Systematic post-translational modification profiling identifies molecular subtypes and therapeutic targets in ulcerative colitis: evidence from multi-omics integration

Systematic post-translational modification profiling identifies molecular subtypes and therapeutic targets in ulcerative colitis: evidence from multi-omics integration | 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 Systematic post-translational modification profiling identifies molecular subtypes and therapeutic targets in ulcerative colitis: evidence from multi-omics integration Lin Li, Zhenhe Jin, Zhe Shen, Mengchen Luo, Xiaohua Ye, Yanlin Hu, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9497163/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 4 You are reading this latest preprint version Abstract Background Ulcerative colitis (UC) is a debilitating inflammatory bowel disease characterized by heterogeneous clinical presentations and variable treatment responses. Post-translational modifications (PTMs) coordinate immune signaling and mucosal barrier function, yet comprehensive PTM profiling to stratify UC patients and guide therapeutic decisions remains limited. Materials and Methods To identify distinct subtypes of UC, we performed consensus clustering based on PTM-related gene expression. We then applied WGCNA to identify module genes and constructed a diagnostic signature using machine learning algorithms. Single-cell and spatial transcriptomics were used to map cellular localization, and virtual knockdown analysis was conducted to elucidate regulatory networks underlying mucosal inflammation. Results Two distinct PTM subtypes of UC were identified, with Cluster1 enriched in inflammatory pathways ( IL-17 signaling, neutrophil extracellular trap formation) and Cluster2 in metabolic processes. Using machine learning, we developed a 7-gene signature ( HIF1A , SPATA2 , USP30 , MYLIP , GALNT2 , ZC3H12A , GALNT8 ) showing robust predictive capabilities (training AUC: 1.000, validation AUC: 0.945/0.673). Single-cell analysis localized hub genes to epithelial and immune compartments, with CellChat revealing remodeled intercellular communication between high-hub and low-hub states. Spatial transcriptomics demonstrated region-restricted expression hotspots, with HIF1A showing broader activation patterns. Virtual knockdown revealed GALNT8 disruption impaired barrier genes ( TFF3 , ITLN1 ), while HIF1A depletion attenuated inflammatory modules. Conclusions Our results demonstrate that PTM-related genes contribute substantially to UC heterogeneity, with the two subtypes predicting distinct immune-metabolic phenotypes. The seven-gene signature we developed exhibits robust predictive capabilities. Single-cell and spatial analyses link hub genes to immune–epithelial contexts, offering insights for UC stratification and therapeutic decision-making. Ulcerative colitis Post-translational modifications Machine learning Single-cell RNA sequencing Spatial transcriptomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Introduction Inflammatory bowel disease (IBD) encompasses chronic, relapsing inflammatory conditions that primarily involve the intestinal tract 1 , 2 . Ulcerative colitis (UC), a major form of IBD, is defined by diffuse and continuous inflammation confined to the colonic mucosa, typically beginning in the rectum and extending proximally for a variable distance 3 . What makes UC particularly challenging is the marked heterogeneity observed in patient responses. Individuals with comparable clinical activity and endoscopic severity can exhibit substantial inter-individual variation in mucosal immune programs, epithelial barrier disruption, and tissue remodeling 4 . This heterogeneity not only complicates mechanistic interpretation from bulk tissue profiles but also underscores the need for molecular frameworks that can delineate disease-relevant mucosal programs beyond conventional clinical descriptors 5 . Post-translational modifications (PTMs) provide a rapid and reversible means of regulating protein function by modulating activity, stability, and subcellular localization. These modifications influence signaling dynamics, proteostasis, and intracellular trafficking that are essential for epithelial barrier integrity and mucosal immune homeostasis in UC 6,7 . Given this regulatory breadth, multiple PTM pathways have been implicated in intestinal inflammation. For instance, ubiquitin-modifying enzymes, shape inflammatory signaling and epithelial stress responses in colitis, and dysregulated ubiquitination has been repeatedly linked to UC pathogenesis, particularly in the context of barrier dysfunction 8 , 9 . Changes in SUMOylation have also been observed in IBD mucosa, suggesting that defective SUMO-dependent control may contribute to epithelial signaling perturbations 7 , 10 . Additionally, proteome-wide studies support the presence of PTM remodeling during experimental colitis, including shifts in lysine acetylation that involve metabolic enzymes and stress-response proteins in DSS-induced models 11 . Moreover, glycosylation programs are increasingly recognized as key regulators of gut homeostasis and inflammation, further reinforcing the view that PTM reprogramming in UC extends beyond a single modification class 12 . Despite these advances, much of the current evidence remains centered on individual PTM types or specific experimental settings, and a unified, transcriptome-informed perspective of PTM-associated regulatory programs and their relationships to cellular compartments in UC remains insufficiently defined. To address this gap, we profiled PTM-associated heterogeneity in UC using an integrative, multi-level framework. Using a curated PTM gene panel spanning 21 modification categories, we integrated and batch-corrected bulk transcriptomic cohorts to define PTM-associated subtypes and characterize their transcriptional and pathway features through differential analysis, WGCNA, and enrichment profiling. We then applied machine-learning approaches to derive a parsimonious seven-gene signature ( HIF1A , SPATA2 , USP30 , MYLIP , GALNT2 , ZC3H12A , and GALNT8 ) and assessed its predictive utility with calibration and decision-curve analyses. Finally, immune infiltration was evaluated and signature expression was mapped across single-cell and spatial transcriptomic layers to place PTM-associated programs within the cellular and spatial context of inflamed UC mucosa. Materials and methods Transcriptomic data acquisition We began our analysis by gathering bulk transcriptomic data from two publicly available datasets of colonic mucosal biopsies: GSE75214 13 (UC = 97, Control = 22) and GSE87466 14 (UC = 87, Control = 21). The two cohorts were merged to form a discovery set, and inter-study batch effects were corrected using the ComBat algorithm in the sva R package 15 . The batch-adjusted expression matrix (UC = 184, Control = 43) was used for differential expression analysis and PTM-related integrative analyses. Additional bulk cohorts (GSE47908 and GSE206285) were compiled for external validation 16,17 . PTM-related genes were curated from published literature covering 21 modification categories (Supplemental Table S1) 18 . The complete analytical workflow is schematized in Fig 1. Consensus clustering Consensus clustering was performed using the ConsensusClusterPlus R package to identify molecular subtypes based on PTM-related gene expression profiles. Clustering stability was evaluated across multiple cluster numbers, and the optimal number was determined by cumulative distribution function (CDF) curves and delta area plots. Principal component analysis (PCA) was applied to validate the separation between identified subtypes. Differential gene analysis To identify differentially expressed genes (DEGs) in bulk transcriptome data, we used the limma R package to screen DEGs from the integrated GSE75214 and GSE87466 datasets. The criteria for DEG selection were\|log₂ fold change\| > 0.585 and adjusted p < 0.05. Furthermore, we used the R package ggplot2 to generate volcano plots of the DEGs, providing a visual representation of their distribution to facilitate subsequent analyses. Weighted gene co-expression network analysis Weighted gene co-expression network analysis (WGCNA) was conducted to identify co-expression modules associated with UC subtypes. Sample outliers were detected through hierarchical clustering, and 43 samples were excluded 19 . A soft-thresholding power of β = 14 was selected to approximate scale-free topology (R² = 0.85). Modules were identified using dynamic tree cutting based on the topological overlap matrix (TOM), and highly similar modules were merged according to eigengene correlations. Module–trait relationships were quantified by correlating module eigengenes with subtype labels, and key subtype-associated modules (cyan and greenyellow) were carried forward for downstream candidate selection and modeling. Functional enrichment and pathway activity analyses GO and KEGG enrichment analyses were performed for the screened PTM genes using the "clusterProfiler" R package, with human gene annotations provided by "org.Hs.eg.db" (keyType = “SYMBOL”). GO enrichment was evaluated across the biological process (BP), cellular component (CC), and molecular function (MF) domains, while KEGG pathway enrichment was implemented using a KEGG Medicus gene-set collection in GMT format (c2.cp.kegg_medicus.v2025.1.Hs.symbols.gmt). During the analysis process, we retained the enriched items according to the significance level (P < 0.05) to ensure the reliability and accuracy of the results. Gene set enrichment analysis (GSEA) was performed to identify enriched pathways between subtypes. Genes were ranked by fold-change, and enrichment was tested against predefined gene sets with permutation-based significance assessment 20 . Gene set variation analysis (GSVA) was applied to estimate sample-level pathway activity scores, converting gene expression matrices into pathway enrichment matrices. Both analyses were conducted to characterize biological programs underlying each subtype 21 . Machine learning-based prediction model A prediction model for UC development was constructed based on PTM-related genes. Candidate genes were identified by intersecting DEGs between subtypes, WGCNA module genes (cyan and greenyellow), and curated PTM-related genes 22 . We next benchmarked multiple machine-learning algorithms and model combinations in the training cohort and two external validation cohorts. Model performance was primarily evaluated using the area under the receiver operating characteristic curve (AUC) 23 . SHAP (SHapley Additive exPlanations) analysis was applied to identify the most influential features from the candidate gene set 24 . A nomogram was constructed for individualized subtype prediction based on the final gene signature. Model reliability was assessed through calibration curves and decision curve analysis. Immune microenvironment analysis The CIBERSORT algorithm was applied to estimate the relative proportions of 22 immune cell types from bulk expression profiles 25 . Immune cell compositions were compared across control, Cluster1, and Cluster2 groups using Wilcoxon rank-sum test or Kruskal-Wallis test. Spearman correlation coefficients were calculated between hub gene expression levels and immune cell proportions. Correlations were visualized using heatmaps and network diagrams, with significance threshold set at P < 0.05. Single-cell RNA sequencing analysis The scRNA-seq dataset GSE116222 26 was processed using Seurat (version 4.1.0). Quality control retained cells with nFeature_RNA between 500 and 4000, nCount_RNA < 6000, and percent.mt < 5%. Data normalization was performed using NormalizeData with default parameters. Highly variable genes (n = 2,000) were identified using FindVariableFeatures (vst method). Expression data were scaled using ScaleData, and principal component analysis was performed on the scaled data. Graph-based clustering was conducted using FindNeighbors and FindClusters (resolution = 0.4). UMAP dimensionality reduction was performed using the first 20 principal components (n.neighbors = 30, min.dist = 0.3, cosine metric). Cell type annotation was performed based on canonical marker gene expression from the CellMarker database. Cell proportions between UC and normal samples were compared using Fisher's exact test ( P < 0.05). Pseudotime trajectory analysis Pseudotime trajectory analysis was performed using Monocle 2 (version 2.26.0). The annotated cell populations were selected for trajectory construction. A CellDataSet object was created from raw UMI counts, with the expressionFamily parameter set to a negative binomial distribution. Size factors and dispersion were estimated using the estimateSizeFactors and estimateDispersions functions. Ordering genes were selected based on mean expression ≥ 0.1 and dispersion_empirical ≥ 1 × dispersion_fit. Dimensionality reduction was performed using reduceDimension with method = "DDRTree" and max_components = 2, followed by cell ordering using the orderCells function. Intercellular ligand-receptor communication analysis The CellChat package (version 1.6.0) was used to analyze intercellular communication among all cell subpopulations. A module score was calculated for each cell based on the identified gene signature using AddModuleScore. Cells were stratified into High_Hub and Low_Hub groups based on median module score. Separate CellChat objects were constructed for each group using normalized expression data. Ligand-receptor interactions were annotated using CellChatDB.human. Communication probabilities were computed using computeCommunProb (type = "triMean", min.cells = 10). By calculating the number of ligands and the weighting values of cell–cell interactions, a weighted, directed network was constructed. Pathway-level communication was aggregated using computeCommunProbPathway. The two CellChat objects were merged for comparative analysis of signaling information flows between High_Hub and Low_Hub groups. Simulated knockout profile of key genes Virtual gene knockout was performed using scTenifoldKnk to predict gene functions. scTenifoldKnk constructs a denoised single-cell gene regulatory network (scGRN) from scRNA-seq data, then replicates the scGRN with target gene out-edges set to zero, generating a pseudo-knockout scGRN. The two scGRNs are mapped into low-dimensional space, and the distance between gene projections reveals the impact of gene knockout: larger disturbances indicate greater importance of the target gene. For each hub gene (HIF1A, GALNT2, GALNT8, ZC3H12A, MYLIP, and SPATA2), raw UMI counts from inflamed mucosal cells were prepared with the top 10,000 highly variable genes as the modeling feature set. Virtual knockout was performed with parameters: qc = TRUE, qc_mtThreshold = 0.1, qc_minLSize = 1000, nc_nNet = 10, nc_nCells = 500. Differentially regulated genes (DRGs) were identified with adjusted P < 0.05. Functional enrichment analysis was performed using clusterProfiler (ont = "BP", pvalueCutoff = 0.05) on the top 20 genes significantly affected by knockout. Spatial transcriptomics Spatial transcriptomic data from the 10x Visium platform were processed using Seurat (version 4.1.0) within an R environment. The filtered count matrix (filtered_feature_bc_matrix.h5) was imported using Load10X_Spatial, and spot-level mitochondrial content was computed. Quality control steps were implemented with the following thresholds: nCount_Spatial > 0, nFeature_Spatial > 200, nFeature_Spatial < 8000, and percent.mt < 20%. Data normalization was performed using the SCTransform function with assay = "Spatial". Subsequently, PCA was applied for dimensionality reduction, and clustering was conducted using the Leiden-based FindClusters function with resolution = 1.2 and the first 10 principal components. To infer spot-level cellular composition, RCTD (Robust Cell Type Decomposition; spacexr) was applied in doublet mode (doublet_mode = "doublet"; max_cores = 10), using a curated single-cell reference prepared by subsampling 500 cells per subtype and excluding rare subtypes with fewer than 25 cells. Predicted cell type annotations (first_type) were transferred to the spatial Seurat