Integrating Machine Learning and Spatial Transcriptomics Uncovers Shared Immunomodulatory Deubiquitinases in MAFLD and HCC

Integrating Machine Learning and Spatial Transcriptomics Uncovers Shared Immunomodulatory Deubiquitinases in MAFLD and HCC | 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 Integrating Machine Learning and Spatial Transcriptomics Uncovers Shared Immunomodulatory Deubiquitinases in MAFLD and HCC Yu-xi Han, Hongze Li, Wendi Xia, Junyi Ma, Jiaqi Zhang, Yiling Li This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6909459/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 10 Apr, 2026 Read the published version in Human Genetics → Version 1 posted 9 You are reading this latest preprint version Abstract Hepatocellular carcinoma (HCC) represents a malignancy with high global mortality. Metabolic dysfunction-associated fatty liver disease (MAFLD) serves as a significant contributory pathogenic factor. Deubiquitinases (DUBs), which regulate protein homeostasis, are implicated in disease progression. This study focused on identifying shared mechanisms between MAFLD and HCC, screening for key DUB genes, and constructing a novel prognostic scoring system termed the Deubiquitination Score (DUBS). The DUBS significantly stratified patient survival, with a high-DUBS indicating poor prognosis and malignant tumor progression. Patients with a low-DUBS demonstrated enhanced responses to immunotherapy and prolonged survival. Their tumors exhibited characteristics of "hot tumors," featuring abundant immune cell infiltration and an active tumor microenvironment, accompanied by higher microsatellite instability（MSI）. The core gene identified, EIF3F, exhibited superior cross-disease diagnostic value between HCC and MAFLD. In vitro and in vivo experiments confirmed that core DUBs significantly promoted malignant behaviors of HCC cells and tumorigenic capacity in vivo. Furthermore, the DUBS revealed associations with sensitivity to chemotherapeutic agents. In summary, this study provided an important molecular tool and mechanistic foundation for the early screening, prognostic assessment, differential diagnosis, and prediction of immunotherapy response in HCC. It also highlighted potential directions for targeted intervention and the development of combination therapeutic strategies. Machine Learning Deubiquitinases Spatial Transcriptome Immune Microenvironment Immunotherapy Prognosis Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Introduction Globally, liver cancer ranks as the seventh most common malignancy and the third leading cause of cancer-related mortality, accounting for approximately 5% of all new malignant tumor cases, and has become a major public health issue [ 1 ] . HCC is the predominant histological subtype of primary liver cancer. Its clinical characteristics include an insidious onset and a lack of specific clinical manifestations in the early stages, posing significant challenges for disease diagnosis and clinical intervention. For patients with advanced HCC, systemic treatment regimens represented by the combination of Atezolizumab and Bevacizumab have been recommended as first-line therapy by the US Food and Drug Administration (FDA) and guidelines from internationally authoritative academic institutions [ 2 ] . However, existing treatment options still exhibit significant limitations, and some patients fail to achieve a desirable clinical response [ 3 ] . This indicates that the optimization of early diagnostic systems for HCC and the development of individualized treatment strategies require further in-depth exploration. A substantial proportion of HCC occurs in the context of multiple coexisting chronic liver diseases. Among these, MAFLD has emerged as a primary driver of the continuously rising global HCC incidence. A meta-analysis incorporating 22 studies revealed that MAFLD was associated with 48.7% of HCC cases [ 4 ] . However, when MAFLD was present as an independent liver disease, it accounted for only 12.4% of the etiological distribution in HCC. This epidemiological evidence indicates that MAFLD acts primarily as a co-carcinogenic factor rather than a significant independent carcinogen in HCC development. Therefore, building upon existing research on MAFLD-related HCC, systematically elucidating the pathogenic synergy between MAFLD and HCC caused by various chronic liver diseases holds significant translational medical value. Ubiquitination, as a crucial mechanism of post-translational protein modification, regulates cellular functions by modulating protein localization, degradation, and interactions [ 5 ] . DUBs confer reversibility to this modification process through the specific hydrolysis of ubiquitin-substrate linkages [ 6 ] . Recent studies have demonstrated that DUBs play key regulatory roles in both MAFLD and HCC: for instance, RPN11 regulates the expression of lipid metabolism-related genes via the METTL3-ACSS3-histone propionylation modification axis; USP family members promote HCC cell proliferation, migration, and stemness through their deubiquitinating activity [ 7 – 10 ] . However, no dual-function DUBs molecule capable of simultaneously regulating both MAFLD progression and HCC development has been identified to date. Identifying such core DUBs regulatory nodes in the MAFLD-HCC comorbidity process will provide a novel theoretical foundation for establishing prognostic biomarker systems and developing targeted therapeutic strategies. This study, based on multi-omics integrated analysis technology, systematically identifies the key DUBs regulatory network in the MAFLD-HCC comorbidity process, develops a DUBS based on 4 core DUBs, and constructs an HCC prognostic stratification system and an immunotherapy efficacy prediction system. Furthermore, pan-cancer analysis is employed to decipher the broad-spectrum biological characteristics of DUBS in tumor biology, and in vitro and in vivo functional validation experiments are combined to elucidate the oncogenic molecular mechanisms of the key DUBs. This research reveals, for the first time, the molecular network mediated by DUBs in MAFLD-HCC, providing novel molecular markers and potential therapeutic targets for early screening, optimization of precision treatment strategies, and dynamic prognostic monitoring in MAFLD-HCC patients. Materials and Methods Data Acquisition and Processing Pipeline Cancer-associated DUBs were obtained from Dewson et al.'s review [11] . Gene expression matrices and relevant clinical information for MAFLD patients and HCC patients were downloaded from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/gds/), The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/), and International Cancer Genome Consortium (ICGC, https://dcc.icgc.org/). We additionally collected data from two distinct patient cohorts receiving anti-PD-1/PD-L1 immune checkpoint inhibitor therapy: (1) Riaz et al. (melanoma patients undergoing nivolumab treatment); (2) Liu et al. (melanoma patients treated with nivolumab or pembrolizumab). Subsequently, we acquired gene expression data and clinical data for TCGA pan-cancer and adjacent normal tissues through the UCSC Xena platform, where RNA sequencing data had been normalized to Fragments Per Kilobase of transcript per Million mapped (FPKM) reads format for subsequent analysis. Screening for Shared DUBs in MAFLD and HCC We performed differential expression analysis of gene expression data using the "limma" R package, with a p-value ＜ 0.05 considered statistically significant. The expression patterns of genes identified by differential analysis were visualized using the "pheatmap" R package. Subsequently, we integrated and batch-corrected the gene expression datasets using the "sva" R package. To screen for MAFLD diagnostic biomarkers, we employed 12 machine learning algorithms: Lasso, Ridge, Stepglm, XGBoost, Random Forest (RF), Elastic Net (Enet), Partial Least Squares Regression for Generalized Linear Models (plsRglm), Generalized Boosted Regression Models (GBM), Naive Bayes, Linear Discriminant Analysis (LDA), Generalized Linear Model Boosting (glmBoost), and Support Vector Machine (SVM). Utilizing the combined data from the GSE135251 and GSE130970 cohorts as the training set, we explored 113 algorithm combinations derived from these 12 algorithms, aiming for variable selection and model development within a 10-fold cross-validation framework. Immediately following this, the GSE48452, GSE63067, and GSE89632 cohorts were used as external validation sets to assess model performance. Based on prior research, we defined the optimal model as the one demonstrating a higher overall mean Area Under the Curve (AUC) and achieving AUC values greater than 0.7 across all training and validation sets. Finally, the genes selected by the optimal model were identified as MAFLD-HCC comorbidity-specific DUBs. DUBS Construction and Survival Stratification We integrated HCC datasets from three independent cohorts (TCGA-LIHC, LIRI-JP, and GSE116174). Data harmonization was performed using the "sva" R package with ComBat algorithm for batch effect correction. Prognosis-associated genes were initially identified through univariate Cox proportional hazards regression analysis for subsequent model construction. The multivariate Cox regression analysis was subsequently employed to establish the DUBS using the following formula: DUBS = Σ [Expression (Gene_i) × β_i] where β_i represents the risk coefficient derived from multivariate regression analysis. Cohort stratification was performed by dichotomizing patients into high- and low-DUBS subgroups using the median score as the cutoff value. Model interpretability was enhanced through SHapley Additive exPlanations (SHAP) analysis implemented via "kernelshap" and "shapviz" packages, with visualization generated using "ggplot2". Survival analyses were conducted with the "survival" R package, including proportional hazards assumption verification through Schoenfeld residuals and Kaplan-Meier survival curve estimation with log-rank tests for prognostic validation. Diagnostic performance was evaluated using time-dependent ROC curves (“pROC” package), with prognostic nomograms constructed via “rms” package. Immunotherapy Efficacy Prediction via Multi-dimensional Tumor Microenvironment Analysis To comprehensively evaluate tumor immune microenvironment characteristics and predict immunotherapy response, we implemented a multi-dimensional computational framework. First, utilizing the Tumor Immune Dysfunction and Exclusion (TIDE) algorithmic platform (http://tide.dfci.harvard.edu/login/), we computed TIDE scores and MSI expression signature (MSI Expr. Sig.) scores for pan-cancer TCGA cohorts. Elevated TIDE scores were interpreted as indicative of enhanced immune evasion potential, serving as predictive biomarkers for immunotherapy resistance. Subsequently, we conducted survival analyses stratified by the TCGA pan-cancer immune classification system established by Thorsson et al. to delineate prognostic disparities across immune subtypes. To further validate the predictive utility of immune checkpoint inhibitors, we integrated Immunophenoscore (IPS) data from The Cancer Immunome Atlas (TCIA, https://tcia.at/home). Comparative analysis of IPS values between anti-CTLA4 and anti-PD1 treatment cohorts (with higher IPS correlating with superior therapeutic efficacy) was performed to assess differential treatment responses. For quantitative characterization of the tumor microenvironment, ESTIMATE scores (including immune/stromal/ESTIMATE composite scores) were computed using the "ESTIMATE" package, systematically deconvoluting immune and stromal compartmentalization within tumor tissues. Finally, the Gene Set Variation Analysis (GSVA) algorithm was employed to interrogate the infiltration landscape of immune cell subsets and functional pathway activation states within the tumor immune niche. Computational Pharmacogenomic Profiling and Molecular Docking Analysis The Genomics for Drug Sensitivity in Cancer (GDSC, https://www.cancerrxgene.org/) and Cancer Therapeutics Response Portal (CTRP, http://portals.broadinstitute.org/ctrp/) databases were utilized to access tumor genomic profiles and corresponding drug sensitivity data. Correlation analyses between target gene expression and FDA-approved drug responses were conducted through the Gene Set Cancer Analysis (GSCA) platform (https://guolab.wchscu.cn/). Drug sensitivity disparities in high/low DUBS subgroups were evaluated using the 'oncopredict' R package with established chemotherapy agents. For molecular docking studies focusing on HCC therapeutics, the 3D structures of clinically used chemotherapeutic agents were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/). Structural data for EIF3F (PDB ID: 6YBD) were acquired from the Protein Data Bank (https://www.rcsb.org/). Molecular docking simulations were performed via the CB-DOCK2 web server (http://clab.labshare.cn/cb-dock2/), a cloud-based platform implementing Autodock Vina algorithms. The docking protocol prioritized binding poses based on Vina scores, with configurations exhibiting the highest absolute binding energy values selected as optimal interaction models. Single-cell RNA Sequencing Analysis and Spatial Transcriptomics Analysis The HCC single-cell RNA-seq dataset (GSE146115) was retrieved from the GEO repository. Raw count matrices underwent rigorous quality control using “Seurat” package. Post-filtering data were normalized via SCTransform with regularized negative binomial regression to correct technical variability. Principal component analysis (PCA) was performed on 3,000 highly variable genes identified by variance-stabilizing transformation, with significant PCs (Elbow method) selected for nonlinear dimensionality reduction using UMAP. Cluster-defining marker genes were identified through Wilcoxon rank-sum test with Bonferroni correction (adj.p value＜0.05, logFC.threshold＞1). Cell type annotation utilized “SingleR” package with HumanPrimaryCellAtlas, BlueprintEncode, MonacoImmuneData, NovershternHematopoieticData, DatabaseImmuneCellExpressionData, ref_Hematopoietic and ref_Human_all reference datasets, followed by “CellChat” package employing a weighted ligand-receptor interaction database and information theory-based communication probability quantification. Spatially resolved HCC transcriptomes (GSE238264) were curated from GSM7661258/GSM7661260 slides exhibiting high target gene expression entropy. 10X Visium data were aligned via Space Ranger (https://support.10xgenomics.com/spatial-gene-expression/software/downloads/latest) using STAR spliced read mapper (GRCh38.p13). Tissue segmentation and spot-by-gene matrices were imported into Seurat for integration with scRNA-seq data through reciprocal PCA-based anchor identification. Following variance stabilization, PCA-reduced dimensions underwent UMAP embedding. Spatial cluster topology was visualized through Seurat 's multimodal plotting functions. Multimodal intersection analysis (MIA) implemented hypergeometric testing to quantify subpopulation colocalization. Cross-platform annotation leveraged probabilistic label transfer via Seurat's FindTransferAnchors method, establishing a joint embedding space for scRNA-seq derived cell states and spatial transcriptomic spots. Pan-cancer Analysis The DUBS core genes were analyzed for pan-cancer expression differences using TCGA data (TOIL-normalized FPKM) acquired from the UCSC Xena platform. Differential expression between tumor and adjacent normal tissues was evaluated through Wilcoxon rank-sum tests with Benjamini-Hochberg false discovery rate (FDR) correction, visualized via log2(FPKM+1)-transformed boxplots. Clinical survival data from TCGA were subsequently integrated to perform Cox proportional hazards regression analysis using the "survival" package, with gene expression-survival correlations visualized through hazard ratio heatmaps generated by "ggplot2". For immunotherapy response evaluation, TIDE scores were computed from TCGA pan-cancer profiles. Comparative analysis of immune evasion potential and predicted immune checkpoint blockade response probabilities between high and low-DUBS subgroups was conducted based on TIDE-derived metrics. Cell Culture Human hepatic stellate cell line LX-2, human HCC cell line HepG2, Huh7, and C3A cells (Shanghai Cell Bank of the Chinese Academy of Sciences) were cultured in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS), with the culture medium replaced every 48 h. All 4 cell lines were maintained in a constant temperature incubator at 37°C with 5% CO₂. Based on prior growth curve analysis, serum starvation synchronization was performed on cells before drug treatment. Stable shRNA Knockdown in HCC Cell Lines HepG2 and Huh7 cells were plated in 6-well plates until reaching 60-70% confluence. Lentiviral particles carrying EIF3F-specific shRNA were complexed with Lipofectamine transfection reagent in serum-free DMEM high-glucose medium. The shRNA-lipid complexes were delivered through spinoculation followed by standard incubation. The medium was changed after 24 h to reduce cytotoxicity and finally the cells were screened with puromycin for 2 weeks to achieve a stable cell line and passaged for protein extraction shEIF3F sequence was: shEIF3F#1: Forward:5’-CCGGCTCTCAAGTGACTTGCAGCAACTCGAGTTGCTGCAAGTCACTTGAGAGTTTTTG-3’ Reverse:5’-AATTCAAAAACTCTCAAGTGACTTGCAGCAACTCGAGTTGCTGCAAGTCACTTGAGAG-3’ shEIF3F#2: Forward:5’-CCGGGTACTACGACACTGAACGCATCTCGAGATGCGTTCAGTGTCGTAGTACTTTTTG-3’ Reverse:5’-AATTCAAAAAGTACTACGACACTGAACGCATCTCGAGATGCGTTCAGTGTCGTAGTAC-3’ Protein Extraction and Western Blot Analysis Cellular protein isolation and immunoblotting were conducted with protocol modifications for adherent cell lines. Cells were rinsed three times with and lysed in pre-chilled RIPA buffer supplemented with a protease inhibitor cocktail (1:100) for 20 min on ice. Lysates were clarified by centrifugation (12,000 × g, 10 min, 4°C), and protein quantification was performed using a BCA assay. Equal protein aliquots were denatured (95°C, 5 min) in Laemmli buffer and electrophoresed on 10% SDS-polyacrylamide gels. Separated proteins were electrotransferred to PVDF membranes, blocked with 5% skim milk for 1 h, and immunoprobed with specific primary antibodies (1:1000 dilution, overnight at 4°C). Following TBST washes, membranes were incubated with HRP-linked secondary antibodies (1:5000 dilution, 1 h, RT) and developed via ECL substrate. All experiments included 3 independent biological replicates to ensure reproducibility. Wound Healing and Transwell Assay The wound healing assays were performed using Huh7 and HepG2 cell lines. Briefly, parallel reference lines were pre-marked on the bottom of 6-well plates. Logarithmic-phase cells were trypsinized, resuspended, and seeded until monolayer confluence was