Breaking Therapeutic Resistance in Hepatocellular Carcinoma: Systems Biology Deconvolution of the B-cell Microenvironment Identifies PDIA6 as a Novel Target

Breaking Therapeutic Resistance in Hepatocellular Carcinoma: Systems Biology Deconvolution of the B-cell Microenvironment Identifies PDIA6 as a Novel Target | 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 Breaking Therapeutic Resistance in Hepatocellular Carcinoma: Systems Biology Deconvolution of the B-cell Microenvironment Identifies PDIA6 as a Novel Target Yang Li, Sinan Cao, Yamei Kuang, Dachuan Shen, Lili Tian This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9213778/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background B cells are integral components of the tumor microenvironment (TME) and play a pivotal role in regulating anti-tumor immunity. However, the systematic classification of B-cell-associated molecular features in hepatocellular carcinoma (HCC) and their value in clinical prognosis and immunotherapy remain elusive. Methods We elucidated the molecular characteristics of B cells within the HCC TME based on single-cell RNA sequencing (scRNA-seq) data and constructed B-cell-associated molecular subtypes by integrating bulk RNA sequencing data. We systematically evaluated differences across subtypes regarding clinical prognosis, biological processes, genomic variations, the immune microenvironment, and immunotherapeutic responses. Furthermore, a B-cell-associated gene signature score (BCAGS) prognostic model was established using 117 combinations of machine learning algorithms. Finally, key genes were selected for in vitro functional validation. Results Based on B cell marker genes, we identified three molecular subtypes with significantly distinct clinical outcomes, validating their robustness across multiple external cohorts. The C2 subtype exhibited higher genomic instability and the poorest prognosis, whereas the C3 subtype presented an "immune-hot" phenotype with predicted sensitivity to immunotherapy. Additionally, the BCAGS was confirmed as an independent risk factor for HCC patients, demonstrating superior predictive performance compared to existing models. Functional experiments further revealed that silencing the key gene, PDIA6 , significantly inhibited the proliferation, migration, and invasion of HCC cells. Conclusions This study systematically unveils the heterogeneity of B-cell-associated molecular subtypes in HCC and their clinical significance, establishing a robust BCAGS prognostic model. Combined with in vitro validation, the results suggest that PDIA6 may play a critical role in promoting HCC progression, providing a novel theoretical basis for precision stratification and the development of potential therapeutic targets in HCC. Hepatocellular carcinoma B cells Molecular subtypes Machine learning Prognostic signature PDIA6 Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Introduction Hepatocellular carcinoma (HCC) remains a leading contributor to global cancer-related mortality, consistently ranking among the top three most aggressive malignancies 1 , 2 . Although multimodal strategies—integrating surgical resection, interventional therapies, and molecularly targeted agents—have markedly improved short-term patient survival, the profound intra- and inter-tumor heterogeneity of HCC continues to pose a formidable barrier to the realization of precision medicine 3 , 4 . Clinical evidence reveals that patients even within the same TNM stage often exhibit strikingly divergent clinical outcomes following standardized treatment 5 . Such heterogeneity underscores the limitations of traditional clinicopathological staging in capturing the complex biological landscape of HCC. Consequently, there is an urgent need to leverage deep molecular profiling to identify more precise prognostic signatures and actionable therapeutic vulnerabilities 6 . The dynamic evolution of the tumor microenvironment (TME) serves as a pivotal determinant of HCC progression 7 , 8 . Historically, investigations into the immune landscape of HCC have predominantly centered on the orchestration of T-cell exhaustion and activation 9 , whereas the functional contribution of B cells—the cornerstone of humoral immunity—within the TME has long remained marginalized. Emerging evidence indicates a paradigm shift: tumor-infiltrating B cells do more than merely mediate anti-tumor effects via antibody secretion; they actively reshape T-cell-mediated immunity by facilitating the assembly of tertiary lymphoid structures 10 – 12 . Nevertheless, the extent to which B-cell-associated molecular signatures define distinct HCC clinical subtypes, and how this heterogeneity dictates the therapeutic efficacy of immune checkpoint inhibitors, has yet to be systematically quantified. The rapid maturation of single-cell RNA sequencing (scRNA-seq) technologies has opened an unprecedented window into the granular deconvolution of intratumoral heterogeneity at single-cell resolution 13 , 14 . Nevertheless, the clinical translation of single-cell findings is often constrained by the limited scale of available cohorts, which hampers the direct derivation of robust predictive models. To bridge this gap, the present study employed a "cross-resolution integration" strategy. By first deciphering the distinct B-cell molecular signatures via scRNA-seq and subsequently projecting these features onto expansive bulk transcriptomic cohorts, we systematically identified and validated three B-cell-associated molecular subtypes characterized by significantly divergent clinical outcomes 15 . To bolster the clinical utility and robustness of our findings, we implemented an integrative ensemble screening strategy to develop a refined B-cell-associated gene signature (BCAGS). This scoring system not only demonstrated superior predictive accuracy across multiple independent, cross-platform cohorts but also highlighted a compelling link between high BCAGS scores and underlying genomic instability 16 . Furthermore, to translate these bioinformatic insights into mechanistic understanding, we performed extensive functional validation of PDIA6 —a core driver gene within the signature—thereby reconciling computational predictions with biological reality 17 , 18 . Collectively, our study enriches the current understanding of immune microenvironment heterogeneity in HCC and provides a robust framework for precision patient stratification and the optimization of personalized immunotherapeutic interventions. Methods 1.1 Data Acquisition and Preprocessing Single-cell RNA sequencing (scRNA-seq) data for the GSE140228 dataset, which comprises six tumor and five adjacent normal samples, were retrieved from the Gene Expression Omnibus (GEO) database. For large-scale genomic analysis, transcriptomic profiles, single-nucleotide variant data, and corresponding clinical annotations for the TCGA-LIHC cohort ( n = 369) were curated from The Cancer Genome Atlas (TCGA). Additionally, two independent validation cohorts were integrated: the GSE14520 dataset ( n = 221) from GEO and the ICGC-LIRI-JP cohort ( n = 232) from the International Cancer Genome Consortium (ICGC) portal. To ensure inter-platform comparability, gene expression matrices were normalized and log2-transformed where appropriate. Comprehensive metadata and characteristics for all included cohorts are detailed in Supplementary Table 1. 1.2 scRNA-seq Data Analysis GSE140228 scRNA-seq data were analyzed using Seurat (v4.2) 19 . Quality control retained genes expressed in ≥ 3 cells and cells with ≥ 200 genes and ≤ 10% mitochondrial content, yielding 3,817 immune cells. After normalization, 2,000 highly variable genes were identified for Principal Component Analysis (PCA). Batch effects were corrected using Harmony 20 , with cell clusters visualized via t-distributed stochastic neighbor embedding (t-SNE). Cluster-specific markers were identified via FindAllMarkers (Wilcoxon rank-sum test), and cell types were annotated using SingleR 21 and canonical lineage markers. clusterProfiler 22 was used for B-cell Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. Intercellular communication and ligand-receptor interactions were inferred via CellChat 23 . 1.3 Identification of B-cell-associated Molecular Subtypes in HCC To delineate the clinical relevance of B-cell heterogeneity, we first identified the marker genes for each B-cell subcluster using the FindAllMarkers function. Subsequently, a univariate Cox proportional hazards regression analysis was conducted to screen for genes significantly associated with the overall survival (OS) of HCC patients, yielding 71 prognostic candidates (Supplementary Table 5). These genes were then utilized as the basis for unsupervised molecular subtyping. Based on the ConsensusClusterPlus R package, consensus clustering was performed using the Partitioning Around Medoids (PAM) algorithm with 1,000 permutations to ensure the stability of the classification 24 . The optimal number of clusters ( k ) was determined by the cumulative distribution function curve and the tracking plot. To further evaluate the robustness and clinical significance of the identified subtypes, Kaplan-Meier survival analysis was employed to compare prognostic outcomes, while t-SNE was utilized to visualize the transcriptional distinctness between the clusters. 1.4 Weighted Gene Co-expression Network Analysis (WGCNA) In order to identify key gene modules and hub genes characteristic of the three identified HCC subtypes, a weighted gene co-expression network was constructed using the WGCNA R package 25 . First, an adjacency matrix was calculated based on the Pearson’s correlation between gene expression profiles. To achieve a scale-free topology, we evaluated a range of soft-thresholding powers ( β ); a power of β = 6 was selected, as it was the lowest power that allowed the scale-free topology fit index ( R 2 ) to reach 0.9. Subsequently, the adjacency matrix was transformed into a Topological Overlap Matrix to estimate the network connectivity. We utilized a hierarchical clustering dendrogram for module detection via the dynamic tree cut method, with the minimum module size set to 40. Finally, highly similar modules were consolidated using the mergeCloseModules function with a height cut-off of 0.25 to yield the final co-expression modules for downstream module-trait correlation analysis. 1.5 Subtype Prediction via Nearest Template Prediction (NTP) We employed a Nearest Template Prediction (NTP) algorithm to facilitate the clinical translation and cross-platform validation of our B-cell-associated molecular subtypes 26 . Unlike traditional clustering methods, the NTP approach allows for the class prediction of individual samples with high flexibility by utilizing a predefined list of signature genes. In this study, the top-ranked marker genes identified from each subtype were used to construct the prediction templates. For each sample in the validation cohorts (TCGA and ICGC), a proximity-based matching was performed against these templates using a cosine distance metric. To ensure the reliability of the classification, a permutation-based P -value was calculated for each prediction, with a threshold of P < 0.05 and a False Discovery Rate (FDR) < 0.1 applied to define high-confidence assignments. 1.6 Functional Enrichment and Pathway Analysis To unravel the biological motifs and signaling pathways underlying the discrete B-cell-associated subtypes, we performed differential expression analysis between each subtype and the remaining cohorts using the limma R package 27 . For each subtype-specific profile, genes were sorted in descending order based on their log2-fold change (log2FC) to generate a pre-ranked gene list. Subsequently, Gene Set Enrichment Analysis (GSEA) was executed via the clusterProfiler package 22 using the GO and KEGG databases as reference sets. A permutation-based approach with 1,000 iterations was applied to estimate the significance of the enrichment. Biological pathways with a normalized enrichment score (NES) absolute value > 1.0 and a Benjamini-Hochberg adjusted P -value < 0.05 were considered significantly enriched. 1.7 Characterization of Genomic Alterations: Somatic Mutations and Copy Number Variations In order to delineate the genomic landscape across different B-cell-associated subtypes, somatic mutation data were processed and visualized using the maftools R package 28 . Mutation Annotation Format files were integrated via the read.maf function, and the overall mutational burden, including variant classifications and transition/transversion ratios, was summarized using the plotmafSummary module. To identify and compare the frequently mutated genes among the subtypes, waterfall plots (oncoplots) were generated to illustrate the mutation frequencies and co-occurrence patterns of top-tier drivers. Furthermore, copy number variation (CNV) profiles were explored using the cBioPortal for Cancer Genomics database 29 . We specifically quantified focal genomic events, focusing on the frequency of high-level amplifications and homozygous deletions for the top 10 most frequently altered genes. 1.8 Characterization of the Tumor Immune Microenvironment (TIME) and Immunotherapy Response To characterize the immune landscape, relative infiltration levels of 28 immune cell subpopulations 30 (Supplementary Table 2) were quantified using the single-sample Gene Set Enrichment Analysis (ssGSEA) algorithm via the GSVA R package 31 . Additionally, the transcriptional profiles of 27 immune checkpoints and 9 Human Leukocyte Antigen (HLA) genes were systematically assessed 32 – 34 (Supplementary Table 3). To predict sensitivity to immune checkpoint inhibitors, three established frameworks were implemented. First, the T-cell-inflamed signature (TIS) score, comprising 18 inflammatory genes, was derived using ssGSEA 35 . Second, Subclass mapping (Submap) analysis was utilized to infer the likelihood of clinical response to anti-PD-1 and anti-CTLA-4 therapies 36 . Finally, the Immunophenoscore (IPS) was calculated via the IOBR package 37 , integrating four functional components: antigen presentation (MHC), effector cells (EC), suppressor cells (SC), and checkpoints (CP) 30 . 1.9 Drug Sensitivity and Connectivity Map (CMap) Analysis To identify potential therapeutic agents, the oncoPredict R package 38 was employed to estimate the half-maximal inhibitory concentration (IC 50 ) of various chemotherapeutic and targeted drugs across HCC subtypes. Concurrently, the Connectivity Map (CMap) database was utilized to screen for candidate small-molecule compounds 39 . Differentially expressed genes (DEGs) between subtypes were identified via limma, and the top 150 up-regulated and 150 down-regulated genes were queried through the CMap L1000 platform. Enrichment scores were calculated to identify therapeutic perturbations associated with the B-cell-related molecular signatures. 1.10 Machine Learning-Based Prognostic Signature Development To construct a robust prognostic signature, 117 machine learning combinations were implemented using the Mime R package, integrating ten distinct algorithms: random survival forest (RSF), elastic net (Enet), stepwise Cox, CoxBoost, partial least squares regression for Cox (plsRcox), supervised principal components (superpc), generalized boosted models (GBM), survival support vector machine (survivalsvm), Ridge, and Lasso 40 . Models were optimized via K-fold cross-validation within the training cohort. The predictive accuracy of each combination was quantified by the Harrell’s concordance index (C-index) across both training and independent validation datasets. The algorithm combination yielding the highest mean C-index was selected for the final model construction. Furthermore, the prognostic performance of our signature was benchmarked against 37 previously published HCC models to evaluate its comparative superiority (Supplementary Table 4). 