object for downstream analyses. Additionally, spot-level enrichment scores for cell-type programs were computed using ssGSEA on the top 10 marker genes per subtype identified from the single-cell reference. For spatially informed ligand-receptor communication analysis, CellChat (version 1.6.1) was employed in spatial mode with CellChatDB.human as reference, incorporating tissue coordinates and spot diameter scaling (spot.size = 65). Communication probabilities were computed using a truncated mean approach with distance constraints (interaction.range = 250, contact.range = 100) and filtered with min.cells = 10. Finally, the spatial expression patterns of the identified hub genes were visualized using SpatialFeaturePlot and vizAllTopics. Results PTM-related gene dysregulation in UC Batch effects were corrected across GSE75214 and GSE87466 datasets to ensure analytical consistency ( Fig. 2A-B ). PTM-related genes were next summarized across 21 PTM categories. From a curated set of 1,427 PTM-related genes covering 21 PTM categories (2–426 genes per category; Fig. 2C ), 1,164 genes were detected in the integrated expression matrix. Differential expression analysis identified 84 PTM-related DEGs (51 upregulated, 33 downregulated; 7.22% of detected PTM genes; Fig. 2D; Supplementary Table S2 ). In the GSE75214 dataset, 1,303 PTM genes were detected, yielding 80 activity-associated DEGs (43 upregulated, 37 downregulated; 6.14%; Fig. 2E; Supplementary Table S3 ). In UC versus control comparisons, upregulated PTM-related genes included S100A8, NOS2, MNDA, SOCS3, IFI16, TRIM22, FGR , and MSN , whereas USP30, USP2, RNF128 , and ASB13 were downregulated. Active versus inactive UC comparisons showed concordant patterns with increased expression of S100A8, NOS2, BIRC3, UBD , and PTPRC and decreased expression of NEDD4L, GALNT12, RNF152 , and B3GNT7 . Seven genes ( S100A8, NOS2, MNDA, SOCS3, TRIM22, MSN, HIF1A ) were consistently dysregulated across both comparisons. Two PTM-associated molecular subtypes identified by consensus clustering GO and KEGG enrichment analyses revealed that PTM-related DEGs were enriched in protein polyubiquitination, ubiquitin ligase complex, and 26S proteasome-mediated protein degradation pathways ( Fig. 3A-B ). To evaluate clustering stability across different cluster numbers, additional consensus matrices for k = 3–6 were provided as supplementary evidence ( Supplementary Fig. S2 ).Consensus clustering of 84 PTM-related DEGs identified an optimal two-cluster solution (k = 2; Fig. 3C-E ), classifying UC samples into Cluster1 and Cluster2. PCA confirmed clear separation between subtypes ( Fig. 3F ). Differential expression analysis between subtypes revealed 1,234 DEGs, with the top DEGs showing reciprocal expression patterns that distinctly segregated samples ( Fig. 3G-H; Supplementary Table S4) . WGCNA identifies subtype–specific co-expression modules Weighted gene co-expression network analysis was performed using a soft-thresholding power of 14 (scale-free topology fit R² = 0.85; Fig. 4A-B ), yielding seven modules after dynamic tree cutting and eigengene-based merging ( Fig. 4C-D ). Module-trait correlation analysis identified the MEcyan module as strongly associated with Cluster1 ( r = 0.82) and negatively with Cluster2 ( r = −0.82). Conversely, the MEgreenyellow module showed positive correlation with Cluster2 ( r = 0.79) and negative correlation with Cluster1 ( r = −0.79; Fig. 4E ). Module membership (MM) in the cyan module correlated negatively with Cluster2 gene significance (GS; cor = −0.988; Fig. 4F ), whereas MM in the greenyellow module correlated positively with Cluster2-GS (cor = 0.984; Fig. 4G ). Genes from both modules were selected for downstream analyses, and the full gene lists of the MEcyan and MEgreenyellow modules are provided in Supplementary Table S5-6 . Distinct biological programs characterize UC subtypes GSEA revealed that Cluster1 was enriched in immune and inflammatory pathways, including IL-17 signaling, cytokine-cytokine receptor interaction, pertussis, viral protein interaction with cytokine receptors, and neutrophil extracellular trap formation ( Fig. 5A ). In contrast, Cluster2 showed enrichment in metabolic pathways such as mineral absorption, drug metabolism via cytochrome P450, xenobiotic metabolism, bile secretion, and pancreatic secretion ( Fig. 5B-C ). GSVA results were consistent with GSEA findings. The heatmap of top 20 differential pathways demonstrated that Cluster1 exhibited elevated activity in innate immune and inflammatory pathways, whereas Cluster2 showed higher activity in metabolic and secretory processes ( Fig. 5D ). Violin plots of representative pathway GSVA scores confirmed these subtype-specific patterns ( Fig. 5E-F ). In addition, GO enrichment analyses supporting these distinct biological programs between subtypes are provided in Supplementary Fig. S4 Seven-gene signature predicts UC risk To develop a minimal predictive gene set, we intersected PTM-related genes with cluster DEGs and WGCNA key modules (cyan and greenyellow), identifying 26 candidate genes ( Fig. 6A; Supplementary Table S8 ). Multiple machine learning algorithms were compared across training and validation cohorts( Supplementary Table S9 ). The Stepglm(both) + Random Forest + Naive Bayes combination achieved optimal performance with AUCs of 1.000 (training set), 0.945 (GSE47908), and 0.673 (GSE206285; mean AUC = 0.873; Fig. 6B ). Based on SHAP feature importance and model stability, seven key genes were selected: HIF1A, SPATA2, USP30, MYLIP, GALNT2, ZC3H12A , and GALNT8 ( Supplementary Table S7) . ROC analysis in the training set demonstrated strong discriminatory power for individual genes, particularly HIF1A, USP30, ZC3H12A , and GALNT2 ( Fig. 6C ). A nomogram was constructed for clinical application ( Fig. 6D ). SHAP analysis identified GALNT2 (22.8%), USP30 (20.6%), GALNT8 (19.0%), and HIF1A (15.9%) as major contributors, followed by ZC3H12A (7.7%), MYLIP (7.1%), and SPATA2 (7.0%; Fig. 6G; Supplementary Fig. S5 ). Calibration curves demonstrated good agreement between predicted and observed probabilities, and decision curve analysis indicated net clinical benefit of the seven-gene model ( Fig. 6E-F ) , with model coefficients provided in Supplementary Table S10 . Immune cell infiltration patterns differ between UC subtypes CIBERSORT analysis estimated 22 immune cell populations across controls, Cluster1, and Cluster2 samples. Stacked bar plots revealed substantial inter-sample variation and subtype-specific immune compositions ( Fig. 7A ). Compared to controls, Cluster1 showed elevated proportions of monocytes, M0 macrophages, activated mast cells, and neutrophils, whereas Cluster2 exhibited relatively higher levels of M2 macrophages, CD8 T cells, and regulatory T cells ( Fig. 7B ). Correlation analysis between the seven signature genes and immune cell infiltration revealed extensive associations with myeloid and granulocyte populations, particularly neutrophils, macrophages, and mast cells ( Fig. 7C; Supplemental Fig. 6A-B ). HIF1A showed positive correlations with neutrophils, CD8 T cells, and activated mast cells (all P < 1 × 10⁻⁶), while USP30 correlated with neutrophils and M2 macrophages ( P < 1 × 10⁻⁶). Single-cell analysis reveals cell type-specific expression patterns The GSE116222 dataset comprised 22,609 cells with 10,470 genes after quality control filtering. Using the top 20 principal components, 15 clusters were identified and annotated into eight major cell types based on canonical markers: enterocytes, crypt cells, epithelial cells, goblet cells, B cells, regulatory CD4 T cells, monocytes, and mast cells ( Fig. 8A-B ). Cell type proportions varied across conditions and individuals ( Fig. 8C ). Supporting plots for quality assessment, principal component selection, and highly variable gene identification are provided in Supplementary Fig. S7 . Mapping of the seven signature genes onto the single-cell atlas revealed distinct cell type-specific expression patterns across epithelial and immune compartments ( Fig. 8D ). Trajectory inference analysis demonstrated pseudotime-dependent regulation, with HIF1A , MYLIP , SPATA2 , and USP30 showing dynamic patterns along immune lineage trajectories, while GALNT2 , GALNT8 , and ZC3H12A exhibited temporal changes along epithelial trajectories ( Fig. 8E-G ). CellChat analysis revealed differential intercellular communication networks between high- and low-signature states, with altered outgoing signaling patterns among cell populations ( Fig. 8H-I ). Virtual knockout analysis predicts gene-specific perturbation networks In silico virtual knockout simulations predicted downstream perturbation networks for six of the seven signature genes ( USP30 was excluded due to insufficient predicted effects). GALNT2 knockout primarily affected translation-related genes ( Fig. 9A ). GALNT8 knockout showed broader effects, perturbing ribosomal genes alongside epithelial markers including ITLN1 , TFF3 , WFDC2 , AGR2 , and SPINT2 ( Fig. 9B ). HIF1A , MYLIP , SPATA2 , and ZC3H12A knockouts predominantly affected translation machinery ( Fig. 9C-F ). Spatial transcriptomics confirms focal gene expression in UC tissue Spatial transcriptomic analysis of UC and control sections revealed tissue- and disease-specific heterogeneity. Immune-related spots showed increased clustering and abundance in UC compared to sparse distribution in controls, where epithelial and crypt compartments dominated ( Fig. 10A-D ). The identified signature genes displayed region-restricted expression with discrete high-signal hotspots in UC sections versus weaker, scattered signals in controls ( Fig. 10E-H ). HIF1A showed broader spatial activation, whereas other genes exhibited more focal expression patterns. The combined seven-gene score demonstrated more prominent spatial activation in UC, delineating regions of elevated transcriptional activity ( Fig. 10I ). Discussion This study employed a PTM-oriented, multi-omics framework to dissect the molecular heterogeneity of UC and identify clinically actionable biomarkers. By systematically integrating bulk transcriptomic profiles with a curated panel of PTM-related genes spanning 21 modification categories, we uncovered consistent perturbations in PTM-associated regulatory programs that distinguish UC from healthy controls and correlate with disease activity states. The convergence of these molecular signals across independent cohorts establishes PTM-related genes as a biologically meaningful feature space for understanding UC pathogenesis and stratifying patient populations based on underlying molecular mechanisms. Consensus clustering of PTM-related differentially expressed genes revealed two molecularly distinct UC subtypes, each characterized by divergent pathway enrichment signatures. Cluster1 exhibited pronounced activation of immune-inflammatory circuits, particularly IL-17 signaling pathways, cytokine-cytokine receptor interactions, and neutrophil extracellular trap formation mechanisms. These molecular features align with accumulating evidence positioning the IL-23/IL-17 axis as a central orchestrator of mucosal inflammation in UC, where it coordinates neutrophil recruitment, promotes tissue damage through NET-mediated mechanisms, and drives chronic inflammatory responses 27 – 29 . The dominance of inflammatory pathways in Cluster1 suggests these patients may harbor an immune-hyperactive type characterized by sustained neutrophil activation and robust cytokine production 29 . In contrast, Cluster2 demonstrated preferential enrichment of metabolic and secretory pathways, including bile secretion, xenobiotic metabolism via cytochrome P450 enzymes, and pancreatic secretion processes. This metabolic signature likely reflects adaptive epithelial responses to chronic inflammation, where mucosal cells recalibrate their metabolic machinery to support barrier repair, maintain secretory functions, and manage cellular stress under inflammatory conditions 30 , 31 . The identification of these two subtypes carries significant translational implications, as immune-dominant versus metabolic-dominant types may exhibit differential therapeutic responsiveness: the former potentially benefiting from targeted immunosuppression, while the latter might require barrier-enhancing or metabolic-supporting interventions 28 , 32 . Notably, the prominence of ubiquitin-related processes across both subtypes, manifesting through enhanced protein polyubiquitination, ubiquitin ligase complex activity, and proteasome-mediated degradation. This suggests fundamental dysregulation of proteostasis networks in UC mucosa 8 , 33 . Ubiquitination serves as a regulatory checkpoint governing inflammatory mediator stability, including NF-κB pathway components and pattern recognition receptors, while orchestrating epithelial stress responses through selective protein degradation 8 , 34 . The convergence of ubiquitin pathway enrichment with PTM-related gene dysregulation supports a conceptual framework wherein proteostasis stress intersects with inflammatory remodeling, representing a mechanistically plausible therapeutic leverage point 6 , 34 . Recent studies demonstrating that specific ubiquitin-modifying enzymes regulate colitis severity provide experimental support for this therapeutic rationale 8 , 9 , 35 . WGCNA provided critical evidence that PTM-associated heterogeneity operates through coordinated transcriptional programs rather than isolated gene alterations. The identification of two key modules exhibiting robust correlations with identified subtypes (| r | > 0.79) and exceptionally tight module membership-gene significance relationships (| r | > 0.98) indicates these co-expression modules capture stable, biologically coherent regulatory circuits. This network-guided organization proved valuable for biomarker discovery, enabling candidate gene prioritization based on embeddedness within subtype-defining transcriptional architectures rather than univariate differential expression alone, enriching for genes with functional roles in UC pathogenesis. Through integration of subtype DEGs, WGCNA modules, and curated PTMRGs, we derived a seven-gene signature ( HIF1A , SPATA2 , USP30 , MYLIP , GALNT2 , ZC3H12A , GALNT8 ) demonstrating robust performance across discovery and validation cohorts. This panel demonstrates mechanistic coherence, careful selection grounded in UC biology. HIF1A , encoding hypoxia-inducible factor 1-α, governs cellular adaptation to oxygen deprivation and has been extensively linked to intestinal barrier preservation during inflammatory stress, coordinating epithelial survival, tight junction integrity, and mucosal healing 36 , 37 . The inflamed UC mucosa experiences localized hypoxia due to increased metabolic demand and immune cell infiltration, creating microenvironmental conditions activating HIF1A -dependent programs 36 . The inclusion of ubiquitin-pathway regulators including USP30 (a mitochondrial deubiquitinase linking inflammation to mitochondrial quality control) and SPATA2 (a scaffolding protein modulating CYLD-dependent TNF/NF-κB signaling), directly reflects the mechanistic importance of ubiquitin-mediated regulation in UC inflammatory circuitry 8 , 34 . GALNT2 and GALNT8 , polypeptide N-acetylgalactosaminyltransferases initiating mucin-type O-glycosylation, regulate epithelial glycobiology programs critical for mucus barrier integrity and host-microbe interactions 12 . SHAP analysis revealed GALNT2 (22.8%), USP30 (20.6%), GALNT8 (19.0%), and HIF1A (15.9%) as dominant contributors, collectively accounting for approximately 80% of model predictions, indicating that glycosylation-related and ubiquitin-related PTM regulators constitute particularly informative UC biomarkers, likely because they integrate converging signals from epithelial stress, inflammatory activation, and barrier remodeling 38 . Immune deconvolution illuminated cellular landscapes underlying molecular subtypes. Cluster1 demonstrated elevated myeloid/granulocytic infiltration (monocytes, M0 macrophages, activated mast cells, neutrophils), while Cluster2 showed enriched M2 macrophages and regulatory T cells 39 . Correlation analyses revealed particularly robust associations between HIF1A / USP30 expression and neutrophil infiltration plus activated mast cells, mechanistically connecting these genes to specific inflammatory programs 34 , 36 , 40 and suggesting PTM-based stratification captures fundamental variations in tissue inflammatory architecture with implications for predicting differential therapeutic responses 32 . Single-cell analysis revealed marked compartmentalization of signature gene expression across the cellular complexity of intestinal mucosa. Inflammation-associated components ( HIF1A , ZC3H12A , MYLIP , SPATA2 ) showed predominant expression in innate immune populations, particularly mast cells and monocytes where they likely calibrate inflammatory output intensity and cytokine production kinetics. In contrast, glycosylation-related genes ( GALNT2 , GALNT8 ) exhibited enrichment in epithelial lineages, especially goblet cells and crypt progenitors responsible for mucin secretion and epithelial renewal 41 , 42 . This compartmentalized expression positions these genes at the nexus of their respective functional programs, namely immune activation and mucus barrier assembly. Pseudotime trajectory analysis added temporal dimensionality, revealing dynamic, state-dependent regulation along immune activation and epithelial differentiation continua, indicating PTM programs respond to fluctuating microenvironmental cues (oxygen tension, cytokine exposure, microbial signals) rather than constitutive activation 36 , 43 . This aligns with UC biology where focal inflammatory hotspots coexist with relatively quiescent regions, creating tissue microdomains with divergent PTM regulatory requirements 32 , 41 . CellChat analysis revealed substantial intercellular communication remodeling between high- and low-signature states, suggesting PTM programs influence not only cell-intrinsic functions but also coordinate multicellular tissue responses through modulation of cell-cell signaling circuits. An important aspect of our study is the application of virtual knockout analysis using scTenifoldKnk, representing a novel approach in IBD research for predicting gene regulatory hierarchies through single-cell network perturbation modeling. Unlike conventional differential expression or correlation analyses that capture static associations, this emerging computational method constructs denoised gene regulatory networks from single-cell data and simulates gene ablation effects by setting target gene edges to zero, revealing downstream perturbation cascades and regulatory dependencies that would be experimentally challenging and resource-intensive to assess systematically 44 . For most hub genes ( HIF1A , MYLIP , SPATA2 , ZC3H12A ), computational ablation primarily affected translational machinery, suggesting coupling to broad biosynthetic and stress-adaptive programs 44 . However, GALNT8 knockout produced a distinctive perturbation signature affecting epithelial-specific functional markers including ITLN1 (antimicrobial lectin), TFF3 (mucus-associated peptide critical for barrier repair), WFDC2 (protease inhibitor), AGR2 (protein disulfide isomerase supporting mucin folding), and SPINT2 (serine protease inhibitor regulating epithelial integrity). This divergent pattern indicates GALNT8 exerts specialized regulatory influence over epithelial functional programs beyond core protein synthesis, positioning it as a potentially critical node coordinating mucus production, antimicrobial defense, and cell adhesion maintenance 12 , 45 . The identification of LRRC75A-AS1 as a recurrently perturbed node across multiple knockouts highlights this long non-coding RNA as a potential integrative hub coordinating cellular responses to diverse PTM signals 46 . While requiring experimental validation, these predictions provide hypothesis-generating insights into gene regulatory hierarchies and functional dependencies unlikely to emerge from static expression profiling, representing an innovative application that advances UC molecular stratification research beyond descriptive associations toward predictive regulatory modeling. Spatial transcriptomics anchored molecular signatures within tissue architecture, revealing region-restricted expression with discrete high-signal hotspots in inflamed tissue contrasting with diffuse low-intensity signals in controls 47 . HIF1A demonstrated broader spatial activation reflecting widespread hypoxic stress, while other genes showed focal patterns corresponding to inflammatory foci or epithelial remodeling sites 36 , 40 . The composite seven-gene score revealed spatial activation gradients demarcating pathological activity zones[47]. This pronounced spatial heterogeneity suggests treatment responses may critically depend on therapeutic penetration into these restricted hotspots, providing rationale for integrating bulk, single-cell, and spatial readouts to comprehensively characterize UC molecular pathology 32 , 47 . Several limitations warrant consideration. Our transcriptome-based analyses require validation through proteomics and modification-specific assays to confirm transcript-level changes reflect actual PTM activity states 6 . Subtype overlap suggests incorporating clinical covariates and longitudinal sampling may sharpen boundaries and improve clinical applicability 5 . Virtual knockout predictions require experimental validation through targeted perturbation in relevant model systems 44 . External validation variability indicates need for larger prospective studies with standardized protocols 5 . Despite these constraints, convergence across multiple analytical modalities strongly supports the biological validity of our PTM-focused framework 6 , 32 . Conclusion In summary, we established the first comprehensive PTM-centric stratification of UC, identifying two molecularly distinct subtypes and developing a mechanistically coherent seven-gene signature for prediction of UC risk. These findings provide foundation for PTM-targeted therapeutic strategies and precision medicine approaches enabling molecular type-based treatment matching in UC. Declarations Clinical trial number not applicable. Conflict of interest: the authors involved declared no conflict of interest. Ethical approval: Not applicable. This study used only publicly available, de-identified datasets; therefore, additional ethical approval and informed consent were not required. Funding Declaration: this work was funded by the National Key R&D Program of China (No 2023YFC2507300 to Z.S.), Zhejiang High-Level Talent Innovation Leadership Project (No 2023R5239 to Z.S), Research on the molecular mechanisms and targeting strategies of STRIP2 in regulating the progression and metastasis of colorectal cancer (NO. JY2024-6-07) and Clinical Medicine Research Special Fund of the Zhejiang Medical Association (No. 2025ZYC-A139). Author Contribution Lin Li and Zhenhe Jin conceived and designed the study. Xiaohua Ye and Zhenhe Jin performed data acquisition/experiments. Mengchen Luo conducted data analysis. Zhe Shen and Jin Ding supervised the project. Yanlin Hu drafted the manuscript, and all authors revised and approved the final version. References GBD 2017 Inflammatory Bowel Disease Collaborators. The global, regional, and national burden of inflammatory bowel disease in 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet Gastroenterol Hepatol . 2020;5(1):17-30. doi:10.1016/S2468-1253(19)30333-4 Neurath MF. Current and emerging therapeutic targets for IBD. Nat Rev Gastroenterol Hepatol . 2017;14(5):269-278. doi:10.1038/nrgastro.2016.208 Ungaro R, Mehandru S, Allen PB, Peyrin-Biroulet L, Colombel JF. Ulcerative colitis. Lancet . 2017;389(10080):1756-1770. doi:10.1016/S0140-6736(16)32126-2 de Souza HSP, Fiocchi C. Immunopathogenesis of IBD: current state of the art. Nat Rev Gastroenterol Hepatol . 2016;13(1):13-27. doi:10.1038/nrgastro.2015.186 Selin KA, Hedin CRH, Villablanca EJ. Immunological Networks Defining the Heterogeneity of Inflammatory Bowel Diseases. J Crohns Colitis . 2021;15(11):1959-1973. doi:10.1093/ecco-jcc/jjab085 Ding J, Wang H, Yang Z, Wang X, Cao Z. Protein acylation in inflammatory diseases: from mechanisms to therapeutic strategies. Cell Commun Signal . 2025;23:488. doi:10.1186/s12964-025-02484-6 Ma XN, Li MY, Qi GQ, Wei LN, Zhang DK. SUMOylation at the crossroads of gut health: insights into physiology and pathology. Cell Commun Signal . 2024;22(1):404. doi:10.1186/s12964-024-01786-5 Ruan J, Schlüter D, Naumann M, Waisman A, Wang X. Ubiquitin-modifying enzymes as regulators of colitis. Trends Mol Med . 2022;28(4):304-318. doi:10.1016/j.molmed.2022.01.006 Liu Z, Liu J, Wei Y, et al. Ubiquitin-specific protease 25 ameliorates ulcerative colitis by regulating the degradation of phosphor-STAT3. Cell Death Dis . 2025;16(1):5. doi:10.1038/s41419-024-07315-z Mustfa SA, Singh M, Suhail A, et al. SUMOylation pathway alteration coupled with downregulation of SUMO E2 enzyme at mucosal epithelium modulates inflammation in inflammatory bowel disease. Open Biol . 2017;7(6):170024. doi:10.1098/rsob.170024 Wang JM, Lin SR, Zhu YB, et al. Proteomic analysis of lysine acetylation reveals that metabolic enzymes and heat shock proteins may be potential targets for DSS-induced mice colitis. Int Immunopharmacol . 2021;101(Pt B):108336. doi:10.1016/j.intimp.2021.108336 Bergstrom K, Xia L. The barrier and beyond: Roles of intestinal mucus and mucin-type O-glycosylation in resistance and tolerance defense strategies guiding host-microbe symbiosis. Gut Microbes . 2022;14(1):2052699. doi:10.1080/19490976.2022.2052699 Wu J, Lv Y, Hao P, et al. Immunological profile of lactylation-related genes in Crohn’s disease: a comprehensive analysis based on bulk and single-cell RNA sequencing data. J Transl Med . 2024;22(1):300. doi:10.1186/s12967-024-05092-z Chen K, Shang S, Yu S, Cui L, Li S, He N. Identification and exploration of pharmacological pyroptosis-related biomarkers of ulcerative colitis. Front Immunol . 2022;13:998470. doi:10.3389/fimmu.2022.998470 Leek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics . 2012;28(6):882-883. doi:10.1093/bioinformatics/bts034 Linggi B, Jairath V, Zou G, et al. Meta-analysis of gene expression disease signatures in colonic biopsy tissue from patients with ulcerative colitis. Sci Rep . 2021;11(1):18243. doi:10.1038/s41598-021-97366-5 Pan Z, Lin H, Fu Y, et al. Identification of gene signatures associated with ulcerative colitis and the association with immune infiltrates in colon cancer. Front Immunol . 2023;14:1086898. doi:10.3389/fimmu.2023.1086898 Wang X, Liu Y, Fu J, Li Y, Zhao M, Tian Q. Systematic post-translational modification genome wide identifies therapeutic targets for Alzheimer’s disease: evidence from multi-cohort analysis. J Prev Alzheimers Dis . Published online October 30, 2025:100422. doi:10.1016/j.tjpad.2025.100422 Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics . 2008;9:559. doi:10.1186/1471-2105-9-559 Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A . 2005;102(43):15545-15550. doi:10.1073/pnas.0506580102 Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics . 2013;14:7. doi:10.1186/1471-2105-14-7 Weighted gene co-expression network analysis identifies specific modules and hub genes related to coronary artery disease | Scientific Reports. Accessed January 21, 2026. https://www.nature.com/articles/s41598-021-86207-0 Fawcett T. An introduction to ROC analysis. Pattern Recognition Letters . 2006;27(8):861-874. doi:10.1016/j.patrec.2005.10.010 Lundberg S, Lee SI. A Unified Approach to Interpreting Model Predictions. arXiv . Preprint posted online November 25, 2017:arXiv:1705.07874. doi:10.48550/arXiv.1705.07874 Hu W, Fang T, Zhou M, Chen X. Identification of hub genes and immune infiltration in ulcerative colitis using bioinformatics. Sci Rep . 2023;13(1):6039. doi:10.1038/s41598-023-33292-y Chai X, Wang H, Wang B, et al. Mendelian randomization integrated with multi-omics analysis identifies TNIK as a key gene in gut microbiota-induced IBD development. Front Immunol . 2025;16:1678444. doi:10.3389/fimmu.2025.1678444 Kvedaraite E. Neutrophil-T cell crosstalk in inflammatory bowel disease. Immunology . 2021;164(4):657-664. doi:10.1111/imm.13391 Noviello D, Mager R, Roda G, Borroni RG, Fiorino G, Vetrano S. The IL23-IL17 Immune Axis in the Treatment of Ulcerative Colitis: Successes, Defeats, and Ongoing Challenges. Front Immunol . 2021;12:611256. doi:10.3389/fimmu.2021.611256 Long D, Mao C, Xu Y, Zhu Y. The emerging role of neutrophil extracellular traps in ulcerative colitis. Front Immunol . 2024;15:1425251. doi:10.3389/fimmu.2024.1425251 Chen L, Jiao T, Liu W, et al. Hepatic cytochrome P450 8B1 and cholic acid potentiate intestinal epithelial injury in colitis by suppressing intestinal stem cell renewal. Cell Stem Cell . 2022;29(9):1366-1381.e9. doi:10.1016/j.stem.2022.08.008 Fleishman JS, Kumar S. Bile acid metabolism and signaling in health and disease: molecular mechanisms and therapeutic targets. Signal Transduct Target Ther . 2024;9(1):97. doi:10.1038/s41392-024-01811-6 Gudiño V, Bartolomé-Casado R, Salas A. Single-cell omics in inflammatory bowel disease: recent insights and future clinical applications. Gut . 2025;74(8):1335-1345. doi:10.1136/gutjnl-2024-334165 Zhu YL, Liu CZ, Li Y. Regulatory Role of Protein Ubiquitination in the Pathogenesis and Progression of Ulcerative Colitis. Journal of Digestive Diseases . 2025;26(9-10):398-405. doi:10.1111/1751-2980.70017 Chen R, Pang X, Li L, Zeng Z, Chen M, Zhang S. Ubiquitin-specific proteases in inflammatory bowel disease-related signalling pathway regulation. Cell Death Dis . 2022;13(2):139. doi:10.1038/s41419-022-04566-6 Liu X, Fang Y, Huang M, et al. Deubiquitinase JOSD2 alleviates colitis by inhibiting inflammation via deubiquitination of IMPDH2 in macrophages. Acta Pharm Sin B . 2025;15(2):1039-1055. doi:10.1016/j.apsb.2024.12.012 Steiner CA, Cartwright IM, Taylor CT, Colgan SP. Hypoxia-inducible factor as a bridge between healthy barrier function, wound healing, and fibrosis. Am J Physiol Cell Physiol . 2022;323(3):C866-C878. doi:10.1152/ajpcell.00227.2022 Zhang Y, Yan M, Yue Y, Cheng Y. Hypoxia-Inducible Factor-1α Modulates the Toll-Like Receptor 4/Nuclear Factor Kappa B Signaling Pathway in Experimental Necrotizing Enterocolitis. Mediators of Inflammation . 2024;2024(1):4811500. doi:10.1155/mi/4811500 Nijhuis L, Peeters JGC, Vastert SJ, van Loosdregt J. Restoring T Cell Tolerance, Exploring the Potential of Histone Deacetylase Inhibitors for the Treatment of Juvenile Idiopathic Arthritis. Front Immunol . 2019;10:151. doi:10.3389/fimmu.2019.00151 Jiang L, Zhang S, Jiang C, et al. Integrative biomarker discovery and immune profiling for ulcerative colitis: a multi-methodological approach. Sci Rep . 2024;14(1):24290. doi:10.1038/s41598-024-75797-0 Tang YY, Wang DC, Wang YQ, Huang AF, Xu WD. Emerging role of hypoxia-inducible factor-1α in inflammatory autoimmune diseases: A comprehensive review. Front Immunol . 2023;13:1073971. doi:10.3389/fimmu.2022.1073971 Smillie CS, Biton M, Ordovas-Montanes J, et al. Intra- and Inter-cellular Rewiring of the Human Colon during Ulcerative Colitis. Cell . 2019;178(3):714-730.e22. doi:10.1016/j.cell.2019.06.029 Parikh K, Antanaviciute A, Fawkner-Corbett D, et al. Colonic epithelial cell diversity in health and inflammatory bowel disease. Nature . 2019;567(7746):49-55. doi:10.1038/s41586-019-0992-y Mitsialis V, Wall S, Liu P, et al. Single-Cell Analyses of Colon and Blood Reveal Distinct Immune Cell Signatures of Ulcerative Colitis and Crohn’s Disease. Gastroenterology . 2020;159(2):591-608.e10. doi:10.1053/j.gastro.2020.04.074 Osorio D, Zhong Y, Li G, et al. scTenifoldKnk: An efficient virtual knockout tool for gene function predictions via single-cell gene regulatory network perturbation. Patterns (N Y) . 2022;3(3):100434. doi:10.1016/j.patter.2022.100434 Patra R, Dey AK, Mukherjee S. Identification of genes critical for inducing ulcerative colitis and exploring their tumorigenic potential in human colorectal carcinoma. PLoS One . 2023;18(8):e0289064. doi:10.1371/journal.pone.0289064 Wang X, Wang H, Zhang R, Li D, Gao MQ. LRRC75A antisense lncRNA1 knockout attenuates inflammatory responses of bovine mammary epithelial cells. Int J Biol Sci . 