achieved. A sterile ruler was used to plan scratch trajectories perpendicular to the reference lines. Standardized scratches were generated along predefined paths using a 10 μl pipette tip, followed by three washes with PBS to remove detached cells. Serum-free medium was applied to minimize proliferation interference. Scratch images were captured at 0 and 48 h post-scratching using an inverted microscope (100× magnification). Wound closure areas were quantified using ImageJ software, with migration rates calculated as: [(Initial area – Final area)/Initial area] × 100%. Experiments were independently repeated 3 times. Intergroup differences were analyzed by independent Student's t-test. For the Transwell assay, HCC Huh7 and HepG2 cells from the NC and sh groups were serum-starved in serum-free medium for 4 h to eliminate serum interference, followed by trypsinization and resuspension in serum-free medium. Cell suspensions were seeded into the upper chambers, while the lower chambers of the 24-well plate were filled with complete medium containing 10% FBS. After 48h incubation at 37°C, cells were washed three times with PBS, fixed with 4% paraformaldehyde (30 min), and stained with crystal violet (30 min). Non-invading cells in the upper chambers were mechanically removed using cotton swabs. Microscopic images were captured at 200× magnification, and cell quantification was performed using Image J software, with triplicate independent experiments conducted. CCK-8 and Colony Formation Assays To evaluate cell proliferation activity, this study employed dual-dimensional analysis through CCK-8 assay and colony formation assay using HCC Huh7 and HepG2 cell lines. For the CCK-8 assay, both cell lines were digested with trypsin-EDTA and resuspended, followed by 3 independent counts using a hemocytometer to obtain mean values. sh-TIMD4-transfected and control Huh7/HepG2 cells were seeded into 96-well plates at a density of 2×10³ cells/well. The experiment included 5 time points (0 h, 24 h, 48 h, 72 h, and 96 h), with 6 replicate wells per time point. At each designated interval, 10 μL of CCK-8 solution was precisely pipetted into each well after gentle mixing. The plate was wrapped in aluminum foil for 4 h dark incubation at 37°C, and absorbance at 450 nm was measured using a multimode microplate reader. For the colony formation assay, sh-TIMD4-treated and control Huh7/HepG2 cells were plated in 6-well plates at 1×10³ cells/well and cultured in high-glucose DMEM medium (Gibco, Cat. No. 11965092) supplemented with 10% fetal bovine serum under 5% CO₂ at 37°C for 7 days. Cells were fixed with 4% paraformaldehyde for 30 min, stained with 0.1% crystal violet for 30 min, and washed 3 times with PBS before air-drying. Colony morphology was documented using a high-resolution imaging system, and colonies containing ≥50 cells were quantified using ImageJ software. All experiments were independently repeated in triplicate. Subcutaneous Xenograft Tumor Model This study established HepG2 HCC xenografts in male BALB/c nude mice (6-week-old, n=10) obtained from China Medical University's accredited animal center with IACUC approval. Viable HepG2 cells were suspended in sterile PBS (5×10⁶ cells/mL) and inoculated subcutaneously (200 μL) into the right axilla. Mice were divided into NC (n=5) and shRNA groups (n=5). Tumor dimensions were measured every 4 days from day 4 post-inoculation using digital calipers, with volumes calculated as 0.52×L×W². The study terminated at day 28 or upon reaching ethical thresholds (>1500 mm³ tumor volume or >20% weight loss). Euthanasia was performed via sodium pentobarbital (50 mg/kg) anesthesia and cervical dislocation, followed by tumor excision/weighing. All procedures were conducted in SPF conditions (22±1°C, 50±5% humidity, 12-h light cycle). Data were analyzed using two-tailed t-tests in GraphPad Prism 9.0 (p value < 0.05). Statistical Analysis and Visualization Statistical analyses were conducted using methodologically validated approaches. Intergroup comparisons employed Student's t-test and Wilcoxon rank-sum test, while survival rate disparities were evaluated through Kaplan-Meier curve analysis. Correlation coefficients were determined via Spearman's rank-order test, with Cox proportional hazards regression modeling providing hazard ratios (HR) and 95% confidence intervals (CI). Significance thresholds were established as *P value < 0.05, **P value < 0.01, and ***P value < 0.001. Analytical workflows were implemented in R (version 4.1.0) and GraphPad Prism 8.0, with graphical representations generated using ImageJ and Adobe Illustrator 2019. Results Shared Gene Identification in MAFLD and HCC Utilizing cross-cohort differential gene integration analysis, we systematically analyzed the regulatory characteristics of DUB genes in MAFLD. Based on cross-cohort differential expression analysis of the TCGA-LIHC and GSE135251 cohorts, 31 shared differentially expressed genes (DEGs) were identified through Venn diagram intersection (Fig. 1A). Heatmap visualization was employed to analyze the expression patterns of these 31 shared DEGs. The results demonstrated significant heterogeneity in their expression across MAFLD samples (Fig. 1B), with the vast majority of genes exhibiting a significant upregulation trend in cancerous tissues (Fig. 1C). To further identify core regulatory factors, this study employed twelve distinct machine learning algorithms to construct feature selection models. By comparing model performance, the Lasso + LDA combined algorithm was ultimately determined to be the optimal model, achieving AUC values greater than 0.7 in both the training and validation sets, and exhibiting the highest overall average AUC (Fig. 1D and S1A). Consequently, 15 DUBs with crucial regulatory potential were selected. This screening outcome reveals the potentially important biological functions these DUBs may play in the pathological progression of both TCGA-LIHC and MAFLD. Establishment of DUBS This study systematically screened the association of 15 DUBs with HCC prognosis. Univariate Cox regression analysis successfully identified 10 DUB molecules with significant prognostic value (Fig. 2A). Based on the multivariate Cox proportional hazards model, the DUBS system comprising USP39, USP32, OTUB2, and EIF3F was further constructed. The calculation formula is as follows (Fig. 2B): DUBS = Exp (USP39) * 0.069 + Exp (USP32) * 0.075 + Exp (OTUB2) * 0.238 + Exp (EIF3F) * 0.014 The SHAP feature interpretation algorithm confirmed a positive correlation between the expression levels of the core genes and prognostic prediction efficacy, revealing the model's interpretability features (Fig. 2C). During the clinical validation phase, survival analysis based on grouping by the median DUBS value demonstrated significantly prolonged overall survival in the low-DUBS group, validating the prognostic assessment efficacy of this scoring system across multiple dimensions (Fig. 2D-I). Notably, pathway enrichment analysis revealed significant biological characteristic differences between the two groups: the high-DUBS group exhibited overactivation of proliferation-related pathways such as cell cycle regulation and DNA replication, which may explain its poor clinical prognosis phenotype. Conversely, the low-DUBS group showed sustained activity in metabolic pathways like lipid metabolism and oxidative phosphorylation, indicating that relatively preserved liver physiological function may constitute a protective prognostic mechanism. These findings not only confirm the clinical application potential of DUBS as a novel prognostic biomarker but also elucidate the underlying mechanisms of heterogeneous HCC progression from the perspective of molecular network regulation. Analysis of DUBS Prognostic Model's Diagnostic Value Across Diseases and Clinical Translation Through systematic investigation of HCC prognostic markers, we have, for the first time, confirmed via univariate Cox regression models that DUBS serves as an independent prognostic risk factor for HCC patients (Fig. 3A), and its predictive efficacy remained robust in multivariate regression analysis (Fig. 3B). In-depth analysis revealed that high-DUBS expression was significantly positively correlated with tumor malignancy (Fig. 3C), providing crucial evidence for elucidating DUBS's pro-oncogenic mechanism. ROC curve analysis optimized by machine learning further expanded the clinical application dimension of DUBS-associated genes. In HCC differential diagnosis: USP39 (AUC=0.965), EIF3F (AUC=0.930), and OTUB2 (AUC=0.887) demonstrated excellent diagnostic performance. And in early MAFLD identification: OTUB2 (AUC=0.870), USP32 (AUC=0.761), and EIF3F (AUC=0.731) also exhibited outstanding performance. Notably, in diagnostic biomarker screening, EIF3F displayed unique cross-disease diagnostic value: this gene not only demonstrated excellent efficacy in HCC differential diagnosis (AUC=0.930) but also emerged as the sole molecular biomarker possessing diagnostic potential in both HCC and MAFLD. Its dual diagnostic characteristics suggest its core regulatory role in liver diseases (Fig. 3D). To advance precision medicine practice, this study innovatively constructed a multi-parameter integrated nomogram model (Fig. 3E). This model organically integrates molecular biomarkers with clinical characteristics through a standardized scoring system, providing a quantitative tool for individualized prognosis management that combines scientific rigor with clinical utility. Analysis of Hot Tumor Phenotype Reveals Mechanisms Underlying Immunotherapy Sensitization in Low-DUBS Group Patients This study systematically revealed the predictive association between DUBS expression levels and the efficacy of immune checkpoint inhibitors by integrating a multi-dimensional biomarker evaluation system. Survival analysis based on two independent anti-PD-1/PD-L1 treatment cohorts demonstrated significantly prolonged overall survival in patients with low-DUBS expression (Fig. 4A). Furthermore, the DUBS score was significantly lower in the complete response/partial response (CR/PR) subgroup compared to the stable disease/progressive disease (SD/PD) group. This survival advantage was further validated in progression-free survival analysis (Fig. 4B). TIDE score analysis indicated that the low-DUBS group had a superior immune evasion prediction value (Fig. 4C), while MSI comparison revealed significantly elevated MSI levels in the low-DUBS group (Fig. 4D), suggesting potential mechanisms for immunotherapy sensitivity. Further analysis of TCGA pan-cancer data based on Thorsson's immune classification revealed that patients with the C3 subtype, characterized by Th17/Th1 signature dominance and lower tumor proliferative activity, had significantly better prognosis than those with the C1 subtype dominated by angiogenic signatures (Fig. 4E), highlighting the independent impact of immune classification features on treatment outcomes. IPS analysis confirmed that patients with low DUBS expression exhibited stronger treatment sensitivity, regardless of targeting PD-1 or CTLA4 (Fig. 4F) [12] . Tumor microenvironment assessment showed significantly elevated Stromal Score, Immune Score, and combined assessment value in the low-DUBS group (Fig. 4G), elucidating the positive regulatory role of the "hot tumor" phenotype characteristics on immunotherapy from the perspective of microenvironment heterogeneity. These findings, supported by multiple independent lines of evidence including survival benefit, molecular characteristics, predictive models, and microenvironment remodeling, collectively endorse the clinical application potential of DUBS as a predictive biomarker for immunotherapy response. Subsequently, through systematic analysis of the immune microenvironment characteristics in the low-DUBS expression group, we revealed the multi-layered mechanisms underlying the formation of its "hot tumor" phenotype. Immune cell infiltration analysis (Fig. 5A) demonstrated significantly increased infiltration abundance of key anti-tumor immune components such as B lymphocytes, CD8+ T cells, cytotoxic effector cells, and dendritic cells in this group, providing the cellular basis for improved prognosis. Feature enrichment analysis (Fig. 5B) further confirmed significant enrichment in core immune signature pathways including B cell activation, immune checkpoint expression, cytolytic activity, and NK cell function in these patients, highlighting the dynamic activated state of their immune microenvironment. At the mechanistic level, the core DUBS genes EIF3F and USP39 not only showed significant positive correlations with the expression of immune checkpoint molecules such as PDCD1, CTLA4, and CD274 (Fig. 5C), suggesting their potential role in stabilizing checkpoint protein expression via deubiquitination, but also exhibited a co-expression pattern with HLA family antigen-presenting genes (Fig. 5D), revealing their dual regulatory mechanism in activating immune responses by enhancing tumor antigen presentation capability. This unique microenvironment signature—characterized by the synergistic action of highly infiltrating immune cells, the enrichment effect of activated pathways, and the molecular cascade of checkpoint-antigen presentation—collectively constitutes the biological basis for immunotherapy sensitivity in the low-DUBS group, elucidating the intrinsic link between the "hot tumor" phenotype and enhanced treatment response from multiple dimensions. Clinical Application Value of DUBS Core Genes in Drug Sensitivity Prediction and Combination Therapy for HCC Based on the comprehensive findings regarding the pivotal role of the DUBS core gene system in disease diagnosis, prognostic assessment of HCC, and immune microenvironment regulation, we postulate its significant research value in predicting drug efficacy. This study reveals multi-dimensional associations between DUBS core gene expression and chemotherapeutic drug sensitivity: association analysis using the GDSC database demonstrated that drug responses to traditional chemotherapeutic agents including methotrexate, vorinostat, and tubastatin A positively correlated with OTUB2 expression but negatively correlated with EIF3F expression (Fig. 6A); CTRP database data further confirmed significant positive correlations between sensitivity to axitinib, doxorubincin, etoposide, mitomycin, and vincristine and OTUB2 expression levels (Fig. 6B). Further analysis within the DUBS grouping model revealed superior therapeutic efficacy for 5-fluorouracil, afatinib, gefitinib, erlotinib, dasatinib, and cyclophosphamide in the low-DUBS group, whereas cisplatin, gemcitabine, axitinib, and irinotecan exhibited enhanced drug responses in the high-DUBS group (Fig. 6C). To validate the reliability of these theoretical predictions, molecular docking simulations were performed via the CB-DOCK2 online platform to model interactions between clinically used drugs (donafenib, doxorubincin, flurouracil, gemcitabine, lenvatinib, and rivoceranib) and the EIF3F protein, with optimal docking models determined by selecting molecular conformations exhibiting the lowest binding free energy followed by visualization (Fig. 6D). Collectively, these findings suggest that a therapeutic strategy combining immunotherapy with small-molecule chemotherapeutic agents may provide critical research directions for enhancing clinical efficacy and improving survival outcomes in HCC patients. Integrating Single-Cell and Spatial Transcriptomics to Decipher the Microenvironmental Regulation and Therapeutic Response Mechanisms of EIF3F in HCC This study systematically deciphered the expression heterogeneity of DUBS core genes within tumor parenchyma and the immune microenvironment of HCC through integrated analysis of single-cell transcriptomic datasets, constructing an HCC single-cell transcriptomic atlas via standardized quality control and UMAP dimensionality reduction that comprised 13 characteristic subpopulations categorized into 5 major cell types (Fig. 7A). Cross-subpopulation profiling revealed significantly elevated EIF3F expression across most cellular lineages, with this pan-cellular distribution pattern dually confirmed through feature gene visualization (Fig. 7B) and spatial expression validation (Fig. 7C-D), collectively indicating its critical regulatory potential in HCC pathogenesis when integrated with prior functional enrichment analyses. To elucidate tumor-microenvironment crosstalk, we established a cell-cell communication network identifying dominant interactions between hepatic parenchymal and immune cells, wherein NK cells exhibited the most active signaling properties (Fig. 7E-F) and the CD74-CXCR4 molecular pair was confirmed as the central signaling hub (Fig. S2A). Spatial transcriptomic in situ analysis further demonstrated significant associations between EIF3F expression abundance and clinical treatment response, revealing lower expression in the combination targeted-immunotherapy sensitive group (ICI nivolumab + TKI cabozantinib) versus marked enrichment in resistant cases, suggesting its role in modulating therapeutic sensitivity and outcomes (Fig. 7G-H), while spatial heterogeneity analysis of microenvironmental infiltration characteristics revealed significant differences in immune cell spatial distribution patterns between response groups (Fig. S2B-D). Through multidimensional omics integration, this study provides the first elucidation of EIF3F's expression dominance at single-cell resolution and its function as an interaction network hub in HCC, revealing spatial heterogeneity-driven regulatory principles governing combination therapy response and uncovering CD74-CXCR4 signaling axis-mediated microenvironmental remodeling mechanisms, thereby offering novel molecular targets and theoretical foundations for precision therapeutic strategies. Pan-Cancer Expression-Prognosis Associations and Immunotherapy Response Prediction of DUBS Core Genes This study systematically evaluated the expression characteristics and clinical predictive value of DUBS core genes across pan-cancer datasets, revealing that OTUB2, USP32, and USP39 were upregulated in most cancer types while EIF3F exhibited specific overexpression in CHOL, COAD, GBM, KIRC, KIRP, LIHC, PRAD, READ, and THCA (Fig. 8A). Survival analysis demonstrated that high EIF3F expression significantly correlated with reduced overall survival in ACC, LIHC, and PCPG; OTUB2 functioned as a risk factor in CESC, LIHC, MESO, PCPG, and SARC; USP32 indicated poor prognosis in KICH, KIRP, LIHC, PCPG, and UCEC; whereas USP39 showed significant risk associations across 11 cancer types including ACC, KICH, KIRC, KIRP, LAML, LGG, LIHC, MESO, PAAD, PCPG, and UCEC (Fig. 8B). Notably, all DUBS core genes correlated with adverse clinical outcomes in PCPG, highlighting