","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":181076,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/d18d1a2078f129959b367b60.docx"},{"id":97349977,"identity":"1ec6d5bf-7e08-4a17-84b4-562f2173e03a","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1282919,"visible":true,"origin":"","legend":"","description":"","filename":"Fig1..jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/6a99bbb67d4a21c81c010def.jpg"},{"id":97349986,"identity":"913f8110-0a10-41b8-94a4-e76335a47162","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7216781,"visible":true,"origin":"","legend":"","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/662a2a64b32005def9f5fc26.jpg"},{"id":97371478,"identity":"d6e3b707-729e-4fb1-8847-8be6a702c541","added_by":"auto","created_at":"2025-12-03 16:29:01","extension":"jpg","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4639941,"visible":true,"origin":"","legend":"","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/ad0a6642414e76d61385494f.jpg"},{"id":97371176,"identity":"890c0d49-2bf4-49af-b7b6-05d0827aeb97","added_by":"auto","created_at":"2025-12-03 16:28:29","extension":"jpg","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1901162,"visible":true,"origin":"","legend":"","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/e3bb5b37d82eadb0e4f7e3f4.jpg"},{"id":97371081,"identity":"4736fa98-eea3-416a-9255-f0b1986c0f91","added_by":"auto","created_at":"2025-12-03 16:28:23","extension":"jpg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6323002,"visible":true,"origin":"","legend":"","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/b539204e7841f18a710a2095.jpg"},{"id":97349990,"identity":"09f72e5f-6776-4315-9e8a-1eb3023349f2","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":990235,"visible":true,"origin":"","legend":"","description":"","filename":"Fig6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/c48d6e8f4a764d1978f22b35.jpg"},{"id":97371079,"identity":"a1487725-a259-4532-8fe8-60f348cc610a","added_by":"auto","created_at":"2025-12-03 16:28:23","extension":"jpg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1970665,"visible":true,"origin":"","legend":"","description":"","filename":"Fig7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/1bdc8b5035b3b94d60a16b63.jpg"},{"id":97371100,"identity":"7f304d1e-bbf1-4903-9e73-59e714f44c09","added_by":"auto","created_at":"2025-12-03 16:28:24","extension":"jpg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2163140,"visible":true,"origin":"","legend":"","description":"","filename":"Fig8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/a286286f201e9afc59f3415b.jpg"},{"id":97371396,"identity":"1bfe9b13-14e8-43a2-8731-bbd39a8313ee","added_by":"auto","created_at":"2025-12-03 16:28:51","extension":"json","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9529,"visible":true,"origin":"","legend":"","description":"","filename":"625199d1ef0348d0a540203a823c159f.json","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/c4f31af896123d259cc492e9.json"},{"id":97350006,"identity":"8dcf0c26-6baa-44a3-918d-e1b7bf7eb5b4","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"pdf","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":173101,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTables.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/e29d0d06161894b14a0a0930.pdf"},{"id":97349999,"identity":"07fe7da5-7b77-4892-b453-0f8e6cccacc6","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":129983,"visible":true,"origin":"","legend":"","description":"","filename":"625199d1ef0348d0a540203a823c159f1enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/6441145f782c321ccaf386a1.xml"},{"id":97370953,"identity":"7399580e-9788-4238-be9a-34e3408fd508","added_by":"auto","created_at":"2025-12-03 16:28:11","extension":"jpg","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1282919,"visible":true,"origin":"","legend":"","description":"","filename":"Fig1..jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/4d63e2077231d6d35fadb34f.jpg"},{"id":97350001,"identity":"95234375-37a7-49d5-a4af-5b8080012565","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7216781,"visible":true,"origin":"","legend":"","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/ce5d268266d9b060cc38fb4a.jpg"},{"id":97370857,"identity":"0905e8fd-9520-40d6-a033-384c1c254299","added_by":"auto","created_at":"2025-12-03 16:28:03","extension":"jpg","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4639941,"visible":true,"origin":"","legend":"","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/7aab4df5e09d1221f46b43ce.jpg"},{"id":97369820,"identity":"7c808f33-b75d-4793-8953-bf27ba3b8b3d","added_by":"auto","created_at":"2025-12-03 16:25:51","extension":"jpg","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1901162,"visible":true,"origin":"","legend":"","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/f90bc76b68f59f428db149a1.jpg"},{"id":97370411,"identity":"cb9f6c7c-d1bc-4118-aea6-b9bcc6d66eb5","added_by":"auto","created_at":"2025-12-03 16:27:18","extension":"jpg","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6323002,"visible":true,"origin":"","legend":"","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/47a7b439049db2cb61adfa6f.jpg"},{"id":97349996,"identity":"82982d65-277b-4dc4-8e7e-d2a772c2a6e0","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":990235,"visible":true,"origin":"","legend":"","description":"","filename":"Fig6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/3c6819b18f1164188b3e673e.jpg"},{"id":97371230,"identity":"439bc56d-221b-463f-aed5-347749d33154","added_by":"auto","created_at":"2025-12-03 16:28:33","extension":"jpg","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1970665,"visible":true,"origin":"","legend":"","description":"","filename":"Fig7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/74d167c65b8545f3dd77447e.jpg"},{"id":97349992,"identity":"19a22fbf-3837-49a9-84d6-d50be8695d19","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2163140,"visible":true,"origin":"","legend":"","description":"","filename":"Fig8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/6ab8def026ebc19fb8e803ee.jpg"},{"id":97350004,"identity":"2ecafb7f-ccd9-4910-866d-0eef88e1b4fb","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpeg","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":20661,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/862d9a8cc4f8a8bc3dc2c6a2.jpeg"},{"id":97350005,"identity":"8a5d5b6d-1a80-4ad8-8333-00aa6e2e70f1","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpeg","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12718,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/21d5b4d67986824c48b23222.jpeg"},{"id":97350007,"identity":"e7abcc09-9a59-4e65-bf51-bf59311ad0fa","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"jpeg","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":15236,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/78f02fcf4b47eaeb7c2d3b69.jpeg"},{"id":97350003,"identity":"5b4d3db9-2ec3-43e5-978b-9bb0710695df","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpeg","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12075,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/14a3f0bce3d64292ea35e947.jpeg"},{"id":97350000,"identity":"be103b68-1ea7-4e1d-9dae-63642a029ee0","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpeg","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13795,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/1ad1e4fc9375bf9466b99a4a.jpeg"},{"id":97371360,"identity":"2f92fdb6-c793-4388-8c8f-1d7ccdd28027","added_by":"auto","created_at":"2025-12-03 16:28:47","extension":"jpeg","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50762,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/44820b8c6450638b8c99f317.jpeg"},{"id":97370418,"identity":"6d562ffb-b15b-4c54-a9a9-e664ec37c86c","added_by":"auto","created_at":"2025-12-03 16:27:19","extension":"jpeg","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":11094,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/918453550bcadf1aeac5faee.jpeg"},{"id":97350023,"identity":"afe8b684-e977-4747-b995-896cb2a4517d","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"jpeg","order_by":27,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9357,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/af187226aeb11f21f8f33b89.jpeg"},{"id":97350024,"identity":"dfbffcbd-6761-484e-b5ef-0151b42a234f","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":28,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":138206,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig1..png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/ed90cd37ef6b6d17832ae4fb.png"},{"id":97350009,"identity":"5d66b23b-50f1-40c8-8f4a-c20da7c44880","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":29,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1318883,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig2.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/8b2422be81cc6ef29a74ca5b.png"},{"id":97350010,"identity":"3b379462-90a6-4edd-9646-f3564770c6c9","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":30,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":905530,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig3.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/5902b8479c53527c9ad9fd99.png"},{"id":97350031,"identity":"aa957ebe-8ea5-41e5-b297-fd5c12f73d13","added_by":"auto","created_at":"2025-12-03 12:17:09","extension":"png","order_by":31,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":296298,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig4.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/9feaf491adbb242cf6e509f5.png"},{"id":97371482,"identity":"7460bb0d-57ba-48d4-9b8c-0a59170c0c64","added_by":"auto","created_at":"2025-12-03 16:29:02","extension":"png","order_by":32,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1329678,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig5.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/f18e3f877d4b54a6b286eb73.png"},{"id":97370838,"identity":"0c1d75ca-3aac-4e0e-9b27-6a0a412ca498","added_by":"auto","created_at":"2025-12-03 16:27:59","extension":"png","order_by":33,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136141,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig6.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/b3afa0a2c60955b5cbc07a25.png"},{"id":97350021,"identity":"c34cdc5e-3460-4788-8886-0a0150336e75","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":34,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":288009,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig7.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/c6935761fe9b0615e7a883e3.png"},{"id":97350013,"identity":"c1c10214-7a51-4498-ba52-02b0ee07c0ab","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":35,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":339307,"visible":true,"origin":"","legend":"","description":"","filename":"OnlineFig8.