2020;16(2):251-263. doi:10.7150/ijbs.38214 Mennillo E, Kim YJ, Lee G, et al. Single-cell and spatial multi-omics highlight effects of anti-integrin therapy across cellular compartments in ulcerative colitis. Nat Commun . 2024;15(1):1493. doi:10.1038/s41467-024-45665-6 Additional Declarations No competing interests reported. Supplementary Files SupplementalTableS1.csv SupplementalTableS2.csv SupplementalTableS3.csv SupplementalFig.S2.pdf SupplementalTableS4.csv SupplementalTableS5.txt SupplementalTableS6.txt SupplementalFig.S4.pdf SupplementalFigS5.pdf SupplementaryTableS7.txt SupplementaryTableS8.txt SupplementaryTableS9.txt SupplementaryTableS10.csv SupplementalFigS6.pdf SupplementalFigS7.pdf Figurelegends.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 24 Apr, 2026 Editor assigned by journal 24 Apr, 2026 Submission checks completed at journal 24 Apr, 2026 First submitted to journal 22 Apr, 2026 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-9497163","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":634089788,"identity":"2471296c-7393-4b99-a02f-68741ec4137d","order_by":0,"name":"Lin Li","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Lin","middleName":"","lastName":"Li","suffix":""},{"id":634089789,"identity":"3665c80f-406e-4b61-872c-0f0173cfab39","order_by":1,"name":"Zhenhe Jin","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Zhenhe","middleName":"","lastName":"Jin","suffix":""},{"id":634089790,"identity":"8178352f-5d4a-4ba5-820c-570e27fa8b0a","order_by":2,"name":"Zhe Shen","email":"","orcid":"","institution":"Zhejiang University","correspondingAuthor":false,"prefix":"","firstName":"Zhe","middleName":"","lastName":"Shen","suffix":""},{"id":634089792,"identity":"b0dd36ca-a02a-4514-a06e-72db637522a0","order_by":3,"name":"Mengchen Luo","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Mengchen","middleName":"","lastName":"Luo","suffix":""},{"id":634089793,"identity":"1f6e0d60-44a0-441c-9b01-cf18ff6f84f9","order_by":4,"name":"Xiaohua Ye","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIie3RsWrDMBCA4RMCeVGSVR1KX0FgSBpS8iwSBmcJmTNkOAh4CmT1YwT6AicEnlS69gm61qV7WjdTJqFuhejfDu6DgwPI5f5hutgT9dun5QTADDNPILKzrg11dYfJRK1LP2o8O9FlTiAzMNq1gvPyld4VbBcWixeKkjmSoV6K8ZSoVhBWFuXGxA9zSK5Vkk8d1oo13qKSOk48Qy+1Ys97GMg5hXQcvDSancQvwQQyPwhwLZlKBageTbcqG7mOk9nDx1f/ef5eTo7BvvW7xf2xCHFylTSXZ4rU/aGC/rCcy+Vyt9QPO+9Ki0wnhg0AAAAASUVORK5CYII=","orcid":"","institution":"Zhejiang University","correspondingAuthor":true,"prefix":"","firstName":"Xiaohua","middleName":"","lastName":"Ye","suffix":""},{"id":634089794,"identity":"4151550a-0397-4e57-9c74-4c9d8666105d","order_by":5,"name":"Yanlin Hu","email":"","orcid":"","institution":"Jinhua Municipal Central Hospital, Shao Xing University","correspondingAuthor":false,"prefix":"","firstName":"Yanlin","middleName":"","lastName":"Hu","suffix":""},{"id":634089795,"identity":"88718ee6-9958-4b05-b6d1-c900bdcdeac8","order_by":6,"name":"Jin Ding","email":"","orcid":"","institution":"Zhejiang University School of Medicine","correspondingAuthor":false,"prefix":"","firstName":"Jin","middleName":"","lastName":"Ding","suffix":""}],"badges":[],"createdAt":"2026-04-22 13:41:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9497163/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9497163/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108820910,"identity":"6fdbc557-82f5-4e53-af96-9eeea022c92a","added_by":"auto","created_at":"2026-05-08 16:43:29","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":252664,"visible":true,"origin":"","legend":"\u003cp\u003eFlow chart of the study design.\u003c/p\u003e","description":"","filename":"Fig1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/559cfde3588dde330be053da.jpg"},{"id":108820890,"identity":"467edbc4-86df-4d2b-9bd9-5bb05e58f1e6","added_by":"auto","created_at":"2026-05-08 16:43:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":122368,"visible":true,"origin":"","legend":"\u003cp\u003eLandscape of PTM-related genes and differential PTM-gene expression in ulcerative colitis. (A) PCA of the merged GSE75214 and GSE87466 dataset before batch correction. (B) PCA after batch-effect correction showing improved dataset integration. (C) Distribution of PTM-related genes across 21 PTM categories (ranked by gene count). (D) Volcano plot of PTM-related differentially expressed genes (DEGs) in UC versus controls in the integrated dataset (1,164 detected PTM genes; 84 PTM-related DEGs). (E) Volcano plot of activity-associated PTM-related DEGs in GSE75214 (1,303 detected PTM genes; 80 PTM-related DEGs).\u003c/p\u003e","description":"","filename":"Binder21.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/e2ccb2e2bb3cb521e9ca15f7.png"},{"id":108820916,"identity":"e4081ad8-3942-47b9-bf12-348c7efd9750","added_by":"auto","created_at":"2026-05-08 16:43:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":194808,"visible":true,"origin":"","legend":"\u003cp\u003eIdentification of two PTM-associated molecular subtypes of UC by consensus clustering. (A–B) GO and KEGG enrichment analyses of PTM-related genes. (C) Consensus matrix heatmap at k = 2 showing two stable PTM-associated clusters. (D) Cumulative distribution function (CDF) curves for consensus clustering. (E) Tracking plot summarizing cluster stability across different k values and supporting k = 2 as the optimal solution. (F) PCA plot showing separation between Cluster1 and Cluster2. (G) Volcano plot of DEGs between Cluster1 and Cluster2. (H) Heatmap of the top DEGs displaying reciprocal expression patterns and clear subtype segregation.\u003c/p\u003e","description":"","filename":"Binder22.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/992cd2dd12f4d0e094afbedc.png"},{"id":108820911,"identity":"8d0c96d7-33ef-406a-a33f-114be706d744","added_by":"auto","created_at":"2026-05-08 16:43:29","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":150067,"visible":true,"origin":"","legend":"\u003cp\u003eWGCNA identifies subtype-associated co-expression modules in UC. (A) Sample clustering dendrogram for outlier detection prior to network construction. (B) Scale independence and mean connectivity analyses used to select the soft-thresholding power (β = 14) to approximate scale-free topology (R² = 0.85). (C) Gene dendrogram and module color assignment generated by dynamic tree cutting and module merging.(D) Heatmap of the correlation between module genes and PTM subtypes (Cluster 1/Cluster 2). (E–F) Relationships between module membership (MM) and gene significance (GS) for the subtype-associated modules.\u003c/p\u003e","description":"","filename":"Binder23.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/66f73306f858a1330767003f.png"},{"id":108820410,"identity":"14d1146c-857c-44a9-b1c0-c5d5bdae6135","added_by":"auto","created_at":"2026-05-08 16:41:20","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":107718,"visible":true,"origin":"","legend":"\u003cp\u003eGSEA and GSVA reveal distinct biological programs between PTM subtypes. (A) GSEA enrichment plots highlighting inflammatory/immune programs enriched in Cluster1.(B) GSEA enrichment plots highlighting metabolic/secretory programs enriched in Cluster2. (C) Summary of differential pathways between Cluster1 and Cluster2 based on GSVA. (D) Heatmap of the top 20 differential pathways (GSVA scores) demonstrating functional divergence between the two subtypes.\u003c/p\u003e","description":"","filename":"Binder24.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/e47159fe7e513d6892b469c9.png"},{"id":108820862,"identity":"a8951015-b1eb-4c50-8826-c7ddb9d9089c","added_by":"auto","created_at":"2026-05-08 16:43:11","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":174537,"visible":true,"origin":"","legend":"\u003cp\u003eMachine-learning pipeline derives a 7-gene signature for subtype prediction. (A) Intersection of subtype DEGs, WGCNA key module genes, and PTM-related genes yielding 26 candidates. (B) Heatmap summarizing AUC performance across multiple algorithm/model combinations in the training cohort and two external validation cohorts. (C) ROC curves of representative single-gene classifiers among the final hub genes. (D) Nomogram for individualized subtype probability estimation. (E) Decision curve analysis evaluating net clinical benefit. (F) Calibration curves assessing agreement between predicted and observed probabilities. (G) SHAP analysis showing feature contributions of the 7-gene signature.\u003c/p\u003e","description":"","filename":"Binder25.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/d564ce023055d5c7a5119b3f.png"},{"id":108820549,"identity":"c6338a5a-3c75-44dc-b99e-6fb1b0383c06","added_by":"auto","created_at":"2026-05-08 16:41:50","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":135325,"visible":true,"origin":"","legend":"\u003cp\u003eImmune infiltration differences between subtypes and immune associations of the 7-gene signature. (A) Stacked bar plot showing the relative proportions of 22 immune cell types estimated by CIBERSORT across controls, Cluster1, and Cluster2. (B) Boxplots comparing immune cell fractions among controls, Cluster1, and Cluster2, highlighting subtype-specific shifts across lymphoid and myeloid populations. (C) Gene–immune correlation heatmap and network visualization summarizing associations between the 7 signature genes and immune cell fractions (edge sign indicates positive/negative correlation; edge width reflects correlation strength).\u003c/p\u003e","description":"","filename":"Binder26.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/e2a071061c03e42d4786759d.png"},{"id":108820899,"identity":"c3791c93-3cf9-44c0-b32b-b59b34e4a531","added_by":"auto","created_at":"2026-05-08 16:43:25","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":251860,"visible":true,"origin":"","legend":"\u003cp\u003eSingle-cell atlas localizes the hub program to specific cellular compartments and dynamic states. (A) UMAP embedding of single-cell transcriptomes annotated into eight major cell populations. (B) Dot plot of canonical marker genes supporting cell-type annotation. (C) Cell-type composition across conditions (Healthy, Non-inflamed, Inflamed). (D) Violin plots showing expression of the 7 hub genes across cell types in inflamed mucosa. (E) Trajectory inference illustrating state transitions within epithelial/immune compartments. (F–G) Pseudotime-dependent expression dynamics of hub genes along trajectories. (H–I) CellChat-based cell–cell communication networks and outgoing signaling patterns comparing high-hub versus low-hub states.\u003c/p\u003e","description":"","filename":"Binder27.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/898f9fc4ad80187d91217d18.png"},{"id":108822146,"identity":"951837fe-0d71-435c-a563-232cfeaeb8d7","added_by":"auto","created_at":"2026-05-08 16:47:32","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":96369,"visible":true,"origin":"","legend":"\u003cp\u003eIn silico virtual knockout predicts downstream perturbed gene networks for hub genes.\u003cstrong\u003e \u003c/strong\u003e(A–F) Predicted perturbed gene networks following virtual knockout of GALNT2 (A), GALNT8 (B), HIF1A (C), MYLIP (D), SPATA2 (E), and ZC3H12A (F). Networks are dominated by translation-associated ribosomal components (RPL/RPS), with additional recurring non-translation nodes (e.g., LRRC75A-AS1, EEF1A1, KLK1).\u003c/p\u003e","description":"","filename":"Binder28.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/3ebf0ad8412a88cef121732f.png"},{"id":108821818,"identity":"ceece4cd-b776-4b39-b7db-b91672d8b9aa","added_by":"auto","created_at":"2026-05-08 16:46:43","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":1895093,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial transcriptomics reveals localized hub-gene activity and spatial heterogeneity. (A–D) Spatial maps of annotated cell types across UC and control sections.(E–H) Spatial feature plots of the seven hub genes across sections. (I) Spatial distribution of the combined 7-gene score.\u003c/p\u003e","description":"","filename":"Binder29.png","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/ce62eb53f107ef992a6107f6.png"},{"id":108822825,"identity":"0e4f87a7-0c28-4b66-9903-a0c36671ed8e","added_by":"auto","created_at":"2026-05-08 16:50:54","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3211892,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/a2e97578-b1ec-40d6-86c4-2f0d7ad57614.pdf"},{"id":108820887,"identity":"11693733-49d1-46e0-b159-a722a8c736ce","added_by":"auto","created_at":"2026-05-08 16:43:24","extension":"csv","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":22135,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS1.csv","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/7bf31e728b1a9cc17df35523.csv"},{"id":108820860,"identity":"a941d671-b73b-425a-af88-b8f3512ce5bf","added_by":"auto","created_at":"2026-05-08 16:43:11","extension":"csv","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":10792,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS2.csv","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/3949f8aca5ba219745c61f3b.csv"},{"id":108820548,"identity":"c2a8146f-a1e6-464d-b6c9-964bcd6a804e","added_by":"auto","created_at":"2026-05-08 16:41:50","extension":"csv","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10320,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS3.csv","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/be1a6a5b13e85ed3fc568780.csv"},{"id":108820547,"identity":"88801030-e050-4a9e-a892-4b990d5c17ae","added_by":"auto","created_at":"2026-05-08 16:41:50","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":69435,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFig.S2.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/2a8614ad862ed541d36c946d.pdf"},{"id":108820546,"identity":"1ff42828-75fc-4f32-9a5d-1c0c0b1ebd4b","added_by":"auto","created_at":"2026-05-08 16:41:49","extension":"csv","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":209957,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS4.csv","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/b958834d1e0fa42b266f4494.csv"},{"id":108820411,"identity":"14c72e11-8aa8-403c-8292-caa1f14add76","added_by":"auto","created_at":"2026-05-08 16:41:24","extension":"txt","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":871,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS5.txt","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/1f30c709f14e08408c6aea57.txt"},{"id":108820407,"identity":"06092cb9-cadb-4d11-800a-4436037570a6","added_by":"auto","created_at":"2026-05-08 16:41:19","extension":"txt","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":4516,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalTableS6.txt","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/a52d99be9e6f72650898255a.txt"},{"id":108820928,"identity":"36648b1c-928c-4005-8a38-2a3faebb2a6f","added_by":"auto","created_at":"2026-05-08 16:43:40","extension":"pdf","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":479135,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFig.S4.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/b1c60a102f9888eeefae65b0.pdf"},{"id":108821609,"identity":"7d30c3c6-6ca1-4dc2-80da-9fe7c1bfb63c","added_by":"auto","created_at":"2026-05-08 16:46:10","extension":"pdf","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":732506,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFigS5.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/b0139d7f9226c2f41d0d085d.pdf"},{"id":108820409,"identity":"e3104f36-ee2f-4a1f-973b-180a5ea96c48","added_by":"auto","created_at":"2026-05-08 16:41:19","extension":"txt","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":250,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS7.txt","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/360bc343c2aba775af4792c7.txt"},{"id":108820931,"identity":"202458de-774c-48c5-b24c-e71f139715dd","added_by":"auto","created_at":"2026-05-08 16:43:40","extension":"txt","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":194,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS8.txt","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/485819d53c03ea4534fef009.txt"},{"id":108820866,"identity":"47f8b695-848e-4baa-aed8-3e82fa55ed55","added_by":"auto","created_at":"2026-05-08 16:43:14","extension":"txt","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":8573,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS9.txt","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/4fdff0fcf42ffd74684fc432.txt"},{"id":108820894,"identity":"d7b1fea5-23e0-4295-8e2a-e66870cd919a","added_by":"auto","created_at":"2026-05-08 16:43:25","extension":"csv","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":1232,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTableS10.csv","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/e4e2fa1c822fc76b6a08f746.csv"},{"id":108820893,"identity":"26c4d10e-ec35-4c6b-a8e1-e896bbf8f5a3","added_by":"auto","created_at":"2026-05-08 16:43:25","extension":"pdf","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":2653848,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFigS6.