1.11 Cell Culture and siRNA Transfection Human HCC cell lines HepG-2 and Huh-7 (Cell Bank of the Chinese Academy of Sciences, Shanghai, China) were cultured in DMEM and RPMI-1640 media, respectively, each supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. All cultures were maintained at 37°C in a humidified incubator with 5% CO 2 . For gene knockdown, specific siRNAs targeting PDIA6 and negative control (NC) sequences (Shanghai Hanhui Biotech) were transfected into cells using Lipofectamine™ RNAiMAX (Invitrogen, USA) following the manufacturer’s protocol. Transfection efficiency was assessed for subsequent functional assays. siRNA sequences are shown in Table 1. 1.12 Quantitative Real-Time PCR Total RNA was extracted from cells using the TRIzol™ Reagent (Invitrogen, Carlsbad, CA, USA) and subsequently reverse-transcribed into cDNA utilizing the PrimeScript™ RT reagent Kit (Takara Bio, Shiga, Japan) strictly following the manufacturers' protocols. Quantitative real-time PCR (RT-qPCR) assays were performed using the SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) on a QuantStudio™ 5 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The relative mRNA expression levels of PDIA6 were calculated using the 2 −ΔΔ Ct method, with GAPDH serving as the endogenous normalization control. The specific primer sequences utilized for amplification are listed in Table 2. 1.13 Cell Proliferation Assay Cell viability was evaluated using the Cell Counting Kit-8 (CCK-8) assay (Beyotime Biotechnology, Shanghai, China). Transfected cells were seeded into 96-well plates at a density of 2–5×10 3 cells per well. At 24, 48, and 72 h post-seeding, 10µL of CCK-8 reagent was added to each well, followed by incubation at 37°C for 1–2 h. The absorbance at 450 nm (OD450) was measured using Synergy H1 microplate reader (BioTek Instruments, Winooski, VT, USA), with background values subtracted from the blank wells. All experiments were performed in triplicate and independently repeated three times to ensure reproducibility. 1.14 Wound Healing Assay Transfected cells were seeded in 6-well plates and cultured to 90%–100% confluence. A linear scratch was created using a sterile 200 µL pipette tip. After washing with PBS to remove cellular debris, the cells were incubated in serum-free medium to minimize the influence of cell proliferation. Microscopic images were captured at 0 and 24 h in the same fields of view using an Olympus IX73 inverted microscope (Olympus, Tokyo, Japan). The wound area was quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA), and the wound closure rate was calculated to evaluate migratory capacity of the cells. All assays were performed in independent triplicates. 1.15 Transwell Migration and Invasion Assays Cell motility was assessed using Transwell chambers equipped with 8-µm pore size polycarbonate membrane inserts (Corning, Corning, NY, USA). For the invasion assay, the upper chamber membranes were pre-coated with Matrigel (Corning, Corning, NY, USA) to simulate the extracellular matrix, whereas non-coated membranes were used for the migration assay. Transfected cells were suspended in 200 µL of serum-free medium (2–5×10 4 cells/well) and seeded into the upper chamber, while 600µL of medium supplemented with 10% FBS was added to the lower chamber as a chemoattractant. After 24 h of incubation at 37°C, cells remaining on the upper surface were removed. Migrated or invaded cells on the lower surface were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The number of migrated or invaded cells was quantified by counting five randomly selected fields per well under a light microscope. All assays were performed in independent triplicates. 1.16 Statistical Analysis All statistical analyses were conducted using R software (version 4.3.1). For continuous variables derived from bioinformatics analyses, the Wilcoxon rank-sum test was employed for comparisons between two groups, while the Kruskal-Wallis test was used for multi-group comparisons. For in vitro experimental data, results were presented as mean ± standard deviation (SD), and statistical significance between two groups was evaluated using Student’s t-test. Survival curves were generated using the Kaplan-Meier method and compared via the log-rank test. Univariate and multivariate Cox proportional hazards regression analyses were implemented using the survival R package to identify independent prognostic factors. Where appropriate, P-values were adjusted for multiple testing using the Benjamini-Hochberg false discovery rate or Bonferroni corrections. All statistical tests were two-tailed, and a P-value < 0.05 was considered statistically significant (ns, not significant; * P < 0.05; ** P < 0.01; *** P < 0.001). Results 2.1 Characterization of the Immune Cell Landscape in HCC Following stringent quality control, a total of 3,817 high-quality immune cells from the GSE140228 dataset were prioritized for analysis. Unsupervised clustering via the Seurat package categorized these cells into 17 distinct subclusters (Fig. 1 a), with their spatial distribution across tumor and adjacent normal tissues visualized via t-SNE (Fig. 1 b). Based on the SingleR algorithm and canonical lineage markers, these cells were annotated into five primary immune lineages: T cells ( CD3D, CD3E ), monocytes ( CD68, CD163 ), natural killer (NK) cells ( KLRC1, NCR1 ), B cells ( CD19, CD79A ), and dendritic cells (DCs; CLEC9A, XCR1 ) (Fig. 1 c; Supplementary Fig. 1a–e). Compositional analysis revealed that T cells constituted the most predominant population across all samples, followed by monocytes and NK cells, while DCs were the least abundant (Fig. 1 d, e). Notably, odds ratio (OR) analysis demonstrated a significant tissue preference, highlighting that B cells were preferentially enriched within the TME compared to normal tissues (Fig. 1 f). Furthermore, the top three highly expressed marker genes for each lineage were identified, reinforcing the accuracy of the cell-type identification (Fig. 1 g). 2.2 Functional Enrichment Analysis of B-Cell Signature Genes To elucidate the biological landscape of the B-cell compartment, we identified subcluster-specific marker genes using the FindAllMarkers function, applying thresholds of log2FC > 1 and P < 0.05. Functional annotation via the clusterProfiler package revealed that these signature genes were predominantly enriched in GO terms associated with humoral immunity and immunoglobulin production. Key biological processes included immunoglobulin-mediated immune responses, B-cell receptor signaling, and lymphocyte-mediated immunity (Fig. 2 a). Furthermore, KEGG pathway analysis highlighted the involvement of these genes in critical immunological and homeostatic pathways, such as the NF-kappa B signaling pathway, cell adhesion molecules, and protein processing in the endoplasmic reticulum (Fig. 2 b). Collectively, these findings reinforce the specialized role of B cells in orchestrating immune responses and proteinostasis within the HCC microenvironment. 2.3 Intercellular Communication Landscape of B Cells We used the CellChat package to map communication between immune cells. Compared with adjacent normal tissue, the tumor microenvironment showed more interactions overall, but the average interaction strength was lower (Fig. 2 c). Differential analysis of cellular cross-talk indicated that B cells in the TME demonstrated markedly higher interaction frequencies and intensities with the other four immune cell lineages compared to their counterparts in normal tissues (Fig. 2 d). Given the robust connectivity of B cells within the TME, we further dissected specific ligand-receptor pairs. Our results revealed that B-cell communication was primarily mediated via MHC-I and MHC-II signaling pathways. Specifically, B cells engaged with T cells predominantly through the HLA-E-CD8A axis, and with monocytes and DCs via the HLA-DRA-CD4 interaction. Furthermore, the B cell-NK cell crosstalk was characterized by the HLA-E-CD94/NKG2A signaling pair (Fig. 2 e). These findings suggest that B cells act as a central hub in orchestrating the immunological status of the HCC microenvironment. 2.4 Identification of B-cell-Associated Molecular Subtypes and Subtype-Specific Gene Modules Univariate Cox regression analysis was initially implemented to evaluate the prognostic significance of the B-cell markers, yielding 71 genes with robust associations with HCC survival ( P < 0.05). Based on these candidates, the ConsensusClusterPlus package with the PAM algorithm was implemented to partition the TCGA-LIHC cohort. Optimal clustering was achieved at k = 3, defining three distinct molecular subtypes (C1, C2, and C3) characterized by high internal consistency and stability (Fig. 3 a–c). The validity of this stratification was further corroborated by t-SNE analysis, which demonstrated clear transcriptomic separation among the three subgroups (Fig. 3 d). Survival analysis revealed significant prognostic heterogeneity; patients in the C1 and C3 groups exhibited superior overall survival and recurrence-free survival compared to those in the C2 group ( P < 0.05; Fig. 3 e, f). Notably, multivariate Cox regression analysis identified the C2 subtype as an independent risk factor for poor prognosis (Supplementary Fig. 2a). Subsequently, we performed WGCNA to extract the unique transcriptional signatures of each subtype. A scale-free co-expression network was constructed with a soft-thresholding power of β = 6, ensuring a scale-free topology fitting index > 0.9 (Fig. 3 g). Using the dynamic tree-cutting method and merging similar modules (Fig. 3 h), we identified 11 co-expression modules. Correlation analysis between modules and clinical traits revealed that the turquoise, blue, and purple modules were most significantly associated with the C1, C2, and C3 subtypes, respectively (Fig. 3 i–k), providing a basis for further biomarker identification. 2.5 External Validation of the B-cell-Associated Molecular Subtypes The stability and robustness of the B-cell-associated molecular subtypes were validated in the GSE14520 and ICGC-LIRI cohorts using NTP analysis. Subtype-specific feature genes were defined by intersecting the upregulated DEGs identified via limma with the core modules previously extracted from WGCNA. Prediction accuracy was strictly controlled by excluding samples with a FDR > 0.05. As anticipated, the three molecular subtypes were effectively recapitulated in both external cohorts, demonstrating high reproducibility of the classification system (Fig. 4 a, b). Prognostic evaluation further corroborated the clinical significance of this stratification. Consistent with the training set findings, the C2 subtype was significantly associated with unfavorable overall survival in both the GSE14520 and ICGC-LIRI datasets (Fig. 4 c. d). Subsequent multivariate Cox regression analysis established the C2 subtype as an independent prognostic indicator for poor clinical outcomes (Fig. 4 e, f). Collectively, these results underscore the high reliability and prognostic value of the B-cell-derived subtyping framework across diverse HCC populations. 2.6 Biological Underpinnings and Functional Landscapes of the Molecular Subtypes The distinct biological functionalities of the three molecular subtypes were elucidated through GSEA using the GO and KEGG repositories. The C1 subtype was predominantly characterized by the enrichment of metabolic and catabolic processes involving various endogenous and exogenous compounds, suggesting a state of metabolic homeostasis (Fig. 5 a). In contrast, the C2 subtype—which was previously associated with the poorest prognosis—exhibited a strong enrichment in oncogenic hallmarks, including the cell cycle, DNA replication, and mitotic cell division signaling pathways (Fig. 5 b). The C3 subtype, conversely, was significantly associated with robust immunological activities, such as immune cell activation, proliferation, and the elimination of aberrant cells, reflecting an "immuno-hot" microenvironment (Fig. 5 c). These divergent functional profiles provide a mechanistic basis for the observed differences in clinical outcomes among the three HCC subtypes. 2.7 Divergent Genomic Landscapes and Mutational Profiles Somatic mutation profiles within the TCGA-LIHC cohort were analyzed using the maftools package, revealing the top 20 most frequently mutated genes (Fig. 6 a). Among these, TP53 and CTNNB1 were identified as primary drivers, consistent with their established roles in HCC pathogenesis. Subtype-specific comparisons further highlighted distinct mutational signatures: the C1 subtype was characterized by a high frequency of CTNNB1 and TTN mutations, whereas TP53 mutations predominated in the C2 subtype (Fig. 6 b). Given that the progressive accumulation of somatic mutations drives tumorigenesis, the relatively lower mutational burden observed in the C3 group aligns with its superior prognosis identified in previous sections. Further investigation into CNV via the cBioPortal database identified significant differences in gene amplification and homozygous deletion across the three subtypes (Fig. 6 c). Patients in the C2 subgroup exhibited a substantially higher CNV burden compared to other groups. Notably, homozygous loss of CSMD1 —a recognized tumor suppressor whose deficiency promotes HCC progression—was prominently observed across all subtypes, with the most marked prevalence in the C2 group (Fig. 6 c). Collectively, these findings underscore that the C2 subtype is defined by extensive genomic alterations and high genomic instability, providing a molecular rationale for its aggressive clinical phenotype. 