the critical role of this regulatory network in neuroendocrine tumors. For immunotherapy response prediction, DUBS expression patterns exhibited significant cancer type-specific variations: patients with BLCA, BRCA, HNSC, KIRC, PRAD, and THCA in the high-DUBS group demonstrated greater immunotherapy sensitivity, whereas GBM, LGG, LUAD, and SARC patients in the low-DUBS group showed superior treatment responses (Fig. 8C). This study confirms DUBS genes as reliable biomarkers for both prognostic assessment and immunotherapy efficacy prediction in LIHC, while their pan-cancer regulatory properties provide a novel theoretical foundation for precision oncology across malignancies, though future multicenter cohort validation and mechanistic investigations remain warranted. Validation of EIF3F Expression Based on the aforementioned screening results, analysis revealed that core DUBs exhibit significant clinical value in HCC diagnosis and prognostic assessment. Notably, EIF3F not only demonstrated pivotal regulatory functions in HCC diagnostic biomarker screening and survival prediction models, but its expression profile also showed significant correlations with MAFLD progression. To validate these findings, this study analyzed the expression profile of EIF3F protein in HCC cell lines using Western Blot, with experimental data revealing significantly elevated expression in Huh7 and HepG2 cell lines compared to hepatic stellate cell, whereas relatively lower expression levels were observed in the C3A cell line (Fig. 9A). Given the high expression characteristics of EIF3F in Huh7 and HepG2 cell lines, these two cellular models were selected for subsequent functional investigations. Effects and Mechanisms of EIF3F Gene Knockdown on Malignant Phenotypes in HCC To elucidate the biological functions of EIF3F in regulating malignant phenotypes of HCC, this study achieved specific EIF3F knockdown in HCC cell lines through construction of shRNA recombinant vectors (shEIF3F#1 and shEIF3F#2), with Western Blot confirming efficient protein expression suppression (Fig. 9B and C). Functional validation demonstrated that the gene knockdown group exhibited significantly reduced cell migration rates compared to controls in wound healing assays, indicating that EIF3F promotes cell motility (Fig. 9D and E). Plate colony formation assays further confirmed that EIF3F knockdown substantially decreased proliferative activity in both Huh7 and HepG2 cells (Fig. 9F and G). To comprehensively validate phenotypic effects, Transwell matrix invasion assays (Fig. 9H and I) and CCK-8 cell viability tests (Fig. 10A and B) were concurrently performed, revealing significantly reduced invasive cell penetration and markedly inhibited cell proliferation within 48 h in interference groups. Collectively, these multidimensional functional evidences demonstrate that EIF3F acts as an oncogene in HCC progression by coordinately regulating malignant phenotypes including migration, invasion, and proliferation, establishing its expression level as a potential diagnostic biomarker with promising targeted therapeutic value. In Vivo Experiment To investigate the in vivo regulatory role of EIF3F knockdown in HCC growth, this study established HepG2 cell lines with stable EIF3F knockdown using shRNA vectors, with Western blot confirming reduced protein expression compared to controls. In male BALB/c nude mouse xenograft models, the shEIF3F group exhibited significantly inhibited growth phenotypes versus NC controls: at the experimental endpoint, tumor volume was reduced and tumor weight demonstrated a 6.8-fold decrease in the knockdown group (Fig. 10C-E). These results indicate that reduced EIF3F expression significantly attenuates in vivo tumor growth potential by suppressing the proliferative capacity of HCC cells. Discussion MAFLD has emerged as a primary causative factor of chronic liver disease globally, with epidemiological studies revealing a significant geographical heterogeneity in its prevalence (6–35%) [ 13 ] . Due to the lack of early specific biomarkers for HCC, approximately 60% of patients are diagnosed at intermediate to advanced stages, missing the opportunity for curative treatment. Despite significant advances in HCC therapeutic strategies, some patients fail to achieve substantial clinical benefits, and the overall prognosis remains suboptimal. Although current targeted-immunotherapy combinations have markedly improved the clinical management landscape of HCC, 30–40% of patients exhibit primary resistance. Notably, aberrant deubiquitinating modifications constitute not only a significant driver of solid tumor progression but also play a crucial regulatory role in MAFLD pathogenesis [ 14 – 17 ] . Emerging evidence indicates that dysregulated deubiquitination deeply participates in the co-regulatory network of MAFLD-HCC comorbidity through mechanisms such as mediating lipid metabolic reprogramming and remodeling the inflammatory microenvironment. Molecular mechanism studies confirm that DUBs participate in disease progression by regulating innate immune signal transduction: for instance, USP21 employs a dual regulatory mechanism—its Lys63-linked deubiquitination suppresses RIG-I-mediated innate immune signaling activation, while simultaneously maintaining GATA3 protein stability in Treg cells via Lys48-linked deubiquitination to shape an immunosuppressive microenvironment [ 18 – 21 ] . Furthermore, USP30 activates the lipogenesis pathway by stabilizing C/EBPβ protein, directly regulating hepatic steatosis progression [ 22 ] . These findings unveil the molecular complexity of the deubiquitination regulatory network within the MAFLD-HCC comorbid process, highlighting the urgent need to establish a DUB activity-based molecular classification system to advance precision therapy. This study, for the first time, completed the systematic identification of HCC-specific DUBs within an MAFLD context, successfully establishing a multi-parameter DUBS evaluation system comprising USP39, USP32, OTUB2, and EIF3F. This system not only shows significant correlation with poor patient prognosis but also precisely quantifies individualized molecular characteristics, providing an objective basis for stratified therapeutic strategies; it further elucidates that specific DUB molecules exert key regulatory functions in the MAFLD-HCC comorbid process through the synergistic regulation of dual mechanisms—lipid metabolic reprogramming and immune microenvironment remodeling. The 4 core DUB molecules identified in this study are associated with MAFLD and various malignancies including HCC. USP39 is recognized as a prognostic marker and potential therapeutic target across multiple cancers, participating in regulating diverse cellular activities including proliferation, migration, invasion, apoptosis, and DNA damage repair [ 23 ] . It is known to maintain hepatocyte autophagy-lipid metabolism homeostasis through regulating alternative RNA splicing mechanisms, and cooperates with the E3 ubiquitin ligase TRIM26 to form a bidirectional regulatory switch for ZEB1 protein ubiquitination modification, decisively controlling HCC proliferation and metastasis [ 9 , 24 ] . USP32 is closely correlated with prognosis in patients with various solid tumors including small cell lung cancer, gastric cancer, and breast cancer. Xiu et al. reported that knocking down USP32 inhibits HCC cell proliferation, colony formation, and migration in vitro while suppressing tumor growth in vivo, with its expression significantly correlating with immune cell infiltration in the tumor microenvironment [ 25 , 26 ] . OTUB2 also plays regulatory roles in the pathogenesis of multiple malignancies, promoting malignant proliferation and migration of HCC cells by increasing PJA1 stability through deubiquitination. Additionally, OTUB2 has been demonstrated to be upregulated in colorectal cancer, where it inhibits the ubiquitination of pyruvate kinase M2 (PKM2) by blocking its interaction with the ubiquitin E3 ligase Parkin, thereby enhancing PKM2 activity, promoting glycolysis, and accelerating colorectal cancer progression [ 27 ] . The eukaryotic translation initiation factor 3 (eIF3) multi-subunit complex regulates core biological processes including cell differentiation and apoptosis through specific binding of its f subunit to target mRNAs [ 28 , 29 ] . Notably, loss of EIF3F expression is significantly associated with pathological processes in various solid tumors and muscular atrophy-related diseases [ 30 ] . Mechanistic studies reveal that EIF3F stabilizes phosphoglycerate dehydrogenase (PHGDH) protein expression through deubiquitination, activating the serine-glycine-one-carbon (SGOC) metabolic pathway to drive malignant progression in colorectal cancer [ 31 ] . Pan-cancer analyses indicate this molecule exhibits significant pro-oncogenic properties in solid tumors including prostate cancer, pancreatic ductal adenocarcinoma, malignant melanoma, and gastric cancer [ 32 – 35 ] . However, the regulatory role and molecular mechanisms of EIF3F in MAFLD and HCC remain undefined. This study first employed multiple machine learning approaches to screen 15 MAFLD core genes including EIF3F from shared DUBs in MAFLD-HCC; then integrated HCC transcriptomic and clinical data to identify 10 genes among them associated with HCC prognosis; subsequently determined EIF3F as a core HCC gene via SHAP analysis (ranking second only to USP39 in importance); finally confirmed through single-cell sequencing, spatial transcriptomics, and in vitro/in vivo validation that EIF3F promotes HCC initiation, growth, and metastasis by regulating malignant phenotypes including proliferation, invasion, and migration of HCC cells. Cancer cell resistance to conventional chemotherapeutic agents has profoundly altered the therapeutic landscape of HCC. While targeted agents and immunotherapeutics developed in recent years have provided novel approaches for HCC treatment, the immunosuppressive tumor microenvironment significantly diminishes immunotherapy efficacy, necessitating the urgent development of novel indicators for predicting the immune microenvironment and drug sensitivity. This study provides a new indicator, DUBS, for HCC treatment through drug sensitivity-gene expression correlation analysis and molecular docking techniques. Results demonstrate that DUBS effectively predicts differential patient responses to chemotherapy and immunotherapy: drug sensitivity analysis revealed superior efficacy of agents such as 5-fluorouracil in the low DUBS group, whereas patients in the high DUBS group exhibited significantly enhanced objective response rates to gemcitabine, providing crucial guidance for personalized treatment selection. Molecular docking-based virtual screening identified candidate drugs including donafenib, doxorubicin, fluorouracil, gemcitabine, lenvatinib, and rivoceranib as having potential therapeutic effects through specific binding to the EIF3F protein; immunotherapy efficacy analysis confirmed DUBS as a reliable predictor (higher DUBS correlates with poorer efficacy), and immune microenvironment analysis indicated that the high DUBS group exhibits immunologically cold tumor characteristics, explaining the inferior immunotherapy outcomes in this group. Therefore, for high-DUBS group patients, targeted therapy combined with chemotherapy may be employed to improve prognosis, while low-DUBS group patients may benefit from immunotherapy combined with targeted therapy or chemotherapy to enhance efficacy. Mechanistic investigation demonstrated that donafenib significantly enhances tumor cell apoptosis and inhibits metastasis through multiple mechanisms, including synergistically inducing reactive oxygen species (ROS) accumulation, promoting ferroptosis-related protein expression, and activating the p53 signaling pathway [ 36 ] . These findings suggest that DUBS-mediated heterogeneity in drug response provides a molecular classification basis for personalized therapy, and that targeting EIF3F combined with immune checkpoint inhibitors may improve HCC patient prognosis, potentially profoundly reversing disease outcomes in the future. Subsequent validation of the clinical utility of DUBS and elucidation of its regulatory network require prospective clinical trials integrated with multi-omics technologies. In summary, the DUBS system developed in this study not only accurately and effectively predicts the prognosis of HCC patients but also provides a novel therapeutic target (specifically EIF3F). However, this study has limitations: neither prospective validation of the DUBS model with clinical samples nor verification of the in vivo therapeutic efficacy of the candidate drugs was performed. Future work necessitates conducting large-scale prospective clinical trials to validate the model's predictive power, in-depth investigation of the mechanisms of action and in vivo efficacy of candidate drugs, and exploration of personalized treatment regimens for patients with both MAFLD and HCC. Conclusion Overall, this study established a DUBS, comprising USP39, USP32, OTUB2, and EIF3F, via multi-omics analysis, and confirmed its role as an independent prognostic marker for HCC. Results demonstrated that patients with low DUBS exhibited significantly prolonged overall survival, enhanced response to immunotherapy, and increased sensitivity to specific chemotherapeutic agents. The core gene EIF3F was identified as a key oncogenic driver, promoting tumor malignancy progression and demonstrating dual diagnostic value for both MAFLD and HCC. Collectively, the DUBS system provides a novel molecular tool for prognostic assessment, therapy response prediction, and individualized treatment strategy formulation in HCC, while highlighting EIF3F as a potential therapeutic target for precision medicine. Future investigations should elucidate the ubiquitination regulatory network mediated by EIF3F and validate, through prospective clinical trials, the translational value of targeting EIF3F combined with immune checkpoint inhibitors (e.g., Donafenib) for reversing chemoresistance in HCC. Abbreviations ACC: Adrenocortical Cancer; AUC: Area Under the Curve; BLCA: Bladder Cancer; BRCA: Breast Cancer; CESC: Cervical Cancer; CHOL: Bile Duct Cancer; CI: confidence intervals; COAD: Colon Cancer; CR: complete response; CTRP: Cancer Therapeutics Response Portal; DFI: Disease Free Interval; DEGs: differentially expressed genes; DLBC: Large B-cell Lymphoma; DSS: Disease Specific Survival; DUB: deubiquitinase; DUBS: Deubiquitination Score; eIF3:eukaryotic translation initiation factor3; ESCA: Esophageal Cancer; FBS: fetal bovine serum; FDA: Food and Drug Administration; FDR: false discovery rate; FPKM: Fragments Per Kilobase Million; GBM: Glioblastoma; GDSC: Genomics of Drug Sensitivity in Cancer; GEO: Gene Expression Omnibus; GSCA: Gene Set Cancer Analysis; GSVA: Gene Set Variation Analysis; GSEA: Gene Set Enrichment Analysis; HCC: hepatocellular carcinoma; HNSC: Head and Neck Cancer; HR: hazardratios; ICGC: International Cancer Genome Consortium; ICIs: Immune Check point Inhibitors; IPS: Immunophenoscore; KICH: Kidney Chromophobe; KIRC/ccRCC: Kidney/Renal Clear Cell Carcinoma; KIRP: Kidney Papillary Cell Carcinoma; KM: Kaplan–Meier; LAML: Acute Myeloid Leukemia; Lasso: Least Absolute Shrinkage and Selection Operator; LGG: Lower Grade Glioma; LIHC: Liver Cancer; LUAD: Lung Adenocarcinoma; LUSC: Lung Squamous Cell Carcinoma; MAFLD: metabolic dysfunction-associated fatty liver disease; MESO: Mesothelioma; MIA: multimodal intersection analysis; MSI: microsatellite instability; Neg.: Negative; OS: Overall Survival; OV: Ovarian Cancer; PAAD: Pancreatic Cancer; PBS: phosphate-buffered saline; PCA: principal component analysis; PD: progressive disease; PHGDH: phosphoglycerate dehydrogenase; PKM2: pyruvate kinase M2; Pos.: Positive; PR: partial response; PRAD: Prostate Cancer; READ: Rectal Cancer; ROC: Receiver Operating Characteristic Curve; ROS: reactive oxygen species; SARC: Sarcoma; SD: stable disease; SGOC: serine-glycine-one-carbon; SHAP: Shapley Additive exPlanations; SKCM: Melanoma; STAD: Stomach Cancer; TCGA: The Cancer Genome Atlas; THCA: Thyroid Cancer; TIDE: Tumor Immune Dysfunction and Exclusion; TMB: Tumor mutational burden; UCEC: Endometrioid Cancer; UVM: Ocular melanomas. Declarations Author Contributions Y.H. wrote and typeset the manuscript for this study; H.L. was responsible for data collection and organization, bioinformatics analysis and study design; W.X. and J.M. were responsible for experiments and editing the full text; J.Z. was responsible for typesetting the manuscript for this study and study design; Y.L. provided corrections and suggestions throughout the study and provided financial support. All authors reviewed the manuscript. Funding This work was supported by Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0508700), Scientific Research Fund of Liaoning Provincial Education Department (LJKMZ20221167)and Supported by Liaoning Revitalization Talents Program（XLYC2412033）. Data availability statement The raw data of this study are freely available from the website TCGA Research Network (https://portal.gdc.cancer.gov/), GEO database(https://www.ncbi.nlm.nih.gov/geo/), ICGC database(https://dcc.icgc.org/). Further inquiries can be directed to the corresponding authors. Conflict of interest The authors declare no conflicts of interest. Ethics approval and consent for participate statement Not applicable. Consent for publication Not applicable. Acknowledgements Not applicable. References Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries[J]. CA: a cancer journal for clinicians, 2021, 71(3): 209-249. Yang X, Yang C, Zhang S, et al. Precision treatment in advanced hepatocellular carcinoma[J]. Cancer Cell, 2024, 42(2): 180-197. Llovet JM, Montal R, Sia D, et al. Molecular therapies and precision medicine for hepatocellular carcinoma[J]. Nature Reviews. Clinical Oncology, 2018, 15(10): 599-616. Crane H, Eslick GD, Gofton C, et al. Global prevalence of metabolic dysfunction-associated fatty liver disease-related hepatocellular carcinoma: A systematic review and meta-analysis[J]. Clinical and Molecular Hepatology, 2024, 30(3): 436-448. Rape M. Ubiquitylation at the crossroads of development and disease[J]. Nature Reviews. Molecular Cell Biology, 2018, 19(1): 59-70. Reyes-Turcu FE, Ventii KH, Wilkinson KD. Regulation and cellular roles of ubiquitin-specific deubiquitinating enzymes[J]. Annual Review of Biochemistry, 2009, 78: 363-397. Wu Q, Qiu Y, Guo J, et al. USP40 promotes hepatocellular carcinoma cell proliferation, migration and stemness by deubiquitinating and stabilizing Claudin1[J]. Biology Direct, 2024, 19(1): 13. Zhou B, Luo Y, Bi H, et al. Amelioration of nonalcoholic fatty liver disease by inhibiting the deubiquitylating enzyme RPN11[J]. Cell