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/39f4d540eccfaf36849368e2.png"},{"id":97350018,"identity":"12c5603b-1923-4205-a1a9-a63430337823","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":36,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36476,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/85a36d80983b6f9e08923833.png"},{"id":97350016,"identity":"69e53b0c-23dc-47d1-b139-d51dd40ef472","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":37,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":12780,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/99ce9265c47d261f21b1d17a.png"},{"id":97350008,"identity":"14fbe8b2-da12-457b-a05c-412f7ffa6905","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":38,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":9608,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/9a7b37c7e57f85484241d15d.png"},{"id":97370713,"identity":"7e4222fe-60bd-471d-8c08-9bf7d6eeba7b","added_by":"auto","created_at":"2025-12-03 16:27:49","extension":"png","order_by":39,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8453,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/4c149c89d6d3c566877a3869.png"},{"id":97350012,"identity":"d85e48ee-e870-4adb-a297-38efd49d76b2","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":40,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13964,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/23831a8a0c179001902639fe.png"},{"id":97371431,"identity":"b9463d87-1300-4ad3-9600-e4bf02e75a74","added_by":"auto","created_at":"2025-12-03 16:28:54","extension":"png","order_by":41,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13763,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/8a456a0a56402dc1faa7570f.png"},{"id":97371089,"identity":"acdb4223-cf6b-4eb2-bf06-ec7f3aa55be4","added_by":"auto","created_at":"2025-12-03 16:28:24","extension":"png","order_by":42,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6932,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/2332f73942acb79d0a788f0d.png"},{"id":97350014,"identity":"8c872391-0c09-45a7-900f-c1ee5798c070","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"png","order_by":43,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6688,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/c52cf86633f00ba384e0aff3.png"},{"id":97371051,"identity":"b8d99a77-71dc-4ff7-a358-78dbb93dcc15","added_by":"auto","created_at":"2025-12-03 16:28:21","extension":"xml","order_by":44,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":129366,"visible":true,"origin":"","legend":"","description":"","filename":"625199d1ef0348d0a540203a823c159f1structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/c0780fcd43767d811449b721.xml"},{"id":97350026,"identity":"b17bb495-1da2-48d8-8aaa-0fcea4caae2c","added_by":"auto","created_at":"2025-12-03 12:17:08","extension":"html","order_by":45,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":141358,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/efe8af0217dbd73c31ed44ca.html"},{"id":97349976,"identity":"148d82fd-142e-41ee-aecf-abf06093c147","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1282919,"visible":true,"origin":"","legend":"\u003cp\u003eSchematic workflow of the research methodology\u003c/p\u003e","description":"","filename":"Fig1..jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/5ae14e152328be066fd97c3e.jpg"},{"id":97349980,"identity":"401dd174-40b1-4db9-ab5f-fef54c44b273","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":7216781,"visible":true,"origin":"","legend":"\u003cp\u003eDifferential gene expression (DEG) in Barrett's Esophagus (BE) and Esophageal Adenocarcinoma (EAC) (FDR \u0026lt; 0.05, p-value \u0026lt; 0.01) (a) Principal component analysis (PCA) plot of normalized BE dataset; (b) PCA plot of standardized EAC dataset; (c) Volcano plot of DEGs in BE (vs. normal controls); (d) Volcano plot of DEGs in EAC (vs. normal controls); (e) Heatmap of DEGs in BE; (f) Heatmap of DEGs in EAC; (g) Venn diagram identifying shared DEGs (SDEGs) between BE and EAC.\u003c/p\u003e","description":"","filename":"Fig2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/458a3972a154b3705ef5f111.jpg"},{"id":97349984,"identity":"91d91ff7-afc8-4b48-86f3-ce77622573e6","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":4639941,"visible":true,"origin":"","legend":"\u003cp\u003eFunctional enrichment and protein interaction network of shared DEGs (SDEGs) (FDR \u0026lt; 0.05, (p - value \u0026lt; 0.05) (a) Bubble plot of Gene Ontology (GO) enrichment analysis for SDEGs; (b) Circle plot of GO enrichment analysis; (c) Bar plot of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis; (d) Circle plot of KEGG pathway analysis; (e) Protein-protein interaction (PPI) network of SDEGs; (f) Top 10 hub genes identified from the PPI network.\u003c/p\u003e","description":"","filename":"Fig3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/037f9fd7b33236f156db9441.jpg"},{"id":97369932,"identity":"fedf09cc-8010-416d-a27a-0b23f2b6d9a5","added_by":"auto","created_at":"2025-12-03 16:26:06","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":1901162,"visible":true,"origin":"","legend":"\u003cp\u003eImmune microenvironment characterization and shared immune-related gene signatures in BE and EAC (FDR \u0026lt; 0.2, p-value \u0026lt; 0.05) (a) Box plot showing immune cell abundance in BE versus healthy controls; (b) Box plot of EAC versus healthy controls; (c) Comparative analysis of 22 immune cell subtypes between BE and controls;(d) Comparative analysis of 22 immune cell subtypes between EAC and controls; (e) Venn diagram of immune-related SDEGs between BE and EAC.\u003c/p\u003e","description":"","filename":"Fig4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/a074c12146aba20cff353855.jpg"},{"id":97349989,"identity":"03d50f40-130c-445f-8b10-613154d1f603","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":6323002,"visible":true,"origin":"","legend":"\u003cp\u003eImmune-correlation profiles and identification of candidate feature genes for BE and EAC (a) Correlation heatmap betwee immune-related SDEGs and 22 immune cell types in BE; (b) Correlation heatmap betwee immune-related SDEGs and immune cells in EAC; (c) Venn diagram showing overlapbetween immune-related SDEGs and core SDEGs; (d) Heatmap of correlations between candidate genes and immune cells in BE; \u0026nbsp;(e) Heatmap of correlations between candidate genes and immune cells in EAC.\u003c/p\u003e","description":"","filename":"Fig5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/6901057c6a21ad59d7a82ec7.jpg"},{"id":97370824,"identity":"84aa800f-a32f-4f91-a1d4-e5ce6fcc8e34","added_by":"auto","created_at":"2025-12-03 16:27:59","extension":"jpg","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":990235,"visible":true,"origin":"","legend":"\u003cp\u003eFeature gene selection via the Least Absolute Shrinkage and Selection Operator (LASSO) regression (a) Optimal lambda (λ) value by 10-fold cross-validation in BE; (b) Optimal lambda (λ) value by 10-fold cross-validation in EAC; (c) LASSO coefficient profiles for BE classification.; (d) LASSO coefficient profiles for EAC classification.\u003c/p\u003e","description":"","filename":"Fig6.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/b349bcbae5426d3e988fddfd.jpg"},{"id":97349991,"identity":"8b6a972a-05e7-4537-9d74-ee77c6282573","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1970665,"visible":true,"origin":"","legend":"\u003cp\u003eFeature gene selection via Random Forest (RF) modeling (a, b) RF plotsof gene importance (IncNodePurity) for BE;(c) out-of-bag (OOB) error rate vs. number of trees for BE; (d) OOB error vs. number of features retained for BE; (e, f) RF plotsof gene importance (IncNodePurity) for EAC; (g) OOB error rate vs. number of trees for EAC; (h) OOB error vs. number of features retained for EAC; (i) Venn diagram of candidate feature genes identified by both LASSO and random forest models.\u003c/p\u003e","description":"","filename":"Fig7.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/44b415be50da86a6f6790eb9.jpg"},{"id":97349994,"identity":"2742e6dd-8d94-417c-bb4a-83f701b18592","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"jpg","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":2163140,"visible":true,"origin":"","legend":"\u003cp\u003eValidation of diagnostic and predictive performance for the two-Gene signature in BE and EAC (a) Receiver operating characteristic (ROC) curve of the BE model in the validation cohort; (b) ROC curve of the EAC model in the validation cohort; (c) Nomogram for predicting BE probability; (d) Nomogram for predicting EAC probability; (e) Calibration curve for the BE nomogram; (f) Calibration curve for the EAC nomogram.\u003c/p\u003e","description":"","filename":"Fig8.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/488ae8c325d1d13d44d96a47.jpg"},{"id":100371445,"identity":"12cdc9a1-ebcb-47e9-9908-0d447c754c57","added_by":"auto","created_at":"2026-01-16 08:10:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":27382490,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/be3f8d7e-c1ce-43c4-8d75-d80a76c85ee9.pdf"},{"id":97349979,"identity":"d360c747-3d07-4913-98d2-21e719f37c1d","added_by":"auto","created_at":"2025-12-03 12:17:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":173101,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTables.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8004374/v1/dbb89d11621e3b5e3f8f740e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development of a Diagnostic Model for Barrett's Esophagus and Esophageal Adenocarcinoma Based on Machine Learning and Immune Infiltration","fulltext":[{"header":"Introduction","content":"\u003cp\u003eEsophageal cancer (EC) ranks as the seventh most prevalent malignant tumor worldwide, exhibiting an exceptionally poor prognosis with a five-year overall survival rate of less than 30%[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e] and ranks as the sixth leading cause of cancer-related mortality[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The clinical manifestations of EC include dysphagia, chest pain or retrosternal burning sensation, dyspepsia or heartburn, vomiting, and hematemesis. Definitive diagnosis requires endoscopic visualization and pathological biopsy confirmation. The two principal histopathological subtypes of EC are esophageal adenocarcinoma (EAC) and esophageal squamous cell carcinoma (ESCC). While ESCC remains the predominant variant globally, the incidence of EAC has increased markedly and steadily over the past four decades[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e], particularly in Western populations. Notably, more than 40% of EAC patients present with metastatic disease at diagnosis, and their five-year survival rate is below 20%[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. EAC originates from metaplastic columnar epithelium (such as Barrett's esophagus[BE]). Chronic exposure to risk factors such as obesity, alcohol consumption, tobacco use, and especially sustained gastroesophageal reflux disease (GERD) induces the replacement of native stratified squamous epithelium with intestinal-type columnar epithelium (intestinal metaplasia, IM) in the distal esophagus, which may then progress to EAC. The treatment of EAC includes surgery (transthoracic esophagectomy with lymph node dissection, endoscopic submucosal dissection [ESD], endoscopic mucosal resection [EMR]), chemoradiotherapy, targeted therapy, and immunotherapy[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eBarrett's esophagus (BE) is a metaplastic transformation of esophageal squamous mucosa to intestinal-type columnar epithelium induced by chronic acid exposure, representing the only established precursor lesion and a major risk factor for EAC[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The diagnosis of BE principally relies on endoscopic examination and histopathological confirmation due to the lack of specific clinical symptoms. Epidemiological data show a significant increase in BE incidence in developed nations in recent decades[\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Risk factors for BE include advancing age, chronic GERD, male sex, tobacco use, Caucasian ethnicity, familial predisposition, and central obesity[\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. The progression from BE to EAC is a stepwise process, starting from non-dysplastic Barrett's esophagus (NDBE), advancing to low-grade dysplasia (LGD), then to high-grade dysplasia (HGD), and ultimately to invasive EAC. Patients with BE have the highest risk of progressing to HGD or EAC within the first 1\u0026ndash;3 months after diagnosis, with a 30- to 125-fold higher incidence of EAC compared to the general population, and individuals with HGD have a significantly higher rate of malignant transformations[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The main treatments for BE include endoscopic eradication therapy (EET), esophagectomy for advanced cases, and proton pump inhibitor (PPI)--based anti-reflux therapy, along with regular endoscopic surveillance[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eEAC often presents with insidious early-stage symptoms that are easily overlooked, posing significant challenges for timely diagnosis. Although BE is a well-established risk factor for EAC, the molecular mechanisms driving this progression remain incompletely understood. Therefore, this study aims to identify shared genetic signatures between BE and EAC using integrative bioinformatics approaches and construct a robust genetic diagnostic model. This will help elucidate the molecular mechanisms of disease evolution and provide new insights for early detection and targeted therapy of Barrett's esophagus-derived esophageal adenocarcinoma.\u003c/p\u003e"},{"header":"Material and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eStudy Design\u003c/h2\u003e\u003cp\u003eThe workflow of this study is schematically presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eData acquisition and preprocessing\u003c/h3\u003e\n\u003cp\u003eThe raw matrix datasets for BE were acquired from the Gene Expression Omnibus (GEO) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/geo/\u003c/span\u003e\u003cspan address=\"https://www.ncbi.nlm.nih.gov/geo/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e). Dataset selection was based on the following specific criteria: (1)data from human subjects, (2)samples from both the BE group and healthy control groups, while excluding EAC group, (3)genome-wide expression profiles. Based on these criteria, four datasets (GSE13083, GSE1420, GSE39491, and GSE36223) were obtained. After annotation and normalization, these datasets were merged to form a combined dataset comprising 78 BE samples and 78 healthy controls. The EAC dataset was obtained from The Cancer Genome Atlas (TCGA) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://portal.gdc.cancer.gov/analysis_page?app=CohortBuilder\u0026amp;tab=general\u003c/span\u003e\u003cspan address=\"https://portal.gdc.cancer.gov/analysis_page?app=CohortBuilder\u0026amp;tab=general\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), which include 81 EAC samples and 13 healthy controls. After annotation and normalization, the data were processed for bioinformatics analysis.Batch effects across samples were corrected using the \u0026ldquo;ComBat\u0026rdquo; algorithm from the R package \u0026ldquo;sva\u0026rdquo;. And variance-stabilizing transformation (VST) was performed using the \u0026ldquo;DESeq2\u0026rdquo; package. For subsequent analysis, both the BE and EAC datasets were randomly divided into a 70% training set and a 30% validation set. Principal component analysis (PCA) was then conducted using the \u0026ldquo;tinyarray\u0026rdquo; and \u0026ldquo;ggplot2\u0026rdquo; packages, and the results were visualized as PCA plots (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea and b).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eDifferentially expressed genes analysis\u003c/h3\u003e\n\u003cp\u003eAll feature selection procedures, including differential gene expression analysis, protein-protein interaction network analysis, immune infiltration assessment, and machine learning-based selection, were performed exclusively on the training sets (70% of the normalized BE and EAC cohorts). The hold-out sets (30%) were used solely for final performance evaluation. To identify differentially expressed genes (DEGs) between disease groups (BE and EAC) and healthy controls, we performed differential expression analysis using both the \u0026ldquo;limma\u0026rdquo; and \u0026ldquo;DESeq2\u0026rdquo; packages on the processed expression data. Significant DEGs were defined with a threshold of adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.01 and absolute log fold change (|log FC|)\u0026thinsp;\u0026gt;\u0026thinsp;1.5 to ensure the inclusion of biologically relevant genes. Volcano plots were generated to visualize the identified DEGs. Subsequently, shared differentially expressed genes (SDEGs) between BE and EAC were identified by taking the intersection of the DEG sets using Venn diagrams implemented in TBtools.\u003c/p\u003e\n\u003ch3\u003eGene ontology and pathway enrichment analyses\u003c/h3\u003e\n\u003cp\u003eFunctional and pathway enrichment analyses of the SDEGs were performed using R with a significance threshold of p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Within the Gene Ontology (GO) framework, genes were systematically classified and functionally annotated three primary categories: biological processes (BP), cellular components (CC), and molecular functions (MF)[\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The Kyoto Encyclopedia of Genes and Genomes (KEGG) database enables systematic analysis of gene functions and the exploration of potential pathways through integrated genomic information[\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e][\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Using the R package \"clusterProfiler\", we performed GO and KEGG enrichment analyses on these SDEGs. Significantly enriched pathways were selected based on thresholds of p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and q-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05. The top 10 most significantly enriched pathways were visualized using bar plots, bubble plots, and circle plots generated by the \"ggplot2\" package in R.\u003c/p\u003e\n\u003ch3\u003eProtein-protein interaction (PPI) network analysis\u003c/h3\u003e\n\u003cp\u003eBased on the identified SDEGs, we constructed a PPI network for BE and EAC using the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING v12.0) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.string-db.org/\u003c/span\u003e\u003cspan address=\"https://www.string-db.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), with an interaction confidence score threshold set at \u0026ge;\u0026thinsp;0.7 (the default medium confidence level in STRING) for filtering interactions(p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Employing the Maximal Clique Centrality (MCC) algorithm, we ranked genes by their centrality within the network. Protein interactions were visualized with \u0026ldquo;Cytoscape v3.10.2\u0026rdquo; software, and its \u0026ldquo;cytoHubba\u0026rdquo; plugin was utilized to identify the top 10 hub genes in the PPI network. These hub genes were subsequently integrated with genes obtained from functional enrichment analysis.\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eImmune infiltration and correlation analysis\u003c/h2\u003e\u003cp\u003eTo elucidate immune cell infiltration patterns in BE, EAC, and healthy control samples, we employed the computational tool \u0026ldquo;CIBERSORT\u0026rdquo; to quantify the relative abundance of 22 human immune cell subtypes (p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Using the LM22 signature matrix, we applied the CIBERSORT algorithm to the intersection of genome-wide profiles between BE and EAC cohorts, comprising 13435 shared genes. This gene set fully encompasses all 547 immune-related genes present in the LM22 signature matrix. With selection criteria of false discovery rate (FDR)\u0026thinsp;\u0026lt;\u0026thinsp;0.2 and an absolute median difference\u0026thinsp;\u0026gt;\u0026thinsp;0.05 between tumor and normal samples, thereby we identified shared immune-related SDEGs. Subsequent correlation analysis was performed using the R package \"pheatmap\" to examine associations between these genes and the 22 immune cell types. The results were visualized as a correlation heatmap generated with the \"ggplot2\" package in R. Finally, we integrated the shared immune-related genes from BE and EAC with the previously identified hub genes and genes derived from functional enrichment analysis.