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/c4a5b96cd539e3a2659aab7d.pdf"},{"id":108820913,"identity":"db362432-5f13-4cee-a765-2437b15ba61a","added_by":"auto","created_at":"2026-05-08 16:43:29","extension":"pdf","order_by":14,"title":"","display":"","copyAsset":false,"role":"supplement","size":1756818,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementalFigS7.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/ac963138a9a4092fb2977707.pdf"},{"id":108820929,"identity":"7285ef65-4d75-4574-92bb-27aea565a454","added_by":"auto","created_at":"2026-05-08 16:43:40","extension":"docx","order_by":15,"title":"","display":"","copyAsset":false,"role":"supplement","size":17307,"visible":true,"origin":"","legend":"","description":"","filename":"Figurelegends.docx","url":"https://assets-eu.researchsquare.com/files/rs-9497163/v1/b582ec307322f531c4fa1f53.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Systematic post-translational modification profiling identifies molecular subtypes and therapeutic targets in ulcerative colitis: evidence from multi-omics integration","fulltext":[{"header":"Introduction","content":"\u003cp\u003eInflammatory bowel disease (IBD) encompasses chronic, relapsing inflammatory conditions that primarily involve the intestinal tract\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Ulcerative colitis (UC), a major form of IBD, is defined by diffuse and continuous inflammation confined to the colonic mucosa, typically beginning in the rectum and extending proximally for a variable distance\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u003c/sup\u003e. What makes UC particularly challenging is the marked heterogeneity observed in patient responses. Individuals with comparable clinical activity and endoscopic severity can exhibit substantial inter-individual variation in mucosal immune programs, epithelial barrier disruption, and tissue remodeling\u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. This heterogeneity not only complicates mechanistic interpretation from bulk tissue profiles but also underscores the need for molecular frameworks that can delineate disease-relevant mucosal programs beyond conventional clinical descriptors\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003ePost-translational modifications (PTMs) provide a rapid and reversible means of regulating protein function by modulating activity, stability, and subcellular localization. These modifications influence signaling dynamics, proteostasis, and intracellular trafficking that are essential for epithelial barrier integrity and mucosal immune homeostasis in UC\u003csup\u003e6,7\u003c/sup\u003e. Given this regulatory breadth, multiple PTM pathways have been implicated in intestinal inflammation. For instance, ubiquitin-modifying enzymes, shape inflammatory signaling and epithelial stress responses in colitis, and dysregulated ubiquitination has been repeatedly linked to UC pathogenesis, particularly in the context of barrier dysfunction\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e. Changes in SUMOylation have also been observed in IBD mucosa, suggesting that defective SUMO-dependent control may contribute to epithelial signaling perturbations\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e. Additionally, proteome-wide studies support the presence of PTM remodeling during experimental colitis, including shifts in lysine acetylation that involve metabolic enzymes and stress-response proteins in DSS-induced models\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. Moreover, glycosylation programs are increasingly recognized as key regulators of gut homeostasis and inflammation, further reinforcing the view that PTM reprogramming in UC extends beyond a single modification class\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Despite these advances, much of the current evidence remains centered on individual PTM types or specific experimental settings, and a unified, transcriptome-informed perspective of PTM-associated regulatory programs and their relationships to cellular compartments in UC remains insufficiently defined.\u003c/p\u003e \u003cp\u003eTo address this gap, we profiled PTM-associated heterogeneity in UC using an integrative, multi-level framework. Using a curated PTM gene panel spanning 21 modification categories, we integrated and batch-corrected bulk transcriptomic cohorts to define PTM-associated subtypes and characterize their transcriptional and pathway features through differential analysis, WGCNA, and enrichment profiling. We then applied machine-learning approaches to derive a parsimonious seven-gene signature (\u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, \u003cem\u003eUSP30\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eGALNT2\u003c/em\u003e, \u003cem\u003eZC3H12A\u003c/em\u003e, and \u003cem\u003eGALNT8\u003c/em\u003e) and assessed its predictive utility with calibration and decision-curve analyses. Finally, immune infiltration was evaluated and signature expression was mapped across single-cell and spatial transcriptomic layers to place PTM-associated programs within the cellular and spatial context of inflamed UC mucosa.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e\u003cstrong\u003eTranscriptomic data acquisition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe began our analysis by gathering bulk transcriptomic data from two publicly available datasets of colonic mucosal biopsies: GSE75214\u003csup\u003e13\u003c/sup\u003e (UC = 97, Control = 22) and GSE87466\u003csup\u003e14\u003c/sup\u003e (UC = 87, Control = 21). The two cohorts were merged to form a discovery set, and inter-study batch effects were corrected using the ComBat algorithm in the sva R package\u003csup\u003e15\u003c/sup\u003e. The batch-adjusted expression matrix (UC = 184, Control = 43) was used for differential expression analysis and PTM-related integrative analyses. Additional bulk cohorts (GSE47908 and GSE206285) were compiled for external validation\u003csup\u003e16,17\u003c/sup\u003e. PTM-related genes were curated from published literature covering 21 modification categories (Supplemental Table S1)\u003csup\u003e18\u003c/sup\u003e. The complete analytical workflow is schematized in Fig 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsensus clustering\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConsensus clustering was performed using the ConsensusClusterPlus R package to identify molecular subtypes based on PTM-related gene expression profiles. Clustering stability was evaluated across multiple cluster numbers, and the optimal number was determined by cumulative distribution function (CDF) curves and delta area plots. Principal component analysis (PCA) was applied to validate the separation between identified subtypes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDifferential gene analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo identify differentially expressed genes (DEGs) in bulk transcriptome data, we used the limma R package to screen DEGs from the integrated GSE75214 and GSE87466 datasets. The criteria for DEG selection were\\|log₂ fold change\\| \u0026gt; 0.585 and adjusted p \u0026lt; 0.05. Furthermore, we used the R package ggplot2 to generate volcano plots of the DEGs, providing a visual representation of their distribution to facilitate subsequent analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWeighted gene co-expression network analysis\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWeighted gene co-expression network analysis (WGCNA) was conducted to identify co-expression modules associated with UC subtypes. Sample outliers were detected through hierarchical clustering, and 43 samples were excluded\u003csup\u003e19\u003c/sup\u003e. A soft-thresholding power of \u0026beta; = 14 was selected to approximate scale-free topology (R\u0026sup2; = 0.85). Modules were identified using dynamic tree cutting based on the topological overlap matrix (TOM), and highly similar modules were merged according to eigengene correlations.\u003c/p\u003e\n\u003cp\u003eModule\u0026ndash;trait relationships were quantified by correlating module eigengenes with subtype labels, and key subtype-associated modules (cyan and greenyellow) were carried forward for downstream candidate selection and modeling.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional enrichment and pathway activity analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGO and KEGG enrichment analyses were performed for the screened PTM genes using the \u0026quot;clusterProfiler\u0026quot; R package, with human gene annotations provided by \u0026quot;org.Hs.eg.db\u0026quot; (keyType = \u0026ldquo;SYMBOL\u0026rdquo;). GO enrichment was evaluated across the biological process (BP), cellular component (CC), and molecular function (MF) domains, while KEGG pathway enrichment was implemented using a KEGG Medicus gene-set collection in GMT format (c2.cp.kegg_medicus.v2025.1.Hs.symbols.gmt). During the analysis process, we retained the enriched items according to the significance level (P \u0026lt; 0.05) to ensure the reliability and accuracy of the results. Gene set enrichment analysis (GSEA) was performed to identify enriched pathways between subtypes. Genes were ranked by fold-change, and enrichment was tested against predefined gene sets with permutation-based significance assessment\u003csup\u003e20\u003c/sup\u003e. Gene set variation analysis (GSVA) was applied to estimate sample-level pathway activity scores, converting gene expression matrices into pathway enrichment matrices. Both analyses were conducted to characterize biological programs underlying each subtype\u003csup\u003e21\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMachine learning-based prediction model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA prediction model for UC development was constructed based on PTM-related genes. Candidate genes were identified by intersecting DEGs between subtypes, WGCNA module genes (cyan and greenyellow), and curated PTM-related genes\u003csup\u003e22\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eWe next benchmarked multiple machine-learning algorithms and model combinations in the training cohort and two external validation cohorts. Model performance was primarily evaluated using the area under the receiver operating characteristic curve (AUC)\u003csup\u003e23\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSHAP (SHapley Additive exPlanations) analysis was applied to identify the most influential features from the candidate gene set\u003csup\u003e24\u003c/sup\u003e. A nomogram was constructed for individualized subtype prediction based on the final gene signature. Model reliability was assessed through calibration curves and decision curve analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmune microenvironment analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CIBERSORT algorithm was applied to estimate the relative proportions of 22 immune cell types from bulk expression profiles\u003csup\u003e25\u003c/sup\u003e. Immune cell compositions were compared across control, Cluster1, and Cluster2 groups using Wilcoxon rank-sum test or Kruskal-Wallis test. Spearman correlation coefficients were calculated between hub gene expression levels and immune cell proportions. Correlations were visualized using heatmaps and network diagrams, with significance threshold set at \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell RNA sequencing analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe scRNA-seq dataset GSE116222\u003csup\u003e26\u003c/sup\u003e was processed using Seurat (version 4.1.0). Quality control retained cells with nFeature_RNA between 500 and 4000, nCount_RNA \u0026lt; 6000, and percent.mt \u0026lt; 5%. Data normalization was performed using NormalizeData with default parameters. Highly variable genes (n = 2,000) were identified using FindVariableFeatures (vst method). Expression data were scaled using ScaleData, and principal component analysis was performed on the scaled data. Graph-based clustering was conducted using FindNeighbors and FindClusters (resolution =\u0026nbsp;0.4). UMAP dimensionality reduction was performed using the first 20 principal components (n.neighbors = 30, min.dist = 0.3, cosine metric). Cell type annotation was performed based on canonical marker gene expression from the CellMarker database. Cell proportions between UC and normal samples were compared using Fisher\u0026apos;s exact test (\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePseudotime trajectory analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePseudotime trajectory analysis was performed using Monocle 2 (version 2.26.0). The annotated cell populations were selected for trajectory construction. A CellDataSet object was created from raw UMI counts, with the expressionFamily parameter set to a negative binomial distribution. Size factors and dispersion were estimated using the estimateSizeFactors and estimateDispersions functions. Ordering genes were selected based on mean expression \u0026ge; 0.1 and dispersion_empirical \u0026ge; 1 \u0026times; dispersion_fit. Dimensionality reduction was performed using reduceDimension with method = \u0026quot;DDRTree\u0026quot; and max_components = 2, followed by cell ordering using the orderCells function.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntercellular ligand-receptor communication analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe CellChat package (version 1.6.0) was used to analyze intercellular communication among all cell subpopulations. A module score was calculated for each cell based on the identified gene signature using AddModuleScore. Cells were stratified into High_Hub and Low_Hub groups based on median module score. Separate CellChat objects were constructed for each group using normalized expression data. Ligand-receptor interactions were annotated using CellChatDB.human. Communication probabilities were computed using computeCommunProb (type = \u0026quot;triMean\u0026quot;, min.cells = 10). By calculating the number of ligands and the weighting values of cell\u0026ndash;cell interactions, a weighted, directed network was constructed. Pathway-level communication was aggregated using computeCommunProbPathway. The two CellChat objects were merged for comparative analysis of signaling information flows between High_Hub and Low_Hub groups.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSimulated knockout profile of key genes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eVirtual gene knockout was performed using scTenifoldKnk to predict gene functions. scTenifoldKnk constructs a denoised single-cell gene regulatory network (scGRN) from scRNA-seq data, then replicates the scGRN with target gene out-edges set to zero, generating a pseudo-knockout scGRN. The two scGRNs are mapped into low-dimensional space, and the distance between gene projections reveals the impact of gene knockout: larger disturbances indicate greater importance of the target gene.\u003c/p\u003e\n\u003cp\u003eFor each hub gene (HIF1A, GALNT2, GALNT8, ZC3H12A, MYLIP, and SPATA2), raw UMI counts from inflamed mucosal cells were prepared with the top 10,000 highly variable genes as the modeling feature set. Virtual knockout was performed with parameters: qc = TRUE, qc_mtThreshold = 0.1, qc_minLSize = 1000, nc_nNet = 10, nc_nCells = 500. Differentially regulated genes (DRGs) were identified with adjusted P \u0026lt; 0.05. Functional enrichment analysis was performed using clusterProfiler (ont = \u0026quot;BP\u0026quot;, pvalueCutoff = 0.05) on the top 20 genes significantly affected by knockout.