Metabolism, 2024, 36(10): 2228-2244.e7. Li X, Yuan J, Song C, et al. Deubiquitinase USP39 and E3 ligase TRIM26 balance the level of ZEB1 ubiquitination and thereby determine the progression of hepatocellular carcinoma[J]. Cell Death and Differentiation, 2021, 28(8): 2315-2332. Dominguez DA, Wong P, Melstrom LG. Existing and emerging biomarkers in hepatocellular carcinoma: relevance in staging, determination of minimal residual disease, and monitoring treatment response: a narrative review[J]. Hepatobiliary Surgery and Nutrition, 2024, 13(1): 39-55. Dewson G, Eichhorn PJA, Komander D. Deubiquitinases in cancer[J]. Nature Reviews. Cancer, 2023, 23(12): 842-862. Thorsson V, Gibbs DL, Brown SD, et al. The Immune Landscape of Cancer[J]. Immunity, 2018, 48(4): 812-830.e14. Huang DQ, El-Serag HB, Loomba R. Global epidemiology of NAFLD-related HCC: trends, predictions, risk factors and prevention[J]. Nature Reviews. Gastroenterology & Hepatology, 2021, 18(4): 223-238. Xue Y, Xue C, Song W. Emerging roles of deubiquitinating enzymes in actin cytoskeleton and tumor metastasis[J]. Cellular Oncology (Dordrecht, Netherlands), 2024, 47(4): 1071-1089. Kong L, Jin X. Dysregulation of deubiquitination in breast cancer[J]. Gene, 2024, 902: 148175. Ren J, Yu P, Liu S, et al. Deubiquitylating Enzymes in Cancer and Immunity[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(36): e2303807. Kong L, Jin X. Dysregulation of deubiquitination in breast cancer[J]. Gene, 2024, 902: 148175. Bech-Otschir D, Kraft R, Huang X, et al. COP9 signalosome-specific phosphorylation targets p53 to degradation by the ubiquitin system[J]. The EMBO journal, 2001, 20(7): 1630-1639. Fan Y, Mao R, Yu Y, et al. USP21 negatively regulates antiviral response by acting as a RIG-I deubiquitinase[J]. The Journal of Experimental Medicine, 2014, 211(2): 313-328. Zhang Q, Chen Z, Tang Q, et al. USP21 promotes self-renewal and tumorigenicity of mesenchymal glioblastoma stem cells by deubiquitinating and stabilizing FOXD1[J]. Cell Death & Disease, 2022, 13(8): 712. Ren J, Yu P, Liu S, et al. Deubiquitylating Enzymes in Cancer and Immunity[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(36): e2303807. Jiang F, Li X, Lin L, et al. NPRC promotes hepatic steatosis via USP30-mediated deubiquitination of C/EBPβ[J]. Metabolism: Clinical and Experimental, 2025, 162: 156050. Y Z, F W, S F, et al. USP39 promote post-translational modifiers to stimulate the progress of cancer[J]. Discover oncology, 2025, 16(1). Cui D, Wang Z, Dang Q, et al. Spliceosome component Usp39 contributes to hepatic lipid homeostasis through the regulation of autophagy[J]. Nature Communications, 2023, 14(1): 7032. Xiu M, Bao W, Wang J, et al. High USP32 expression contributes to cancer progression and is correlated with immune infiltrates in hepatocellular carcinoma[J]. BMC Cancer, 2023, 23: 1105. S L, Y S, K W, et al. USP32 deubiquitinase: cellular functions, regulatory mechanisms, and potential as a cancer therapy target[J]. Cell death discovery, 2023, 9(1). S Y, W Z, Y Q, et al. Deubiquitinase OTUB2 exacerbates the progression of colorectal cancer by promoting PKM2 activity and glycolysis[J]. Oncogene, 2022, 41(1). Lee ASY, Kranzusch PJ, Cate JHD. eIF3 targets cell-proliferation messenger RNAs for translational activation or repression[J]. Nature, 2015, 522(7554): 111-114. Masutani M, Sonenberg N, Yokoyama S, et al. Reconstitution reveals the functional core of mammalian eIF3[J]. The EMBO journal, 2007, 26(14): 3373-3383. Marchione R, Leibovitch SA, Lenormand JL. The translational factor eIF3f: the ambivalent eIF3 subunit[J]. Cellular and molecular life sciences: CMLS, 2013, 70(19): 3603-3616. Pan Q, Yu F, Jin H, et al. eIF3f Mediates SGOC Pathway Reprogramming by Enhancing Deubiquitinating Activity in Colorectal Cancer[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(27): e2300759. Li J, Yu W, Ge J, et al. Targeting eIF3f Suppresses the Growth of Prostate Cancer Cells by Inhibiting Akt Signaling[J]. OncoTargets and Therapy, 2020, 13: 3739-3750. Doldan A, Chandramouli A, Shanas R, et al. Loss of the eukaryotic initiation factor 3f in melanoma[J]. Molecular Carcinogenesis, 2008, 47(10): 806-813. A D, A C, R S, et al. Loss of the eukaryotic initiation factor 3f in pancreatic cancer[J]. Molecular carcinogenesis, 2008, 47(3). Cheng Y, Zhou J, Li H. Clinicopathologic Implications of Eukaryotic Initiation Factor 3f and Her-2/neu Expression in Gastric Cancer[J]. Clinical and Translational Science, 2015, 8(4): 320-325. Liang J, Chen M, Yan G, et al. Donafenib activates the p53 signaling pathway in hepatocellular carcinoma, induces ferroptosis, and enhances cell apoptosis[J]. Clinical and Experimental Medicine, 2025, 25(1): 29. Additional Declarations No competing interests reported. Supplementary Files SupplementaryFigureLegend.docx SupplementaryTables.xlsx FigureS1.tif Figure S1: Diagnostic performance evaluation and functional enrichment profiling. (A) Diagnostic ROC trajectories of Lasso-LDA integration in each validation cohort.(B) GSEA results based on GO gene sets in the DUBS-high group.(C) GSEA results based on GO gene sets in the DUBS-low group.(D) GSEA results based on KEGG gene sets in the DUBS-high group.(E) GSEA results based on KEGG gene sets in the DUBS-low group. FigureS2.tif Figure S2: Cell-cell communication network profiling and spatial immune contexture analysis. (A) Bubble plot visualization of intercellular ligand-receptor interaction networks.(B) Spatially resolved immune cell infiltration quantification in responders versus non-responders to targeted-immunotherapy combination using transcriptome deconvolution.(C-D) Compartment-specific infiltration profiling of immune microenvironmental cells across spatial clusters. EIF3Fexpression1.tif EIF3Fexpression2and3.tif huh7tubulin1.tif huh7sh2.tif huh7sh3.tif huh7sh1.tif hepG2sh2.tif hepG2sh1.tif hepG2sh3.tif hepg2tubulin1.tif hepg2tubulin2.tif huh7tubulin2.tif huh7tubulin3.tif hepg2tubulin3.tif Cite Share Download PDF Status: Published Journal Publication published 10 Apr, 2026 Read the published version in Human Genetics → Version 1 posted Editorial decision: Revision requested 17 Sep, 2025 Reviews received at journal 05 Sep, 2025 Reviewers agreed at journal 15 Aug, 2025 Reviews received at journal 15 Aug, 2025 Reviewers agreed at journal 11 Aug, 2025 Reviewers invited by journal 23 Jun, 2025 Editor assigned by journal 17 Jun, 2025 Submission checks completed at journal 17 Jun, 2025 First submitted to journal 16 Jun, 2025 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-6909459","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":475596165,"identity":"23953e2d-df82-4182-81e7-fdf69c8e3c39","order_by":0,"name":"Yu-xi Han","email":"","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yu-xi","middleName":"","lastName":"Han","suffix":""},{"id":475596166,"identity":"0a06a8c2-30bf-42b5-bb5c-38aadbe4c127","order_by":1,"name":"Hongze Li","email":"","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Hongze","middleName":"","lastName":"Li","suffix":""},{"id":475596167,"identity":"53319bb1-c55e-4dca-a744-3d847c4c93e1","order_by":2,"name":"Wendi Xia","email":"","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Wendi","middleName":"","lastName":"Xia","suffix":""},{"id":475596168,"identity":"276389f8-3bb9-454e-824c-a169fc3c3ad8","order_by":3,"name":"Junyi Ma","email":"","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Junyi","middleName":"","lastName":"Ma","suffix":""},{"id":475596169,"identity":"b56e7a52-7c14-4816-a6fa-713a657bc3cf","order_by":4,"name":"Jiaqi Zhang","email":"","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jiaqi","middleName":"","lastName":"Zhang","suffix":""},{"id":475596170,"identity":"d3e70315-654a-43f6-bc8e-590e2e582899","order_by":5,"name":"Yiling Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAArUlEQVRIiWNgGAWjYBACPgbmBuY/FTY8/PwNRGphY2BsYOA5kyYjOeMAKVp42w7bGDQkEKtFIrGBQYLtPI8BwwHGDx9ziNViwHObxxzoJ8mZ24jT0v4jQeI2j2XDATZmXiK1NDAcMDjHY3AggQQtjA0JB0jRwvOwgZnhQDKP5IyDzcT5hZ89+QAz4z87e37+5oMfPhKjhUEgAcYCRhBxgP8AkQpHwSgYBaNg5AIAB4EwqZ1HxHQAAAAASUVORK5CYII=","orcid":"","institution":"The First Hospital of China Medical University","correspondingAuthor":true,"prefix":"","firstName":"Yiling","middleName":"","lastName":"Li","suffix":""}],"badges":[],"createdAt":"2025-06-17 02:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6909459/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6909459/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1007/s00439-026-02833-6","type":"published","date":"2026-04-10T15:59:18+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":85616676,"identity":"73aca0dc-d1fe-460a-859d-686d7a24ec07","added_by":"auto","created_at":"2025-06-29 14:39:04","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":539097,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eScreening of common genes between MAFLD and HCC. \u003c/strong\u003e(A) Intersection of differentially expressed genes in MAFLD and HCC.(B) Heatmap of expression differences of intersection genes in MAFLD.(C) Heatmap of expression differences of intersection genes in HCC.(D) AUC values of 113 machine learning algorithms.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/125902ca08b9ce8204f40329.png"},{"id":85616697,"identity":"c1cb2250-2f0b-43b4-a589-77bbbf681104","added_by":"auto","created_at":"2025-06-29 14:39:06","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":188709,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEstablishment of DUBS and validation of its prognostic prediction efficacy. \u003c/strong\u003e(A) Forest plot of univariate regression analysis for MAFLD-characteristic DUBs in HCC.(B) Coefficients of the DUBS multivariate regression model.(C) SHAP analysis coefficient honeycomb plot.(D) Survival analysis based on the TCGA-LIHC cohort.(E) Survival analysis based on the ICGC cohort.(F) Survival analysis based on the GSE116174 cohort.(G) Survival status dot plot and heatmap of DUBS core gene expression based on the TCGA-LIHC cohort.(H) Survival status dot plot and heatmap of DUBS core gene expression based on the ICGC cohort.(I) Survival status dot plot and heatmap of DUBS core gene expression based on the GSE116174 cohort.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/2d09f0705e02703c99c26039.png"},{"id":85617633,"identity":"1513f614-6d24-4102-9837-de0024f20b19","added_by":"auto","created_at":"2025-06-29 14:47:05","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":224529,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePrognostic and diagnostic analysis of DUBS. \u003c/strong\u003e(A) Univariate regression analysis of clinical characteristics and DUBS.(B) Multivariate regression analysis of clinical characteristics and DUBS.(C) Distribution of clinical characteristics in high/low DUBS groups.(D) AUC values of diagnostic efficacy for DUBS core genes in HCC and MAFLD cohorts.(E) \u0026nbsp;Nomogram for HCC prognostic prediction.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/2fcffb48ac708e018a24edeb.png"},{"id":85616709,"identity":"be204373-2c16-47ec-b565-901f39ae35bc","added_by":"auto","created_at":"2025-06-29 14:39:06","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":155875,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRole of DUBS in predicting immunotherapy efficacy. \u003c/strong\u003e(A) Comparison of overall survival \u0026nbsp;differences and immunotherapy efficacy in the GSE91061 cohort.(B) Comparison of progression-free survival differences and immunotherapy efficacy in the Liu et al. cohort.(C) Comparison of TIDE scores and immunotherapy prediction efficacy between high/low DUBS groups.(D) Comparison of MSI differences between high/low DUBS groups.(E)Survival analysis across different immune subtypes.(F)Comparison of IPS score differences between high/low DUBS groups.(G) Comparison of immune microenvironment score differences between high/low DUBS groups.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/d2ccc4359d9032b05b694df5.png"},{"id":85616688,"identity":"8139aebe-a1a4-44bc-9a72-7dd414f4651e","added_by":"auto","created_at":"2025-06-29 14:39:05","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":268067,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAnalysis of the immune microenvironment in HCC patients. \u003c/strong\u003e(A) Differences in immune cell infiltration between high/low DUBS groups in HCC patients.(B)Differences in immune function enrichment between high/low DUBS groups in HCC patients.(C) Correlation heatmap between immunosuppressive genes and DUBS core gene expression.(D) Correlation heatmap between immune-activating genes and DUBS core gene expression.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/e37c06829bf92e1fb555977b.png"},{"id":85617667,"identity":"74a2ce96-1487-4396-af26-49c5303a3eb6","added_by":"auto","created_at":"2025-06-29 14:47:08","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":294088,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eTargeted drug screening for DUBS core genes. \u003c/strong\u003e(A) Analysis of the correlation between drug sensitivity and gene expression based on the GDSC database.(B) Analysis of the correlation between drug sensitivity and gene expression based on the CTRP database.(C) Differences in sensitivity to common chemotherapeutic drugs between high/low DUBS groups.(D) Molecular docking simulation of EIF3F with common chemotherapeutic drugs.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/4a2d0ab3d32ece508fd00c05.png"},{"id":85616722,"identity":"9e54b331-4e6f-4a2b-8ec4-8c67a961fee9","added_by":"auto","created_at":"2025-06-29 14:39:07","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":341530,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-cell and spatial transcriptomic profiling of DUBS core genes. \u003c/strong\u003e(A) UMAP clustering visualization of cellular subpopulations.(B) Cluster-specific expression patterns of DUBS core genes visualized by bubble plot.(C) Spatial distribution map of DUBS core gene expression.(D) Cell-type annotation mapping integrated with UMAP clustering in HCC specimens.(E) Quantitative assessment of intercellular communication networks within tumor microenvironment.(F) Signal flux intensity analysis of cell-cell crosstalk in tumor microenvironment.(G) Spatial transcriptome mapping of EIF3F expression in responders to targeted-immunotherapy combinatorial regimen.(H) Spatial transcriptome mapping of EIF3F expression in non-responders to targeted-immunotherapy combinatorial regimen.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/abef09b253af41f63169e2d2.png"},{"id":85616670,"identity":"8ec6f889-ec49-4ca0-8a76-fad146db7364","added_by":"auto","created_at":"2025-06-29 14:39:03","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":296480,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ePan-cancer landscape of DUBS core gene signatures. \u003c/strong\u003e(A) Pan-cancer transcriptomic landscape of DUBS core gene expression across 33 TCGA malignancies.(B) Prognostic association mapping via univariate Cox regression across 33 cancer types.(C) Differential immunotherapy responsiveness stratification between DUBS-high and DUBS-low subgroups in multi-cancer cohorts.\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/f9e4ab88896eecf89974831f.png"},{"id":85616735,"identity":"d5d982c5-d2df-4084-89ae-1bd233145fda","added_by":"auto","created_at":"2025-06-29 14:39:07","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":344780,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIn vitro functional characterization of EIF3F. \u003c/strong\u003e(A)\u003cstrong\u003e \u003c/strong\u003eEIF3F expression levels in human normal hepatocyte versus HCC cell lines.(B)Western blot analysis of EIF3F knockdown efficiency mediated by shRNA in Huh7 cells.(C) Western blot confirmation of shEIF3F-induced EIF3F depletion in HepG2 cells.(D-E) Wound healing assays comparing migration capacities of (D) Huh7 and (E) HepG2 cells post-EIF3F silencing.(F-G) Colony formation assays assessing proliferative abilities of (F) Huh7 and (G) HepG2 cells with EIF3F knockdown.(H-I) Transwell migration assays quantifying migratory potential of (H) EIF3F-silenced Huh7 and (I) HepG2 cells.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/c7cdb0c094a0837c216c2589.png"},{"id":85616734,"identity":"8bcbbecc-7410-4bac-92db-ae202778b1b1","added_by":"auto","created_at":"2025-06-29 14:39:07","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":154057,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eProliferation assessment of EIF3F-silenced HCC cells in vitro and in vivo. \u003c/strong\u003e(A-B) CCK-8 assays assessing proliferation rates of (A) Huh7 and (B) HepG2 cells with EIF3F knockdown.(C-E) Xenograft tumor analysis of HepG2 cells with EIF3F knockdown.(C) Representative tumor images, (D) Quantitative tumor volume measurements, (E) Final tumor weight comparison.\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/4f8e530d9c85f3d999f6ce42.png"},{"id":106809336,"identity":"9606b225-70fc-42f2-ba49-dc82e1acb722","added_by":"auto","created_at":"2026-04-13 16:09:38","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3764147,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/060b7668-3f31-43f9-be6c-2f1090e83b02.pdf"},{"id":85617628,"identity":"1bc5b273-7d8c-47a6-818f-f2a66c96d0c4","added_by":"auto","created_at":"2025-06-29 14:47:04","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1679126,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigureLegend.docx","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/1bf01a22c668d3f107cf1854.docx"},{"id":85616689,"identity":"9c97b9a3-a630-41cb-97ad-ae82030bcceb","added_by":"auto","created_at":"2025-06-29 14:39:05","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":288906,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTables.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/1d69264b8209ff895a736f57.xlsx"},{"id":85616678,"identity":"04c0feae-599a-4995-ab1a-60e3a574625b","added_by":"auto","created_at":"2025-06-29 14:39:04","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":3712860,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure S1: Diagnostic performance evaluation and functional enrichment profiling. \u003c/strong\u003e(A) Diagnostic ROC trajectories of Lasso-LDA integration in each validation cohort.(B) GSEA results based on GO gene sets in the DUBS-high group.(C) GSEA results based on GO gene sets in the DUBS-low group.(D) GSEA results based on KEGG gene sets in the DUBS-high group.(E) GSEA results based on KEGG gene sets in the DUBS-low group.