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eMachine learning and feature genes selection\u003c/h3\u003e\n\u003cp\u003eTo optimize feature genes selection, we implemented two machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO) regression and Random Forest (RF). LASSO employs L1 regularization to compress the variable space, making it suitable for sparse feature selection in high-dimensional data, whereas RF effectively captures non-linear relationships and interactions among variables through the integration of multiple decision trees. Their combination can thereby complement and optimize the biomarker identification pipeline. Using the six candidate genes previously identified through functional enrichment, PPI, and immune infiltration analysis as inputs, LASSO and RF analyses were performed separately. The RF model was implemented using the R package \u0026ldquo;randomForest\u0026rdquo; [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. The number of trees (ntree) was set to 500, and the number of features considered at each split (mtry) was set to \u0026radic;p, where p represents the total number of features minus one. Bootstrap sampling was enabled with out-of-bag (OOB) samples retained for every tree. Model stability was evaluated by plotting the OOB error against the number of trees. Feature importance was ranked based on the MeanDecreaseGini (also known as IncNodePurity) metric [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. To determine the feature selection threshold, a stepwise elimination strategy was applied: features with lower importance were removed sequentially, and the corresponding OOB error was computed after each removal. The relationship between the number of retained features and the OOB error was analyzed to identify a critical subset of features. Additionally, LASSO regression was also employed for feature selection. The LASSO regression was carried out using the R package \u0026ldquo;glmnet\u0026rdquo;[\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e] with 10-fold cross-validation to determine the optimal penalty coefficient λ (lambda). Genes with non-zero coefficients at the selected λ value were ultimately identified as potential feature genes. This procedure represents the standard approach for selecting the λ value, aimed at minimizing the cross-validation error of the model[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Ultimately, the overlapping genes identified by both algorithms were selected as a core set of feature genes for subsequent validation of their diagnostic potential. All key parameters, filtering thresholds, and dataset partitioning information related to the machine learning analyses in this study are summarized in Supplementary Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e.\u003c/p\u003e\n\u003ch3\u003eLogistic regression validation\u003c/h3\u003e\n\u003cp\u003eTo assess the clinical diagnostic value of these potential feature genes identified by LASSO regression and RF algorithms, we partitioned the BE and EAC datasets into a 70% training set and a 30% validation set. Subsequently, we employed logistic regression modeling based on the selected feature genes and evaluated its performance using the validation set. The model was visualized through diagnostic plots. Using the validation set, we generated the receiver operating characteristic (ROC) curves with the R package \u0026ldquo;pROC\u0026rdquo; and calculated the area under the curve (AUC) to evaluate the model\u0026rsquo;s predictive performance. The contribution of the model to disease risk was visualized using nomograms generated by the R package \u0026ldquo;nomogramFormula\u0026rdquo;. To assess model calibration, we performed the Hosmer-Lemeshow goodness-of-fit test (H-L test) and generated calibration curves using the R package \u0026ldquo;ResourceSelection\u0026rdquo;, which validated the model's discrimination and calibration. Detailed performance metrics are summarized in Supplementary Table S2 and S3.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eIdentification of shared differentially expressed genes\u003c/h2\u003e\u003cp\u003eDifferential expression analysis of BE and EAC was performed using thresholds of adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.01 and |logFC| \u0026gt;1.5. The identified DEGs were visualized in volcano plots (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec and d), while heatmaps illustrated expression patterns across multiple samples and genes (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee and f).The analysis revealed 609 DEGs in BE (352 upregulated, 257 downregulated) and 4,042 DEGs in EAC (1,618 upregulated, 2,424 downregulated). Intersection analysis of both disease groups identified 101 shared differentially expressed genes (SDEGs), visualized in a Venn diagram (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eg). Complete gene lists are summarized in Supplementary Table S4.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eGO and KEGG enrichment analyses results\u003c/h2\u003e\u003cp\u003eThe results of GO and KEGG enrichment analyses were visualized using bar plots and bubble plots to represent the significantly enriched terms and pathways. The GO enrichment analysis of the 101 SDEGs revealed significant associations with multiple processes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea and b). In the biological process (BP) category, these genes were predominantly enriched in digestive functions (13 genes), gastrointestinal epithelium maintenance, epithelial structure maintenance, non-protein amino acid biosynthesis, cellular aldehyde metabolism, and sodium/potassium ion homeostasis; For cellular components (CC), the SDEGs were primarily localized to the apical plasma membrane, cell apical part, Golgi lumen, collagen-containing extracellular matrix, granzyme, cornified envelope; molecular function (MF) analysis demonstrated these genes were significantly involved in branched-chain amino acid transmembrane transporter activity, glycosaminoglycan binding, ATPase-coupled transmembrane transporter activity, extracellular matrix structural constituency. These findings collectively indicate that the identified SDEGs are functionally associated with: nutrient digestion and absorption, maintenance of epithelial barriers, transmembrane transport-driven ion, and molecular homeostasis.KEGG pathway analysis revealed significant enrichment patterns among the SDEGs (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec and d). Five genes were notably enriched in the digestive system-related pathway \"gastric acid secretion.\" Additionally, these SDEGs demonstrated prominent enrichment in several key pathways including \"Collecting duct acid secretion\", \"Nitrogen metabolism\", \"Amino acid biosynthesis\", \"IL-17 signaling pathway\", \"Rheumatoid arthritis\", etc. The core commonalities among these pathways suggest their involvement in the regulation of inflammatory responses, immune functions, and metabolic processes. By integrating the results from both GO and KEGG analyses, we identified 13 pathway-associated SDEGs that showed consistent enrichment across these functional categories (with the threshold of p - value\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003ePPI network analysis and Hub genes selection\u003c/h2\u003e\u003cp\u003eTo identify shared molecular signatures between BE and EAC, we selected shared SDEGs based on stringent criteria (adjusted p - value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and |logFC| \u0026gt;1.5). This approach ensured a focus on genes with significant differential expression, which are more likely to play pivotal roles in BE and EAC pathogenesis. After that, we delved into the exploration of their intricate PPI network. To ensure the reliability and relevance of the PPI network analysis, we removed nodes that were either unconnected or weakly connected to the main network, which might represent spurious interactions or genes with limited functional relevance in the context of BE and EAC. Subsequently, we used the \u0026ldquo;CytoHubba\u0026rdquo; plugin, a well-established tool in network analysis, to identify critical nodes and subnetworks within the refined PPI network. Through this in-depth analysis, we were able to pinpoint the top 10 hub genes, namely \u003cem\u003eVIL1\u003c/em\u003e, \u003cem\u003eCDH17\u003c/em\u003e, \u003cem\u003eMUC2\u003c/em\u003e, \u003cem\u003eOLFM4\u003c/em\u003e, \u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003eATP4B\u003c/em\u003e, \u003cem\u003eGKN1\u003c/em\u003e, \u003cem\u003eLIPF\u003c/em\u003e, \u003cem\u003eATP4A\u003c/em\u003e, and \u003cem\u003ePGC\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee and f). These hub genes are likely to serve as key regulators in the molecular pathways underlying BE and EAC, given their central positions in the PPI network. Ultimately, by the combination of the 10 hub genes with 13 genes from functional enrichment analyses, we identified a set of 18 core regulatory SDEGs strongly implicated in the development and progression of BE and EAC, including \u003cem\u003eMUC4\u003c/em\u003e, \u003cem\u003eMDK\u003c/em\u003e, \u003cem\u003eGCNT3\u003c/em\u003e, \u003cem\u003eADRA2A\u003c/em\u003e, \u003cem\u003eMUC2\u003c/em\u003e, \u003cem\u003eVIL1\u003c/em\u003e, \u003cem\u003eSST\u003c/em\u003e, \u003cem\u003eCAPN9\u003c/em\u003e, \u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003ePBLD\u003c/em\u003e, \u003cem\u003eSI\u003c/em\u003e, \u003cem\u003ePGC\u003c/em\u003e, \u003cem\u003eGKN1\u003c/em\u003e, \u003cem\u003eCDH17\u003c/em\u003e, \u003cem\u003eOLFM4\u003c/em\u003e, \u003cem\u003eATP4B\u003c/em\u003e, \u003cem\u003eLIPF\u003c/em\u003e, and \u003cem\u003eATP4A\u003c/em\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eImmune infiltration and correlation analysis\u003c/h2\u003e\u003cp\u003eGiven the significant enrichment of SDEGs in immune-related pathways identified in functional analyses, we further investigated these genes from an immune perspective. Box plots demonstrated significantly elevated immune cell infiltration in both BE and EAC groups compared to healthy controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea and b). Analysis of 22 immune cell subtypes revealed distinct patterns: in BE, increased plasma cells and mast cells but decreased T lymphocytes and dendritic cells compared to controls; in EAC, increased T lymphocytes, macrophages, and dendritic cells but reduced B lymphocytes, plasma cells, and mast cells relative to healthy controls (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec and d). Applying stringent thresholds (FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.2 and absolute median difference |Tumor_Median - Normal_Median| \u0026gt;0.05), we identified 100 immune-related SDEGs in BE and 40 immune-related SDEGs in EAC (p - value\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The intersection of the two yielded 40 immune-associated DEGs shared by BE and EAC (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee). Correlation heatmaps illustrated associations between these 40 immune-related SDEGs and various immune cell types (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea and b). By integrating these with the previously identified 18 core regulatory genes, we identified 6 candidate feature genes: \u003cem\u003eGCNT3\u003c/em\u003e, \u003cem\u003eADRA2A\u003c/em\u003e, \u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003ePBLD\u003c/em\u003e, \u003cem\u003eSI\u003c/em\u003e, and \u003cem\u003eOLFM4\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec). Their distinct correlation patterns with different immune cell populations are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ed and e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eMachine learning modeling to screen for signature genes\u003c/h2\u003e\u003cp\u003eThe LASSO and RF algorithms were employed to screen the final characteristic genes from the six candidate key genes. For the LASSO regression, genes with non-zero coefficients were selected at the optimal λ value (λ =0.019 for BE, λ\u0026thinsp;=\u0026thinsp;0.012 for EAC)(Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea and b). This resulted in the identification of two genes (\u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD)\u003c/em\u003e for BE (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec) and four genes (\u003cem\u003eADRA2A\u003c/em\u003e, \u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003ePBLD\u003c/em\u003e, and \u003cem\u003eOLFM4)\u003c/em\u003e for EAC (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed). In the RF model, the OOB error decreased progressively with an increasing number of trees and stabilized between 400 and 500 trees (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec and g). Feature importance was then ranked by the mean decrease Gini index (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea, b, e and f). Using a stepwise feature elimination strategy, we was employed evaluated how different feature subsets affected the OOB error. The results showed a clear inflection point in the OOB error as features were sequentially removed (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ed and h). Based on this criterion, two genes (\u003cem\u003ePBLD\u003c/em\u003e and \u003cem\u003eMUC13)\u003c/em\u003e were selected as the most important features for BE, and three genes (\u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003ePBLD\u003c/em\u003e, and \u003cem\u003eGCNT3)\u003c/em\u003e for EAC. Finally, by integrating the results from both algorithms, we identified two shared feature genes: \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ei). Differential expression analysis indicated that both genes were up-regulated in BE, while in EAC, \u003cem\u003eMUC13\u003c/em\u003e was up-regulated and \u003cem\u003ePBLD\u003c/em\u003e was down-regulated.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eLogistic regression validation\u003c/h2\u003e\u003cp\u003eTo preliminarily evaluate the diagnostic performance of the feature genes \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e, we performed logistic regression analysis on the internal validation set, which comprised 30% of samples randomly selected from the normalized and merged dataset. The discriminatory ability of the model was assessed by plotting ROC curves and calculating the AUC values, which range from 0.5 (random guessing) to 1 (perfect classification). The internal validation results showed AUC values of 0.94 (95% CI: 0.858\u0026ndash;1.000) for BE and 0.98 (95% CI: 0.957\u0026ndash;1.000) for EAC (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ea and b), indicating that the diagnostic\u003c/p\u003e\u003cp\u003emodel based on these two feature genes exhibits high sensitivity and specificity. We also constructed nomograms to visualize the model's ability to predict individual disease risk (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ec and d). The substantial contributions of both genes to disease risk, as indicated by their positions on the horizontal axes, confirm the model's strong predictive power. To evaluate the calibration between predicted probabilities and observed outcomes, we performed the Hosmer-Lemeshow (H-L) goodness-of-fit test (detailed metrics are provided in Supplementary Table S2) and plotted calibration curves (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ee and f). The H-L test showed no significant deviation (BE: p\u0026thinsp;\u0026gt;\u0026thinsp;0.9; EAC: p\u0026thinsp;\u0026gt;\u0026thinsp;0.9), and the calibration curves closely followed the diagonal reference line, indicating good model fit and high calibration accuracy, with predicted risks aligning well with actual observations. These validation results demonstrate that the diagnostic model based on \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e has favorable diagnostic value and potential for distinguishing EAC arising from BE. This suggests that both genes may play critical roles in BE-to-EAC progression and could serve as biomarkers for early diagnosis and targeted intervention.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eEsophageal adenocarcinoma (EAC) is a highly lethal malignancy, with studies projecting a continued global increase in its incidence in the coming years[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. EAC typically presents with nonspecific early symptoms, and most patients are diagnosed at advanced stages. Although Barrett's esophagus (BE) is the only known precursor lesion of EAC, reliable risk stratification tools and clinical monitoring strategies remain limited. Therefore, the development of novel early diagnostic biomarkers is critical to advancing precision prevention and early therapeutic intervention for esophageal cancer.\u003c/p\u003e\u003cp\u003eIn this study, we employed multi-modal bioinformatics approaches to systematically investigate shared molecular pathways in BE and EAC pathogenesis. Differential expression analysis of training datasets identified 101 SDEGs common to both conditions. Subsequent functional annotation through GO and KEGG enrichment analyses, combined with PPI network analysis, revealed 10 hub genes. Integration of these results yielded a core set of 18 key regulatory genes. Furthermore, immune infiltration analysis identified 40 shared immune-related SDEGs. Intersection analysis of these gene sets ultimately pinpointed six high-priority candidate feature genes: \u003cem\u003eGCNT3\u003c/em\u003e, \u003cem\u003eADRA2A\u003c/em\u003e, \u003cem\u003eMUC13\u003c/em\u003e, \u003cem\u003ePBLD\u003c/em\u003e, \u003cem\u003eSI\u003c/em\u003e, and \u003cem\u003eOLFM4\u003c/em\u003e. Through machine learning approaches employing LASSO regression and RF algorithms, we identified two feature genes: \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e. Subsequent validation using the hold-out sets (30% of normalized sets) demonstrated strong diagnostic performance, as evidenced by: ROC curve analysis showing high discriminative power (AUC\u0026thinsp;=\u0026thinsp;0.94 for BE, and AUC\u0026thinsp;=\u0026thinsp;0.98 for EAC), nomogram analysis confirming accurate disease risk prediction, and Hosmer-Lemeshow goodness-of-fit test indicating good calibration (close alignment with ideal curves). These comprehensive validation results confirm the accuracy and clinical diagnostic potential of this genetic model.\u003c/p\u003e\u003cp\u003eGO and KEGG enrichment analyses of SDEGs in BE and EAC, combined with PPI network analysis, revealed significant associations with inflammatory responses, immune regulation, and epithelial barrier function. Both BE and EAC develop from reflux-induced esophageal mucosal inflammation, with established inflammation-to-cancer transformation mechanisms. Cytokine storms and proinflammatory microenvironments play pivotal roles in this process, where interleukin-1β (IL-1β) and tumor necrosis factor-α (TNF- α) promote immune cell chemotaxis through nuclear factor kappa B (NF-κB) signaling pathway activation[\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. The interplay between the prostaglandin/epithelial growth factor/peroxisome proliferator-activated receptor-γ (PG/EGF/PPARγ) system and proinflammatory cytokines, TNFα and IL-1β, drives the progression from chronic inflammation to BE and subsequent EAC[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. The reciprocal activation between the interleukin-6/signal transducer and activator of transcription 3 (IL-6/STAT3) and NF-κB signaling pathways fosters an immunosuppressive microenvironment and accelerates the malignant progression of BE[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Reflux-induced dysbiosis promotes EAC progression by amplifying procarcinogenic signals through the toll-like receptor (TLR) /NF-κB/tumor protein p53 (TP53) signaling cascade while suppressing antitumor immune responses[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Studies have demonstrated aberrant activation of the Wingless/Integrated (Wnt)/β-catenin pathway during BE carcinogenesis[\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Hyperactive Wnt/β-catenin signaling induces immune exclusion and impairs T cell-mediated antitumor immunity, while enhancing resistance to immunotherapy[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Recent evidence indicates that poly ADP-ribose glycohydrolase (PARG) activates the Wnt/β-catenin pathway to promote epithelial-mesenchymal transition (EMT), thereby facilitating EC cell growth, metastasis, and invasion[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Research indicates that chronic inflammatory exposure in the esophagus promotes EMT process through AP endonuclease 1 (APE1) redox-dependent E-cadherin cleavage, establishing a persistent positive feedback loop between inflammatory microenvironment and EMT that drives EAC initiation and metastasis[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Additionally, acid exposure activates the mitogen-activated protein kinase/extracellular signal-regulated kinase (MAPK/ERK) pathway, enhancing proliferation and survival while suppressing apoptosis in human Barrett's adenocarcinoma cell line (SEG-1), thereby accelerating BE malignant progression[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. These established mechanisms of BE-to-EAC progression are consistent with the functional enrichment results observed in our SDEG analysis.