2.8 Characterization of Immune Infiltration and Assessment of Immunotherapy Response The relative abundance of 28 immune cell types within the TME was quantified using ssGSEA. The C3 subtype exhibited a markedly higher abundance of immune cell infiltration compared to the other two subgroups (Fig. 7 a). This "immuno-hot" phenotype in C3 patients was characterized by a dense infiltration of various immune effectors, including activated B cells, activated CD4 T cells, activated CD8 T cells, and immature B cells ( P < 0.05; Supplementary Fig. 2c). Further profiling of immune checkpoint molecules and HLA genes revealed significant transcriptomic divergence across the subtypes. Key checkpoints, such as CD27 , PDCD1 ( PD-1 ), CTLA4 , and LAG3 , as well as HLA family members, were significantly upregulated in the C3 subtype, suggesting heightened immune activation and enhanced antigen-presenting potential in these patients (Fig. 7 b, c; Supplementary Fig. 2b). The potential clinical response to immunotherapy was evaluated through a multi-algorithmic approach integrating TIS, IPS, and Submap analyses. Submap analysis indicated that patients in the C3 group were more likely to respond to anti-PD-1 therapy (Bonferroni-corrected P < 0.05; Fig. 7 d). Additionally, the significantly higher TIS scores observed in the C3 subtype suggested a greater likelihood of clinical benefit from immune checkpoint inhibitors (Fig. 7 e). IPS evaluation further corroborated these findings, demonstrating that C3 patients possessed superior antigen recognition and presentation capacities, higher effector cell activity, and reduced risks of immune suppression and escape (Fig. 7 f–i). Consistency of these observations was verified within the ICGC-LIRI cohort, where the C3 subtype's advantages in immune infiltration and therapeutic sensitivity were successfully recapitulated (Supplementary Fig. 2d–i). Collectively, these findings establish the C3 subtype as the optimal candidate for precision immunotherapy in HCC. 2.9 Pharmacogenomic Profiling and Identification of Candidate Small-Molecule Perturbagens The therapeutic sensitivity of HCC tumor cells was predicted using the oncoPredict package, where lower half-maximal inhibitory concentration (IC 50 ) values indicate heightened drug sensitivity. Significant differences in drug responsiveness were observed among the three molecular subtypes. Specifically, the C1 subtype demonstrated enhanced sensitivity to dihydrorotenone and sinularin (Fig. 8 a, b). In contrast, the C2 subtype exhibited more pronounced responses to lapatinib and sepantronium bromide (Fig. 8 c, d). Furthermore, patients in the C3 subgroup showed superior sensitivity to alisertib and olaparib (Fig. 8 e, f). To identify potential small-molecule compounds targeting each subtype, the unique DEGs of each group were utilized as input signatures for the CMap database. As illustrated in Fig. 8 g, the top candidate drugs for the C1 subtype included amsacrine, MEK1-2 inhibitor, GDC-0879, lestaurtinib, avrainvillamide-analog-5, ochratoxin A, and TPCA-1. For the C2 subtype, the analysis highlighted MDL-11939, GW-843682X, nocodazole, SA-792574, ABT-751, vinorelbine, MK-1775, and ALW-II-49-7 as prospective agents. Meanwhile, candidates for the C3 subtype comprised cimaterol, benperidol, gatifloxacin, importazole, atenolol, and MDL-11939. Mapping the targeted pathways of these candidate drugs provided further insights into the development of multi-drug or combination therapeutic strategies (Fig. 8 h). Collectively, these findings pinpoint subtype-specific therapeutic vulnerabilities and offer a blueprint for personalized treatment strategies in HCC. 2.10 Construction and Validation of the B-cell-associated Gene Signature A comprehensive machine learning framework integrating 10 distinct algorithms was employed to construct 117 prognostic models based on the expression profiles of the 71 B-cell-related prognostic genes. These algorithms included RSF, Enet, StepCox, CoxBoost, plsRcox, superpc, GBM, survivalsvm, Ridge, and Lasso. The C-index for each model was rigorously evaluated across the TCGA-LIHC discovery cohort and two independent external cohorts (ICGC-LIRI and GSE14520) (Fig. 9 a). The "StepCox [forward] + Ridge" combination emerged as the optimal integration, demonstrating the highest average C-index across all three datasets (Fig. 9 b). This integrative approach prioritized 21 core genes to establish the BCAGS. Each patient’s risk score was calculated, enabling their stratification into high- or low-risk groups. Kaplan-Meier survival analysis revealed that the high-risk group suffered from significantly inferior clinical outcomes compared to the low-risk group across all cohorts (Fig. 9 c–e). Time-dependent Receiver Operating Characteristic analysis was further implemented to evaluate the predictive accuracy of the BCAGS. In the TCGA, ICGC, and GSE cohorts, the Area Under the Curve values reached 0.779, 0.714, and 0.677 for 1-year survival; 0.746, 0.733, and 0.633 for 3-year survival; and 0.67, 0.761, and 0.58 for 5-year survival, respectively (Fig. 9 f–h), confirming the robust predictive performance and generalizability of the BCAGS. To further solidify the prognostic significance of the BCAGS, a meta-analysis based on univariate Cox regression was conducted, identifying the BCAGS as a consistent risk factor for HCC across all datasets (Fig. 10 a, b). Notably, the BCAGS demonstrated superior stability and predictive power when benchmarked against 37 previously published HCC prognostic models (Fig. 10 c). These findings highlight the potential of the BCAGS as a highly reliable tool for individual survival prediction in HCC patients. 2.11 Functional Validation of PDIA6 in HCC Cell Lines To further refine the signature and identify the most critical drivers of HCC progression, we implemented eight distinct machine learning algorithms to screen the 71 B-cell-associated prognostic genes. This integrative approach prioritized the top 10 feature genes based on their selection frequency across the various model combinations (Fig. 11 a). Among these, PDIA6 emerged as a top-ranking candidate with high stability. However, the specific functional role and underlying molecular mechanisms of PDIA6 in hepatocellular carcinoma have not yet been fully elucidated, necessitating further experimental investigation. To this end, Loss-of-function experiments were conducted in two human HCC cell lines (HepG-2 and Huh-7) to elucidate the biological role of PDIA6 in tumor progression. A marked decrease in PDIA6 gene expression at the mRNA level was observed following transfection with PDIA6 -specific siRNA, as determined by RT-qPCR (Fig. 11 b). Functional assessments using the CCK-8 assay demonstrated that silencing PDIA6 markedly suppressed the proliferative capacity of HepG-2 and Huh-7 cells, suggesting its essential role in promoting HCC cell growth (Fig. 11 c). Wound-healing assays further indicated that PDIA6 knockdown significantly impeded the migratory potential of HCC cells compared to the negative control groups (Fig. 11 d). Consistently, Transwell migration and Matrigel invasion assays revealed a substantial decrease in the number of migrated and invaded cells following PDIA6 depletion in both cell lines (Fig. 11 e). Collectively, these in vitro findings demonstrate that PDIA6 facilitates the proliferation, migration, and invasion of HCC cells, reinforcing its role as an oncogenic driver and validating the clinical significance of our computational modeling. Discussion The high heterogeneity of hepatocellular carcinoma (HCC) remains a formidable barrier to effective prognosis and precision therapy. While the dichotomy of "immune-hot" and "immune-cold" tumors has guided recent therapeutic strategies, the specific contribution of B cells—often overshadowed by T cells—to the HCC landscape remains underappreciated. Single-cell sequencing technologies have recently enabled the high-resolution characterization of such specific cell populations, providing insights that traditional bulk sequencing could not achieve 41 , 42 . In this study, we bridged single-cell resolution with bulk transcriptomics to deconstruct the B-cell-associated architecture of HCC. Our multi-omics approach not only delineated three clinically distinct molecular subtypes with unique genomic and varying immune profiles but also established a robust machine-learning-based prognostic signature, termed BCAGS. Crucially, we validated PDIA6 as a functional driver linking B-cell-associated signatures to tumor aggression, offering a novel translational bridge between computational stratification and biological intervention. Our scRNA-seq analysis provides compelling evidence that B cells are not merely bystanders but active orchestrators within the TME. Contrary to the classical view of T cell-dominated immunity, we observed that B cells in HCC exhibit enhanced intercellular communication, particularly through the MHC-I and MHC-II axes. The robust interaction pairs identified, such as HLA-E-CD8A and HLA-DRA-CD4, suggest that B cells may function as pivotal antigen-presenting cells. This aligns with emerging concepts that MHC-I expression initiates CD8 + T cell cytotoxicity 43 , while MHC-II expression on tumor or immune cells can recruit and prime CD4 + helper T cells to enhance the anti-tumor immune loop 44 . The enrichment of pathways related to immunoglobulin production and receptor signaling further implies that the humoral arm of immunity is actively engaged, albeit often suppressed, within the HCC milieu. A key translational output of our work is the identification of three stable molecular subtypes (C1, C2, C3), which effectively stratify patients based on biological distinctness rather than just clinical staging. The C3 subtype represents a classic "immune-hot" phenotype, characterized by dense infiltration of activated B and T cells and, paradoxically, elevated expression of exhaustion markers ( PD-1, CTLA-4 ). This expression pattern is the hallmark of patients who are most likely to respond to immune checkpoint blockade, as confirmed by our TIS, IPS, and Submap predictive analyses. In this context, the high expression of HLA molecules in C3 suggests that the machinery for antigen presentation is intact, making these tumors "visible" to the immune system once the checkpoint brakes are released. In stark contrast, the C2 subtype epitomizes the aggressive, "immune-cold" phenotype associated with poor clinical outcomes. Our genomic landscaping revealed that C2 is driven by profound genomic instability, highlighted by a high frequency of TP53 mutations and significant CNVs. Previous studies have established that TP53 mutations not only drive genomic instability but also foster an immunosuppressive microenvironment conducive to immune evasion, which explains the dismal prognosis observed in this subgroup 45 . Furthermore, the homozygous deletion of CSMD1 in this subtype parallels findings that link the loss of this tumor suppressor to HCC progression and chromosomal instability 46 . This stratification suggests that C2 patients might be refractory to standard immunotherapy and may instead require strategies targeting cell cycle dysregulation or specific metabolic vulnerabilities, such as the sensitivity to lapatinib identified in our drug screening. To translate these complex biological patterns into a clinical tool, we developed the BCAGS using an ensemble of 117 machine-learning algorithms. The superiority of the "StepCox [forward] + Ridge" model lies in its ability to distill high-dimensional data into a pragmatic risk score that remains robust across independent cohorts (TCGA, ICGC, GSE14520). Unlike single-biomarker approaches, the BCAGS integrates the multifaceted nature of B cell biology, serving as an independent prognostic factor that outperforms widespread clinical models. This score holds potential utility for oncologists to identify high-risk patients who may benefit from intensified adjuvant therapies. Finally, we moved beyond purely associative computational findings to provide experimental validation. Among the genes constituting our signature, PDIA6 emerged as a critical link between the B-cell-associated risk profile and tumor cell capability. PDIA6 serves as a key regulator of endoplasmic reticulum protein folding, and its upregulation has been implicated in the proliferation and invasion of various malignancies 47 , 48 . Consistent with these observations, our in vitro loss-of-function assays in HepG2 and Huh-7 lines unequivocally demonstrate its role in promoting proliferation, migration, and invasion in HCC. This suggests that PDIA6 may not only serve as a biomarker but also as a therapeutic target. While our in vitro data unequivocally pinpoint PDIA6 as a driver of HCC aggressiveness, we must acknowledge that Petri dishes cannot replicate the complex dialogue between tumor cells and the immune system. It's like examining the seed without the soil. A critical missing piece is verifying whether PDIA6 directly dictates the recruitment or exhaustion of B cells—a question our current cell line models simply cannot answer. Therefore, our next step is non-negotiable: we must move to immunocompetent mouse models. Only then can we confirm if targeting PDIA6 truly remodels the tumor microenvironment or merely suppresses tumor growth in isolation. Conclusions This study provides a comprehensive characterization of the B-cell-associated molecular landscape in hepatocellular carcinoma (HCC) through the integration of single-cell and bulk transcriptomic analyses. The identification of three distinct B-cell-related subtypes—C1 (metabolic), C2 (proliferative), and C3 (immuno-hot)—unveils the profound heterogeneity of the tumor microenvironment and its critical impact on clinical outcomes. Leveraging an ensemble of 117 machine learning combinations, the established BCAGS prognostic model demonstrates superior stability and predictive accuracy across multiple independent cohorts, outperforming existing clinical benchmarks. Moreover, experimental validation identifies PDIA6 as a novel oncogenic driver that facilitates HCC progression, reinforcing the clinical relevance of our computational findings. Collectively, these results offer a refined framework for prognostic stratification and provide actionable insights for personalized immunotherapy and targeted treatment strategies in HCC patients. Declarations Ethics approval and consent to participate The research was conducted on public datasets (TCGA, GEO, and ICGC) and established human cell lines. Ethical approval for the analysis of public data was waived by the local ethics committee. The human HCC cell lines (HepG-2 and Huh-7) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), and their use did not require additional ethical approval. Consent for publication Not applicable. Availability of data and materials The datasets generated and/or analyzed during the current study are available in the following repositories: GSE140228 (scRNA-seq data): Gene Expression Omnibus (GEO) database https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140228. TCGA-LIHC (Transcriptomic and clinical data): The Cancer Genome Atlas (TCGA) program via the GDC Data Portal https://portal.gdc.cancer.gov/. GSE14520 (Validation cohort): Gene Expression Omnibus (GEO) database https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14520. ICGC-LIRI-JP (Validation cohort): International Cancer Genome Consortium (ICGC) Data Portal https://dcc.icgc.org/projects/LIRI-JP. Publicly available portal cBioPortal https://www.cbioportal.org/ was used for CNV analysis, and Connectivity Map (CMap) https://clue.io/ was used for drug screening. Competing interests The authors declare that they have no competing interests. Funding The authors declare that no funds, grants, or other support were received during the preparation of this manuscript. Authors' contributions Yang Li and Sinan Cao contributed equally to this work and share first authorship. Yang Li performed the bioinformatics analyses, integrated multi-omics data, conducted statistical analyses, and drafted the manuscript. Sinan Cao performed the experimental validation, including cell culture, gene knockdown, and RT-qPCR assays, and contributed to data interpretation. Yamei Kuang assisted with data collection, data processing, and figure preparation. Dachuan Shen participated in result interpretation and manuscript revision. Lili Tian conceived and supervised the study, provided critical revisions to the manuscript, and approved the final version for submission. All authors read and approved the final manuscript. Acknowledgements We thank the contributors of the TCGA, GEO, and ICGC databases for sharing the sequencing data. References Sung H, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. Cancer J Clin. 2021;71:209–49. https://doi.org/10.3322/caac.21660 . Llovet JM, et al. Hepatocellular carcinoma. Nat Rev Dis Primers. 2021;7:6. https://doi.org/10.1038/s41572-020-00240-3 . Zheng R, et al. Liver cancer incidence and mortality in China: Temporal trends and projections to 2030. Chin J Cancer Res. 2018;30:571–9. https://doi.org/10.21147/j.issn.1000-9604.2018.06.01 . Beaufrère A, et al. Gene expression signature as a surrogate marker of microvascular invasion on routine hepatocellular carcinoma biopsies. J Hepatol. 2022;76:343–52. https://doi.org/10.1016/j.jhep.2021.09.034 . Chew V, et al. Delineation of an immunosuppressive gradient in hepatocellular carcinoma using high-dimensional proteomic and transcriptomic analyses. Proc Natl Acad Sci U S A. 2017;114:E5900–9. https://doi.org/10.1073/pnas.1706559114 . Zheng Y, et al. Immune suppressive landscape in the human esophageal squamous cell carcinoma microenvironment. Nat Commun. 