\u003c/p\u003e","description":"","filename":"FigureS1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/1f015b399d6cf6fcfcbd207d.tif"},{"id":85616717,"identity":"3797744c-2710-4a53-bf05-941146f00e32","added_by":"auto","created_at":"2025-06-29 14:39:07","extension":"tif","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":10411068,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eFigure S2: Cell-cell communication network profiling and spatial immune contexture analysis. \u003c/strong\u003e(A) Bubble plot visualization of intercellular ligand-receptor interaction networks.(B) Spatially resolved immune cell infiltration quantification in responders versus non-responders to targeted-immunotherapy combination using transcriptome deconvolution.(C-D) Compartment-specific infiltration profiling of immune microenvironmental cells across spatial clusters.\u003c/p\u003e","description":"","filename":"FigureS2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/c1ac0a3ec49b30a60bb639ec.tif"},{"id":85616669,"identity":"79173e56-e06a-44ab-b16d-ba37a373c7b3","added_by":"auto","created_at":"2025-06-29 14:39:03","extension":"tif","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":871012,"visible":true,"origin":"","legend":"","description":"","filename":"EIF3Fexpression1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/de3bfcc2b615c6b2fd509e14.tif"},{"id":85617654,"identity":"4bd25f75-fdd8-4a9b-b69c-902e650825b3","added_by":"auto","created_at":"2025-06-29 14:47:06","extension":"tif","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":2653876,"visible":true,"origin":"","legend":"","description":"","filename":"EIF3Fexpression2and3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/211f79115e46f1807a31b4f9.tif"},{"id":85616671,"identity":"0c7233a2-54c9-4f2d-a611-c2213c9b80ab","added_by":"auto","created_at":"2025-06-29 14:39:04","extension":"tif","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":1737592,"visible":true,"origin":"","legend":"","description":"","filename":"huh7tubulin1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/5b0c6b98bb0ae590c6d1456f.tif"},{"id":85616681,"identity":"89b58650-cb9f-4dbc-a13c-a539bd30bf15","added_by":"auto","created_at":"2025-06-29 14:39:05","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":2331596,"visible":true,"origin":"","legend":"","description":"","filename":"huh7sh2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/1cd5d7dbedda986a20e9c1d6.tif"},{"id":85619170,"identity":"07b61da2-0f08-4428-8d05-5d73823770d5","added_by":"auto","created_at":"2025-06-29 14:55:08","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":1557728,"visible":true,"origin":"","legend":"","description":"","filename":"huh7sh3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/ebdc5e642aa7a14cf807bef3.tif"},{"id":85617656,"identity":"b0e55965-f184-46f4-b545-bb01c7babed9","added_by":"auto","created_at":"2025-06-29 14:47:06","extension":"tif","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":1657724,"visible":true,"origin":"","legend":"","description":"","filename":"huh7sh1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/88f225981b7d9bf049edfa8a.tif"},{"id":85617631,"identity":"0fb9b82f-789d-4f23-88a0-212d0745fd94","added_by":"auto","created_at":"2025-06-29 14:47:05","extension":"tif","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":2110904,"visible":true,"origin":"","legend":"","description":"","filename":"hepG2sh2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/6c1b0feb373f8a55edc58f33.tif"},{"id":85617634,"identity":"5f6d126d-7936-4214-b1f3-59c973ec5dfa","added_by":"auto","created_at":"2025-06-29 14:47:05","extension":"tif","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":2226600,"visible":true,"origin":"","legend":"","description":"","filename":"hepG2sh1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/dd793ff101db90056ab0bf6b.tif"},{"id":85617664,"identity":"07d661cb-2e6c-4403-910b-1572f95a4045","added_by":"auto","created_at":"2025-06-29 14:47:07","extension":"tif","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":1841076,"visible":true,"origin":"","legend":"","description":"","filename":"hepG2sh3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/75add075a726e924beb2736d.tif"},{"id":85616682,"identity":"965cd590-33fb-4d85-9e20-ef342846c7e2","added_by":"auto","created_at":"2025-06-29 14:39:05","extension":"tif","order_by":14,"title":"","display":"","copyAsset":false,"role":"supplement","size":3251352,"visible":true,"origin":"","legend":"","description":"","filename":"hepg2tubulin1.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/d65a243fbee067dcf7f74c1d.tif"},{"id":85616692,"identity":"ebd44eb6-9ba6-4e3e-81dc-816e334c6be6","added_by":"auto","created_at":"2025-06-29 14:39:05","extension":"tif","order_by":15,"title":"","display":"","copyAsset":false,"role":"supplement","size":1392388,"visible":true,"origin":"","legend":"","description":"","filename":"hepg2tubulin2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/c0eea69566b0a0d564f6840f.tif"},{"id":85616765,"identity":"0424547e-efe7-49ad-9c7b-43d425ce5a82","added_by":"auto","created_at":"2025-06-29 14:39:09","extension":"tif","order_by":16,"title":"","display":"","copyAsset":false,"role":"supplement","size":6426716,"visible":true,"origin":"","legend":"","description":"","filename":"huh7tubulin2.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/f5364815b0ce40bda7a97112.tif"},{"id":85616742,"identity":"8acff303-e98b-4192-98cd-df452f6bd21a","added_by":"auto","created_at":"2025-06-29 14:39:08","extension":"tif","order_by":17,"title":"","display":"","copyAsset":false,"role":"supplement","size":5790584,"visible":true,"origin":"","legend":"","description":"","filename":"huh7tubulin3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/7e308eea291a44e7ac02ea4d.tif"},{"id":85616728,"identity":"9e2c54c8-5975-455b-8e13-3939860da7a1","added_by":"auto","created_at":"2025-06-29 14:39:07","extension":"tif","order_by":18,"title":"","display":"","copyAsset":false,"role":"supplement","size":5790000,"visible":true,"origin":"","legend":"","description":"","filename":"hepg2tubulin3.tif","url":"https://assets-eu.researchsquare.com/files/rs-6909459/v1/3cd5ea4fc7a6f5644ecf054f.tif"}],"financialInterests":"No competing interests reported.","formattedTitle":"Integrating Machine Learning and Spatial Transcriptomics Uncovers Shared Immunomodulatory Deubiquitinases in MAFLD and HCC","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGlobally, liver cancer ranks as the seventh most common malignancy and the third leading cause of cancer-related mortality, accounting for approximately 5% of all new malignant tumor cases, and has become a major public health issue\u003csup\u003e[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]\u003c/sup\u003e. HCC is the predominant histological subtype of primary liver cancer. Its clinical characteristics include an insidious onset and a lack of specific clinical manifestations in the early stages, posing significant challenges for disease diagnosis and clinical intervention. For patients with advanced HCC, systemic treatment regimens represented by the combination of Atezolizumab and Bevacizumab have been recommended as first-line therapy by the US Food and Drug Administration (FDA) and guidelines from internationally authoritative academic institutions\u003csup\u003e[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]\u003c/sup\u003e. However, existing treatment options still exhibit significant limitations, and some patients fail to achieve a desirable clinical response\u003csup\u003e[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]\u003c/sup\u003e. This indicates that the optimization of early diagnostic systems for HCC and the development of individualized treatment strategies require further in-depth exploration.\u003c/p\u003e \u003cp\u003eA substantial proportion of HCC occurs in the context of multiple coexisting chronic liver diseases. Among these, MAFLD has emerged as a primary driver of the continuously rising global HCC incidence. A meta-analysis incorporating 22 studies revealed that MAFLD was associated with 48.7% of HCC cases \u003csup\u003e[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]\u003c/sup\u003e. However, when MAFLD was present as an independent liver disease, it accounted for only 12.4% of the etiological distribution in HCC. This epidemiological evidence indicates that MAFLD acts primarily as a co-carcinogenic factor rather than a significant independent carcinogen in HCC development. Therefore, building upon existing research on MAFLD-related HCC, systematically elucidating the pathogenic synergy between MAFLD and HCC caused by various chronic liver diseases holds significant translational medical value.\u003c/p\u003e \u003cp\u003eUbiquitination, as a crucial mechanism of post-translational protein modification, regulates cellular functions by modulating protein localization, degradation, and interactions\u003csup\u003e[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/sup\u003e. DUBs confer reversibility to this modification process through the specific hydrolysis of ubiquitin-substrate linkages\u003csup\u003e[\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]\u003c/sup\u003e. Recent studies have demonstrated that DUBs play key regulatory roles in both MAFLD and HCC: for instance, RPN11 regulates the expression of lipid metabolism-related genes via the METTL3-ACSS3-histone propionylation modification axis; USP family members promote HCC cell proliferation, migration, and stemness through their deubiquitinating activity\u003csup\u003e[\u003cspan additionalcitationids=\"CR8 CR9\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]\u003c/sup\u003e. However, no dual-function DUBs molecule capable of simultaneously regulating both MAFLD progression and HCC development has been identified to date. Identifying such core DUBs regulatory nodes in the MAFLD-HCC comorbidity process will provide a novel theoretical foundation for establishing prognostic biomarker systems and developing targeted therapeutic strategies.\u003c/p\u003e \u003cp\u003eThis study, based on multi-omics integrated analysis technology, systematically identifies the key DUBs regulatory network in the MAFLD-HCC comorbidity process, develops a DUBS based on 4 core DUBs, and constructs an HCC prognostic stratification system and an immunotherapy efficacy prediction system. Furthermore, pan-cancer analysis is employed to decipher the broad-spectrum biological characteristics of DUBS in tumor biology, and in vitro and in vivo functional validation experiments are combined to elucidate the oncogenic molecular mechanisms of the key DUBs. This research reveals, for the first time, the molecular network mediated by DUBs in MAFLD-HCC, providing novel molecular markers and potential therapeutic targets for early screening, optimization of precision treatment strategies, and dynamic prognostic monitoring in MAFLD-HCC patients.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e\u003cstrong\u003eData Acquisition and Processing Pipeline\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCancer-associated DUBs were obtained from Dewson et al.'s review\u003csup\u003e[11]\u003c/sup\u003e. Gene expression matrices and relevant clinical information for MAFLD patients and HCC patients were downloaded from the Gene Expression Omnibus (GEO,\u0026nbsp;https://www.ncbi.nlm.nih.gov/gds/), The Cancer Genome Atlas (TCGA,\u0026nbsp;https://portal.gdc.cancer.gov/), and International Cancer Genome Consortium (ICGC,\u0026nbsp;https://dcc.icgc.org/). We additionally collected data from two distinct patient cohorts receiving anti-PD-1/PD-L1 immune checkpoint inhibitor therapy: (1) Riaz et al. (melanoma patients undergoing nivolumab treatment); (2) Liu et al. (melanoma patients treated with nivolumab or pembrolizumab). Subsequently, we acquired gene expression data and clinical data for TCGA pan-cancer and adjacent normal tissues through the UCSC Xena platform, where RNA sequencing data had been normalized to Fragments Per Kilobase of transcript per Million mapped (FPKM) reads format for subsequent analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eScreening for\u003c/strong\u003e \u003cstrong\u003eShared DUBs in MAFLD and HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe performed differential expression analysis of gene expression data using the \"limma\" R package, with a p-value\u0026nbsp;＜\u0026nbsp;0.05 considered statistically significant. The expression patterns of genes identified by differential analysis were visualized using the \"pheatmap\" R package. Subsequently, we integrated and batch-corrected the gene expression datasets using the \"sva\" R package. To screen for MAFLD diagnostic biomarkers, we employed 12 machine learning algorithms: Lasso, Ridge, Stepglm, XGBoost, Random Forest (RF), Elastic Net (Enet), Partial Least Squares Regression for Generalized Linear Models (plsRglm), Generalized Boosted Regression Models (GBM), Naive Bayes, Linear Discriminant Analysis (LDA), Generalized Linear Model Boosting (glmBoost), and Support Vector Machine (SVM). Utilizing the combined data from the GSE135251 and GSE130970 cohorts as the training set, we explored 113 algorithm combinations derived from these 12 algorithms, aiming for variable selection and model development within a 10-fold cross-validation framework. Immediately following this, the GSE48452, GSE63067, and GSE89632 cohorts were used as external validation sets to assess model performance. Based on prior research, we defined the optimal model as the one demonstrating a higher overall mean Area Under the Curve (AUC) and achieving AUC values greater than 0.7 across all training and validation sets. Finally, the genes selected by the optimal model were identified as MAFLD-HCC comorbidity-specific DUBs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDUBS Construction and Survival Stratification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe integrated HCC datasets from three independent cohorts (TCGA-LIHC, LIRI-JP, and GSE116174). Data harmonization was performed using the \"sva\" R package with ComBat algorithm for batch effect correction. Prognosis-associated genes were initially identified through univariate Cox proportional hazards regression analysis for subsequent model construction. The multivariate Cox regression analysis was subsequently employed to establish the DUBS using the following formula:\u003c/p\u003e\n\u003cp\u003eDUBS = Σ [Expression (Gene_i) × β_i]\u003c/p\u003e\n\u003cp\u003ewhere β_i represents the risk coefficient derived from multivariate regression analysis. Cohort stratification was performed by dichotomizing patients into high- and low-DUBS subgroups using the median score as the cutoff value. Model interpretability was enhanced through SHapley Additive exPlanations (SHAP) analysis implemented via \"kernelshap\" and \"shapviz\" packages, with visualization generated using \"ggplot2\". Survival analyses were conducted with the \"survival\" R package, including proportional hazards assumption verification through Schoenfeld residuals and Kaplan-Meier survival curve estimation with log-rank tests for prognostic validation. Diagnostic performance was evaluated using time-dependent ROC curves (“pROC” package), with prognostic nomograms constructed via\u0026nbsp;“rms” package.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunotherapy Efficacy Prediction via Multi-dimensional Tumor Microenvironment Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo comprehensively evaluate tumor immune microenvironment characteristics and predict immunotherapy response, we implemented a multi-dimensional computational framework. First, utilizing the Tumor Immune Dysfunction and Exclusion (TIDE) algorithmic platform (http://tide.dfci.harvard.edu/login/), we computed TIDE scores and MSI expression signature (MSI Expr. Sig.) scores for pan-cancer TCGA cohorts. Elevated TIDE scores were interpreted as indicative of enhanced immune evasion potential, serving as predictive biomarkers for immunotherapy resistance. Subsequently, we conducted survival analyses stratified by the TCGA pan-cancer immune classification system established by Thorsson et al. to delineate prognostic disparities across immune subtypes. To further validate the predictive utility of immune checkpoint inhibitors, we integrated Immunophenoscore (IPS) data from The Cancer Immunome Atlas (TCIA,\u0026nbsp;https://tcia.at/home). Comparative analysis of IPS values between anti-CTLA4 and anti-PD1 treatment cohorts (with higher IPS correlating with superior therapeutic efficacy) was performed to assess differential treatment responses. For quantitative characterization of the tumor microenvironment, ESTIMATE scores (including immune/stromal/ESTIMATE composite scores) were computed using the \"ESTIMATE\" package, systematically deconvoluting immune and stromal compartmentalization within tumor tissues. Finally, the Gene Set Variation Analysis (GSVA) algorithm was employed to interrogate the infiltration landscape of immune cell subsets and functional pathway activation states within the tumor immune niche.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComputational Pharmacogenomic Profiling and Molecular Docking Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Genomics for Drug Sensitivity in Cancer (GDSC,\u0026nbsp;https://www.cancerrxgene.org/) and Cancer Therapeutics Response Portal (CTRP,\u0026nbsp;http://portals.broadinstitute.org/ctrp/) databases were utilized to access tumor genomic profiles and corresponding drug sensitivity data. Correlation analyses between target gene expression and FDA-approved drug responses were conducted through the Gene Set Cancer Analysis (GSCA) platform (https://guolab.wchscu.cn/). Drug sensitivity disparities in high/low DUBS subgroups were evaluated using the 'oncopredict' R package with established chemotherapy agents. For molecular docking studies focusing on HCC therapeutics, the 3D structures of clinically used chemotherapeutic agents were retrieved from PubChem (https://pubchem.ncbi.nlm.nih.gov/). Structural data for EIF3F (PDB ID: 6YBD) were acquired from the Protein Data Bank (https://www.rcsb.org/). Molecular docking simulations were performed via the CB-DOCK2 web server (http://clab.labshare.cn/cb-dock2/), a cloud-based platform implementing Autodock Vina algorithms. The docking protocol prioritized binding poses based on Vina scores, with configurations exhibiting the highest absolute binding energy values selected as optimal interaction models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSingle-cell RNA Sequencing Analysis and Spatial Transcriptomics Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe HCC single-cell RNA-seq dataset (GSE146115) was retrieved from the GEO repository. Raw count matrices underwent rigorous quality control using\u0026nbsp;“Seurat” package. Post-filtering data were normalized via SCTransform with regularized negative binomial regression to correct technical variability. Principal component analysis (PCA) was performed on 3,000 highly variable genes identified by variance-stabilizing transformation, with significant PCs (Elbow method) selected for nonlinear dimensionality reduction using UMAP. Cluster-defining marker genes were identified through Wilcoxon rank-sum test with Bonferroni correction (adj.p value＜0.05, logFC.threshold＞1). Cell type annotation utilized\u0026nbsp;“SingleR” package with HumanPrimaryCellAtlas, BlueprintEncode, MonacoImmuneData, NovershternHematopoieticData, DatabaseImmuneCellExpressionData, ref_Hematopoietic and ref_Human_all reference datasets, followed by\u0026nbsp;“CellChat” package employing a weighted ligand-receptor interaction database and information theory-based communication probability quantification.