\u003c/p\u003e\u003cp\u003eThe progression from BE to EAC represents a classic example of inflammation\u0026ndash;cancer transformation. Interestingly, key pathways including Wnt/β-catenin, NF-κB, and EMT do not operate in isolation, but rather form a highly interconnected and mutually reinforcing signaling network. This robust tripartite alliance collaboratively drives malignant progression. (i) The inflammatory microenvironment induced by chronic reflux serves as the central driver of this process. Refluxates such as bile acids and gastric acid continuously stimulate the lower esophageal mucosa, activating the NF-κB signaling pathway and upregulating a series of pro-inflammatory cytokines (e.g., TNF-α, IL-1β, IL-6, IL-8) and cyclooxygenase-2 (COX-2). This creates a microenvironment rich in growth factors and genotoxic reactive oxygen species (ROS), leading to DNA damage, promoting epithelial cell proliferation, inhibiting apoptosis, and facilitating the activation of other oncogenic pathways. These events represent the initial steps of inflammation\u0026ndash;cancer transformation[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e][\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. (ii) The extensive bidirectional crosstalk between NF-κB and Wnt signaling represents a critical mechanism in the malignant transformation of BE. Inflammatory cytokines such as TNF-α upregulate the expression of Wnt ligands (e.g., Wnt 1, Wnt 2) while suppressing Wnt pathway antagonists (e.g., DKK1) in an NF-κB-dependent manner. This leads to nuclear accumulation and transcriptional activation of β-catenin. Activated β-catenin physically interacts with key components of the NF-κB pathway, such as p65/RelA, thereby enhancing NF-κB transcriptional activity. This interaction establishes a positive feedback loop that perpetuates pro-inflammatory and pro-proliferative signaling[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. (iii) EMT is a critical step in the malignant progression of BE, conferring invasive and metastatic capabilities. Both Wnt/β-catenin and NF-κB signaling pathways are centrally involved in the core regulation of EMT and act synergistically to drive this process. The NF-κB complex and β-catenin/TCF complex can cooperatively bind to the promoter regions of EMT-related transcription factors such as SNAIL, SLUG, and ZEB1, thereby upregulating their expression[\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. These transcription factors subsequently suppress epithelial markers (e.g., E-cadherin) and promote mesenchymal markers (e.g., Vimentin, N-cadherin), initiating the EMT program. Concurrently, under conditions of oxidative stress, APE1 cleaves E-cadherin, compromising epithelial integrity, while factors such as PARG further enhance β-catenin activity. Additionally, reflux-related acid and bile acid stimuli induce EMT activation through NF-κB and MAPK/ERK pathways[\u003cspan additionalcitationids=\"CR27\" citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], collectively accelerating the EMT process and promoting tumor cell invasion.\u003c/p\u003e\u003cp\u003eGiven the pivotal involvement of inflammation-immune regulation and tumor metastasis pathways in BE- to- EAC progression, we further investigated the functional mechanisms underlying the two potential biomarkers, \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e. \u003cem\u003eMucin 13\u003c/em\u003e (\u003cem\u003eMUC13\u003c/em\u003e), a member of the transmembrane mucin family, is typically expressed at low levels across epithelial surfaces of the gastrointestinal tract, respiratory system, and reproductive tract, where it contributes to barrier function maintenance. Substantial evidence indicates that \u003cem\u003eMUC13\u003c/em\u003e overexpression disrupts normal cellular polarity, promoting malignant transformation through enhanced tumorigenesis, metastasis, invasive capacity, and colony formation[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Studies have demonstrated aberrant overexpression of \u003cem\u003eMUC13\u003c/em\u003e across multiple cancer types, where it significantly correlates with adenocarcinoma initiation, progression, and poor clinical outcomes[\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]. The mechanistic investigations are listed as follows: In digestive system cancers, in esophageal cancer (EC), immunohistochemical (IHC) analysis revealed significant overexpression of \u003cem\u003eMUC13\u003c/em\u003e in tumor tissues compared with adjacent non-tumor tissues. Silencing \u003cem\u003eMUC13\u003c/em\u003e expression suppressed cell proliferation, reduced colony formation, induced cell cycle arrest, and promoted apoptosis. It also markedly decreased tumor volume and tumor weight. These findings suggest that \u003cem\u003eMUC13\u003c/em\u003e acts as a key regulator of O-glycosylation processes and contributes to esophageal cancer progression[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]. \u003cem\u003eMUC13\u003c/em\u003e promotes cancer progression by activating the Wnt/β-catenin pathway in colorectal cancer (CRC)[\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e] and hepatocellular carcinoma (HCC)[\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e], enhancing cell proliferation, EMT, immune evasion, and anti-tumor immune responses; additionally, \u003cem\u003eMUC13\u003c/em\u003e suppresses apoptosis and facilitates metastasis in colon cancer cells by augmenting TNF-induced NF-κB activation[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. In pancreatic cancer, MUC13 activates pro-oncogenic signaling by modulating human epidermal growth factor receptor-2 (HER2) kinase activity, inducing NF-κB Inhibitory Protein (IκB ) phosphorylation and NF-κB p65 nuclear translocation[\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. In gastric cancer (GC), both miR-361-3p and miR-132-3p can suppress tumor progression by targeting and inhibiting \u003cem\u003eMUC13\u003c/em\u003e expression[\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]; furthermore, Takahiro Shimamura et al investigated \u003cem\u003eMUC13\u003c/em\u003e expression in gastric tissues using immunohistochemistry and RT-PCR, demonstrating its progressive upregulation across different tissue types (absent in normal gastric mucosa, detectable in intestinal metaplasia, and highly expressed in 64.9% of GC specimens); and, notably, \u003cem\u003eMUC13\u003c/em\u003e showed particularly prominent overexpression in the intestinal-type GC[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e]. In lung cancer, \u003cem\u003eMUC13\u003c/em\u003e promotes tumor progression by activating the MAPK/ERK signaling pathway[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. In esophageal squamous cell carcinoma (ESCC), Wang H et al performed immunohistochemical analysis of 186 ESCC specimens from patients receiving neoadjuvant chemotherapy and demonstrated that the \u003cem\u003eMUC13/MUC20\u003c/em\u003e combination may serve as a potential prognostic biomarker for ESCC patients following neoadjuvant chemoradiotherapy[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e]. In ovarian cancer, studies have demonstrated elevated \u003cem\u003eMUC13\u003c/em\u003e expression in tumor tissues; \u003cem\u003eMUC13\u003c/em\u003e-transfected cell lines exhibited malignant phenotypes and xenograft models showed upregulated HER2, p21-Activated Kinase 1 (PAK1), and Tumor Protein p38 Mitogen-Activated Protein Kinase (p38) signaling pathways coupled with enhanced tumor growth in response to \u003cem\u003eMUC13\u003c/em\u003e expression[\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e]. These findings collectively establish \u003cem\u003eMUC13\u003c/em\u003e's oncogenic role and promoting cancer cell proliferation, invasion/ metastasis.\u003c/p\u003e\u003cp\u003eThe phenazine biosynthesis-like domain-containing protein (\u003cem\u003ePBLD\u003c/em\u003e), the sole representative of the phenazine biosynthesis-like protein family, has been shown to exhibit reduced expression in multiple cancers. Accumulating evidence indicates that \u003cem\u003ePBLD\u003c/em\u003e functions as a tumor suppressor across various cancer types, ameliorating inflammatory responses, enhancing immune responses, and suppressing tumorigenesis and invasion. In the inflammation-cancer transformation of colorectal tissues, studies have revealed that vitexin modulates macrophage polarization through the Vitamin D Receptor (VDR)/\u003cem\u003ePBLD\u003c/em\u003e pathway, thereby attenuating malignant progression in chronic colitis[\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. Chen S et al have demonstrated reduced mRNA and protein levels of \u003cem\u003ePBLD\u003c/em\u003e in ulcerative colitis (UC) tissues, and \u003cem\u003ePBLD\u003c/em\u003e deficiency has been found to exacerbate immune cell infiltration in intestinal epithelial cells, while mechanistic studies have revealed that \u003cem\u003ePBLD\u003c/em\u003e ameliorates intestinal inflammation and enhances barrier function through suppression of NF-κB signaling transduction[\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]; further studies have demonstrated that rapamycin promotes intestinal barrier repair in UC through activation of the Mechanistic Target of Rapamycin (mTOR)/ \u003cem\u003ePBLD\u003c/em\u003e/Angiomotin (AMOT) signaling pathway[\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. In gastric cancer (GC), studies have confirmed that \u003cem\u003ePBLD\u003c/em\u003e suppresses tumor progression and invasion by inhibiting the Transforming Growth Factor β (TGF-β) signaling-mediated differentiation process and EMT program[\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]. In hepatocellular carcinoma (HCC), \u003cem\u003ePBLD\u003c/em\u003e has been identified as a critical regulator of tumor angiogenic microenvironment remodeling, exerting its inhibitory effects on HCC angiogenesis through the Extracellular Signal-Regulated Kinase (ERK)/Hypoxia-Inducible Factor 1α (HIF-1α)/Vascular Endothelial Growth Factor (VEGF) signaling axis[\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]; furthermore, studies have demonstrated that the \u003cem\u003ePBLD\u003c/em\u003e activator, Cedrelone, suppresses EMT phenotypes and HCC progression via \u003cem\u003ePBLD\u003c/em\u003e upregulation, while \u003cem\u003ePBLD\u003c/em\u003e overexpression promotes apoptosis through activation of Ras and Ras-proximate-1 (Rap1) signaling pathways[\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. In breast cancer, studies have revealed that Circular RNA Lysine Demethylase 4C (circKDM4C) suppresses tumor progression and attenuates doxorubicin resistance by modulating the miR-548p/\u003cem\u003ePBLD\u003c/em\u003e axis[\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eFurthermore, the clinical translational value of \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e as potential biomarkers for BE-EAC progression lies primarily in their feasibility across two commonly used diagnostic modalities: tissue biopsy and liquid biopsy. The most straightforward strategy, compatible with current BE surveillance practices, involves detecting their protein expression via immunohistochemical (IHC) analysis in endoscopic biopsy samples. As a transmembrane mucin, \u003cem\u003eMUC13\u003c/em\u003e is technically suitable for IHC detection. Although direct evidence in EAC remains to be further substantiated, existing studies have demonstrated its consistent detection in FFPE tissue specimens of various gastrointestinal adenocarcinomas\u0026mdash;including esophageal[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] and colorectal cancer[\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]\u0026mdash;indicating that the associated antibody-based detection system is reasonably reliable. Similarly, \u003cem\u003ePBLD\u003c/em\u003e protein has also been validated in FFPE tissues from other cancer types such as hepatocellular carcinoma[\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e][\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e] and breast cancer[\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e]. These studies from other malignancies provide a methodological reference for preliminary assay development. Therefore, future exploration of a dual-marker IHC assay based on \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e for esophageal biopsy specimens may represent a translatable hypothesis worthy of further validation in prospective clinical cohorts.