2020;11:6268. https://doi.org/10.1038/s41467-020-20019-0 . Llovet JM, et al. Immunotherapies for hepatocellular carcinoma. Nat Rev Clin Oncol. 2022;19:151–72. https://doi.org/10.1038/s41571-021-00573-2 . Kurebayashi Y, et al. Landscape of immune microenvironment in hepatocellular carcinoma and its additional impact on histological and molecular classification. Hepatology. 2018;68:1025–41. https://doi.org/10.1002/hep.29904 . Sangro B, Sarobe P, Hervás-Stubbs S, Melero. I. Advances in immunotherapy for hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol. 2021;18:525–43. https://doi.org/10.1038/s41575-021-00438-0 . Cabrita R, et al. Author Correction: Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature. 2020;580:E1. https://doi.org/10.1038/s41586-020-2155-6 . Helmink BA, et al. B cells and tertiary lymphoid structures promote immunotherapy response. Nature. 2020;577:549–55. https://doi.org/10.1038/s41586-019-1922-8 . Garnelo M, et al. Interaction between tumour-infiltrating B cells and T cells controls the progression of hepatocellular carcinoma. Gut. 2017;66:342–51. https://doi.org/10.1136/gutjnl-2015-310814 . Zhang Q, et al. Landscape and Dynamics of Single Immune Cells in Hepatocellular Carcinoma. Cell. 2019;179:829–e845820. https://doi.org/10.1016/j.cell.2019.10.003 . Ma L, et al. Tumor Cell Biodiversity Drives Microenvironmental Reprogramming in Liver Cancer. Cancer Cell. 2019;36:418–e430416. https://doi.org/10.1016/j.ccell.2019.08.007 . Newman AM, et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol. 2019;37:773–82. https://doi.org/10.1038/s41587-019-0114-2 . Schumacher TN, Schreiber RD. Neoantigens in cancer immunotherapy. Science. 2015;348:69–74. https://doi.org/10.1126/science.aaa4971 . Ma Y et al. PDIA6 promotes pancreatic cancer progression and immune escape through CSN5-mediated deubiquitination of β-catenin and PD-L1. Neoplasia 23, 912–928 (2021). https://doi.org/10.1016/j.neo.2021.07.004 Bai Y, et al. PDIA6 modulates apoptosis and autophagy of non-small cell lung cancer cells via the MAP4K1/JNK signaling pathway. EBioMedicine. 2019;42:311–25. https://doi.org/10.1016/j.ebiom.2019.03.045 . Butler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol. 2018;36:411–20. https://doi.org/10.1038/nbt.4096 . Korsunsky I, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods. 2019;16:1289–96. https://doi.org/10.1038/s41592-019-0619-0 . Aran D, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol. 2019;20:163–72. https://doi.org/10.1038/s41590-018-0276-y . Xu S, et al. Using clusterProfiler to characterize multiomics data. Nat Protoc. 2024. https://doi.org/10.1038/s41596-024-01020-z . Jin S, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12:1088. https://doi.org/10.1038/s41467-021-21246-9 . Wilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinf (Oxford England). 2010;26:1572–3. https://doi.org/10.1093/bioinformatics/btq170 . Langfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. https://doi.org/10.1186/1471-2105-9-559 . Hoshida Y. Nearest template prediction: a single-sample-based flexible class prediction with confidence assessment. PLoS ONE. 2010;5:e15543. https://doi.org/10.1371/journal.pone.0015543 . Ritchie ME et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research 43, e47 (2015). https://doi.org/10.1093/nar/gkv007 Mayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28:1747–56. https://doi.org/10.1101/gr.239244.118 . Gao J et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. Science signaling 6, pl1 (2013). https://doi.org/10.1126/scisignal.2004088 Charoentong P, et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell Rep. 2017;18:248–62. https://doi.org/10.1016/j.celrep.2016.12.019 . Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7. https://doi.org/10.1186/1471-2105-14-7 . Shi JY, et al. Exploring prognostic indicators in the pathological images of hepatocellular carcinoma based on deep learning. Gut. 2021;70:951–61. https://doi.org/10.1136/gutjnl-2020-320930 . Gonzalez LC, et al. A coreceptor interaction between the CD28 and TNF receptor family members B and T lymphocyte attenuator and herpesvirus entry mediator. Proc Natl Acad Sci USA. 2005;102:1116–21. https://doi.org/10.1073/pnas.0409071102 . Liu DH, Mou FF, An M, Xia P. Human leukocyte antigen and tumor immunotherapy (Review). Int J Oncol. 2023;62. https://doi.org/10.3892/ijo.2023.5516 . Ayers M, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Investig. 2017;127:2930–40. https://doi.org/10.1172/jci91190 . Hoshida Y, Brunet JP, Tamayo P, Golub TR, Mesirov JP. Subclass mapping: identifying common subtypes in independent disease data sets. PLoS ONE. 2007;2:e1195. https://doi.org/10.1371/journal.pone.0001195 . Zeng D, et al. Multi-Omics Immuno-Oncology Biological Research to Decode Tumor Microenvironment and Signatures. Front Immunol. 2021;12:687975. https://doi.org/10.3389/fimmu.2021.687975 . Maeser D, Gruener RF, Huang RS. oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data. Brief Bioinform. 2021;22. https://doi.org/10.1093/bib/bbab260 . Lamb J, et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Sci (New York N Y). 2006;313:1929–35. https://doi.org/10.1126/science.1132939 . Liu H, et al. A flexible machine-learning framework to construct and visualize models for clinical characteristics prediction and feature selection. Comput Struct Biotechnol J. 2024;23:2798–810. https://doi.org/10.1016/j.csbj.2024.06.035 . Mime. Shaw R, Tian X, Xu J. Single-Cell Transcriptome Analysis in Plants: Advances and Challenges. Mol Plant. 2021;14:115–26. https://doi.org/10.1016/j.molp.2020.10.012 . Wang Y, et al. Changing Technologies of RNA Sequencing and Their Applications in Clinical Oncology. Front Oncol. 2020;10:447. https://doi.org/10.3389/fonc.2020.00447 . Raskov H, Orhan A, Christensen JP, Gögenur I. Cytotoxic CD8 + T cells in cancer and cancer immunotherapy. Br J Cancer. 2021;124:359–67. https://doi.org/10.1038/s41416-020-01048-4 . Axelrod ML, Cook RS, Johnson DB, Balko JM. Biological Consequences of MHC-II Expression by Tumor Cells in Cancer. Clin cancer research: official J Am Association Cancer Res. 2019;25:2392–402. https://doi.org/10.1158/1078-0432.Ccr-18-3200 . Ko KP, et al. Key Genetic Determinants Driving Esophageal Squamous Cell Carcinoma Initiation and Immune Evasion. Gastroenterology. 2023;165:613–e628620. https://doi.org/10.1053/j.gastro.2023.05.030 . Midorikawa Y, et al. Allelic imbalances and homozygous deletion on 8p23.2 for stepwise progression of hepatocarcinogenesis. Hepatology (Baltimore MD). 2009;49:513–22. https://doi.org/10.1002/hep.22698 . Wang P, et al. PDIA6, which is regulated by TRPM2-AS/miR-424-5p axis, promotes endometrial cancer progression via TGF-beta pathway. Cell Death Dis. 2023;14:829. https://doi.org/10.1038/s41419-023-06297-8 . Xiang P, et al. circSETD3 confers radiotherapy resistance in nasopharyngeal carcinoma by attenuating ER stress-induced autophagy and apoptosis via PDIA6 upregulation. Oncogene. 2025. https://doi.org/10.1038/s41388-025-03652-1 . Tables Tables 1 and 2 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table1.xlsx Table2.xlsx SupplementaryTable2.xlsx SupplementaryTable1.xlsx SupplementaryTable4.xlsx Supplementarytable5.xlsx SupplementaryTable3.xlsx SupplementaryFigure1.tif SupplementaryFigure2.tif SupplementaryFigurelegends.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9213778","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":611477417,"identity":"ee6a9f0e-3b2c-49e6-93c0-1f84c8cdabd9","order_by":0,"name":"Yang Li","email":"","orcid":"","institution":"The First Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yang","middleName":"","lastName":"Li","suffix":""},{"id":611477418,"identity":"2e10d40b-be88-454a-8e14-ca9b7d348d85","order_by":1,"name":"Sinan Cao","email":"","orcid":"","institution":"Henan University People's Hospital, Zhengzhou University People's Hospital)","correspondingAuthor":false,"prefix":"","firstName":"Sinan","middleName":"","lastName":"Cao","suffix":""},{"id":611477419,"identity":"0b9f179c-7672-4b0e-9d92-d027bdc49879","order_by":2,"name":"Yamei Kuang","email":"","orcid":"","institution":"Henan University People's Hospital, Zhengzhou University People's Hospital)","correspondingAuthor":false,"prefix":"","firstName":"Yamei","middleName":"","lastName":"Kuang","suffix":""},{"id":611477420,"identity":"123ff259-c3d7-470e-ba53-8e90f1a7a730","order_by":3,"name":"Dachuan Shen","email":"","orcid":"","institution":"The First Hospital of Dalian Medical University","correspondingAuthor":false,"prefix":"","firstName":"Dachuan","middleName":"","lastName":"Shen","suffix":""},{"id":611477421,"identity":"ac6d4a7e-a89e-4071-9dfb-2be51ccc7797","order_by":4,"name":"Lili Tian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4klEQVRIiWNgGAWjYDACZhBhAGIwH3zwoUJCjp+QDh64Fna2ZMMZZyyMJRsIaYGz+HnMpDnbKhI3ENJiz8787OGPgjuJ/cw8xsaM8yQYNzAwP3x0A6/D2MyNeQyeJc5sZit8XLhNgtmcgc3YOAe/X8ykGQwOJ244zLzZeOY2CTbLBh42afxa2L9J/gBrAerlnSPBY3CAoBYeM6AykBYWoJYGCQnCWg7zlEkDtRgD/QIM5GMSBpLNBPzC3n98m+SPP4dl+/kPA6Oypq6+n7354WN8WrAAZtKUj4JRMApGwSjAAgA9/kO+v3tATQAAAABJRU5ErkJggg==","orcid":"","institution":"The First Hospital of Dalian Medical University","correspondingAuthor":true,"prefix":"","firstName":"Lili","middleName":"","lastName":"Tian","suffix":""}],"badges":[],"createdAt":"2026-03-24 15:11:49","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9213778/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9213778/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105465921,"identity":"0d9c5f38-4846-49a8-b7bd-bfd627210f3a","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":440213,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-cell landscape and distribution of immune cells in hepatocellular carcinoma (HCC)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, \u003c/strong\u003et-SNE visualization of 3,817 immune cells from GSE140228, clustered into 17 distinct groups. \u003cstrong\u003eb,\u003c/strong\u003eDistribution of immune cells originating from tumor (orange) and normal (blue) tissues. \u003cstrong\u003ec,\u003c/strong\u003e Annotation of five major immune cell types based on canonical markers: T cells (\u003cem\u003eCD3D, CD3E\u003c/em\u003e), monocytes (\u003cem\u003eCD68, CD163\u003c/em\u003e), NK cells (\u003cem\u003eKLRC1, NCR1\u003c/em\u003e), B cells (\u003cem\u003eCD19, CD79A\u003c/em\u003e), and DCs (\u003cem\u003eCLEC9A, XCR1\u003c/em\u003e). \u003cstrong\u003ed,\u003c/strong\u003e Bar plot comparing the proportions of the five immune cell types between tumor and normal tissues. \u003cstrong\u003ee,\u003c/strong\u003e Sample-specific cell proportions and absolute cell counts across the 11 included samples. \u003cstrong\u003ef,\u003c/strong\u003e Tissue preference of immune cell types calculated by odds ratio (OR), showing B cell enrichment in tumor tissues. \u003cstrong\u003eg,\u003c/strong\u003e Dot plot displaying the top 3 marker genes for each immune cell lineage.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/671ea3faaf73591a9371c44a.png"},{"id":105465923,"identity":"2e59f5bb-2f97-41cb-a36b-68a8a623f32a","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":354692,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eBiological functions and cell-cell communication of tumor-infiltrating B cells\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, b,\u003c/strong\u003e GO (A) and KEGG (B) enrichment analysis of B cell marker genes, highlighting pathways involved in immunoglobulin production and B-cell receptor signaling. \u003cstrong\u003ec,\u003c/strong\u003e Comparison of the total number and strength of cell-cell interactions between normal and tumor tissues. \u003cstrong\u003ed,\u003c/strong\u003e Differential interaction networks showing increased communication between B cells and other immune cells in HCC. \u003cstrong\u003ee,\u003c/strong\u003eSignificant ligand-receptor pairs mediating communication between B cells and T cells, NK cells, Monocytes, and DCs, predominantly via MHC-I and MHC-II signaling pathways.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/eab06224c2ee17c41677d957.png"},{"id":105465934,"identity":"40f54567-8fa4-4608-abc8-d3a5817f9b08","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":569708,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eIdentification and characterization of B-cell-associated molecular subtypes in HCC\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea–c,\u003c/strong\u003e Consensus clustering analysis of the TCGA-LIHC cohort based on 71 B-cell-associated prognostic genes, showing optimal stability at \u003cem\u003ek\u003c/em\u003e = 3. \u003cstrong\u003ed,\u003c/strong\u003e t-SNE plot validating the separation of three molecular subtypes (C1, C2, and C3). \u003cstrong\u003ee, f,\u003c/strong\u003eKaplan-Meier curves for overall survival \u003cstrong\u003e(e)\u003c/strong\u003e and recurrence-free survival \u003cstrong\u003e(f)\u003c/strong\u003e among the three subtypes. \u003cstrong\u003eg,\u003c/strong\u003e Network topology analysis for various soft-thresholding powers in WGCNA. \u003cstrong\u003eh,\u003c/strong\u003e Heatmap of the correlation between gene modules and molecular subtypes. \u003cstrong\u003ei–k,\u003c/strong\u003e Scatter plots showing the correlation between module membership and gene significance for turquoise (C1), blue (C2), and purple (C3) modules.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/9b9af3b9045ed8727e63e190.png"},{"id":105566130,"identity":"d9a83cef-a540-4591-838b-f13c2132c7e5","added_by":"auto","created_at":"2026-03-27 12:55:24","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":988079,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExternal validation of B-cell-associated molecular subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, b,\u003c/strong\u003e Nearest Template Prediction (NTP) heatmaps validating the three molecular subtypes in the GSE14520 \u003cstrong\u003e(a)\u003c/strong\u003e and ICGC-LIRI \u003cstrong\u003e(b)\u003c/strong\u003e cohorts. \u003cstrong\u003ec, d,\u003c/strong\u003e Kaplan-Meier survival analysis confirming the significantly worse prognosis of the C2 subtype in GSE14520 \u003cstrong\u003e(c)\u003c/strong\u003e and ICGC-LIRI \u003cstrong\u003e(d)\u003c/strong\u003e cohorts. \u003cstrong\u003ee, f,\u003c/strong\u003eMultivariate Cox regression analysis identifying the C2 subtype as an independent risk factor for OS in external validation cohorts.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/efce84b7067d8ce422c8681c.png"},{"id":105465927,"identity":"9b9c586a-7e2c-4c15-b8be-ec11e3e6cb79","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":523772,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistinct biological processes and pathways associated with molecular subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea–c,\u003c/strong\u003e Gene Set Enrichment Analysis (GSEA) showing representative GO terms and KEGG pathways enriched in C1 (\u003cstrong\u003ea,\u003c/strong\u003emetabolism-related), C2 (\u003cstrong\u003eb,\u003c/strong\u003e cell cycle and DNA replication), and C3 (\u003cstrong\u003ec,\u003c/strong\u003eimmune activation and T cell proliferation) subtypes.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/3501f6d150df85e395d04213.png"},{"id":105465925,"identity":"db96f444-ddc3-4fa9-8635-67e02f873f76","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":452676,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGenomic landscape and somatic mutation profiles across subtypes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, \u003c/strong\u003eOncoplot of the top 20 most frequently mutated genes in the TCGA-LIHC cohort. \u003cstrong\u003eb,\u003c/strong\u003e Comparison of mutation frequencies among C1, C2, and C3 subtypes, showing high \u003cem\u003eTP53\u003c/em\u003emutation in C2. \u003cstrong\u003ec,\u003c/strong\u003e Comparison of copy number variations (CNV), including gene amplification and homozygous deletion, highlighting the higher genomic instability in the C2 subtype.