\u003cbr\u003eSpatially resolved HCC transcriptomes (GSE238264) were curated from GSM7661258/GSM7661260 slides exhibiting high target gene expression entropy. 10X Visium data were aligned via Space Ranger (https://support.10xgenomics.com/spatial-gene-expression/software/downloads/latest) using STAR spliced read mapper (GRCh38.p13). Tissue segmentation and spot-by-gene matrices were imported into \u003cem\u003eSeurat\u003c/em\u003e for integration with scRNA-seq data through reciprocal PCA-based anchor identification. Following variance stabilization, PCA-reduced dimensions underwent UMAP embedding. Spatial cluster topology was visualized through \u003cem\u003eSeurat\u003c/em\u003e's multimodal plotting functions. Multimodal intersection analysis (MIA) implemented hypergeometric testing to quantify subpopulation colocalization. Cross-platform annotation leveraged probabilistic label transfer via\u0026nbsp;Seurat's FindTransferAnchors method, establishing a joint embedding space for scRNA-seq derived cell states and spatial transcriptomic spots.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePan-cancer Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe DUBS core genes were analyzed for pan-cancer expression differences using TCGA data (TOIL-normalized FPKM) acquired from the UCSC Xena platform. Differential expression between tumor and adjacent normal tissues was evaluated through Wilcoxon rank-sum tests with Benjamini-Hochberg false discovery rate (FDR) correction, visualized via log2(FPKM+1)-transformed boxplots. Clinical survival data from TCGA were subsequently integrated to perform Cox proportional hazards regression analysis using the \"survival\" package, with gene expression-survival correlations visualized through hazard ratio heatmaps generated by \"ggplot2\". For immunotherapy response evaluation, TIDE scores were computed from TCGA pan-cancer profiles. Comparative analysis of immune evasion potential and predicted immune checkpoint blockade response probabilities between high and low-DUBS subgroups was conducted based on TIDE-derived metrics.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell Culture\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHuman hepatic stellate cell line LX-2, human HCC cell line HepG2, Huh7, and C3A cells (Shanghai Cell Bank of the Chinese Academy of Sciences) were cultured in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS), with the culture medium replaced every 48 h. All 4 cell lines were maintained in a constant temperature incubator at 37°C with 5% CO₂. Based on prior growth curve analysis, serum starvation synchronization was performed on cells before drug treatment.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStable shRNA Knockdown in HCC Cell Lines\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eHepG2 and Huh7 cells were plated in 6-well plates until reaching 60-70% confluence. Lentiviral particles carrying EIF3F-specific shRNA\u0026nbsp;were complexed with Lipofectamine transfection reagent in serum-free DMEM high-glucose medium. The shRNA-lipid complexes were delivered through spinoculation followed by standard incubation. The medium was changed after 24 h to reduce cytotoxicity and finally the cells were screened with puromycin for 2 weeks to achieve a stable cell line and passaged for protein extraction shEIF3F sequence was:\u003c/p\u003e\n\u003cp\u003eshEIF3F#1:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eForward:5’-CCGGCTCTCAAGTGACTTGCAGCAACTCGAGTTGCTGCAAGTCACTTGAGAGTTTTTG-3’\u003c/p\u003e\n\u003cp\u003eReverse:5’-AATTCAAAAACTCTCAAGTGACTTGCAGCAACTCGAGTTGCTGCAAGTCACTTGAGAG-3’\u003c/p\u003e\n\u003cp\u003eshEIF3F#2:\u003c/p\u003e\n\u003cp\u003eForward:5’-CCGGGTACTACGACACTGAACGCATCTCGAGATGCGTTCAGTGTCGTAGTACTTTTTG-3’\u003c/p\u003e\n\u003cp\u003eReverse:5’-AATTCAAAAAGTACTACGACACTGAACGCATCTCGAGATGCGTTCAGTGTCGTAGTAC-3’\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtein Extraction and Western Blot Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCellular protein isolation and immunoblotting were conducted with protocol modifications for adherent cell lines. Cells were rinsed three times with \u0026nbsp; and lysed in pre-chilled RIPA buffer supplemented with a protease inhibitor cocktail (1:100) for 20 min on ice. Lysates were clarified by centrifugation (12,000 × g, 10 min, 4°C), and protein quantification was performed using a BCA assay. Equal protein aliquots were denatured (95°C, 5 min) in Laemmli buffer and electrophoresed on 10% SDS-polyacrylamide gels. Separated proteins were electrotransferred to PVDF membranes, blocked with 5% skim milk for 1 h, and immunoprobed with specific primary antibodies (1:1000 dilution, overnight at 4°C). Following TBST washes, membranes were incubated with HRP-linked secondary antibodies (1:5000 dilution, 1 h, RT) and developed via ECL substrate. All experiments included 3 independent biological replicates to ensure reproducibility.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWound Healing and Transwell Assay\u003cbr\u003e\u003c/strong\u003eThe wound healing assays were performed using Huh7 and HepG2 cell lines. Briefly, parallel reference lines were pre-marked on the bottom of 6-well plates. Logarithmic-phase cells were trypsinized, resuspended, and seeded until monolayer confluence was achieved. A sterile ruler was used to plan scratch trajectories perpendicular to the reference lines. Standardized scratches were generated along predefined paths using a 10 μl pipette tip, followed by three washes with PBS to remove detached cells. Serum-free medium was applied to minimize proliferation interference. Scratch images were captured at 0 and 48 h post-scratching using an inverted microscope (100× magnification). Wound closure areas were quantified using ImageJ software, with migration rates calculated as: [(Initial area – Final area)/Initial area] × 100%. Experiments were independently repeated 3 times. Intergroup differences were analyzed by independent Student's t-test.\u003c/p\u003e\n\u003cp\u003eFor the Transwell assay, HCC Huh7 and HepG2 cells from the NC and sh groups were serum-starved in serum-free medium for 4 h to eliminate serum interference, followed by trypsinization and resuspension in serum-free medium. Cell suspensions were seeded into the upper chambers, while the lower chambers of the 24-well plate were filled with complete medium containing 10% FBS. After 48h incubation at 37°C, cells were washed three times with PBS, fixed with 4% paraformaldehyde (30 min), and stained with crystal violet (30 min). Non-invading cells in the upper chambers were mechanically removed using cotton swabs. Microscopic images were captured at 200× magnification, and cell quantification was performed using Image J software, with triplicate independent experiments conducted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCCK-8 and Colony Formation Assays\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate cell proliferation activity, this study employed dual-dimensional analysis through CCK-8 assay and colony formation assay using HCC Huh7 and HepG2 cell lines. For the CCK-8 assay, both cell lines were digested with trypsin-EDTA and resuspended, followed by 3 independent counts using a hemocytometer to obtain mean values. sh-TIMD4-transfected and control Huh7/HepG2 cells were seeded into 96-well plates at a density of 2×10³ cells/well. The experiment included 5 time points (0 h, 24 h, 48 h, 72 h, and\u0026nbsp;96 h), with 6 replicate wells per time point. At each designated interval, 10 μL of CCK-8 solution was precisely pipetted into each well after gentle mixing. The plate was wrapped in aluminum foil for 4 h dark incubation at 37°C, and absorbance at 450 nm was measured using a multimode microplate reader.\u003c/p\u003e\n\u003cp\u003eFor the colony formation assay, sh-TIMD4-treated and control Huh7/HepG2 cells were plated in 6-well plates at 1×10³ cells/well and cultured in high-glucose DMEM medium (Gibco, Cat. No. 11965092) supplemented with 10% fetal bovine serum under 5% CO₂ at 37°C for 7 days. Cells were fixed with 4% paraformaldehyde for 30 min, stained with 0.1% crystal violet for 30 min, and washed 3 times with PBS before air-drying. Colony morphology was documented using a high-resolution imaging system, and colonies containing ≥50 cells were quantified using ImageJ software. All experiments were independently repeated in triplicate.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSubcutaneous Xenograft Tumor Model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study established HepG2 HCC xenografts in male BALB/c nude mice (6-week-old, n=10) obtained from China Medical University's accredited animal center with IACUC approval. Viable HepG2 cells were suspended in sterile PBS (5×10⁶ cells/mL) and inoculated subcutaneously (200 μL) into the right axilla. Mice were divided into NC (n=5) and shRNA groups (n=5). Tumor dimensions were measured every 4 days from day 4 post-inoculation using digital calipers, with volumes calculated as 0.52×L×W². The study terminated at day 28 or upon reaching ethical thresholds (\u0026gt;1500 mm³ tumor volume or \u0026gt;20% weight loss). Euthanasia was performed via sodium pentobarbital (50 mg/kg) anesthesia and cervical dislocation, followed by tumor excision/weighing. All procedures were conducted in SPF conditions (22±1°C, 50±5% humidity, 12-h light cycle). Data were analyzed using two-tailed t-tests in GraphPad Prism 9.0 (p value \u0026lt; 0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis and Visualization\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were conducted using methodologically validated approaches. Intergroup comparisons employed Student's t-test and Wilcoxon rank-sum test, while survival rate disparities were evaluated through Kaplan-Meier curve analysis. Correlation coefficients were determined via Spearman's rank-order test, with Cox proportional hazards regression modeling providing hazard ratios (HR) and 95% confidence intervals (CI). Significance thresholds were established as *P value \u0026lt; 0.05, **P value \u0026lt; 0.01, and ***P value \u0026lt; 0.001. Analytical workflows were implemented in R (version 4.1.0) and GraphPad Prism 8.0, with graphical representations generated using ImageJ and Adobe Illustrator 2019.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eShared Gene Identification in MAFLD and HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUtilizing cross-cohort differential gene integration analysis, we systematically analyzed the regulatory characteristics of DUB genes in MAFLD. Based on cross-cohort differential expression analysis of the TCGA-LIHC and GSE135251 cohorts, 31 shared differentially expressed genes (DEGs) were identified through Venn diagram intersection (Fig. 1A). Heatmap visualization was employed to analyze the expression patterns of these 31 shared DEGs. The results demonstrated significant heterogeneity in their expression across MAFLD samples (Fig. 1B), with the vast majority of genes exhibiting a significant upregulation trend in cancerous tissues (Fig. 1C). To further identify core regulatory factors, this study employed twelve distinct machine learning algorithms to construct feature selection models. By comparing model performance, the Lasso + LDA combined algorithm was ultimately determined to be the optimal model, achieving AUC values greater than 0.7 in both the training and validation sets, and exhibiting the highest overall average AUC (Fig. 1D and S1A). Consequently, 15 DUBs with crucial regulatory potential were selected. This screening outcome reveals the potentially important biological functions these DUBs may play in the pathological progression of both TCGA-LIHC and MAFLD.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEstablishment of DUBS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study systematically screened the association of 15 DUBs with HCC prognosis. Univariate Cox regression analysis successfully identified 10 DUB molecules with significant prognostic value (Fig. 2A). Based on the multivariate Cox proportional hazards model, the DUBS system comprising USP39, USP32, OTUB2, and EIF3F was further constructed. The calculation formula is as follows (Fig. 2B):\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDUBS = Exp (USP39) * 0.069 + Exp (USP32) * 0.075 + Exp (OTUB2) * 0.238 + Exp (EIF3F) * 0.014\u003c/p\u003e\n\u003cp\u003eThe SHAP feature interpretation algorithm confirmed a positive correlation between the expression levels of the core genes and prognostic prediction efficacy, revealing the model's interpretability features (Fig. 2C). During the clinical validation phase, survival analysis based on grouping by the median DUBS value demonstrated significantly prolonged overall survival in the low-DUBS group, validating the prognostic assessment efficacy of this scoring system across multiple dimensions (Fig. 2D-I). Notably, pathway enrichment analysis revealed significant biological characteristic differences between the two groups: the high-DUBS group exhibited overactivation of proliferation-related pathways such as cell cycle regulation and DNA replication, which may explain its poor clinical prognosis phenotype. Conversely, the low-DUBS group showed sustained activity in metabolic pathways like lipid metabolism and oxidative phosphorylation, indicating that relatively preserved liver physiological function may constitute a protective prognostic mechanism. These findings not only confirm the clinical application potential of DUBS as a novel prognostic biomarker but also elucidate the underlying mechanisms of heterogeneous HCC progression from the perspective of molecular network regulation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of DUBS Prognostic Model's Diagnostic Value Across Diseases and Clinical Translation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThrough systematic investigation of HCC prognostic markers, we have, for the first time, confirmed via univariate Cox regression models that DUBS serves as an independent prognostic risk factor for HCC patients (Fig. 3A), and its predictive efficacy remained robust in multivariate regression analysis (Fig. 3B). In-depth analysis revealed that high-DUBS expression was significantly positively correlated with tumor malignancy (Fig. 3C), providing crucial evidence for elucidating DUBS's pro-oncogenic mechanism. ROC curve analysis optimized by machine learning further expanded the clinical application dimension of DUBS-associated genes. In HCC differential diagnosis: USP39 (AUC=0.965), EIF3F (AUC=0.930), and OTUB2 (AUC=0.887) demonstrated excellent diagnostic performance. And in early MAFLD identification: OTUB2 (AUC=0.870), USP32 (AUC=0.761), and EIF3F (AUC=0.731) also exhibited outstanding performance. Notably, in diagnostic biomarker screening, EIF3F displayed unique cross-disease diagnostic value: this gene not only demonstrated excellent efficacy in HCC differential diagnosis (AUC=0.930) but also emerged as the sole molecular biomarker possessing diagnostic potential in both HCC and MAFLD. Its dual diagnostic characteristics suggest its core regulatory role in liver diseases (Fig. 3D). To advance precision medicine practice, this study innovatively constructed a multi-parameter integrated nomogram model (Fig. 3E). This model organically integrates molecular biomarkers with clinical characteristics through a standardized scoring system, providing a quantitative tool for individualized prognosis management that combines scientific rigor with clinical utility.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnalysis of Hot Tumor Phenotype Reveals Mechanisms Underlying Immunotherapy Sensitization in Low-DUBS Group Patients\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study systematically revealed the predictive association between DUBS expression levels and the efficacy of immune checkpoint inhibitors by integrating a multi-dimensional biomarker evaluation system. Survival analysis based on two independent anti-PD-1/PD-L1 treatment cohorts demonstrated significantly prolonged overall survival in patients with low-DUBS expression (Fig. 4A). Furthermore, the DUBS score was significantly lower in the complete response/partial response (CR/PR) subgroup compared to the stable disease/progressive disease (SD/PD) group. This survival advantage was further validated in progression-free survival analysis (Fig. 4B). TIDE score analysis indicated that the low-DUBS group had a superior immune evasion prediction value (Fig. 4C), while MSI comparison revealed significantly elevated MSI levels in the low-DUBS group (Fig. 4D), suggesting potential mechanisms for immunotherapy sensitivity. Further analysis of TCGA pan-cancer data based on Thorsson's immune classification revealed that patients with the C3 subtype, characterized by Th17/Th1 signature dominance and lower tumor proliferative activity, had significantly better prognosis than those with the C1 subtype dominated by angiogenic signatures (Fig. 4E), highlighting the independent impact of immune classification features on treatment outcomes. IPS analysis confirmed that patients with low DUBS expression exhibited stronger treatment sensitivity, regardless of targeting PD-1 or CTLA4 (Fig. 4F)\u003csup\u003e[12]\u003c/sup\u003e. Tumor microenvironment assessment showed significantly elevated Stromal Score, Immune Score, and combined assessment value in the low-DUBS group (Fig. 4G), elucidating the positive regulatory role of the \"hot tumor\" phenotype characteristics on immunotherapy from the perspective of microenvironment heterogeneity. These findings, supported by multiple independent lines of evidence including survival benefit, molecular characteristics, predictive models, and microenvironment remodeling, collectively endorse the clinical application potential of DUBS as a predictive biomarker for immunotherapy response.