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial transcriptomics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpatial transcriptomic data from the 10x Visium platform were processed using Seurat (version 4.1.0) within an R environment. The filtered count matrix (filtered_feature_bc_matrix.h5) was imported using Load10X_Spatial, and spot-level mitochondrial content was computed. Quality control steps were implemented with the following thresholds: nCount_Spatial \u0026gt; 0, nFeature_Spatial \u0026gt; 200, nFeature_Spatial \u0026lt; 8000, and percent.mt \u0026lt; 20%. Data normalization was performed using the SCTransform function with assay = \u0026quot;Spatial\u0026quot;. Subsequently, PCA was applied for dimensionality reduction, and clustering was conducted using the Leiden-based FindClusters function with resolution = 1.2 and the first 10 principal components. To infer spot-level cellular composition, RCTD (Robust Cell Type Decomposition; spacexr) was applied in doublet mode (doublet_mode = \u0026quot;doublet\u0026quot;; max_cores = 10), using a curated single-cell reference prepared by subsampling 500 cells per subtype and excluding rare subtypes with fewer than 25 cells. Predicted cell type annotations (first_type) were transferred to the spatial Seurat object for downstream analyses. Additionally, spot-level enrichment scores for cell-type programs were computed using ssGSEA on the top 10 marker genes per subtype identified from the single-cell reference. For spatially informed ligand-receptor communication analysis, CellChat (version 1.6.1) was employed in spatial mode with CellChatDB.human as reference, incorporating tissue coordinates and spot diameter scaling (spot.size = 65). Communication probabilities were computed using a truncated mean approach with distance constraints (interaction.range = 250, contact.range = 100) and filtered with min.cells = 10. Finally, the spatial expression patterns of the identified hub genes were visualized using SpatialFeaturePlot and vizAllTopics.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003ePTM-related gene dysregulation in UC\u0026nbsp;\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBatch effects were corrected across GSE75214 and GSE87466 datasets to ensure analytical consistency\u0026nbsp;(\u003cstrong\u003eFig. 2A-B\u003c/strong\u003e). PTM-related genes were next summarized across 21 PTM categories. From a curated set of 1,427 PTM-related genes covering 21 PTM categories (2\u0026ndash;426 genes per category;\u003cstrong\u003e\u0026nbsp;Fig. 2C\u003c/strong\u003e), 1,164 genes were detected in the integrated expression matrix. Differential expression analysis identified 84 PTM-related DEGs (51 upregulated, 33 downregulated; 7.22% of detected PTM genes; \u003cstrong\u003eFig. 2D; Supplementary Table S2\u003c/strong\u003e). In the GSE75214 dataset, 1,303 PTM genes were detected, yielding 80 activity-associated DEGs (43 upregulated, 37 downregulated; 6.14%; \u003cstrong\u003eFig. 2E; Supplementary Table S3\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eIn UC versus control comparisons, upregulated PTM-related genes included \u003cem\u003eS100A8, NOS2, MNDA, SOCS3, IFI16, TRIM22, FGR\u003c/em\u003e, and \u003cem\u003eMSN\u003c/em\u003e, whereas \u003cem\u003eUSP30, USP2, RNF128\u003c/em\u003e, and \u003cem\u003eASB13\u003c/em\u003e were downregulated. Active versus inactive UC comparisons showed concordant patterns with increased expression of \u003cem\u003eS100A8, NOS2, BIRC3, UBD\u003c/em\u003e, and \u003cem\u003ePTPRC\u003c/em\u003e and decreased expression of \u003cem\u003eNEDD4L, GALNT12, RNF152\u003c/em\u003e, and \u003cem\u003eB3GNT7\u003c/em\u003e. Seven genes (\u003cem\u003eS100A8, NOS2, MNDA, SOCS3, TRIM22, MSN, HIF1A\u003c/em\u003e) were consistently dysregulated across both comparisons.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTwo PTM-associated molecular subtypes identified by consensus clustering\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGO and KEGG enrichment analyses revealed that PTM-related DEGs were enriched in protein polyubiquitination, ubiquitin ligase complex, and 26S proteasome-mediated protein degradation pathways (\u003cstrong\u003eFig. 3A-B\u003c/strong\u003e). To evaluate clustering stability across different cluster numbers, additional consensus matrices for \u003cem\u003ek\u003c/em\u003e = 3\u0026ndash;6 were provided as supplementary evidence (\u003cstrong\u003eSupplementary Fig. S2\u003c/strong\u003e).Consensus clustering of 84 PTM-related DEGs identified an optimal two-cluster solution (k = 2; \u003cstrong\u003eFig. 3C-E\u003c/strong\u003e), classifying UC samples into Cluster1 and Cluster2. PCA confirmed clear separation between subtypes (\u003cstrong\u003eFig. 3F\u003c/strong\u003e). Differential expression analysis between subtypes revealed 1,234 DEGs, with the top DEGs showing reciprocal expression patterns that distinctly segregated samples (\u003cstrong\u003eFig. 3G-H; Supplementary Table S4)\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWGCNA identifies subtype\u0026ndash;specific co-expression modules\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWeighted gene co-expression network analysis was performed using a soft-thresholding power of 14 (scale-free topology fit \u003cem\u003eR\u0026sup2;\u003c/em\u003e = 0.85; \u003cstrong\u003eFig. 4A-B\u003c/strong\u003e), yielding seven modules after dynamic tree cutting and eigengene-based merging (\u003cstrong\u003eFig. 4C-D\u003c/strong\u003e). Module-trait correlation analysis identified the MEcyan module as strongly associated with Cluster1 (\u003cem\u003er\u003c/em\u003e = 0.82) and negatively with Cluster2 (\u003cem\u003er\u003c/em\u003e = \u0026minus;0.82). Conversely, the MEgreenyellow module showed positive correlation with Cluster2 (\u003cem\u003er\u003c/em\u003e = 0.79) and negative correlation with Cluster1 (\u003cem\u003er\u003c/em\u003e = \u0026minus;0.79; \u003cstrong\u003eFig. 4E\u003c/strong\u003e). Module membership (MM) in the cyan module correlated negatively with Cluster2 gene significance (GS; cor = \u0026minus;0.988; \u003cstrong\u003eFig. 4F\u003c/strong\u003e), whereas MM in the greenyellow module correlated positively with Cluster2-GS (cor = 0.984; \u003cstrong\u003eFig. 4G\u003c/strong\u003e). Genes from both modules were selected for downstream analyses, and the full gene lists of the MEcyan and MEgreenyellow modules are provided in \u003cstrong\u003eSupplementary Table S5-6\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDistinct biological programs characterize UC subtypes\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eGSEA revealed that Cluster1 was enriched in immune and inflammatory pathways, including IL-17 signaling, cytokine-cytokine receptor interaction, pertussis, viral protein interaction with cytokine receptors, and neutrophil extracellular trap formation (\u003cstrong\u003eFig. 5A\u003c/strong\u003e). In contrast, Cluster2 showed enrichment in metabolic pathways such as mineral absorption, drug metabolism via cytochrome P450, xenobiotic metabolism, bile secretion, and pancreatic secretion (\u003cstrong\u003eFig. 5B-C\u003c/strong\u003e). GSVA results were consistent with GSEA findings. The heatmap of top 20 differential pathways demonstrated that Cluster1 exhibited elevated activity in innate immune and inflammatory pathways, whereas Cluster2 showed higher activity in metabolic and secretory processes (\u003cstrong\u003eFig. 5D\u003c/strong\u003e). Violin plots of representative pathway GSVA scores confirmed these subtype-specific patterns (\u003cstrong\u003eFig. 5E-F\u003c/strong\u003e). In addition, GO enrichment analyses supporting these distinct biological programs between subtypes are provided in\u0026nbsp;\u003cstrong\u003eSupplementary Fig. S4\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSeven-gene signature predicts UC risk\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo develop a minimal predictive gene set, we intersected PTM-related genes with cluster DEGs and WGCNA key modules (cyan and greenyellow), identifying 26 candidate genes (\u003cstrong\u003eFig. 6A; Supplementary Table S8\u003c/strong\u003e). Multiple machine learning algorithms were compared across training and validation cohorts(\u003cstrong\u003eSupplementary Table S9\u003c/strong\u003e). The Stepglm(both) + Random Forest + Naive Bayes combination achieved optimal performance with AUCs of 1.000 (training set), 0.945 (GSE47908), and 0.673 (GSE206285; mean AUC = 0.873; \u003cstrong\u003eFig. 6B\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eBased on SHAP feature importance and model stability, seven key genes were selected: \u003cem\u003eHIF1A, SPATA2, USP30, MYLIP, GALNT2, ZC3H12A\u003c/em\u003e, and \u003cem\u003eGALNT8\u003c/em\u003e(\u003cstrong\u003eSupplementary Table S7)\u003c/strong\u003e. ROC analysis in the training set demonstrated strong discriminatory power for individual genes, particularly \u003cem\u003eHIF1A, USP30, ZC3H12A\u003c/em\u003e, and \u003cem\u003eGALNT2\u003c/em\u003e (\u003cstrong\u003eFig. 6C\u003c/strong\u003e). A nomogram was constructed for clinical application (\u003cstrong\u003eFig. 6D\u003c/strong\u003e). SHAP analysis identified \u003cem\u003eGALNT2\u003c/em\u003e (22.8%), \u003cem\u003eUSP30\u003c/em\u003e (20.6%), \u003cem\u003eGALNT8\u003c/em\u003e (19.0%), and \u003cem\u003eHIF1A\u003c/em\u003e (15.9%) as major contributors, followed by \u003cem\u003eZC3H12A\u003c/em\u003e (7.7%), \u003cem\u003eMYLIP\u003c/em\u003e (7.1%), and \u003cem\u003eSPATA2\u003c/em\u003e (7.0%; \u003cstrong\u003eFig. 6G; Supplementary Fig. S5\u003c/strong\u003e). Calibration curves demonstrated good agreement between predicted and observed probabilities, and decision curve analysis indicated net clinical benefit of the seven-gene model (\u003cstrong\u003eFig. 6E-F\u003c/strong\u003e) , with model coefficients provided in \u003cstrong\u003eSupplementary Table S10\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmune cell infiltration patterns differ between UC subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eCIBERSORT analysis estimated 22 immune cell populations across controls, Cluster1, and Cluster2 samples. Stacked bar plots revealed substantial inter-sample variation and subtype-specific immune compositions (\u003cstrong\u003eFig. 7A\u003c/strong\u003e). Compared to controls, Cluster1 showed elevated proportions of monocytes, M0 macrophages, activated mast cells, and neutrophils, whereas Cluster2 exhibited relatively higher levels of M2 macrophages, CD8 T cells, and regulatory T cells (\u003cstrong\u003eFig. 7B\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003eCorrelation analysis between the seven signature genes and immune cell infiltration revealed extensive associations with myeloid and granulocyte populations, particularly neutrophils, macrophages, and mast cells (\u003cstrong\u003eFig. 7C; Supplemental Fig. 6A-B\u003c/strong\u003e). HIF1A showed positive correlations with neutrophils, CD8 T cells, and activated mast cells (all \u003cem\u003eP\u003c/em\u003e \u0026lt; 1 \u0026times; 10⁻⁶), while USP30 correlated with neutrophils and M2 macrophages (\u003cem\u003eP\u003c/em\u003e \u0026lt; 1 \u0026times; 10⁻⁶).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell analysis reveals cell type-specific expression patterns\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe GSE116222 dataset comprised 22,609 cells with 10,470 genes after quality control filtering. Using the top 20 principal components, 15 clusters were identified and annotated into eight major cell types based on canonical markers: enterocytes, crypt cells, epithelial cells, goblet cells, B cells, regulatory CD4 T cells, monocytes, and mast cells (\u003cstrong\u003eFig. 8A-B\u003c/strong\u003e). Cell type proportions varied across conditions and individuals (\u003cstrong\u003eFig. 8C\u003c/strong\u003e). Supporting plots for quality assessment, principal component selection, and highly variable gene identification are provided in \u003cstrong\u003eSupplementary Fig. S7\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eMapping of the seven signature genes onto the single-cell atlas revealed distinct cell type-specific expression patterns across epithelial and immune compartments (\u003cstrong\u003eFig. 8D\u003c/strong\u003e). Trajectory inference analysis demonstrated pseudotime-dependent regulation, with \u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, and \u003cem\u003eUSP30\u003c/em\u003e showing dynamic patterns along immune lineage trajectories, while \u003cem\u003eGALNT2\u003c/em\u003e, \u003cem\u003eGALNT8\u003c/em\u003e, and \u003cem\u003eZC3H12A\u003c/em\u003e exhibited temporal changes along epithelial trajectories (\u003cstrong\u003eFig. 8E-G\u003c/strong\u003e). CellChat analysis revealed differential intercellular communication networks between high- and low-signature states, with altered outgoing signaling patterns among cell populations (\u003cstrong\u003eFig. 8H-I\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVirtual knockout analysis predicts gene-specific perturbation networks\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn silico virtual knockout simulations predicted downstream perturbation networks for six of the seven signature genes (\u003cem\u003eUSP30\u003c/em\u003e was excluded due to insufficient predicted effects). \u003cem\u003eGALNT2\u003c/em\u003e knockout primarily affected translation-related genes (\u003cstrong\u003eFig. 9A\u003c/strong\u003e). GALNT8 knockout showed broader effects, perturbing ribosomal genes alongside epithelial markers including \u003cem\u003eITLN1\u003c/em\u003e, \u003cem\u003eTFF3\u003c/em\u003e, \u003cem\u003eWFDC2\u003c/em\u003e, \u003cem\u003eAGR2\u003c/em\u003e, and \u003cem\u003eSPINT2\u0026nbsp;\u003c/em\u003e(\u003cstrong\u003eFig. 9B\u003c/strong\u003e). \u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, and \u003cem\u003eZC3H12A\u003c/em\u003e knockouts predominantly affected translation machinery (\u003cstrong\u003eFig. 9C-F\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSpatial transcriptomics confirms focal gene expression in UC tissue\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSpatial transcriptomic analysis of UC and control sections revealed tissue- and disease-specific heterogeneity. Immune-related spots showed increased clustering and abundance in UC compared to sparse distribution in controls, where epithelial and crypt compartments dominated (\u003cstrong\u003eFig. 10A-D\u003c/strong\u003e). The identified signature genes displayed region-restricted expression with discrete high-signal hotspots in UC sections versus weaker, scattered signals in controls (\u003cstrong\u003eFig. 10E-H\u003c/strong\u003e). \u003cem\u003eHIF1A\u003c/em\u003e showed broader spatial activation, whereas other genes exhibited more focal expression patterns. The combined seven-gene score demonstrated more prominent spatial activation in UC, delineating regions of elevated transcriptional activity (\u003cstrong\u003eFig. 10I\u003c/strong\u003e).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study employed a PTM-oriented, multi-omics framework to dissect the molecular heterogeneity of UC and identify clinically actionable biomarkers. By systematically integrating bulk transcriptomic profiles with a curated panel of PTM-related genes spanning 21 modification categories, we uncovered consistent perturbations in PTM-associated regulatory programs that distinguish UC from healthy controls and correlate with disease activity states. The convergence of these molecular signals across independent cohorts establishes PTM-related genes as a biologically meaningful feature space for understanding UC pathogenesis and stratifying patient populations based on underlying molecular mechanisms.