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/5a670e5446213269241728a5.png"},{"id":105465939,"identity":"8bd867fb-d2d5-4d35-9e83-5388cd55a393","added_by":"auto","created_at":"2026-03-26 10:49:56","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":632980,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eEvaluation of immune infiltration and immunotherapy response\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Heatmap displaying the relative abundance of 28 immune cell types across subtypes calculated by ssGSEA. \u003cstrong\u003eb,\u003c/strong\u003e Expression profiles of 27 immune checkpoint molecules, with C3 showing higher levels of \u003cem\u003ePDCD1\u003c/em\u003e and \u003cem\u003eCTLA4\u003c/em\u003e. \u003cstrong\u003ec,\u003c/strong\u003e Expression of human leukocyte antigen (HLA) molecules across subtypes. \u003cstrong\u003ed,\u003c/strong\u003e Submap analysis predicting the response to anti-PD-1 and anti-CTLA-4 therapy. \u003cstrong\u003ee,\u003c/strong\u003eComparison of T-cell inflammation (TIS) scores. \u003cstrong\u003ef–i,\u003c/strong\u003e Immunophenoscore (IPS) analysis (MHC, EC, SC, and CP scores) evaluating the immunotherapeutic potential of each subtype.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/0927e6a9198db97442020601.png"},{"id":105465931,"identity":"68215e7f-8815-410f-ab41-56a2100f3476","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":308247,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDrug sensitivity analysis and screening of candidate small molecules\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea–f,\u003c/strong\u003e Box plots showing the predicted IC\u003csub\u003e50\u003c/sub\u003e values for subtype-specific drugs (e.g., Alisertib and Olaparib for C3). \u003cstrong\u003eg,\u003c/strong\u003e Heatmap of candidate small molecules for each subtype identified via Connectivity Map (CMap) analysis. \u003cstrong\u003eh,\u003c/strong\u003e Target pathways and mechanisms of action for the identified candidate small molecules.\u003c/p\u003e","description":"","filename":"Figure8.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/fafc57f8daad9a8089a2d617.png"},{"id":105566035,"identity":"268529ef-26c5-4b32-bf32-b872b41fdb60","added_by":"auto","created_at":"2026-03-27 12:55:07","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":640120,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDevelopment and prognostic evaluation of the B-cell-associated gene signature (BCAGS)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea,\u003c/strong\u003e Performance (C-index) of 117 machine learning algorithm combinations in training and validation cohorts. \u003cstrong\u003eb,\u003c/strong\u003e C-index of the \"StepCox + Ridge\" model across three cohorts. \u003cstrong\u003ec–e,\u003c/strong\u003eKaplan-Meier curves for OS between high- and low-BCAGS groups in TCGA-LIHC \u003cstrong\u003e(c)\u003c/strong\u003e, ICGC-LIRI \u003cstrong\u003e(d)\u003c/strong\u003e, and GSE14520 \u003cstrong\u003e(e)\u003c/strong\u003e cohorts. \u003cstrong\u003ef–h,\u003c/strong\u003e Time-dependent Receiver Operating Characteristic curves predicting 1-, 3-, and 5-year OS in the three cohorts.\u003c/p\u003e","description":"","filename":"Figure9.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/2d59912185845eb2007e2317.png"},{"id":105465933,"identity":"8e2f1d15-452b-4bc0-b0a2-4319bc2393d7","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":457542,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRobustness and comparative analysis of the BCAGS model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, \u003c/strong\u003eMeta-analysis of univariate Cox regression across multiple cohorts identifying BCAGS as a consistent risk factor. \u003cstrong\u003eb,\u003c/strong\u003e Univariate Cox regression comparison of BCAGS with other clinical parameters. \u003cstrong\u003ec,\u003c/strong\u003e Comparison of the C-index between BCAGS and 37 previously published HCC prognostic models.\u003c/p\u003e","description":"","filename":"Figure10.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/b984047c5ac65ed0900a65a9.png"},{"id":105465937,"identity":"e49dbf08-3cb9-42bc-9c17-11a2c949625c","added_by":"auto","created_at":"2026-03-26 10:49:56","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":507428,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003ePDIA6\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e promotes HCC cell proliferation, migration, and invasion in vitro\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ea, \u003c/strong\u003eSelection frequency of the top 10 feature genes across eight machine learning algorithms, identifying \u003cem\u003ePDIA6\u003c/em\u003eas a core prognostic driver. \u003cstrong\u003eb,\u003c/strong\u003e RT-qPCR analysis showing that transfection with \u003cem\u003ePDIA6\u003c/em\u003e-specific siRNA (si-\u003cem\u003ePDIA6\u003c/em\u003e) markedly reduced \u003cem\u003ePDIA6\u003c/em\u003e mRNA expression in HepG-2 and Huh-7 cells compared with the negative control (si-NC). \u003cstrong\u003ec,\u003c/strong\u003e CCK-8 assays demonstrating that silencing \u003cem\u003ePDIA6\u003c/em\u003e significantly inhibited the proliferation of HepG-2 and Huh-7 cells. \u003cstrong\u003ed,\u003c/strong\u003e Wound-healing assays revealing that \u003cem\u003ePDIA6\u003c/em\u003eknockdown markedly impaired the migratory ability of hepatocellular carcinoma cells; representative images displayed delayed wound closure with wider gaps in the si-\u003cem\u003ePDIA6\u003c/em\u003e group, while quantitative analysis confirmed a significantly lower wound closure rate compared to the control. \u003cstrong\u003ee,\u003c/strong\u003eTranswell migration and invasion assays further indicating that the numbers of migrating and invading cells were significantly reduced after \u003cem\u003ePDIA6\u003c/em\u003esilencing in both HepG-2 and Huh-7 cell lines. Data are presented as mean ± SD from three independent experiments. Statistical significance was assessed using Student’s t-test (ns, not significant; *\u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05, **\u003cem\u003eP \u003c/em\u003e\u0026lt; 0.01, ***\u003cem\u003eP \u003c/em\u003e\u0026lt; 0.001).\u003c/p\u003e","description":"","filename":"Figure11.png","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/57db8c2e03e7958bddf092c2.png"},{"id":105842948,"identity":"4eaf7d27-2652-4b25-abc7-f5d62bcc84f1","added_by":"auto","created_at":"2026-03-31 17:10:57","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6805271,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/5aacf1cb-b235-4b3b-8766-07b437af8b28.pdf"},{"id":105566034,"identity":"e4ac222a-81f4-43ab-b3dc-7c75ef0a96e8","added_by":"auto","created_at":"2026-03-27 12:55:07","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":9979,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/974ff980e7af98a0962c5c65.xlsx"},{"id":105566832,"identity":"b3130aee-758d-4ec2-bebc-3a514f653ddf","added_by":"auto","created_at":"2026-03-27 12:57:29","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10069,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/5820a5d5cfbbe62679eea7bc.xlsx"},{"id":105465924,"identity":"9b6f972c-c019-4f17-8ab2-bbb95c4405ce","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":25644,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/7678f3d4d9530c1b1e2acabd.xlsx"},{"id":105565678,"identity":"68db3676-b4b9-43a4-9842-f482b5341fe3","added_by":"auto","created_at":"2026-03-27 12:54:01","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":11720,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/4447f0d9c9226ea996cf72ef.xlsx"},{"id":105728225,"identity":"a9191f27-9699-4713-ae21-fe9804fc14f4","added_by":"auto","created_at":"2026-03-30 11:10:59","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":20331,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/3ae7dd1523766a5573494a46.xlsx"},{"id":105566651,"identity":"77c8f44d-196c-491b-8fcd-40f824e7aa63","added_by":"auto","created_at":"2026-03-27 12:56:53","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":14817,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarytable5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/cc0102416a04ff5d490cc909.xlsx"},{"id":105566258,"identity":"0be4364b-e702-45ba-9ac6-976dd9c5a940","added_by":"auto","created_at":"2026-03-27 12:55:55","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":9705,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable3.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/b2aecc0924f7805735086316.xlsx"},{"id":105465936,"identity":"9b287a37-536e-4774-a058-47fa2c1c21d9","added_by":"auto","created_at":"2026-03-26 10:49:56","extension":"tif","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":994836,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure1.tif","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/322d6a0246ce12541d00e32b.tif"},{"id":105465938,"identity":"f8c5a912-00ca-41f2-8726-aa1650bd7a1f","added_by":"auto","created_at":"2026-03-26 10:49:56","extension":"tif","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":2661452,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigure2.tif","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/6637651ef1a2e90b2b5a596e.tif"},{"id":105465932,"identity":"8a90902e-3b98-472f-b3fc-866d6e8e4ece","added_by":"auto","created_at":"2026-03-26 10:49:55","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":15799,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigurelegends.docx","url":"https://assets-eu.researchsquare.com/files/rs-9213778/v1/b29ffc69ac0ed91a86b25841.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Breaking Therapeutic Resistance in Hepatocellular Carcinoma: Systems Biology Deconvolution of the B-cell Microenvironment Identifies PDIA6 as a Novel Target","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHepatocellular carcinoma (HCC) remains a leading contributor to global cancer-related mortality, consistently ranking among the top three most aggressive malignancies \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Although multimodal strategies\u0026mdash;integrating surgical resection, interventional therapies, and molecularly targeted agents\u0026mdash;have markedly improved short-term patient survival, the profound intra- and inter-tumor heterogeneity of HCC continues to pose a formidable barrier to the realization of precision medicine \u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Clinical evidence reveals that patients even within the same TNM stage often exhibit strikingly divergent clinical outcomes following standardized treatment \u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u003c/sup\u003e. Such heterogeneity underscores the limitations of traditional clinicopathological staging in capturing the complex biological landscape of HCC. Consequently, there is an urgent need to leverage deep molecular profiling to identify more precise prognostic signatures and actionable therapeutic vulnerabilities \u003csup\u003e\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe dynamic evolution of the tumor microenvironment (TME) serves as a pivotal determinant of HCC progression \u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Historically, investigations into the immune landscape of HCC have predominantly centered on the orchestration of T-cell exhaustion and activation \u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e, whereas the functional contribution of B cells\u0026mdash;the cornerstone of humoral immunity\u0026mdash;within the TME has long remained marginalized. Emerging evidence indicates a paradigm shift: tumor-infiltrating B cells do more than merely mediate anti-tumor effects via antibody secretion; they actively reshape T-cell-mediated immunity by facilitating the assembly of tertiary lymphoid structures \u003csup\u003e\u003cspan additionalcitationids=\"CR11\" citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. Nevertheless, the extent to which B-cell-associated molecular signatures define distinct HCC clinical subtypes, and how this heterogeneity dictates the therapeutic efficacy of immune checkpoint inhibitors, has yet to be systematically quantified.\u003c/p\u003e \u003cp\u003eThe rapid maturation of single-cell RNA sequencing (scRNA-seq) technologies has opened an unprecedented window into the granular deconvolution of intratumoral heterogeneity at single-cell resolution \u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e. Nevertheless, the clinical translation of single-cell findings is often constrained by the limited scale of available cohorts, which hampers the direct derivation of robust predictive models. To bridge this gap, the present study employed a \"cross-resolution integration\" strategy. By first deciphering the distinct B-cell molecular signatures via scRNA-seq and subsequently projecting these features onto expansive bulk transcriptomic cohorts, we systematically identified and validated three B-cell-associated molecular subtypes characterized by significantly divergent clinical outcomes \u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eTo bolster the clinical utility and robustness of our findings, we implemented an integrative ensemble screening strategy to develop a refined B-cell-associated gene signature (BCAGS). This scoring system not only demonstrated superior predictive accuracy across multiple independent, cross-platform cohorts but also highlighted a compelling link between high BCAGS scores and underlying genomic instability \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Furthermore, to translate these bioinformatic insights into mechanistic understanding, we performed extensive functional validation of \u003cem\u003ePDIA6\u003c/em\u003e\u0026mdash;a core driver gene within the signature\u0026mdash;thereby reconciling computational predictions with biological reality \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Collectively, our study enriches the current understanding of immune microenvironment heterogeneity in HCC and provides a robust framework for precision patient stratification and the optimization of personalized immunotherapeutic interventions.