\u003c/p\u003e\n\u003cp\u003eSubsequently, through systematic analysis of the immune microenvironment characteristics in the low-DUBS expression group, we revealed the multi-layered mechanisms underlying the formation of its \"hot tumor\" phenotype. Immune cell infiltration analysis (Fig. 5A) demonstrated significantly increased infiltration abundance of key anti-tumor immune components such as B lymphocytes, CD8+ T cells, cytotoxic effector cells, and dendritic cells in this group, providing the cellular basis for improved prognosis. Feature enrichment analysis (Fig. 5B) further confirmed significant enrichment in core immune signature pathways including B cell activation, immune checkpoint expression, cytolytic activity, and NK cell function in these patients, highlighting the dynamic activated state of their immune microenvironment. At the mechanistic level, the core DUBS genes EIF3F and USP39 not only showed significant positive correlations with the expression of immune checkpoint molecules such as PDCD1, CTLA4, and CD274 (Fig. 5C), suggesting their potential role in stabilizing checkpoint protein expression via deubiquitination, but also exhibited a co-expression pattern with HLA family antigen-presenting genes (Fig. 5D), revealing their dual regulatory mechanism in activating immune responses by enhancing tumor antigen presentation capability. This unique microenvironment signature—characterized by the synergistic action of highly infiltrating immune cells, the enrichment effect of activated pathways, and the molecular cascade of checkpoint-antigen presentation—collectively constitutes the biological basis for immunotherapy sensitivity in the low-DUBS group, elucidating the intrinsic link between the \"hot tumor\" phenotype and enhanced treatment response from multiple dimensions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical Application Value of DUBS Core Genes in Drug Sensitivity Prediction and Combination Therapy for HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on the comprehensive findings regarding the pivotal role of the DUBS core gene system in disease diagnosis, prognostic assessment of HCC, and immune microenvironment regulation, we postulate its significant research value in predicting drug efficacy. This study reveals multi-dimensional associations between DUBS core gene expression and chemotherapeutic drug sensitivity: association analysis using the GDSC database demonstrated that drug responses to traditional chemotherapeutic agents including methotrexate, vorinostat, and tubastatin A positively correlated with OTUB2 expression but negatively correlated with EIF3F expression (Fig. 6A); CTRP database data further confirmed significant positive correlations between sensitivity to axitinib, doxorubincin, etoposide, mitomycin, and vincristine and OTUB2 expression levels (Fig. 6B). Further analysis within the DUBS grouping model revealed superior therapeutic efficacy for 5-fluorouracil, afatinib, gefitinib, erlotinib, dasatinib, and cyclophosphamide in the low-DUBS group, whereas cisplatin, gemcitabine, axitinib, and irinotecan exhibited enhanced drug responses in the high-DUBS group (Fig. 6C). To validate the reliability of these theoretical predictions, molecular docking simulations were performed via the CB-DOCK2 online platform to model interactions between clinically used drugs (donafenib, doxorubincin, flurouracil, gemcitabine, lenvatinib, and rivoceranib) and the EIF3F protein, with optimal docking models determined by selecting molecular conformations exhibiting the lowest binding free energy followed by visualization (Fig. 6D). Collectively, these findings suggest that a therapeutic strategy combining immunotherapy with small-molecule chemotherapeutic agents may provide critical research directions for enhancing clinical efficacy and improving survival outcomes in HCC patients.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIntegrating Single-Cell and Spatial Transcriptomics to Decipher the Microenvironmental Regulation and Therapeutic Response Mechanisms of EIF3F in HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study systematically deciphered the expression heterogeneity of DUBS core genes within tumor parenchyma and the immune microenvironment of HCC through integrated analysis of single-cell transcriptomic datasets, constructing an HCC single-cell transcriptomic atlas via standardized quality control and UMAP dimensionality reduction that comprised 13 characteristic subpopulations categorized into 5 major cell types (Fig. 7A). Cross-subpopulation profiling revealed significantly elevated EIF3F expression across most cellular lineages, with this pan-cellular distribution pattern dually confirmed through feature gene visualization (Fig. 7B) and spatial expression validation (Fig. 7C-D), collectively indicating its critical regulatory potential in HCC pathogenesis when integrated with prior functional enrichment analyses. To elucidate tumor-microenvironment crosstalk, we established a cell-cell communication network identifying dominant interactions between hepatic parenchymal and immune cells, wherein NK cells exhibited the most active signaling properties (Fig. 7E-F) and the CD74-CXCR4 molecular pair was confirmed as the central signaling hub (Fig. S2A). Spatial transcriptomic in situ analysis further demonstrated significant associations between EIF3F expression abundance and clinical treatment response, revealing lower expression in the combination targeted-immunotherapy sensitive group (ICI nivolumab + TKI cabozantinib) versus marked enrichment in resistant cases, suggesting its role in modulating therapeutic sensitivity and outcomes (Fig. 7G-H), while spatial heterogeneity analysis of microenvironmental infiltration characteristics revealed significant differences in immune cell spatial distribution patterns between response groups (Fig. S2B-D). Through multidimensional omics integration, this study provides the first elucidation of EIF3F's expression dominance at single-cell resolution and its function as an interaction network hub in HCC, revealing spatial heterogeneity-driven regulatory principles governing combination therapy response and uncovering CD74-CXCR4 signaling axis-mediated microenvironmental remodeling mechanisms, thereby offering novel molecular targets and theoretical foundations for precision therapeutic strategies.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePan-Cancer Expression-Prognosis Associations and Immunotherapy Response Prediction of DUBS Core Genes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study systematically evaluated the expression characteristics and clinical predictive value of DUBS core genes across pan-cancer datasets, revealing that OTUB2, USP32, and USP39 were upregulated in most cancer types while EIF3F exhibited specific overexpression in CHOL, COAD, GBM, KIRC, KIRP, LIHC, PRAD, READ, and THCA (Fig. 8A). Survival analysis demonstrated that high EIF3F expression significantly correlated with reduced overall survival in ACC, LIHC, and PCPG; OTUB2 functioned as a risk factor in CESC, LIHC, MESO, PCPG, and SARC; USP32 indicated poor prognosis in KICH, KIRP, LIHC, PCPG, and UCEC; whereas USP39 showed significant risk associations across 11 cancer types including ACC, KICH, KIRC, KIRP, LAML, LGG, LIHC, MESO, PAAD, PCPG, and UCEC (Fig. 8B). Notably, all DUBS core genes correlated with adverse clinical outcomes in PCPG, highlighting the critical role of this regulatory network in neuroendocrine tumors. For immunotherapy response prediction, DUBS expression patterns exhibited significant cancer type-specific variations: patients with BLCA, BRCA, HNSC, KIRC, PRAD, and THCA in the high-DUBS group demonstrated greater immunotherapy sensitivity, whereas GBM, LGG, LUAD, and SARC patients in the low-DUBS group showed superior treatment responses (Fig. 8C). This study confirms DUBS genes as reliable biomarkers for both prognostic assessment and immunotherapy efficacy prediction in LIHC, while their pan-cancer regulatory properties provide a novel theoretical foundation for precision oncology across malignancies, though future multicenter cohort validation and mechanistic investigations remain warranted.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eValidation of EIF3F Expression\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBased on the aforementioned screening results, analysis revealed that core DUBs exhibit significant clinical value in HCC diagnosis and prognostic assessment. Notably, EIF3F not only demonstrated pivotal regulatory functions in HCC diagnostic biomarker screening and survival prediction models, but its expression profile also showed significant correlations with MAFLD progression. To validate these findings, this study analyzed the expression profile of EIF3F protein in HCC cell lines using Western Blot, with experimental data revealing significantly elevated expression in Huh7 and HepG2 cell lines compared to hepatic stellate cell, whereas relatively lower expression levels were observed in the C3A cell line (Fig. 9A). Given the high expression characteristics of EIF3F in Huh7 and HepG2 cell lines, these two cellular models were selected for subsequent functional investigations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEffects and Mechanisms of EIF3F Gene Knockdown on Malignant Phenotypes in HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo elucidate the biological functions of EIF3F in regulating malignant phenotypes of HCC, this study achieved specific EIF3F knockdown in HCC cell lines through construction of shRNA recombinant vectors (shEIF3F#1 and shEIF3F#2), with Western Blot confirming efficient protein expression suppression (Fig. 9B and C). Functional validation demonstrated that the gene knockdown group exhibited significantly reduced cell migration rates compared to controls in wound healing assays, indicating that EIF3F promotes cell motility (Fig. 9D and E). Plate colony formation assays further confirmed that EIF3F knockdown substantially decreased proliferative activity in both Huh7 and HepG2 cells (Fig. 9F and G). To comprehensively validate phenotypic effects, Transwell matrix invasion assays (Fig. 9H and I) and CCK-8 cell viability tests (Fig. 10A and B) were concurrently performed, revealing significantly reduced invasive cell penetration and markedly inhibited cell proliferation within 48 h in interference groups. Collectively, these multidimensional functional evidences demonstrate that EIF3F acts as an oncogene in HCC progression by coordinately regulating malignant phenotypes including migration, invasion, and proliferation, establishing its expression level as a potential diagnostic biomarker with promising targeted therapeutic value.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn Vivo Experiment\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate the in vivo regulatory role of EIF3F knockdown in HCC growth, this study established HepG2 cell lines with stable EIF3F knockdown using shRNA vectors, with Western blot confirming reduced protein expression compared to controls. In male BALB/c nude mouse xenograft models, the shEIF3F group exhibited significantly inhibited growth phenotypes versus NC controls: at the experimental endpoint, tumor volume was reduced and tumor weight demonstrated a 6.8-fold decrease in the knockdown group (Fig. 10C-E). These results indicate that reduced EIF3F expression significantly attenuates in vivo tumor growth potential by suppressing the proliferative capacity of HCC cells.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eMAFLD has emerged as a primary causative factor of chronic liver disease globally, with epidemiological studies revealing a significant geographical heterogeneity in its prevalence (6\u0026ndash;35%)\u003csup\u003e[\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/sup\u003e. Due to the lack of early specific biomarkers for HCC, approximately 60% of patients are diagnosed at intermediate to advanced stages, missing the opportunity for curative treatment. Despite significant advances in HCC therapeutic strategies, some patients fail to achieve substantial clinical benefits, and the overall prognosis remains suboptimal. Although current targeted-immunotherapy combinations have markedly improved the clinical management landscape of HCC, 30\u0026ndash;40% of patients exhibit primary resistance. Notably, aberrant deubiquitinating modifications constitute not only a significant driver of solid tumor progression but also play a crucial regulatory role in MAFLD pathogenesis\u003csup\u003e[\u003cspan additionalcitationids=\"CR15 CR16\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/sup\u003e. Emerging evidence indicates that dysregulated deubiquitination deeply participates in the co-regulatory network of MAFLD-HCC comorbidity through mechanisms such as mediating lipid metabolic reprogramming and remodeling the inflammatory microenvironment. Molecular mechanism studies confirm that DUBs participate in disease progression by regulating innate immune signal transduction: for instance, USP21 employs a dual regulatory mechanism\u0026mdash;its Lys63-linked deubiquitination suppresses RIG-I-mediated innate immune signaling activation, while simultaneously maintaining GATA3 protein stability in Treg cells via Lys48-linked deubiquitination to shape an immunosuppressive microenvironment\u003csup\u003e[\u003cspan additionalcitationids=\"CR19 CR20\" citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/sup\u003e. Furthermore, USP30 activates the lipogenesis pathway by stabilizing C/EBPβ protein, directly regulating hepatic steatosis progression\u003csup\u003e[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/sup\u003e. These findings unveil the molecular complexity of the deubiquitination regulatory network within the MAFLD-HCC comorbid process, highlighting the urgent need to establish a DUB activity-based molecular classification system to advance precision therapy. This study, for the first time, completed the systematic identification of HCC-specific DUBs within an MAFLD context, successfully establishing a multi-parameter DUBS evaluation system comprising USP39, USP32, OTUB2, and EIF3F. This system not only shows significant correlation with poor patient prognosis but also precisely quantifies individualized molecular characteristics, providing an objective basis for stratified therapeutic strategies; it further elucidates that specific DUB molecules exert key regulatory functions in the MAFLD-HCC comorbid process through the synergistic regulation of dual mechanisms\u0026mdash;lipid metabolic reprogramming and immune microenvironment remodeling.\u003c/p\u003e \u003cp\u003eThe 4 core DUB molecules identified in this study are associated with MAFLD and various malignancies including HCC. USP39 is recognized as a prognostic marker and potential therapeutic target across multiple cancers, participating in regulating diverse cellular activities including proliferation, migration, invasion, apoptosis, and DNA damage repair\u003csup\u003e[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]\u003c/sup\u003e. It is known to maintain hepatocyte autophagy-lipid metabolism homeostasis through regulating alternative RNA splicing mechanisms, and cooperates with the E3 ubiquitin ligase TRIM26 to form a bidirectional regulatory switch for ZEB1 protein ubiquitination modification, decisively controlling HCC proliferation and metastasis \u003csup\u003e[\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]\u003c/sup\u003e. USP32 is closely correlated with prognosis in patients with various solid tumors including small cell lung cancer, gastric cancer, and breast cancer. Xiu et al. reported that knocking down USP32 inhibits HCC cell proliferation, colony formation, and migration in vitro while suppressing tumor growth in vivo, with its expression significantly correlating with immune cell infiltration in the tumor microenvironment\u003csup\u003e[\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]\u003c/sup\u003e. OTUB2 also plays regulatory roles in the pathogenesis of multiple malignancies, promoting malignant proliferation and migration of HCC cells by increasing PJA1 stability through deubiquitination. Additionally, OTUB2 has been demonstrated to be upregulated in colorectal cancer, where it inhibits the ubiquitination of pyruvate kinase M2 (PKM2) by blocking its interaction with the ubiquitin E3 ligase Parkin, thereby enhancing PKM2 activity, promoting glycolysis, and accelerating colorectal cancer progression\u003csup\u003e[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]\u003c/sup\u003e. The eukaryotic translation initiation factor 3 (eIF3) multi-subunit complex regulates core biological processes including cell differentiation and apoptosis through specific binding of its f subunit to target mRNAs\u003csup\u003e[\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]\u003c/sup\u003e. Notably, loss of EIF3F expression is significantly associated with pathological processes in various solid tumors and muscular atrophy-related diseases\u003csup\u003e[\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]\u003c/sup\u003e. Mechanistic studies reveal that EIF3F stabilizes phosphoglycerate dehydrogenase (PHGDH) protein expression through deubiquitination, activating the serine-glycine-one-carbon (SGOC) metabolic pathway to drive malignant progression in colorectal cancer\u003csup\u003e[\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]\u003c/sup\u003e. Pan-cancer analyses indicate this molecule exhibits significant pro-oncogenic properties in solid tumors including prostate cancer, pancreatic ductal adenocarcinoma, malignant melanoma, and gastric cancer\u003csup\u003e[\u003cspan additionalcitationids=\"CR33 CR34\" citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e]\u003c/sup\u003e. However, the regulatory role and molecular mechanisms of EIF3F in MAFLD and HCC remain undefined. This study first employed multiple machine learning approaches to screen 15 MAFLD core genes including EIF3F from shared DUBs in MAFLD-HCC; then integrated HCC transcriptomic and clinical data to identify 10 genes among them associated with HCC prognosis; subsequently determined EIF3F as a core HCC gene via SHAP analysis (ranking second only to USP39 in importance); finally confirmed through single-cell sequencing, spatial transcriptomics, and in vitro/in vivo validation that EIF3F promotes HCC initiation, growth, and metastasis by regulating malignant phenotypes including proliferation, invasion, and migration of HCC cells.