\u003c/p\u003e\u003cp\u003eIn summary, current evidence suggests that \u003cem\u003eMUC13\u003c/em\u003e is frequently highly expressed across multiple malignant tumors, including esophageal cancer and other adenocarcinoma subtypes, whereas \u003cem\u003ePBLD\u003c/em\u003e often shows reduced expression. In other cancer types, they have been implicated in potential oncogenic and tumor-suppressive functions, respectively. However, their specific roles in Barrett\u0026rsquo;s esophagus and esophageal adenocarcinoma remain limited. It is noteworthy that several signaling pathways associated with these genes\u0026mdash;such as Wnt/β-catenin, NF-κB, MAPK/ERK, and EMT\u0026mdash;have been widely reported to play critical roles in the malignant progression from BE to EAC. These include NF-κB-mediated regulation of the inflammatory-immune microenvironment[\u003cspan additionalcitationids=\"CR20 CR21 CR22\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e], aberrant activation of the Wnt/β-catenin pathway[\u003cspan additionalcitationids=\"CR25\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], the EMT process[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e], MAPK/ERK-driven proliferation and anti-apoptotic mechanisms[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e], as well as crosstalk among these pathways[\u003cspan additionalcitationids=\"CR30 CR31 CR32\" citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Combined with the differential expression results from this study\u0026mdash;which showed significant upregulation of \u003cem\u003eMUC13\u003c/em\u003e in both BE and EAC tissues, while \u003cem\u003ePBLD\u003c/em\u003e was mildly elevated in non-dysplastic BE but markedly downregulated in EAC\u0026mdash;we propose that \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e may act as key regulatory molecules during BE-EAC progression, potentially through interactive mechanisms involving the aforementioned pathways. This hypothesis warrants further experimental validation. Such efforts could provide new research directions for early diagnosis, mechanistic insight, and targeted intervention in BE malignant transformation.\u003c/p\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eStudy Limitations\u003c/h2\u003e\u003cp\u003eThis study identified a two-gene diagnostic signature through integrated bioinformatic analyses and demonstrated good performance in internal validation (AUC: 0.94\u0026ndash;0.98). However, two key limitations must be emphasized. First, all reported performance metrics are derived from internal validation only. The limited number of normal samples in the EAC cohort increases the risk of overfitting and may have led to optimistic performance estimates. Second, the study lacks experimental validation. The biological and diagnostic relevance of MUC13 and PBLD, while supported by bioinformatic evidence and previous literature, requires confirmation through experimental methods such as immunohistochemistry, Western blot, or functional assays, as well as validation in independent external cohorts. Therefore, our findings should be regarded as preliminary yet highly promising. The clinical translational potential of these two gene biomarkers warrants further confirmation through well-designed multi-center prospective studies and external validation efforts.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, our integrated bioinformatic study identifies \u003cem\u003eMUC13\u003c/em\u003e and \u003cem\u003ePBLD\u003c/em\u003e as potential diagnostic biomarker for Barrett's esophagus (BE) and esophageal adenocarcinoma (EAC). Their strong discriminatory power throughout the BE-to-EAC sequence, coupled with their involvement in regulating the inflammatory-immune microenvironment and tumor invasion, suggests their potential involvement in malignant progression. These findings provide a foundation for developing early diagnostic, risk stratification and management strategies for BE patients, as well as for elucidating the molecular mechanisms of EAC progression. Further validation of these biomarkers in independent external cohorts and through functional assays is needed to confirm their clinical utility and translational value for improving patient outcomes.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe are grateful to public databases for providing valuable data, and we sincerely appreciate the contributions and efforts of R software and R package developers for their convenience.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research was supported by the Hunan Provincial Natural Science Foundation Science and Health Joint Project, grant number 2023JJ60044; Hunan Province Traditional Chinese Medicine Research Program Project, grant number B2023079; Hunan Province Clinical Medical Technology Innovation Guidance Project, grant number 2021SK51413; Excellent Youth Project of the Research Program of Hunan Provincial Department of Education, grant number 21B0389. The APC was funded by the Project for Advancing Clinical Evidence-Based Capacity of Traditional Chinese Medicine in Treating Advantageous Diseases, grant number czxm-kyb-2025001.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFH conceived and designed the study.\u0026nbsp;TCS\u0026nbsp;and\u0026nbsp;ZM\u0026nbsp;participated in data collection.\u0026nbsp;LJY\u0026nbsp;and XJ performed data analysis and visualization. ZDX and ZT conducted methodological validation.\u0026nbsp;FH\u0026nbsp;and LJY contributed to manuscript drafting and writing. XY provided overall research guidance and secured funding. All authors critically reviewed the manuscript and approved the final version, taking full responsibility for the content.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author(s) declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analysed during the current study are available in Gene Expression Omnibus (GSE13083, GSE1420, GSE39491, and GSE36223) and The Cancer Genome Atlas (TCGA-EAC). These datasets are accessible at https://www.ncbi.nlm.nih.gov/geo/ and https://portal.gdc.cancer.gov/analysis_page?app=CohortBuilder\u0026amp;tab=general. The data used to support the findings of this study are available from the corresponding author upon request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to participate/Consent to publish\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eXia C, et al. Cancer statistics in China and United States, 2022: profiles, trends, and determinants. Chin Med J (Engl). 2022;135:584\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSantucci C, et al. Global trends in esophageal cancer mortality with predictions to 2025, and in incidence by histotype. Cancer Epidemiol. 2023;87:102486.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBray F, et al. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68:394\u0026ndash;424.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSiegel RL, Miller KD, Jemal A. Cancer statistics, 2020. CA Cancer J Clin. 2020;70:7\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eObermannov\u0026aacute; R, et al. Oesophageal cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol. 2022;33:992\u0026ndash;1004.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCurtius K, Rubenstein JH, Chak A, Inadomi JM. Computational modelling suggests that barrett\u0026rsquo;s oesophagus may be the precursor of all oesophageal adenocarcinomas. Gut. 2020;70:1435\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThrift AP. Global burden and epidemiology of barrett oesophagus and oesophageal cancer. Nat Rev Gastroenterol Hepatol. 2021;18:432\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBeydoun AS, Stabenau KA, Altman KW, Johnston N. Cancer risk in barrett\u0026rsquo;s esophagus: a clinical review. Int J Mol Sci. 2023;24:6018.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHolmberg D, Ness-Jensen E, Mattsson F, El-Serag HB, Lagergren J. Risk of oesophageal adenocarcinoma in individuals with barrett\u0026rsquo;s oesophagus. Eur J Cancer. 2017;75:41\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeusten BLAM, et al. Diagnosis and management of barrett esophagus: european society of gastrointestinal endoscopy (ESGE) guideline. Endoscopy. 2023;55:1124\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eThe Gene Ontology Consortium. The gene ontology resource: 20 years and still GOing strong. Nucleic Acids Res. 2019;47:D330\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKanehisa M, Sato Y, Kawashima M, Furumichi M, Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016;44:D457\u0026ndash;462.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28:27\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEllis K, et al. A random forest classifier for the prediction of energy expenditure and type of physical activity from wrist and hip accelerometers. Physiol Meas. 2014;35:2191\u0026ndash;203.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCutler A, Stevens JR. Random forests for microarrays. Methods Enzymol, 422\u0026ndash;32 (2006).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFriedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33:1\u0026ndash;22.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTibshirani R. The lasso method for variable selection in the cox model. \u003cem\u003eStat. Med\u003c/em\u003e. 16, 385\u0026ndash;395 (1997).Liu, C.-Q. Epidemiology of esophageal cancer in 2020 and projections to 2030 and 2040. \u003cem\u003eThorac. Cancer\u003c/em\u003e 14, 3\u0026ndash;11 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu C-Q, et al. Epidemiology of esophageal cancer in 2020 and projections to 2030 and 2040. Thorac Cancer. 