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e1.1 Data Acquisition and Preprocessing\u003c/h2\u003e \u003cp\u003eSingle-cell RNA sequencing (scRNA-seq) data for the GSE140228 dataset, which comprises six tumor and five adjacent normal samples, were retrieved from the Gene Expression Omnibus (GEO) database. For large-scale genomic analysis, transcriptomic profiles, single-nucleotide variant data, and corresponding clinical annotations for the TCGA-LIHC cohort (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;369) were curated from The Cancer Genome Atlas (TCGA). Additionally, two independent validation cohorts were integrated: the GSE14520 dataset (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;221) from GEO and the ICGC-LIRI-JP cohort (\u003cem\u003en\u0026thinsp;=\u003c/em\u003e\u0026thinsp;232) from the International Cancer Genome Consortium (ICGC) portal. To ensure inter-platform comparability, gene expression matrices were normalized and log2-transformed where appropriate. Comprehensive metadata and characteristics for all included cohorts are detailed in Supplementary Table\u0026nbsp;1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e1.2 scRNA-seq Data Analysis\u003c/h2\u003e \u003cp\u003eGSE140228 scRNA-seq data were analyzed using Seurat (v4.2) \u003csup\u003e19\u003c/sup\u003e. Quality control retained genes expressed in \u0026ge;\u0026thinsp;3 cells and cells with \u0026ge;\u0026thinsp;200 genes and \u0026le;\u0026thinsp;10% mitochondrial content, yielding 3,817 immune cells. After normalization, 2,000 highly variable genes were identified for Principal Component Analysis (PCA). Batch effects were corrected using Harmony \u003csup\u003e\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e, with cell clusters visualized via t-distributed stochastic neighbor embedding (t-SNE). Cluster-specific markers were identified via FindAllMarkers (Wilcoxon rank-sum test), and cell types were annotated using SingleR \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e and canonical lineage markers. clusterProfiler \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e was used for B-cell Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses. Intercellular communication and ligand-receptor interactions were inferred via CellChat \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e1.3 Identification of B-cell-associated Molecular Subtypes in HCC\u003c/h2\u003e \u003cp\u003eTo delineate the clinical relevance of B-cell heterogeneity, we first identified the marker genes for each B-cell subcluster using the FindAllMarkers function. Subsequently, a univariate Cox proportional hazards regression analysis was conducted to screen for genes significantly associated with the overall survival (OS) of HCC patients, yielding 71 prognostic candidates (Supplementary Table\u0026nbsp;5). These genes were then utilized as the basis for unsupervised molecular subtyping. Based on the ConsensusClusterPlus R package, consensus clustering was performed using the Partitioning Around Medoids (PAM) algorithm with 1,000 permutations to ensure the stability of the classification \u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e. The optimal number of clusters (\u003cem\u003ek\u003c/em\u003e) was determined by the cumulative distribution function curve and the tracking plot. To further evaluate the robustness and clinical significance of the identified subtypes, Kaplan-Meier survival analysis was employed to compare prognostic outcomes, while t-SNE was utilized to visualize the transcriptional distinctness between the clusters.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e1.4 Weighted Gene Co-expression Network Analysis (WGCNA)\u003c/h2\u003e \u003cp\u003eIn order to identify key gene modules and hub genes characteristic of the three identified HCC subtypes, a weighted gene co-expression network was constructed using the WGCNA R package \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e. First, an adjacency matrix was calculated based on the Pearson\u0026rsquo;s correlation between gene expression profiles. To achieve a scale-free topology, we evaluated a range of soft-thresholding powers (\u003cem\u003eβ\u003c/em\u003e); a power of \u003cem\u003eβ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;6 was selected, as it was the lowest power that allowed the scale-free topology fit index (\u003cem\u003eR\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e) to reach 0.9. Subsequently, the adjacency matrix was transformed into a Topological Overlap Matrix to estimate the network connectivity. We utilized a hierarchical clustering dendrogram for module detection via the dynamic tree cut method, with the minimum module size set to 40. Finally, highly similar modules were consolidated using the mergeCloseModules function with a height cut-off of 0.25 to yield the final co-expression modules for downstream module-trait correlation analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e1.5 Subtype Prediction via Nearest Template Prediction (NTP)\u003c/h2\u003e \u003cp\u003eWe employed a Nearest Template Prediction (NTP) algorithm to facilitate the clinical translation and cross-platform validation of our B-cell-associated molecular subtypes \u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e. Unlike traditional clustering methods, the NTP approach allows for the class prediction of individual samples with high flexibility by utilizing a predefined list of signature genes. In this study, the top-ranked marker genes identified from each subtype were used to construct the prediction templates. For each sample in the validation cohorts (TCGA and ICGC), a proximity-based matching was performed against these templates using a cosine distance metric. To ensure the reliability of the classification, a permutation-based \u003cem\u003eP\u003c/em\u003e-value was calculated for each prediction, with a threshold of \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and a False Discovery Rate (FDR)\u0026thinsp;\u0026lt;\u0026thinsp;0.1 applied to define high-confidence assignments.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e1.6 Functional Enrichment and Pathway Analysis\u003c/h2\u003e \u003cp\u003eTo unravel the biological motifs and signaling pathways underlying the discrete B-cell-associated subtypes, we performed differential expression analysis between each subtype and the remaining cohorts using the limma R package \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e. For each subtype-specific profile, genes were sorted in descending order based on their log2-fold change (log2FC) to generate a pre-ranked gene list. Subsequently, Gene Set Enrichment Analysis (GSEA) was executed via the clusterProfiler package \u003csup\u003e\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e using the GO and KEGG databases as reference sets. A permutation-based approach with 1,000 iterations was applied to estimate the significance of the enrichment. Biological pathways with a normalized enrichment score (NES) absolute value\u0026thinsp;\u0026gt;\u0026thinsp;1.0 and a Benjamini-Hochberg adjusted \u003cem\u003eP\u003c/em\u003e-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered significantly enriched.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e1.7 Characterization of Genomic Alterations: Somatic Mutations and Copy Number Variations\u003c/h2\u003e \u003cp\u003eIn order to delineate the genomic landscape across different B-cell-associated subtypes, somatic mutation data were processed and visualized using the maftools R package \u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e. Mutation Annotation Format files were integrated via the read.maf function, and the overall mutational burden, including variant classifications and transition/transversion ratios, was summarized using the plotmafSummary module. To identify and compare the frequently mutated genes among the subtypes, waterfall plots (oncoplots) were generated to illustrate the mutation frequencies and co-occurrence patterns of top-tier drivers. Furthermore, copy number variation (CNV) profiles were explored using the cBioPortal for Cancer Genomics database \u003csup\u003e\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e\u003c/sup\u003e. We specifically quantified focal genomic events, focusing on the frequency of high-level amplifications and homozygous deletions for the top 10 most frequently altered genes.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e1.8 Characterization of the Tumor Immune Microenvironment (TIME) and Immunotherapy Response\u003c/h2\u003e \u003cp\u003eTo characterize the immune landscape, relative infiltration levels of 28 immune cell subpopulations \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e (Supplementary Table\u0026nbsp;2) were quantified using the single-sample Gene Set Enrichment Analysis (ssGSEA) algorithm via the GSVA R package \u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e. Additionally, the transcriptional profiles of 27 immune checkpoints and 9 Human Leukocyte Antigen (HLA) genes were systematically assessed \u003csup\u003e\u003cspan additionalcitationids=\"CR33\" citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e (Supplementary Table\u0026nbsp;3). To predict sensitivity to immune checkpoint inhibitors, three established frameworks were implemented. First, the T-cell-inflamed signature (TIS) score, comprising 18 inflammatory genes, was derived using ssGSEA \u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Second, Subclass mapping (Submap) analysis was utilized to infer the likelihood of clinical response to anti-PD-1 and anti-CTLA-4 therapies \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e. Finally, the Immunophenoscore (IPS) was calculated via the IOBR package \u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e, integrating four functional components: antigen presentation (MHC), effector cells (EC), suppressor cells (SC), and checkpoints (CP) \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e1.9 Drug Sensitivity and Connectivity Map (CMap) Analysis\u003c/h2\u003e \u003cp\u003eTo identify potential therapeutic agents, the oncoPredict R package \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e was employed to estimate the half-maximal inhibitory concentration (IC\u003csub\u003e50\u003c/sub\u003e) of various chemotherapeutic and targeted drugs across HCC subtypes. Concurrently, the Connectivity Map (CMap) database was utilized to screen for candidate small-molecule compounds \u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u003c/sup\u003e. Differentially expressed genes (DEGs) between subtypes were identified via limma, and the top 150 up-regulated and 150 down-regulated genes were queried through the CMap L1000 platform. Enrichment scores were calculated to identify therapeutic perturbations associated with the B-cell-related molecular signatures.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e1.10 Machine Learning-Based Prognostic Signature Development\u003c/h2\u003e \u003cp\u003eTo construct a robust prognostic signature, 117 machine learning combinations were implemented using the Mime R package, integrating ten distinct algorithms: random survival forest (RSF), elastic net (Enet), stepwise Cox, CoxBoost, partial least squares regression for Cox (plsRcox), supervised principal components (superpc), generalized boosted models (GBM), survival support vector machine (survivalsvm), Ridge, and Lasso \u003csup\u003e\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e. Models were optimized via K-fold cross-validation within the training cohort. The predictive accuracy of each combination was quantified by the Harrell\u0026rsquo;s concordance index (C-index) across both training and independent validation datasets. The algorithm combination yielding the highest mean C-index was selected for the final model construction. Furthermore, the prognostic performance of our signature was benchmarked against 37 previously published HCC models to evaluate its comparative superiority (Supplementary Table\u0026nbsp;4).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e1.11 Cell Culture and siRNA Transfection\u003c/h2\u003e \u003cp\u003eHuman HCC cell lines HepG-2 and Huh-7 (Cell Bank of the Chinese Academy of Sciences, Shanghai, China) were cultured in DMEM and RPMI-1640 media, respectively, each supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin. All cultures were maintained at 37\u0026deg;C in a humidified incubator with 5% CO\u003csub\u003e2\u003c/sub\u003e. For gene knockdown, specific siRNAs targeting \u003cem\u003ePDIA6\u003c/em\u003e and negative control (NC) sequences (Shanghai Hanhui Biotech) were transfected into cells using Lipofectamine\u0026trade; RNAiMAX (Invitrogen, USA) following the manufacturer\u0026rsquo;s protocol. Transfection efficiency was assessed for subsequent functional assays. siRNA sequences are shown in Table\u0026nbsp;1.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e1.12 Quantitative Real-Time PCR\u003c/h2\u003e \u003cp\u003eTotal RNA was extracted from cells using the TRIzol\u0026trade; Reagent (Invitrogen, Carlsbad, CA, USA) and subsequently reverse-transcribed into cDNA utilizing the PrimeScript\u0026trade; RT reagent Kit (Takara Bio, Shiga, Japan) strictly following the manufacturers' protocols. Quantitative real-time PCR (RT-qPCR) assays were performed using the SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) on a QuantStudio\u0026trade; 5 Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The relative mRNA expression levels of \u003cem\u003ePDIA6\u003c/em\u003e were calculated using the 2\u003csup\u003e\u0026minus;ΔΔ\u003cem\u003eCt\u003c/em\u003e\u003c/sup\u003e method, with \u003cem\u003eGAPDH\u003c/em\u003e serving as the endogenous normalization control. The specific primer sequences utilized for amplification are listed in Table\u0026nbsp;2.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e1.13 Cell Proliferation Assay\u003c/h2\u003e \u003cp\u003eCell viability was evaluated using the Cell Counting Kit-8 (CCK-8) assay (Beyotime Biotechnology, Shanghai, China). Transfected cells were seeded into 96-well plates at a density of 2\u0026ndash;5\u0026times;10\u003csup\u003e3\u003c/sup\u003e cells per well. At 24, 48, and 72 h post-seeding, 10\u0026micro;L of CCK-8 reagent was added to each well, followed by incubation at 37\u0026deg;C for 1\u0026ndash;2 h. The absorbance at 450 nm (OD450) was measured using Synergy H1 microplate reader (BioTek Instruments, Winooski, VT, USA), with background values subtracted from the blank wells. All experiments were performed in triplicate and independently repeated three times to ensure reproducibility.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e1.14 Wound Healing Assay\u003c/h2\u003e \u003cp\u003eTransfected cells were seeded in 6-well plates and cultured to 90%\u0026ndash;100% confluence. A linear scratch was created using a sterile 200 \u0026micro;L pipette tip. After washing with PBS to remove cellular debris, the cells were incubated in serum-free medium to minimize the influence of cell proliferation. Microscopic images were captured at 0 and 24 h in the same fields of view using an Olympus IX73 inverted microscope (Olympus, Tokyo, Japan). The wound area was quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA), and the wound closure rate was calculated to evaluate migratory capacity of the cells. All assays were performed in independent triplicates.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e1.15 Transwell Migration and Invasion Assays\u003c/h2\u003e \u003cp\u003eCell motility was assessed using Transwell chambers equipped with 8-\u0026micro;m pore size polycarbonate membrane inserts (Corning, Corning, NY, USA). For the invasion assay, the upper chamber membranes were pre-coated with Matrigel (Corning, Corning, NY, USA) to simulate the extracellular matrix, whereas non-coated membranes were used for the migration assay. Transfected cells were suspended in 200 \u0026micro;L of serum-free medium (2\u0026ndash;5\u0026times;10\u003csup\u003e4\u003c/sup\u003e cells/well) and seeded into the upper chamber, while 600\u0026micro;L of medium supplemented with 10% FBS was added to the lower chamber as a chemoattractant. After 24 h of incubation at 37\u0026deg;C, cells remaining on the upper surface were removed. Migrated or invaded cells on the lower surface were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The number of migrated or invaded cells was quantified by counting five randomly selected fields per well under a light microscope. All assays were performed in independent triplicates.