\u003c/p\u003e \u003cp\u003eCancer cell resistance to conventional chemotherapeutic agents has profoundly altered the therapeutic landscape of HCC. While targeted agents and immunotherapeutics developed in recent years have provided novel approaches for HCC treatment, the immunosuppressive tumor microenvironment significantly diminishes immunotherapy efficacy, necessitating the urgent development of novel indicators for predicting the immune microenvironment and drug sensitivity. This study provides a new indicator, DUBS, for HCC treatment through drug sensitivity-gene expression correlation analysis and molecular docking techniques. Results demonstrate that DUBS effectively predicts differential patient responses to chemotherapy and immunotherapy: drug sensitivity analysis revealed superior efficacy of agents such as 5-fluorouracil in the low DUBS group, whereas patients in the high DUBS group exhibited significantly enhanced objective response rates to gemcitabine, providing crucial guidance for personalized treatment selection. Molecular docking-based virtual screening identified candidate drugs including donafenib, doxorubicin, fluorouracil, gemcitabine, lenvatinib, and rivoceranib as having potential therapeutic effects through specific binding to the EIF3F protein; immunotherapy efficacy analysis confirmed DUBS as a reliable predictor (higher DUBS correlates with poorer efficacy), and immune microenvironment analysis indicated that the high DUBS group exhibits immunologically cold tumor characteristics, explaining the inferior immunotherapy outcomes in this group. Therefore, for high-DUBS group patients, targeted therapy combined with chemotherapy may be employed to improve prognosis, while low-DUBS group patients may benefit from immunotherapy combined with targeted therapy or chemotherapy to enhance efficacy. Mechanistic investigation demonstrated that donafenib significantly enhances tumor cell apoptosis and inhibits metastasis through multiple mechanisms, including synergistically inducing reactive oxygen species (ROS) accumulation, promoting ferroptosis-related protein expression, and activating the p53 signaling pathway\u003csup\u003e[\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e]\u003c/sup\u003e. These findings suggest that DUBS-mediated heterogeneity in drug response provides a molecular classification basis for personalized therapy, and that targeting EIF3F combined with immune checkpoint inhibitors may improve HCC patient prognosis, potentially profoundly reversing disease outcomes in the future. Subsequent validation of the clinical utility of DUBS and elucidation of its regulatory network require prospective clinical trials integrated with multi-omics technologies.\u003c/p\u003e \u003cp\u003eIn summary, the DUBS system developed in this study not only accurately and effectively predicts the prognosis of HCC patients but also provides a novel therapeutic target (specifically EIF3F). However, this study has limitations: neither prospective validation of the DUBS model with clinical samples nor verification of the in vivo therapeutic efficacy of the candidate drugs was performed. Future work necessitates conducting large-scale prospective clinical trials to validate the model's predictive power, in-depth investigation of the mechanisms of action and in vivo efficacy of candidate drugs, and exploration of personalized treatment regimens for patients with both MAFLD and HCC.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOverall, this study established a DUBS, comprising USP39, USP32, OTUB2, and EIF3F, via multi-omics analysis, and confirmed its role as an independent prognostic marker for HCC. Results demonstrated that patients with low DUBS exhibited significantly prolonged overall survival, enhanced response to immunotherapy, and increased sensitivity to specific chemotherapeutic agents. The core gene EIF3F was identified as a key oncogenic driver, promoting tumor malignancy progression and demonstrating dual diagnostic value for both MAFLD and HCC. Collectively, the DUBS system provides a novel molecular tool for prognostic assessment, therapy response prediction, and individualized treatment strategy formulation in HCC, while highlighting EIF3F as a potential therapeutic target for precision medicine. Future investigations should elucidate the ubiquitination regulatory network mediated by EIF3F and validate, through prospective clinical trials, the translational value of targeting EIF3F combined with immune checkpoint inhibitors (e.g., Donafenib) for reversing chemoresistance in HCC.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eACC: Adrenocortical Cancer; AUC: Area Under the Curve; BLCA: Bladder Cancer; BRCA: Breast Cancer; CESC: Cervical Cancer; CHOL: Bile Duct Cancer; CI: confidence intervals; COAD: Colon Cancer; CR: complete response; CTRP: Cancer Therapeutics Response Portal; DFI: Disease Free \u0026nbsp;Interval; DEGs: differentially expressed genes; DLBC: Large B-cell Lymphoma; DSS: Disease Specific Survival; DUB: deubiquitinase; DUBS: Deubiquitination Score; eIF3:eukaryotic translation initiation factor3; ESCA: Esophageal Cancer; FBS: fetal bovine serum; FDA: Food and Drug Administration; FDR: false discovery rate; FPKM: Fragments Per Kilobase Million; GBM: Glioblastoma; GDSC: Genomics of Drug Sensitivity in Cancer; GEO: Gene Expression Omnibus; GSCA: Gene Set Cancer Analysis; GSVA: Gene Set Variation Analysis; GSEA: Gene Set Enrichment Analysis; HCC: hepatocellular carcinoma; HNSC: Head and Neck Cancer; HR: hazardratios; ICGC: International Cancer Genome Consortium; ICIs: Immune Check point Inhibitors; IPS: Immunophenoscore; KICH: Kidney Chromophobe; KIRC/ccRCC: Kidney/Renal Clear Cell Carcinoma; KIRP: Kidney Papillary Cell Carcinoma; KM: Kaplan–Meier; LAML: Acute Myeloid Leukemia; Lasso: Least Absolute Shrinkage and Selection Operator; LGG: Lower Grade Glioma; LIHC: Liver Cancer; LUAD: Lung Adenocarcinoma; LUSC: Lung Squamous Cell Carcinoma; MAFLD: metabolic dysfunction-associated fatty liver disease; MESO: Mesothelioma; MIA: multimodal intersection analysis; MSI: microsatellite instability; Neg.: Negative; OS: Overall Survival; OV: Ovarian Cancer; PAAD: Pancreatic Cancer; PBS: phosphate-buffered saline; PCA: principal component analysis; PD: progressive disease; PHGDH: phosphoglycerate dehydrogenase; PKM2: pyruvate kinase M2; Pos.: Positive; PR: partial response; PRAD: Prostate Cancer; READ: Rectal Cancer; ROC: Receiver Operating Characteristic Curve; ROS: reactive oxygen species; SARC: Sarcoma; SD: stable disease; SGOC: serine-glycine-one-carbon; SHAP: Shapley Additive exPlanations; SKCM: Melanoma; STAD: Stomach Cancer; TCGA: The Cancer Genome Atlas; THCA: Thyroid Cancer; TIDE: Tumor Immune Dysfunction and Exclusion; TMB: Tumor mutational burden; UCEC: Endometrioid Cancer; UVM: Ocular melanomas.\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eY.H. wrote and typeset the manuscript for this study; H.L. was responsible for data collection and organization, bioinformatics analysis and study design; W.X. and J.M. were responsible for experiments and editing the full text;\u0026nbsp;J.Z. was responsible for typesetting the manuscript for this study and study design;\u0026nbsp;Y.L. provided corrections and suggestions throughout the study and provided financial support. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0508700), Scientific Research Fund of Liaoning Provincial Education Department (LJKMZ20221167)and Supported by Liaoning Revitalization Talents Program（XLYC2412033）.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data of this study are freely available from the website TCGA Research Network (https://portal.gdc.cancer.gov/), GEO database(https://www.ncbi.nlm.nih.gov/geo/), ICGC database(https://dcc.icgc.org/). Further inquiries can be directed to the corresponding authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflicts of interest.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent for participate statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries[J]. CA: a cancer journal for clinicians, 2021, 71(3): 209-249.\u003c/li\u003e\n\u003cli\u003eYang X, Yang C, Zhang S, et al. Precision treatment in advanced hepatocellular carcinoma[J]. Cancer Cell, 2024, 42(2): 180-197.\u003c/li\u003e\n\u003cli\u003eLlovet JM, Montal R, Sia D, et al. Molecular therapies and precision medicine for hepatocellular carcinoma[J]. Nature Reviews. Clinical Oncology, 2018, 15(10): 599-616.\u003c/li\u003e\n\u003cli\u003eCrane H, Eslick GD, Gofton C, et al. Global prevalence of metabolic dysfunction-associated fatty liver disease-related hepatocellular carcinoma: A systematic review and meta-analysis[J]. Clinical and Molecular Hepatology, 2024, 30(3): 436-448.\u003c/li\u003e\n\u003cli\u003eRape M. Ubiquitylation at the crossroads of development and disease[J]. Nature Reviews. Molecular Cell Biology, 2018, 19(1): 59-70.\u003c/li\u003e\n\u003cli\u003eReyes-Turcu FE, Ventii KH, Wilkinson KD. Regulation and cellular roles of ubiquitin-specific deubiquitinating enzymes[J]. Annual Review of Biochemistry, 2009, 78: 363-397.\u003c/li\u003e\n\u003cli\u003eWu Q, Qiu Y, Guo J, et al. USP40 promotes hepatocellular carcinoma cell proliferation, migration and stemness by deubiquitinating and stabilizing Claudin1[J]. Biology Direct, 2024, 19(1): 13.\u003c/li\u003e\n\u003cli\u003eZhou B, Luo Y, Bi H, et al. Amelioration of nonalcoholic fatty liver disease by inhibiting the deubiquitylating enzyme RPN11[J]. Cell Metabolism, 2024, 36(10): 2228-2244.e7.\u003c/li\u003e\n\u003cli\u003eLi X, Yuan J, Song C, et al. Deubiquitinase USP39 and E3 ligase TRIM26 balance the level of ZEB1 ubiquitination and thereby determine the progression of hepatocellular carcinoma[J]. Cell Death and Differentiation, 2021, 28(8): 2315-2332.\u003c/li\u003e\n\u003cli\u003eDominguez DA, Wong P, Melstrom LG. Existing and emerging biomarkers in hepatocellular carcinoma: relevance in staging, determination of minimal residual disease, and monitoring treatment response: a narrative review[J]. Hepatobiliary Surgery and Nutrition, 2024, 13(1): 39-55.\u003c/li\u003e\n\u003cli\u003eDewson G, Eichhorn PJA, Komander D. Deubiquitinases in cancer[J]. Nature Reviews. Cancer, 2023, 23(12): 842-862.\u003c/li\u003e\n\u003cli\u003eThorsson V, Gibbs DL, Brown SD, et al. The Immune Landscape of Cancer[J]. Immunity, 2018, 48(4): 812-830.e14.\u003c/li\u003e\n\u003cli\u003eHuang DQ, El-Serag HB, Loomba R. Global epidemiology of NAFLD-related HCC: trends, predictions, risk factors and prevention[J]. Nature Reviews. Gastroenterology \u0026amp; Hepatology, 2021, 18(4): 223-238.\u003c/li\u003e\n\u003cli\u003eXue Y, Xue C, Song W. Emerging roles of deubiquitinating enzymes in actin cytoskeleton and tumor metastasis[J]. Cellular Oncology (Dordrecht, Netherlands), 2024, 47(4): 1071-1089.\u003c/li\u003e\n\u003cli\u003eKong L, Jin X. Dysregulation of deubiquitination in breast cancer[J]. Gene, 2024, 902: 148175.\u003c/li\u003e\n\u003cli\u003eRen J, Yu P, Liu S, et al. Deubiquitylating Enzymes in Cancer and Immunity[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(36): e2303807.\u003c/li\u003e\n\u003cli\u003eKong L, Jin X. Dysregulation of deubiquitination in breast cancer[J]. Gene, 2024, 902: 148175.\u003c/li\u003e\n\u003cli\u003eBech-Otschir D, Kraft R, Huang X, et al. COP9 signalosome-specific phosphorylation targets p53 to degradation by the ubiquitin system[J]. The EMBO journal, 2001, 20(7): 1630-1639.\u003c/li\u003e\n\u003cli\u003eFan Y, Mao R, Yu Y, et al. USP21 negatively regulates antiviral response by acting as a RIG-I deubiquitinase[J]. The Journal of Experimental Medicine, 2014, 211(2): 313-328.\u003c/li\u003e\n\u003cli\u003eZhang Q, Chen Z, Tang Q, et al. USP21 promotes self-renewal and tumorigenicity of mesenchymal glioblastoma stem cells by deubiquitinating and stabilizing FOXD1[J]. Cell Death \u0026amp; Disease, 2022, 13(8): 712.\u003c/li\u003e\n\u003cli\u003eRen J, Yu P, Liu S, et al. Deubiquitylating Enzymes in Cancer and Immunity[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(36): e2303807.\u003c/li\u003e\n\u003cli\u003eJiang F, Li X, Lin L, et al. NPRC promotes hepatic steatosis via USP30-mediated deubiquitination of C/EBP\u0026beta;[J]. Metabolism: Clinical and Experimental, 2025, 162: 156050.\u003c/li\u003e\n\u003cli\u003eY Z, F W, S F, et al. USP39 promote post-translational modifiers to stimulate the progress of cancer[J]. Discover oncology, 2025, 16(1).\u003c/li\u003e\n\u003cli\u003eCui D, Wang Z, Dang Q, et al. Spliceosome component Usp39 contributes to hepatic lipid homeostasis through the regulation of autophagy[J]. Nature Communications, 2023, 14(1): 7032.\u003c/li\u003e\n\u003cli\u003eXiu M, Bao W, Wang J, et al. High USP32 expression contributes to cancer progression and is correlated with immune infiltrates in hepatocellular carcinoma[J]. BMC Cancer, 2023, 23: 1105.\u003c/li\u003e\n\u003cli\u003eS L, Y S, K W, et al. USP32 deubiquitinase: cellular functions, regulatory mechanisms, and potential as a cancer therapy target[J]. Cell death discovery, 2023, 9(1).\u003c/li\u003e\n\u003cli\u003eS Y, W Z, Y Q, et al. Deubiquitinase OTUB2 exacerbates the progression of colorectal cancer by promoting PKM2 activity and glycolysis[J]. Oncogene, 2022, 41(1).\u003c/li\u003e\n\u003cli\u003eLee ASY, Kranzusch PJ, Cate JHD. eIF3 targets cell-proliferation messenger RNAs for translational activation or repression[J]. Nature, 2015, 522(7554): 111-114.\u003c/li\u003e\n\u003cli\u003eMasutani M, Sonenberg N, Yokoyama S, et al. Reconstitution reveals the functional core of mammalian eIF3[J]. The EMBO journal, 2007, 26(14): 3373-3383.\u003c/li\u003e\n\u003cli\u003eMarchione R, Leibovitch SA, Lenormand JL. The translational factor eIF3f: the ambivalent eIF3 subunit[J]. Cellular and molecular life sciences: CMLS, 2013, 70(19): 3603-3616.\u003c/li\u003e\n\u003cli\u003ePan Q, Yu F, Jin H, et al. eIF3f Mediates SGOC Pathway Reprogramming by Enhancing Deubiquitinating Activity in Colorectal Cancer[J]. Advanced Science (Weinheim, Baden-Wurttemberg, Germany), 2023, 10(27): e2300759.\u003c/li\u003e\n\u003cli\u003eLi J, Yu W, Ge J, et al. Targeting eIF3f Suppresses the Growth of Prostate Cancer Cells by Inhibiting Akt Signaling[J]. OncoTargets and Therapy, 2020, 13: 3739-3750.\u003c/li\u003e\n\u003cli\u003eDoldan A, Chandramouli A, Shanas R, et al. Loss of the eukaryotic initiation factor 3f in melanoma[J]. Molecular Carcinogenesis, 2008, 47(10): 806-813.\u003c/li\u003e\n\u003cli\u003eA D, A C, R S, et al. Loss of the eukaryotic initiation factor 3f in pancreatic cancer[J]. Molecular carcinogenesis, 2008, 47(3).\u003c/li\u003e\n\u003cli\u003eCheng Y, Zhou J, Li H. Clinicopathologic Implications of Eukaryotic Initiation Factor 3f and Her-2/neu Expression in Gastric Cancer[J]. Clinical and Translational Science, 2015, 8(4): 320-325.\u003c/li\u003e\n\u003cli\u003eLiang J, Chen M, Yan G, et al. Donafenib activates the p53 signaling pathway in hepatocellular carcinoma, induces ferroptosis, and enhances cell apoptosis[J]. Clinical and Experimental Medicine, 2025, 25(1): 29.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"human-genetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"huge","sideBox":"Learn more about [Human Genetics](https://www.springer.com/journal/439)","snPcode":"439","submissionUrl":"https://submission.nature.com/new-submission/439/3","title":"Human Genetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Machine Learning, Deubiquitinases, Spatial Transcriptome, Immune Microenvironment, Immunotherapy, Prognosis","lastPublishedDoi":"10.21203/rs.3.rs-6909459/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6909459/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eHepatocellular carcinoma (HCC) represents a malignancy with high global mortality. Metabolic dysfunction-associated fatty liver disease (MAFLD) serves as a significant contributory pathogenic factor. Deubiquitinases (DUBs), which regulate protein homeostasis, are implicated in disease progression. This study focused on identifying shared mechanisms between MAFLD and HCC, screening for key DUB genes, and constructing a novel prognostic scoring system termed the Deubiquitination Score (DUBS). The DUBS significantly stratified patient survival, with a high-DUBS indicating poor prognosis and malignant tumor progression. Patients with a low-DUBS demonstrated enhanced responses to immunotherapy and prolonged survival. Their tumors exhibited characteristics of \"hot tumors,\" featuring abundant immune cell infiltration and an active tumor microenvironment, accompanied by higher microsatellite instability（MSI）. The core gene identified, EIF3F, exhibited superior cross-disease diagnostic value between HCC and MAFLD. In vitro and in vivo experiments confirmed that core DUBs significantly promoted malignant behaviors of HCC cells and tumorigenic capacity in vivo. Furthermore, the DUBS revealed associations with sensitivity to chemotherapeutic agents. In summary, this study provided an important molecular tool and mechanistic foundation for the early screening, prognostic assessment, differential diagnosis, and prediction of immunotherapy response in HCC. It also highlighted potential directions for targeted intervention and the development of combination therapeutic strategies.\u003c/p\u003e","manuscriptTitle":"Integrating Machine Learning and Spatial Transcriptomics Uncovers Shared Immunomodulatory Deubiquitinases in MAFLD and HCC","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-29 14:38:57","doi":"10.21203/rs.3.rs-6909459/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-17T15:09:51+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-05T15:18:40+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"150710658240279385916667808526636245181","date":"2025-08-15T23:48:22+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-15T12:13:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"304792725439049790281170531949973666265","date":"2025-08-12T02:31:12+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-06-23T17:00:48+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-06-18T00:47:54+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-06-18T00:46:21+00:00","index":"","fulltext":""},{"type":"submitted","content":"Human Genetics","date":"2025-06-17T02:19:22+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"human-genetics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"huge","sideBox":"Learn more about [Human Genetics](https://www.springer.com/journal/439)","snPcode":"439","submissionUrl":"https://submission.nature.com/new-submission/439/3","title":"Human Genetics","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"dbcf55d5-32df-458c-8d5f-e1ba79592d4a","owner":[],"postedDate":"June 29th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-04-13T16:06:53+00:00","versionOfRecord":{"articleIdentity":"rs-6909459","link":"https://doi.org/10.1007/s00439-026-02833-6","journal":{"identity":"human-genetics","isVorOnly":false,"title":"Human Genetics"},"publishedOn":"2026-04-10 15:59:18","publishedOnDateReadable":"April 10th, 2026"},"versionCreatedAt":"2025-06-29 14:38:57","video":"","vorDoi":"10.1007/s00439-026-02833-6","vorDoiUrl":"https://doi.org/10.1007/s00439-026-02833-6","workflowStages":[]},"version":"v1","identity":"rs-6909459","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6909459","identity":"rs-6909459","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}