\u003c/p\u003e \u003cp\u003eConsensus clustering of PTM-related differentially expressed genes revealed two molecularly distinct UC subtypes, each characterized by divergent pathway enrichment signatures. Cluster1 exhibited pronounced activation of immune-inflammatory circuits, particularly IL-17 signaling pathways, cytokine-cytokine receptor interactions, and neutrophil extracellular trap formation mechanisms. These molecular features align with accumulating evidence positioning the IL-23/IL-17 axis as a central orchestrator of mucosal inflammation in UC, where it coordinates neutrophil recruitment, promotes tissue damage through NET-mediated mechanisms, and drives chronic inflammatory responses\u003csup\u003e\u003cspan additionalcitationids=\"CR28\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. The dominance of inflammatory pathways in Cluster1 suggests these patients may harbor an immune-hyperactive type characterized by sustained neutrophil activation and robust cytokine production\u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. In contrast, Cluster2 demonstrated preferential enrichment of metabolic and secretory pathways, including bile secretion, xenobiotic metabolism via cytochrome P450 enzymes, and pancreatic secretion processes. This metabolic signature likely reflects adaptive epithelial responses to chronic inflammation, where mucosal cells recalibrate their metabolic machinery to support barrier repair, maintain secretory functions, and manage cellular stress under inflammatory conditions\u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. The identification of these two subtypes carries significant translational implications, as immune-dominant versus metabolic-dominant types may exhibit differential therapeutic responsiveness: the former potentially benefiting from targeted immunosuppression, while the latter might require barrier-enhancing or metabolic-supporting interventions\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eNotably, the prominence of ubiquitin-related processes across both subtypes, manifesting through enhanced protein polyubiquitination, ubiquitin ligase complex activity, and proteasome-mediated degradation. This suggests fundamental dysregulation of proteostasis networks in UC mucosa\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e\u003c/sup\u003e. Ubiquitination serves as a regulatory checkpoint governing inflammatory mediator stability, including NF-κB pathway components and pattern recognition receptors, while orchestrating epithelial stress responses through selective protein degradation\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. The convergence of ubiquitin pathway enrichment with PTM-related gene dysregulation supports a conceptual framework wherein proteostasis stress intersects with inflammatory remodeling, representing a mechanistically plausible therapeutic leverage point\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. Recent studies demonstrating that specific ubiquitin-modifying enzymes regulate colitis severity provide experimental support for this therapeutic rationale\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWGCNA provided critical evidence that PTM-associated heterogeneity operates through coordinated transcriptional programs rather than isolated gene alterations. The identification of two key modules exhibiting robust correlations with identified subtypes (|\u003cem\u003er\u003c/em\u003e| \u0026gt; 0.79) and exceptionally tight module membership-gene significance relationships (|\u003cem\u003er\u003c/em\u003e| \u0026gt; 0.98) indicates these co-expression modules capture stable, biologically coherent regulatory circuits. This network-guided organization proved valuable for biomarker discovery, enabling candidate gene prioritization based on embeddedness within subtype-defining transcriptional architectures rather than univariate differential expression alone, enriching for genes with functional roles in UC pathogenesis.\u003c/p\u003e \u003cp\u003eThrough integration of subtype DEGs, WGCNA modules, and curated PTMRGs, we derived a seven-gene signature (\u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, \u003cem\u003eUSP30\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eGALNT2\u003c/em\u003e, \u003cem\u003eZC3H12A\u003c/em\u003e, \u003cem\u003eGALNT8\u003c/em\u003e) demonstrating robust performance across discovery and validation cohorts. This panel demonstrates mechanistic coherence, careful selection grounded in UC biology. \u003cem\u003eHIF1A\u003c/em\u003e, encoding hypoxia-inducible factor 1-α, governs cellular adaptation to oxygen deprivation and has been extensively linked to intestinal barrier preservation during inflammatory stress, coordinating epithelial survival, tight junction integrity, and mucosal healing\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e. The inflamed UC mucosa experiences localized hypoxia due to increased metabolic demand and immune cell infiltration, creating microenvironmental conditions activating \u003cem\u003eHIF1A\u003c/em\u003e-dependent programs\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. The inclusion of ubiquitin-pathway regulators including \u003cem\u003eUSP30\u003c/em\u003e (a mitochondrial deubiquitinase linking inflammation to mitochondrial quality control) and \u003cem\u003eSPATA2\u003c/em\u003e (a scaffolding protein modulating CYLD-dependent TNF/NF-κB signaling), directly reflects the mechanistic importance of ubiquitin-mediated regulation in UC inflammatory circuitry\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. \u003cem\u003eGALNT2\u003c/em\u003e and \u003cem\u003eGALNT8\u003c/em\u003e, polypeptide N-acetylgalactosaminyltransferases initiating mucin-type O-glycosylation, regulate epithelial glycobiology programs critical for mucus barrier integrity and host-microbe interactions\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. SHAP analysis revealed \u003cem\u003eGALNT2\u003c/em\u003e (22.8%), \u003cem\u003eUSP30\u003c/em\u003e (20.6%), \u003cem\u003eGALNT8\u003c/em\u003e (19.0%), and \u003cem\u003eHIF1A\u003c/em\u003e (15.9%) as dominant contributors, collectively accounting for approximately 80% of model predictions, indicating that glycosylation-related and ubiquitin-related PTM regulators constitute particularly informative UC biomarkers, likely because they integrate converging signals from epithelial stress, inflammatory activation, and barrier remodeling\u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eImmune deconvolution illuminated cellular landscapes underlying molecular subtypes. Cluster1 demonstrated elevated myeloid/granulocytic infiltration (monocytes, M0 macrophages, activated mast cells, neutrophils), while Cluster2 showed enriched M2 macrophages and regulatory T cells\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. Correlation analyses revealed particularly robust associations between \u003cem\u003eHIF1A\u003c/em\u003e/\u003cem\u003eUSP30\u003c/em\u003e expression and neutrophil infiltration plus activated mast cells, mechanistically connecting these genes to specific inflammatory programs\u003csup\u003e\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e,\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e and suggesting PTM-based stratification captures fundamental variations in tissue inflammatory architecture with implications for predicting differential therapeutic responses\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSingle-cell analysis revealed marked compartmentalization of signature gene expression across the cellular complexity of intestinal mucosa. Inflammation-associated components (\u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eZC3H12A\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e) showed predominant expression in innate immune populations, particularly mast cells and monocytes where they likely calibrate inflammatory output intensity and cytokine production kinetics. In contrast, glycosylation-related genes (\u003cem\u003eGALNT2\u003c/em\u003e, \u003cem\u003eGALNT8\u003c/em\u003e) exhibited enrichment in epithelial lineages, especially goblet cells and crypt progenitors responsible for mucin secretion and epithelial renewal\u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. This compartmentalized expression positions these genes at the nexus of their respective functional programs, namely immune activation and mucus barrier assembly. Pseudotime trajectory analysis added temporal dimensionality, revealing dynamic, state-dependent regulation along immune activation and epithelial differentiation continua, indicating PTM programs respond to fluctuating microenvironmental cues (oxygen tension, cytokine exposure, microbial signals) rather than constitutive activation\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e. This aligns with UC biology where focal inflammatory hotspots coexist with relatively quiescent regions, creating tissue microdomains with divergent PTM regulatory requirements\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e. CellChat analysis revealed substantial intercellular communication remodeling between high- and low-signature states, suggesting PTM programs influence not only cell-intrinsic functions but also coordinate multicellular tissue responses through modulation of cell-cell signaling circuits.\u003c/p\u003e \u003cp\u003eAn important aspect of our study is the application of virtual knockout analysis using scTenifoldKnk, representing a novel approach in IBD research for predicting gene regulatory hierarchies through single-cell network perturbation modeling. Unlike conventional differential expression or correlation analyses that capture static associations, this emerging computational method constructs denoised gene regulatory networks from single-cell data and simulates gene ablation effects by setting target gene edges to zero, revealing downstream perturbation cascades and regulatory dependencies that would be experimentally challenging and resource-intensive to assess systematically\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. For most hub genes (\u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, \u003cem\u003eZC3H12A\u003c/em\u003e), computational ablation primarily affected translational machinery, suggesting coupling to broad biosynthetic and stress-adaptive programs\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. However, \u003cem\u003eGALNT8\u003c/em\u003e knockout produced a distinctive perturbation signature affecting epithelial-specific functional markers including \u003cem\u003eITLN1\u003c/em\u003e (antimicrobial lectin), \u003cem\u003eTFF3\u003c/em\u003e (mucus-associated peptide critical for barrier repair), \u003cem\u003eWFDC2\u003c/em\u003e (protease inhibitor), \u003cem\u003eAGR2\u003c/em\u003e (protein disulfide isomerase supporting mucin folding), and \u003cem\u003eSPINT2\u003c/em\u003e (serine protease inhibitor regulating epithelial integrity). This divergent pattern indicates \u003cem\u003eGALNT8\u003c/em\u003e exerts specialized regulatory influence over epithelial functional programs beyond core protein synthesis, positioning it as a potentially critical node coordinating mucus production, antimicrobial defense, and cell adhesion maintenance\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e,\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. The identification of LRRC75A-AS1 as a recurrently perturbed node across multiple knockouts highlights this long non-coding RNA as a potential integrative hub coordinating cellular responses to diverse PTM signals\u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. While requiring experimental validation, these predictions provide hypothesis-generating insights into gene regulatory hierarchies and functional dependencies unlikely to emerge from static expression profiling, representing an innovative application that advances UC molecular stratification research beyond descriptive associations toward predictive regulatory modeling.\u003c/p\u003e \u003cp\u003eSpatial transcriptomics anchored molecular signatures within tissue architecture, revealing region-restricted expression with discrete high-signal hotspots in inflamed tissue contrasting with diffuse low-intensity signals in controls\u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. \u003cem\u003eHIF1A\u003c/em\u003e demonstrated broader spatial activation reflecting widespread hypoxic stress, while other genes showed focal patterns corresponding to inflammatory foci or epithelial remodeling sites\u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. The composite seven-gene score revealed spatial activation gradients demarcating pathological activity zones[47]. This pronounced spatial heterogeneity suggests treatment responses may critically depend on therapeutic penetration into these restricted hotspots, providing rationale for integrating bulk, single-cell, and spatial readouts to comprehensively characterize UC molecular pathology\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e,\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSeveral limitations warrant consideration. Our transcriptome-based analyses require validation through proteomics and modification-specific assays to confirm transcript-level changes reflect actual PTM activity states\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Subtype overlap suggests incorporating clinical covariates and longitudinal sampling may sharpen boundaries and improve clinical applicability\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Virtual knockout predictions require experimental validation through targeted perturbation in relevant model systems\u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. External validation variability indicates need for larger prospective studies with standardized protocols\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Despite these constraints, convergence across multiple analytical modalities strongly supports the biological validity of our PTM-focused framework\u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e,\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, we established the first comprehensive PTM-centric stratification of UC, identifying two molecularly distinct subtypes and developing a mechanistically coherent seven-gene signature for prediction of UC risk. These findings provide foundation for PTM-targeted therapeutic strategies and precision medicine approaches enabling molecular type-based treatment matching in UC.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eClinical trial number\u003c/h2\u003e\n\u003cp\u003enot applicable.\u003c/p\u003e\n\u003ch2\u003eConflict of interest:\u003c/h2\u003e\n\u003cp\u003ethe authors involved declared no conflict of interest.\u003c/p\u003e\n\u003ch2\u003eEthical approval:\u003c/h2\u003e\n\u003cp\u003eNot applicable. This study used only publicly available, de-identified datasets; therefore, additional ethical approval and informed consent were not required.\u003c/p\u003e\n\u003ch2\u003eFunding\u0026nbsp;Declaration:\u003c/h2\u003e\n\u003cp\u003ethis work was funded by the National Key R\u0026amp;D Program of China (No 2023YFC2507300 to Z.S.), Zhejiang High-Level Talent Innovation Leadership Project (No 2023R5239 to Z.S), Research on the molecular mechanisms and targeting strategies of STRIP2 in regulating the progression and metastasis of colorectal cancer (NO. JY2024-6-07) and Clinical Medicine Research Special Fund of the Zhejiang Medical Association (No. 2025ZYC-A139).\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eLin Li and Zhenhe Jin conceived and designed the study. Xiaohua Ye and Zhenhe Jin performed data acquisition/experiments. Mengchen Luo conducted data analysis. Zhe Shen and Jin Ding supervised the project. Yanlin Hu drafted the manuscript, and all authors revised and approved the final version.