2023;14:3\u0026ndash;11.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eO\u0026rsquo;Riordan JM, et al. Proinflammatory cytokine and nuclear factor kappa-B expression along the inflammation-metaplasia-dysplasia-adenocarcinoma sequence in the esophagus. Am J Gastroenterol. 2005;100:1257\u0026ndash;64.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDunbar KB, et al. Association of acute gastroesophageal reflux disease with esophageal histologic changes. JAMA. 2016;315:2104\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMajka J, et al. Interaction of epidermal growth factor with COX-2 products and peroxisome proliferator-activated receptor-γ system in experimental rat barrett\u0026rsquo;s esophagus. Am J Physiol Gastrointest Liver Physiol. 2020;318:G375\u0026ndash;89.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDvorak K, Dvorak B. Role of interleukin-6 in barrett\u0026rsquo;s esophagus pathogenesis. World J Gastroenterol. 2013;19:2307\u0026ndash;12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWeh KM, et al. Prebiotic proanthocyanidins inhibit bile reflux-induced esophageal adenocarcinoma through reshaping the gut microbiome and esophageal metabolome. JCI Insight. 2024;9:e168112.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLyros O, et al. Dickkopf-1 (DKK1) promotes tumor growth via akt-phosphorylation and independently of wnt-axis in barrett\u0026rsquo;s associated esophageal adenocarcinoma. Am J Cancer Res. 2019;9:330\u0026ndash;46.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhou Y, et al. Wnt signaling pathway in cancer immunotherapy. Cancer Lett. 2022;525:84\u0026ndash;96.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYan J, Zhou Y, Wang Y, Liu Y. PARG promotes esophagus cancer cell metastasis by activation of the wnt/β-catenin pathway. Biochem Genet. 2024;62:761\u0026ndash;74.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLu H, et al. Reflux conditions induce E-cadherin cleavage and EMT via APE1 redox function in oesophageal adenocarcinoma. Gut. 2023;73:47\u0026ndash;62.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSrinivas BK, Shivamadhu MC, Jayarama S. Musa acuminata lectin exerts anti-cancer effects on HeLa and EAC cells via activation of caspase and inhibitions of akt, erk, and jnk pathway expression and suppresses the neoangiogenesis in in-vivo models. Int J Biol Macromol. 2021;166:1173\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAhmad S, et al. Long non-coding RNAs regulated NF-κB signaling in cancer metastasis: micromanaging by not so small non-coding RNAs. Semin Cancer Biol. 2022;85:155\u0026ndash;63.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXia Y, Shen S, Verma IM. NF-κB, an active player in human cancers. Cancer Immunol Res. 2014;2:823\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa B, Hottiger MO. Crosstalk between wnt/β-catenin and NF-κB signaling pathway during inflammation. Front Immunol. 2016;7:378.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOh A, et al. NF-κB signaling in neoplastic transition from epithelial to mesenchymal phenotype. Cell Commun Signal : CCS. 2023;21:291.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSolanas G, et al. E-cadherin controls beta-catenin and NF-kappaB transcriptional activity in mesenchymal gene expression. J Cell Sci. 2008;121:2224\u0026ndash;34.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDai Y, et al. Overexpression of MUC13, a poor prognostic predictor, promotes cell growth by activating wnt signaling in hepatocellular carcinoma. Am J Pathol. 2018;188:378\u0026ndash;91.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJonckheere N, Skrypek N, Van Seuningen I. Mucins and tumor resistance to chemotherapeutic drugs. Biochim Biophys Acta. 2014;1846:142\u0026ndash;51.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHan Y, et al. MUC13 promotes the development of esophageal cancer by upregulating the expression of O-glycan process-related molecules. Discov Oncol. 2023;14:123.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen X, et al. RUNX1-MUC13 interaction activates wnt/β-catenin signaling implications for colorectal cancer metastasis. Int J Biol Sci. 2024;20:4999\u0026ndash;5026.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoxtater K, et al. MUC13 drives cancer aggressiveness and metastasis through the YAP1-dependent pathway. Life Sci Alliance. 2023;6:e202301975.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSheng YH, et al. MUC13 protects colorectal cancer cells from death by activating the NF-κB pathway and is a potential therapeutic target. Oncogene. 2017;36:700\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKumari S, et al. MUC13 contributes to rewiring of glucose metabolism in pancreatic cancer. Oncogenesis. 2018;7:19.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCai T, Peng B, Hu J, He Y. Long noncoding RNA BBOX1-AS1 promotes the progression of gastric cancer by regulating the miR-361-3p/mucin 13 signaling axis. Bioengineered. 2022;13:13407\u0026ndash;21.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHe L, Qu L, Wei L, Chen Y, Suo J. Reduction of miR-132-3p contributes to gastric cancer proliferation by targeting MUC13. Mol Med Rep. 2017;15:3055\u0026ndash;61.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShimamura T, et al. Overexpression of MUC13 is associated with intestinal-type gastric cancer. Cancer Sci. 2005;96:265\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePang Y, et al. MUC13 promotes lung cancer development and progression by activating ERK signaling. Oncol Lett. 2022;23:37.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang H, et al. The expression and prognostic significance of mucin 13 and mucin 20 in esophageal squamous cell carcinoma. J Cancer Res Ther. 2015;11(Suppl 1):C74\u0026ndash;79.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMalik S, et al. Emerging role of mucins in antibody drug conjugates for ovarian cancer therapy. J Ovarian Res. 2024;17:161.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen Y, et al. Discovery of vitexin as a novel VDR agonist that mitigates the transition from chronic intestinal inflammation to colorectal cancer. Mol Cancer. 2024;23:196.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen S, et al. Epithelial PBLD attenuates intestinal inflammatory response and improves intestinal barrier function by inhibiting NF-κB signaling. Cell Death Dis. 2021;12:563.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXu Y, et al. Rapamycin promotes the intestinal barrier repair in ulcerative colitis via the mTOR/PBLD/AMOT signaling pathway. Biochim Biophys Acta Mol Basis Dis. 2024;1870:167287.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi D, et al. Mitogen-activated protein kinase activator with WD40 repeats (MAWD) and MAWD-binding protein induce cell differentiation in gastric cancer. BMC Cancer. 2015;15:637.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHan L, et al. PBLD inhibits angiogenesis via impeding VEGF/VEGFR2-mediated microenvironmental cross-talk between HCC cells and endothelial cells. Oncogene. 2022;41:1851\u0026ndash;65.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWu J, Niu Q, Yuan J, Xu X, Cao L. Novel compound cedrelone inhibits hepatocellular carcinoma progression via PBLD and ras/Rap1. Exp Ther Med. 2019;18:4209\u0026ndash;20.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCorrection. circKDM4C suppresses tumor progression and attenuates doxorubicin resistance by regulating miR-548p/PBLD axis in breast cancer - PubMed. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pubmed.ncbi.nlm.nih.gov/33767441/\u003c/span\u003e\u003cspan address=\"https://pubmed.ncbi.nlm.nih.gov/33767441/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi A, et al. Decreased expression of PBLD correlates with poor prognosis and functions as a tumor suppressor in human hepatocellular carcinoma. Oncotarget. 2016;7:524\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Barrett’s Esophagus, Esophageal Adenocarcinoma, biomarker, diagnostic model, MUC13, PBLD, transcriptomics, immune infiltration, machine learning","lastPublishedDoi":"10.21203/rs.3.rs-8004374/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8004374/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e\u003cp\u003eEsophageal adenocarcinoma (EAC) is a highly lethal cancer, with Barrett's esophagus (BE) as its only known precursor. Early diagnosis is challenging, and the key biomarkers and mechanisms driving the BE-to-EAC progression are not fully understood.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eWe integrated transcriptomic data from BE (from the Gene Expression Omnibus (GEO)) and EAC (from The Cancer Genome Atlas (TCGA)) to identify shared differentially expressed genes (SDEGs). We then performed functional enrichment, protein-protein interaction, and immune infiltration analyses. Machine learning algorithms, including Least Absolute Shrinkage and Selection Operator (LASSO) and Random Forest, were applied to screen for diagnostic biomarkers.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eWe identified 101 SDEGs. Subsequent analysis highlighted immune-related regulatory genes. Machine learning pinpointed MUC13 and PBLD as robust diagnostic biomarkers. The diagnostic model exhibited excellent performance, with area under the curve (AUC) values of 0.94 for BE and 0.98 for EAC, supported by nomograms and calibration curves (p\u0026thinsp;\u0026gt;\u0026thinsp;0.9). Mechanistically, MUC13 promotes tumor progression via NF-κB, Wnt/β-catenin, MAPK/ERK, HIF-1α/VEGF, and epithelial-mesenchymal transition (EMT) pathways, driving inflammation-immune crosstalk and metastasis. In contrast, PBLD appears to suppress these processes.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eMUC13 and PBLD are identified as potential biomarkers for the progression from BE to EAC. They play opposing roles in regulating key oncogenic pathways, immune response, and metastasis, offering significant potential for improving early diagnosis and developing targeted therapies.\u003c/p\u003e","manuscriptTitle":"Development of a Diagnostic Model for Barrett's Esophagus and Esophageal Adenocarcinoma Based on Machine Learning and Immune Infiltration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-03 12:17:02","doi":"10.21203/rs.3.rs-8004374/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"25cbc156-37ee-47e6-a974-b7a8acde20b6","owner":[],"postedDate":"December 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-14T10:39:34+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-03 12:17:02","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8004374","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8004374","identity":"rs-8004374","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}