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e1.16 Statistical Analysis\u003c/h2\u003e \u003cp\u003eAll statistical analyses were conducted using R software (version 4.3.1). For continuous variables derived from bioinformatics analyses, the Wilcoxon rank-sum test was employed for comparisons between two groups, while the Kruskal-Wallis test was used for multi-group comparisons. For in vitro experimental data, results were presented as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation (SD), and statistical significance between two groups was evaluated using Student\u0026rsquo;s t-test. Survival curves were generated using the Kaplan-Meier method and compared via the log-rank test. Univariate and multivariate Cox proportional hazards regression analyses were implemented using the survival R package to identify independent prognostic factors. Where appropriate, P-values were adjusted for multiple testing using the Benjamini-Hochberg false discovery rate or Bonferroni corrections. All statistical tests were two-tailed, and a P-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05 was considered statistically significant (ns, not significant; * \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05; ** \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.01; *** \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Characterization of the Immune Cell Landscape in HCC\u003c/h2\u003e \u003cp\u003eFollowing stringent quality control, a total of 3,817 high-quality immune cells from the GSE140228 dataset were prioritized for analysis. Unsupervised clustering via the Seurat package categorized these cells into 17 distinct subclusters (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea), with their spatial distribution across tumor and adjacent normal tissues visualized via t-SNE (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb). Based on the SingleR algorithm and canonical lineage markers, these cells were annotated into five primary immune lineages: T cells (\u003cem\u003eCD3D, CD3E\u003c/em\u003e), monocytes (\u003cem\u003eCD68, CD163\u003c/em\u003e), natural killer (NK) cells (\u003cem\u003eKLRC1, NCR1\u003c/em\u003e), B cells (\u003cem\u003eCD19, CD79A\u003c/em\u003e), and dendritic cells (DCs; \u003cem\u003eCLEC9A, XCR1\u003c/em\u003e) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ec; Supplementary Fig.\u0026nbsp;1a\u0026ndash;e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eCompositional analysis revealed that T cells constituted the most predominant population across all samples, followed by monocytes and NK cells, while DCs were the least abundant (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed, e). Notably, odds ratio (OR) analysis demonstrated a significant tissue preference, highlighting that B cells were preferentially enriched within the TME compared to normal tissues (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ef). Furthermore, the top three highly expressed marker genes for each lineage were identified, reinforcing the accuracy of the cell-type identification (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eg).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Functional Enrichment Analysis of B-Cell Signature Genes\u003c/h2\u003e \u003cp\u003eTo elucidate the biological landscape of the B-cell compartment, we identified subcluster-specific marker genes using the FindAllMarkers function, applying thresholds of log2FC\u0026thinsp;\u0026gt;\u0026thinsp;1 and \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Functional annotation via the clusterProfiler package revealed that these signature genes were predominantly enriched in GO terms associated with humoral immunity and immunoglobulin production. Key biological processes included immunoglobulin-mediated immune responses, B-cell receptor signaling, and lymphocyte-mediated immunity (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). Furthermore, KEGG pathway analysis highlighted the involvement of these genes in critical immunological and homeostatic pathways, such as the NF-kappa B signaling pathway, cell adhesion molecules, and protein processing in the endoplasmic reticulum (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). Collectively, these findings reinforce the specialized role of B cells in orchestrating immune responses and proteinostasis within the HCC microenvironment.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Intercellular Communication Landscape of B Cells\u003c/h2\u003e \u003cp\u003eWe used the CellChat package to map communication between immune cells. Compared with adjacent normal tissue, the tumor microenvironment showed more interactions overall, but the average interaction strength was lower (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). Differential analysis of cellular cross-talk indicated that B cells in the TME demonstrated markedly higher interaction frequencies and intensities with the other four immune cell lineages compared to their counterparts in normal tissues (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed). Given the robust connectivity of B cells within the TME, we further dissected specific ligand-receptor pairs. Our results revealed that B-cell communication was primarily mediated via MHC-I and MHC-II signaling pathways. Specifically, B cells engaged with T cells predominantly through the HLA-E-CD8A axis, and with monocytes and DCs via the HLA-DRA-CD4 interaction. Furthermore, the B cell-NK cell crosstalk was characterized by the HLA-E-CD94/NKG2A signaling pair (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). These findings suggest that B cells act as a central hub in orchestrating the immunological status of the HCC microenvironment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Identification of B-cell-Associated Molecular Subtypes and Subtype-Specific Gene Modules\u003c/h2\u003e \u003cp\u003eUnivariate Cox regression analysis was initially implemented to evaluate the prognostic significance of the B-cell markers, yielding 71 genes with robust associations with HCC survival (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Based on these candidates, the ConsensusClusterPlus package with the PAM algorithm was implemented to partition the TCGA-LIHC cohort. Optimal clustering was achieved at \u003cem\u003ek\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3, defining three distinct molecular subtypes (C1, C2, and C3) characterized by high internal consistency and stability (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea\u0026ndash;c). The validity of this stratification was further corroborated by t-SNE analysis, which demonstrated clear transcriptomic separation among the three subgroups (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). Survival analysis revealed significant prognostic heterogeneity; patients in the C1 and C3 groups exhibited superior overall survival and recurrence-free survival compared to those in the C2 group (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee, f). Notably, multivariate Cox regression analysis identified the C2 subtype as an independent risk factor for poor prognosis (Supplementary Fig.\u0026nbsp;2a). Subsequently, we performed WGCNA to extract the unique transcriptional signatures of each subtype. A scale-free co-expression network was constructed with a soft-thresholding power of \u003cem\u003eβ\u003c/em\u003e\u0026thinsp;=\u0026thinsp;6, ensuring a scale-free topology fitting index\u0026thinsp;\u0026gt;\u0026thinsp;0.9 (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg). Using the dynamic tree-cutting method and merging similar modules (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh), we identified 11 co-expression modules. Correlation analysis between modules and clinical traits revealed that the turquoise, blue, and purple modules were most significantly associated with the C1, C2, and C3 subtypes, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ei\u0026ndash;k), providing a basis for further biomarker identification.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e2.5 External Validation of the B-cell-Associated Molecular Subtypes\u003c/h2\u003e \u003cp\u003eThe stability and robustness of the B-cell-associated molecular subtypes were validated in the GSE14520 and ICGC-LIRI cohorts using NTP analysis. Subtype-specific feature genes were defined by intersecting the upregulated DEGs identified via limma with the core modules previously extracted from WGCNA. Prediction accuracy was strictly controlled by excluding samples with a FDR\u0026thinsp;\u0026gt;\u0026thinsp;0.05. As anticipated, the three molecular subtypes were effectively recapitulated in both external cohorts, demonstrating high reproducibility of the classification system (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea, b). Prognostic evaluation further corroborated the clinical significance of this stratification. Consistent with the training set findings, the C2 subtype was significantly associated with unfavorable overall survival in both the GSE14520 and ICGC-LIRI datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec. d). Subsequent multivariate Cox regression analysis established the C2 subtype as an independent prognostic indicator for poor clinical outcomes (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee, f). Collectively, these results underscore the high reliability and prognostic value of the B-cell-derived subtyping framework across diverse HCC populations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Biological Underpinnings and Functional Landscapes of the Molecular Subtypes\u003c/h2\u003e \u003cp\u003eThe distinct biological functionalities of the three molecular subtypes were elucidated through GSEA using the GO and KEGG repositories. The C1 subtype was predominantly characterized by the enrichment of metabolic and catabolic processes involving various endogenous and exogenous compounds, suggesting a state of metabolic homeostasis (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). In contrast, the C2 subtype\u0026mdash;which was previously associated with the poorest prognosis\u0026mdash;exhibited a strong enrichment in oncogenic hallmarks, including the cell cycle, DNA replication, and mitotic cell division signaling pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb). The C3 subtype, conversely, was significantly associated with robust immunological activities, such as immune cell activation, proliferation, and the elimination of aberrant cells, reflecting an \"immuno-hot\" microenvironment (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec). These divergent functional profiles provide a mechanistic basis for the observed differences in clinical outcomes among the three HCC subtypes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e2.7 Divergent Genomic Landscapes and Mutational Profiles\u003c/h2\u003e \u003cp\u003eSomatic mutation profiles within the TCGA-LIHC cohort were analyzed using the maftools package, revealing the top 20 most frequently mutated genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). Among these, \u003cem\u003eTP53\u003c/em\u003e and \u003cem\u003eCTNNB1\u003c/em\u003e were identified as primary drivers, consistent with their established roles in HCC pathogenesis. Subtype-specific comparisons further highlighted distinct mutational signatures: the C1 subtype was characterized by a high frequency of \u003cem\u003eCTNNB1\u003c/em\u003e and \u003cem\u003eTTN\u003c/em\u003e mutations, whereas \u003cem\u003eTP53\u003c/em\u003e mutations predominated in the C2 subtype (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). Given that the progressive accumulation of somatic mutations drives tumorigenesis, the relatively lower mutational burden observed in the C3 group aligns with its superior prognosis identified in previous sections.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFurther investigation into CNV via the cBioPortal database identified significant differences in gene amplification and homozygous deletion across the three subtypes (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Patients in the C2 subgroup exhibited a substantially higher CNV burden compared to other groups. Notably, homozygous loss of \u003cem\u003eCSMD1\u003c/em\u003e\u0026mdash;a recognized tumor suppressor whose deficiency promotes HCC progression\u0026mdash;was prominently observed across all subtypes, with the most marked prevalence in the C2 group (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Collectively, these findings underscore that the C2 subtype is defined by extensive genomic alterations and high genomic instability, providing a molecular rationale for its aggressive clinical phenotype.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e2.8 Characterization of Immune Infiltration and Assessment of Immunotherapy Response\u003c/h2\u003e \u003cp\u003eThe relative abundance of 28 immune cell types within the TME was quantified using ssGSEA. The C3 subtype exhibited a markedly higher abundance of immune cell infiltration compared to the other two subgroups (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea). This \"immuno-hot\" phenotype in C3 patients was characterized by a dense infiltration of various immune effectors, including activated B cells, activated CD4 T cells, activated CD8 T cells, and immature B cells (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05; Supplementary Fig.\u0026nbsp;2c).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFurther profiling of immune checkpoint molecules and HLA genes revealed significant transcriptomic divergence across the subtypes. Key checkpoints, such as \u003cem\u003eCD27\u003c/em\u003e, \u003cem\u003ePDCD1\u003c/em\u003e (\u003cem\u003ePD-1\u003c/em\u003e), \u003cem\u003eCTLA4\u003c/em\u003e, and \u003cem\u003eLAG3\u003c/em\u003e, as well as HLA family members, were significantly upregulated in the C3 subtype, suggesting heightened immune activation and enhanced antigen-presenting potential in these patients (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb, c; Supplementary Fig.\u0026nbsp;2b).\u003c/p\u003e \u003cp\u003eThe potential clinical response to immunotherapy was evaluated through a multi-algorithmic approach integrating TIS, IPS, and Submap analyses. Submap analysis indicated that patients in the C3 group were more likely to respond to anti-PD-1 therapy (Bonferroni-corrected \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05; Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ed). Additionally, the significantly higher TIS scores observed in the C3 subtype suggested a greater likelihood of clinical benefit from immune checkpoint inhibitors (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ee). IPS evaluation further corroborated these findings, demonstrating that C3 patients possessed superior antigen recognition and presentation capacities, higher effector cell activity, and reduced risks of immune suppression and escape (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ef\u0026ndash;i).\u003c/p\u003e \u003cp\u003eConsistency of these observations was verified within the ICGC-LIRI cohort, where the C3 subtype's advantages in immune infiltration and therapeutic sensitivity were successfully recapitulated (Supplementary Fig.\u0026nbsp;2d\u0026ndash;i). Collectively, these findings establish the C3 subtype as the optimal candidate for precision immunotherapy in HCC.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e2.9 Pharmacogenomic Profiling and Identification of Candidate Small-Molecule Perturbagens\u003c/h2\u003e \u003cp\u003eThe therapeutic sensitivity of HCC tumor cells was predicted using the oncoPredict package, where lower half-maximal inhibitory concentration (IC\u003csub\u003e50\u003c/sub\u003e) values indicate heightened drug sensitivity. Significant differences in drug responsiveness were observed among the three molecular subtypes. Specifically, the C1 subtype demonstrated enhanced sensitivity to dihydrorotenone and sinularin (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ea, b). In contrast, the C2 subtype exhibited more pronounced responses to lapatinib and sepantronium bromide (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ec, d). Furthermore, patients in the C3 subgroup showed superior sensitivity to alisertib and olaparib (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003ee, f).