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGBD 2017 Inflammatory Bowel Disease Collaborators. The global, regional, and national burden of inflammatory bowel disease in 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. \u003cem\u003eLancet Gastroenterol Hepatol\u003c/em\u003e. 2020;5(1):17-30. doi:10.1016/S2468-1253(19)30333-4\u003c/li\u003e\n\u003cli\u003eNeurath MF. Current and emerging therapeutic targets for IBD. \u003cem\u003eNat Rev Gastroenterol Hepatol\u003c/em\u003e. 2017;14(5):269-278. doi:10.1038/nrgastro.2016.208\u003c/li\u003e\n\u003cli\u003eUngaro R, Mehandru S, Allen PB, Peyrin-Biroulet L, Colombel JF. Ulcerative colitis. \u003cem\u003eLancet\u003c/em\u003e. 2017;389(10080):1756-1770. doi:10.1016/S0140-6736(16)32126-2\u003c/li\u003e\n\u003cli\u003ede Souza HSP, Fiocchi C. Immunopathogenesis of IBD: current state of the art. \u003cem\u003eNat Rev Gastroenterol Hepatol\u003c/em\u003e. 2016;13(1):13-27. doi:10.1038/nrgastro.2015.186\u003c/li\u003e\n\u003cli\u003eSelin KA, Hedin CRH, Villablanca EJ. Immunological Networks Defining the Heterogeneity of Inflammatory Bowel Diseases. \u003cem\u003eJ Crohns Colitis\u003c/em\u003e. 2021;15(11):1959-1973. doi:10.1093/ecco-jcc/jjab085\u003c/li\u003e\n\u003cli\u003eDing J, Wang H, Yang Z, Wang X, Cao Z. Protein acylation in inflammatory diseases: from mechanisms to therapeutic strategies. \u003cem\u003eCell Commun Signal\u003c/em\u003e. 2025;23:488. doi:10.1186/s12964-025-02484-6\u003c/li\u003e\n\u003cli\u003eMa XN, Li MY, Qi GQ, Wei LN, Zhang DK. SUMOylation at the crossroads of gut health: insights into physiology and pathology. \u003cem\u003eCell Commun Signal\u003c/em\u003e. 2024;22(1):404. doi:10.1186/s12964-024-01786-5\u003c/li\u003e\n\u003cli\u003eRuan J, Schl\u0026uuml;ter D, Naumann M, Waisman A, Wang X. Ubiquitin-modifying enzymes as regulators of colitis. \u003cem\u003eTrends Mol Med\u003c/em\u003e. 2022;28(4):304-318. doi:10.1016/j.molmed.2022.01.006\u003c/li\u003e\n\u003cli\u003eLiu Z, Liu J, Wei Y, et al. Ubiquitin-specific protease 25 ameliorates ulcerative colitis by regulating the degradation of phosphor-STAT3. \u003cem\u003eCell Death Dis\u003c/em\u003e. 2025;16(1):5. doi:10.1038/s41419-024-07315-z\u003c/li\u003e\n\u003cli\u003eMustfa SA, Singh M, Suhail A, et al. SUMOylation pathway alteration coupled with downregulation of SUMO E2 enzyme at mucosal epithelium modulates inflammation in inflammatory bowel disease. \u003cem\u003eOpen Biol\u003c/em\u003e. 2017;7(6):170024. doi:10.1098/rsob.170024\u003c/li\u003e\n\u003cli\u003eWang JM, Lin SR, Zhu YB, et al. Proteomic analysis of lysine acetylation reveals that metabolic enzymes and heat shock proteins may be potential targets for DSS-induced mice colitis. \u003cem\u003eInt Immunopharmacol\u003c/em\u003e. 2021;101(Pt B):108336. doi:10.1016/j.intimp.2021.108336\u003c/li\u003e\n\u003cli\u003eBergstrom K, Xia L. The barrier and beyond: Roles of intestinal mucus and mucin-type O-glycosylation in resistance and tolerance defense strategies guiding host-microbe symbiosis. \u003cem\u003eGut Microbes\u003c/em\u003e. 2022;14(1):2052699. doi:10.1080/19490976.2022.2052699\u003c/li\u003e\n\u003cli\u003eWu J, Lv Y, Hao P, et al. Immunological profile of lactylation-related genes in Crohn\u0026rsquo;s disease: a comprehensive analysis based on bulk and single-cell RNA sequencing data. \u003cem\u003eJ Transl Med\u003c/em\u003e. 2024;22(1):300. doi:10.1186/s12967-024-05092-z\u003c/li\u003e\n\u003cli\u003eChen K, Shang S, Yu S, Cui L, Li S, He N. Identification and exploration of pharmacological pyroptosis-related biomarkers of ulcerative colitis. \u003cem\u003eFront Immunol\u003c/em\u003e. 2022;13:998470. doi:10.3389/fimmu.2022.998470\u003c/li\u003e\n\u003cli\u003eLeek JT, Johnson WE, Parker HS, Jaffe AE, Storey JD. The sva package for removing batch effects and other unwanted variation in high-throughput experiments. \u003cem\u003eBioinformatics\u003c/em\u003e. 2012;28(6):882-883. doi:10.1093/bioinformatics/bts034\u003c/li\u003e\n\u003cli\u003eLinggi B, Jairath V, Zou G, et al. Meta-analysis of gene expression disease signatures in colonic biopsy tissue from patients with ulcerative colitis. \u003cem\u003eSci Rep\u003c/em\u003e. 2021;11(1):18243. doi:10.1038/s41598-021-97366-5\u003c/li\u003e\n\u003cli\u003ePan Z, Lin H, Fu Y, et al. Identification of gene signatures associated with ulcerative colitis and the association with immune infiltrates in colon cancer. \u003cem\u003eFront Immunol\u003c/em\u003e. 2023;14:1086898. doi:10.3389/fimmu.2023.1086898\u003c/li\u003e\n\u003cli\u003eWang X, Liu Y, Fu J, Li Y, Zhao M, Tian Q. Systematic post-translational modification genome wide identifies therapeutic targets for Alzheimer\u0026rsquo;s disease: evidence from multi-cohort analysis. \u003cem\u003eJ Prev Alzheimers Dis\u003c/em\u003e. Published online October 30, 2025:100422. doi:10.1016/j.tjpad.2025.100422\u003c/li\u003e\n\u003cli\u003eLangfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e. 2008;9:559. doi:10.1186/1471-2105-9-559\u003c/li\u003e\n\u003cli\u003eSubramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e. 2005;102(43):15545-15550. doi:10.1073/pnas.0506580102\u003c/li\u003e\n\u003cli\u003eH\u0026auml;nzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. \u003cem\u003eBMC Bioinformatics\u003c/em\u003e. 2013;14:7. doi:10.1186/1471-2105-14-7\u003c/li\u003e\n\u003cli\u003eWeighted gene co-expression network analysis identifies specific modules and hub genes related to coronary artery disease | Scientific Reports. Accessed January 21, 2026. https://www.nature.com/articles/s41598-021-86207-0\u003c/li\u003e\n\u003cli\u003eFawcett T. An introduction to ROC analysis. \u003cem\u003ePattern Recognition Letters\u003c/em\u003e. 2006;27(8):861-874. doi:10.1016/j.patrec.2005.10.010\u003c/li\u003e\n\u003cli\u003eLundberg S, Lee SI. A Unified Approach to Interpreting Model Predictions. \u003cem\u003earXiv\u003c/em\u003e. Preprint posted online November 25, 2017:arXiv:1705.07874. doi:10.48550/arXiv.1705.07874\u003c/li\u003e\n\u003cli\u003eHu W, Fang T, Zhou M, Chen X. Identification of hub genes and immune infiltration in ulcerative colitis using bioinformatics. \u003cem\u003eSci Rep\u003c/em\u003e. 2023;13(1):6039. doi:10.1038/s41598-023-33292-y\u003c/li\u003e\n\u003cli\u003eChai X, Wang H, Wang B, et al. Mendelian randomization integrated with multi-omics analysis identifies TNIK as a key gene in gut microbiota-induced IBD development. \u003cem\u003eFront Immunol\u003c/em\u003e. 2025;16:1678444. doi:10.3389/fimmu.2025.1678444\u003c/li\u003e\n\u003cli\u003eKvedaraite E. Neutrophil-T cell crosstalk in inflammatory bowel disease. \u003cem\u003eImmunology\u003c/em\u003e. 2021;164(4):657-664. doi:10.1111/imm.13391\u003c/li\u003e\n\u003cli\u003eNoviello D, Mager R, Roda G, Borroni RG, Fiorino G, Vetrano S. The IL23-IL17 Immune Axis in the Treatment of Ulcerative Colitis: Successes, Defeats, and Ongoing Challenges. \u003cem\u003eFront Immunol\u003c/em\u003e. 2021;12:611256. doi:10.3389/fimmu.2021.611256\u003c/li\u003e\n\u003cli\u003eLong D, Mao C, Xu Y, Zhu Y. The emerging role of neutrophil extracellular traps in ulcerative colitis. \u003cem\u003eFront Immunol\u003c/em\u003e. 2024;15:1425251. doi:10.3389/fimmu.2024.1425251\u003c/li\u003e\n\u003cli\u003eChen L, Jiao T, Liu W, et al. Hepatic cytochrome P450 8B1 and cholic acid potentiate intestinal epithelial injury in colitis by suppressing intestinal stem cell renewal. \u003cem\u003eCell Stem Cell\u003c/em\u003e. 2022;29(9):1366-1381.e9. doi:10.1016/j.stem.2022.08.008\u003c/li\u003e\n\u003cli\u003eFleishman JS, Kumar S. Bile acid metabolism and signaling in health and disease: molecular mechanisms and therapeutic targets. \u003cem\u003eSignal Transduct Target Ther\u003c/em\u003e. 2024;9(1):97. doi:10.1038/s41392-024-01811-6\u003c/li\u003e\n\u003cli\u003eGudi\u0026ntilde;o V, Bartolom\u0026eacute;-Casado R, Salas A. Single-cell omics in inflammatory bowel disease: recent insights and future clinical applications. \u003cem\u003eGut\u003c/em\u003e. 2025;74(8):1335-1345. doi:10.1136/gutjnl-2024-334165\u003c/li\u003e\n\u003cli\u003eZhu YL, Liu CZ, Li Y. Regulatory Role of Protein Ubiquitination in the Pathogenesis and Progression of Ulcerative Colitis. \u003cem\u003eJournal of Digestive Diseases\u003c/em\u003e. 2025;26(9-10):398-405. doi:10.1111/1751-2980.70017\u003c/li\u003e\n\u003cli\u003eChen R, Pang X, Li L, Zeng Z, Chen M, Zhang S. Ubiquitin-specific proteases in inflammatory bowel disease-related signalling pathway regulation. \u003cem\u003eCell Death Dis\u003c/em\u003e. 2022;13(2):139. doi:10.1038/s41419-022-04566-6\u003c/li\u003e\n\u003cli\u003eLiu X, Fang Y, Huang M, et al. Deubiquitinase JOSD2 alleviates colitis by inhibiting inflammation via deubiquitination of IMPDH2 in macrophages. \u003cem\u003eActa Pharm Sin B\u003c/em\u003e. 2025;15(2):1039-1055. doi:10.1016/j.apsb.2024.12.012\u003c/li\u003e\n\u003cli\u003eSteiner CA, Cartwright IM, Taylor CT, Colgan SP. Hypoxia-inducible factor as a bridge between healthy barrier function, wound healing, and fibrosis. \u003cem\u003eAm J Physiol Cell Physiol\u003c/em\u003e. 2022;323(3):C866-C878. doi:10.1152/ajpcell.00227.2022\u003c/li\u003e\n\u003cli\u003eZhang Y, Yan M, Yue Y, Cheng Y. Hypoxia-Inducible Factor-1\u0026alpha; Modulates the Toll-Like Receptor 4/Nuclear Factor Kappa B Signaling Pathway in Experimental Necrotizing Enterocolitis. \u003cem\u003eMediators of Inflammation\u003c/em\u003e. 2024;2024(1):4811500. doi:10.1155/mi/4811500\u003c/li\u003e\n\u003cli\u003eNijhuis L, Peeters JGC, Vastert SJ, van Loosdregt J. Restoring T Cell Tolerance, Exploring the Potential of Histone Deacetylase Inhibitors for the Treatment of Juvenile Idiopathic Arthritis. \u003cem\u003eFront Immunol\u003c/em\u003e. 2019;10:151. doi:10.3389/fimmu.2019.00151\u003c/li\u003e\n\u003cli\u003eJiang L, Zhang S, Jiang C, et al. Integrative biomarker discovery and immune profiling for ulcerative colitis: a multi-methodological approach. \u003cem\u003eSci Rep\u003c/em\u003e. 2024;14(1):24290. doi:10.1038/s41598-024-75797-0\u003c/li\u003e\n\u003cli\u003eTang YY, Wang DC, Wang YQ, Huang AF, Xu WD. Emerging role of hypoxia-inducible factor-1\u0026alpha; in inflammatory autoimmune diseases: A comprehensive review. \u003cem\u003eFront Immunol\u003c/em\u003e. 2023;13:1073971. doi:10.3389/fimmu.2022.1073971\u003c/li\u003e\n\u003cli\u003eSmillie CS, Biton M, Ordovas-Montanes J, et al. Intra- and Inter-cellular Rewiring of the Human Colon during Ulcerative Colitis. \u003cem\u003eCell\u003c/em\u003e. 2019;178(3):714-730.e22. doi:10.1016/j.cell.2019.06.029\u003c/li\u003e\n\u003cli\u003eParikh K, Antanaviciute A, Fawkner-Corbett D, et al. Colonic epithelial cell diversity in health and inflammatory bowel disease. \u003cem\u003eNature\u003c/em\u003e. 2019;567(7746):49-55. doi:10.1038/s41586-019-0992-y\u003c/li\u003e\n\u003cli\u003eMitsialis V, Wall S, Liu P, et al. Single-Cell Analyses of Colon and Blood Reveal Distinct Immune Cell Signatures of Ulcerative Colitis and Crohn\u0026rsquo;s Disease. \u003cem\u003eGastroenterology\u003c/em\u003e. 2020;159(2):591-608.e10. doi:10.1053/j.gastro.2020.04.074\u003c/li\u003e\n\u003cli\u003eOsorio D, Zhong Y, Li G, et al. scTenifoldKnk: An efficient virtual knockout tool for gene function predictions via single-cell gene regulatory network perturbation. \u003cem\u003ePatterns (N Y)\u003c/em\u003e. 2022;3(3):100434. doi:10.1016/j.patter.2022.100434\u003c/li\u003e\n\u003cli\u003ePatra R, Dey AK, Mukherjee S. Identification of genes critical for inducing ulcerative colitis and exploring their tumorigenic potential in human colorectal carcinoma. \u003cem\u003ePLoS One\u003c/em\u003e. 2023;18(8):e0289064. doi:10.1371/journal.pone.0289064\u003c/li\u003e\n\u003cli\u003eWang X, Wang H, Zhang R, Li D, Gao MQ. LRRC75A antisense lncRNA1 knockout attenuates inflammatory responses of bovine mammary epithelial cells. \u003cem\u003eInt J Biol Sci\u003c/em\u003e. 2020;16(2):251-263. doi:10.7150/ijbs.38214\u003c/li\u003e\n\u003cli\u003eMennillo E, Kim YJ, Lee G, et al. Single-cell and spatial multi-omics highlight effects of anti-integrin therapy across cellular compartments in ulcerative colitis. \u003cem\u003eNat Commun\u003c/em\u003e. 2024;15(1):1493. doi:10.1038/s41467-024-45665-6\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"immunologic-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"imre","sideBox":"Learn more about [Immunologic Research](http://link.springer.com/journal/12026)","snPcode":"12026","submissionUrl":"https://submission.nature.com/new-submission/12026/3","title":"Immunologic Research","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Ulcerative colitis, Post-translational modifications, Machine learning, Single-cell RNA sequencing, Spatial transcriptomics","lastPublishedDoi":"10.21203/rs.3.rs-9497163/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9497163/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eUlcerative colitis (UC) is a debilitating inflammatory bowel disease characterized by heterogeneous clinical presentations and variable treatment responses. Post-translational modifications (PTMs) coordinate immune signaling and mucosal barrier function, yet comprehensive PTM profiling to stratify UC patients and guide therapeutic decisions remains limited.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003eTo identify distinct subtypes of UC, we performed consensus clustering based on PTM-related gene expression. We then applied WGCNA to identify module genes and constructed a diagnostic signature using machine learning algorithms. Single-cell and spatial transcriptomics were used to map cellular localization, and virtual knockdown analysis was conducted to elucidate regulatory networks underlying mucosal inflammation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eTwo distinct PTM subtypes of UC were identified, with Cluster1 enriched in inflammatory pathways (\u003cem\u003eIL-17\u003c/em\u003e signaling, neutrophil extracellular trap formation) and Cluster2 in metabolic processes. Using machine learning, we developed a 7-gene signature (\u003cem\u003eHIF1A\u003c/em\u003e, \u003cem\u003eSPATA2\u003c/em\u003e, \u003cem\u003eUSP30\u003c/em\u003e, \u003cem\u003eMYLIP\u003c/em\u003e, \u003cem\u003eGALNT2\u003c/em\u003e, \u003cem\u003eZC3H12A\u003c/em\u003e, \u003cem\u003eGALNT8\u003c/em\u003e) showing robust predictive capabilities (training AUC: 1.000, validation AUC: 0.945/0.673). Single-cell analysis localized hub genes to epithelial and immune compartments, with CellChat revealing remodeled intercellular communication between high-hub and low-hub states. Spatial transcriptomics demonstrated region-restricted expression hotspots, with \u003cem\u003eHIF1A\u003c/em\u003e showing broader activation patterns. Virtual knockdown revealed \u003cem\u003eGALNT8\u003c/em\u003e disruption impaired barrier genes (\u003cem\u003eTFF3\u003c/em\u003e, \u003cem\u003eITLN1\u003c/em\u003e), while \u003cem\u003eHIF1A\u003c/em\u003e depletion attenuated inflammatory modules.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eOur results demonstrate that PTM-related genes contribute substantially to UC heterogeneity, with the two subtypes predicting distinct immune-metabolic phenotypes. The seven-gene signature we developed exhibits robust predictive capabilities. Single-cell and spatial analyses link hub genes to immune\u0026ndash;epithelial contexts, offering insights for UC stratification and therapeutic decision-making.\u003c/p\u003e","manuscriptTitle":"Systematic post-translational modification profiling identifies molecular subtypes and therapeutic targets in ulcerative colitis: evidence from multi-omics integration","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-08 16:22:11","doi":"10.21203/rs.3.rs-9497163/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2026-04-24T07:43:05+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-24T04:13:20+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-04-24T04:12:58+00:00","index":"","fulltext":""},{"type":"submitted","content":"Immunologic Research","date":"2026-04-22T13:20:45+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"immunologic-research","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"imre","sideBox":"Learn more about [Immunologic Research](http://link.springer.com/journal/12026)","snPcode":"12026","submissionUrl":"https://submission.nature.com/new-submission/12026/3","title":"Immunologic Research","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"0b60283e-2e19-40cf-8ec6-20664963e8ee","owner":[],"postedDate":"May 8th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-08T16:22:12+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-08 16:22:11","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9497163","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9497163","identity":"rs-9497163","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}