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eTo identify potential small-molecule compounds targeting each subtype, the unique DEGs of each group were utilized as input signatures for the CMap database. As illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eg, the top candidate drugs for the C1 subtype included amsacrine, MEK1-2 inhibitor, GDC-0879, lestaurtinib, avrainvillamide-analog-5, ochratoxin A, and TPCA-1. For the C2 subtype, the analysis highlighted MDL-11939, GW-843682X, nocodazole, SA-792574, ABT-751, vinorelbine, MK-1775, and ALW-II-49-7 as prospective agents. Meanwhile, candidates for the C3 subtype comprised cimaterol, benperidol, gatifloxacin, importazole, atenolol, and MDL-11939.\u003c/p\u003e \u003cp\u003eMapping the targeted pathways of these candidate drugs provided further insights into the development of multi-drug or combination therapeutic strategies (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003eh). Collectively, these findings pinpoint subtype-specific therapeutic vulnerabilities and offer a blueprint for personalized treatment strategies in HCC.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e2.10 Construction and Validation of the B-cell-associated Gene Signature\u003c/h2\u003e \u003cp\u003eA comprehensive machine learning framework integrating 10 distinct algorithms was employed to construct 117 prognostic models based on the expression profiles of the 71 B-cell-related prognostic genes. These algorithms included RSF, Enet, StepCox, CoxBoost, plsRcox, superpc, GBM, survivalsvm, Ridge, and Lasso. The C-index for each model was rigorously evaluated across the TCGA-LIHC discovery cohort and two independent external cohorts (ICGC-LIRI and GSE14520) (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ea).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eThe \"StepCox [forward] + Ridge\" combination emerged as the optimal integration, demonstrating the highest average C-index across all three datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003eb). This integrative approach prioritized 21 core genes to establish the BCAGS. Each patient\u0026rsquo;s risk score was calculated, enabling their stratification into high- or low-risk groups. Kaplan-Meier survival analysis revealed that the high-risk group suffered from significantly inferior clinical outcomes compared to the low-risk group across all cohorts (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ec\u0026ndash;e).\u003c/p\u003e \u003cp\u003eTime-dependent Receiver Operating Characteristic analysis was further implemented to evaluate the predictive accuracy of the BCAGS. In the TCGA, ICGC, and GSE cohorts, the Area Under the Curve values reached 0.779, 0.714, and 0.677 for 1-year survival; 0.746, 0.733, and 0.633 for 3-year survival; and 0.67, 0.761, and 0.58 for 5-year survival, respectively (Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003ef\u0026ndash;h), confirming the robust predictive performance and generalizability of the BCAGS.\u003c/p\u003e \u003cp\u003eTo further solidify the prognostic significance of the BCAGS, a meta-analysis based on univariate Cox regression was conducted, identifying the BCAGS as a consistent risk factor for HCC across all datasets (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003ea, b). Notably, the BCAGS demonstrated superior stability and predictive power when benchmarked against 37 previously published HCC prognostic models (Fig.\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003ec). These findings highlight the potential of the BCAGS as a highly reliable tool for individual survival prediction in HCC patients.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section2\"\u003e \u003ch2\u003e2.11 Functional Validation of \u003cem\u003ePDIA6\u003c/em\u003e in HCC Cell Lines\u003c/h2\u003e \u003cp\u003eTo further refine the signature and identify the most critical drivers of HCC progression, we implemented eight distinct machine learning algorithms to screen the 71 B-cell-associated prognostic genes. This integrative approach prioritized the top 10 feature genes based on their selection frequency across the various model combinations (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003ea). Among these, \u003cem\u003ePDIA6\u003c/em\u003e emerged as a top-ranking candidate with high stability. However, the specific functional role and underlying molecular mechanisms of \u003cem\u003ePDIA6\u003c/em\u003e in hepatocellular carcinoma have not yet been fully elucidated, necessitating further experimental investigation. To this end, Loss-of-function experiments were conducted in two human HCC cell lines (HepG-2 and Huh-7) to elucidate the biological role of \u003cem\u003ePDIA6\u003c/em\u003e in tumor progression. A marked decrease in \u003cem\u003ePDIA6\u003c/em\u003e gene expression at the mRNA level was observed following transfection with \u003cem\u003ePDIA6\u003c/em\u003e-specific siRNA, as determined by RT-qPCR (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003eb).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eFunctional assessments using the CCK-8 assay demonstrated that silencing \u003cem\u003ePDIA6\u003c/em\u003e markedly suppressed the proliferative capacity of HepG-2 and Huh-7 cells, suggesting its essential role in promoting HCC cell growth (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003ec). Wound-healing assays further indicated that \u003cem\u003ePDIA6\u003c/em\u003e knockdown significantly impeded the migratory potential of HCC cells compared to the negative control groups (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003ed). Consistently, Transwell migration and Matrigel invasion assays revealed a substantial decrease in the number of migrated and invaded cells following \u003cem\u003ePDIA6\u003c/em\u003e depletion in both cell lines (Fig.\u0026nbsp;\u003cspan refid=\"Fig11\" class=\"InternalRef\"\u003e11\u003c/span\u003ee).\u003c/p\u003e \u003cp\u003eCollectively, these in vitro findings demonstrate that \u003cem\u003ePDIA6\u003c/em\u003e facilitates the proliferation, migration, and invasion of HCC cells, reinforcing its role as an oncogenic driver and validating the clinical significance of our computational modeling.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe high heterogeneity of hepatocellular carcinoma (HCC) remains a formidable barrier to effective prognosis and precision therapy. While the dichotomy of \"immune-hot\" and \"immune-cold\" tumors has guided recent therapeutic strategies, the specific contribution of B cells\u0026mdash;often overshadowed by T cells\u0026mdash;to the HCC landscape remains underappreciated. Single-cell sequencing technologies have recently enabled the high-resolution characterization of such specific cell populations, providing insights that traditional bulk sequencing could not achieve \u003csup\u003e\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e,\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e. In this study, we bridged single-cell resolution with bulk transcriptomics to deconstruct the B-cell-associated architecture of HCC. Our multi-omics approach not only delineated three clinically distinct molecular subtypes with unique genomic and varying immune profiles but also established a robust machine-learning-based prognostic signature, termed BCAGS. Crucially, we validated \u003cem\u003ePDIA6\u003c/em\u003e as a functional driver linking B-cell-associated signatures to tumor aggression, offering a novel translational bridge between computational stratification and biological intervention.\u003c/p\u003e \u003cp\u003eOur scRNA-seq analysis provides compelling evidence that B cells are not merely bystanders but active orchestrators within the TME. Contrary to the classical view of T cell-dominated immunity, we observed that B cells in HCC exhibit enhanced intercellular communication, particularly through the MHC-I and MHC-II axes. The robust interaction pairs identified, such as HLA-E-CD8A and HLA-DRA-CD4, suggest that B cells may function as pivotal antigen-presenting cells. This aligns with emerging concepts that MHC-I expression initiates CD8\u003csup\u003e+\u003c/sup\u003eT cell cytotoxicity \u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e, while MHC-II expression on tumor or immune cells can recruit and prime CD4\u003csup\u003e+\u003c/sup\u003e helper T cells to enhance the anti-tumor immune loop \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. The enrichment of pathways related to immunoglobulin production and receptor signaling further implies that the humoral arm of immunity is actively engaged, albeit often suppressed, within the HCC milieu.\u003c/p\u003e \u003cp\u003eA key translational output of our work is the identification of three stable molecular subtypes (C1, C2, C3), which effectively stratify patients based on biological distinctness rather than just clinical staging. The C3 subtype represents a classic \"immune-hot\" phenotype, characterized by dense infiltration of activated B and T cells and, paradoxically, elevated expression of exhaustion markers (\u003cem\u003ePD-1, CTLA-4\u003c/em\u003e). This expression pattern is the hallmark of patients who are most likely to respond to immune checkpoint blockade, as confirmed by our TIS, IPS, and Submap predictive analyses. In this context, the high expression of HLA molecules in C3 suggests that the machinery for antigen presentation is intact, making these tumors \"visible\" to the immune system once the checkpoint brakes are released.\u003c/p\u003e \u003cp\u003eIn stark contrast, the C2 subtype epitomizes the aggressive, \"immune-cold\" phenotype associated with poor clinical outcomes. Our genomic landscaping revealed that C2 is driven by profound genomic instability, highlighted by a high frequency of \u003cem\u003eTP53\u003c/em\u003e mutations and significant CNVs. Previous studies have established that \u003cem\u003eTP53\u003c/em\u003e mutations not only drive genomic instability but also foster an immunosuppressive microenvironment conducive to immune evasion, which explains the dismal prognosis observed in this subgroup \u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e. Furthermore, the homozygous deletion of \u003cem\u003eCSMD1\u003c/em\u003e in this subtype parallels findings that link the loss of this tumor suppressor to HCC progression and chromosomal instability \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e\u003c/sup\u003e. This stratification suggests that C2 patients might be refractory to standard immunotherapy and may instead require strategies targeting cell cycle dysregulation or specific metabolic vulnerabilities, such as the sensitivity to lapatinib identified in our drug screening.\u003c/p\u003e \u003cp\u003eTo translate these complex biological patterns into a clinical tool, we developed the BCAGS using an ensemble of 117 machine-learning algorithms. The superiority of the \"StepCox [forward] + Ridge\" model lies in its ability to distill high-dimensional data into a pragmatic risk score that remains robust across independent cohorts (TCGA, ICGC, GSE14520). Unlike single-biomarker approaches, the BCAGS integrates the multifaceted nature of B cell biology, serving as an independent prognostic factor that outperforms widespread clinical models. This score holds potential utility for oncologists to identify high-risk patients who may benefit from intensified adjuvant therapies.\u003c/p\u003e \u003cp\u003eFinally, we moved beyond purely associative computational findings to provide experimental validation. Among the genes constituting our signature, \u003cem\u003ePDIA6\u003c/em\u003e emerged as a critical link between the B-cell-associated risk profile and tumor cell capability. \u003cem\u003ePDIA6\u003c/em\u003e serves as a key regulator of endoplasmic reticulum protein folding, and its upregulation has been implicated in the proliferation and invasion of various malignancies \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e,\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e\u003c/sup\u003e. Consistent with these observations, our in vitro loss-of-function assays in HepG2 and Huh-7 lines unequivocally demonstrate its role in promoting proliferation, migration, and invasion in HCC. This suggests that \u003cem\u003ePDIA6\u003c/em\u003e may not only serve as a biomarker but also as a therapeutic target. While our \u003cem\u003ein vitro\u003c/em\u003e data unequivocally pinpoint \u003cem\u003ePDIA6\u003c/em\u003e as a driver of HCC aggressiveness, we must acknowledge that Petri dishes cannot replicate the complex dialogue between tumor cells and the immune system. It's like examining the seed without the soil. A critical missing piece is verifying whether \u003cem\u003ePDIA6\u003c/em\u003e directly dictates the recruitment or exhaustion of B cells\u0026mdash;a question our current cell line models simply cannot answer. Therefore, our next step is non-negotiable: we must move to immunocompetent mouse models. Only then can we confirm if targeting \u003cem\u003ePDIA6\u003c/em\u003e truly remodels the tumor microenvironment or merely suppresses tumor growth in isolation.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study provides a comprehensive characterization of the B-cell-associated molecular landscape in hepatocellular carcinoma (HCC) through the integration of single-cell and bulk transcriptomic analyses. The identification of three distinct B-cell-related subtypes\u0026mdash;C1 (metabolic), C2 (proliferative), and C3 (immuno-hot)\u0026mdash;unveils the profound heterogeneity of the tumor microenvironment and its critical impact on clinical outcomes. Leveraging an ensemble of 117 machine learning combinations, the established BCAGS prognostic model demonstrates superior stability and predictive accuracy across multiple independent cohorts, outperforming existing clinical benchmarks. Moreover, experimental validation identifies \u003cem\u003ePDIA6\u003c/em\u003e as a novel oncogenic driver that facilitates HCC progression, reinforcing the clinical relevance of our computational findings. Collectively, these results offer a refined framework for prognostic stratification and provide actionable insights for personalized immunotherapy and targeted treatment strategies in HCC patients.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe research was conducted on public datasets (TCGA, GEO, and ICGC) and established human cell lines. Ethical approval for the analysis of public data was waived by the local ethics committee. The human HCC cell lines (HepG-2 and Huh-7) were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China), and their use did not require additional ethical approval.\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\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analyzed during the current study are available in the following repositories:\u003c/p\u003e\n\u003cp\u003eGSE140228 (scRNA-seq data): Gene Expression Omnibus (GEO) database https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140228.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTCGA-LIHC (Transcriptomic and clinical data): The Cancer Genome Atlas (TCGA) program via the GDC Data Portal https://portal.gdc.cancer.gov/.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGSE14520 (Validation cohort): Gene Expression Omnibus (GEO) database https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14520.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eICGC-LIRI-JP (Validation cohort): International Cancer Genome Consortium (ICGC) Data Portal https://dcc.icgc.org/projects/LIRI-JP. Publicly available portal cBioPortal https://www.cbioportal.org/ was used for CNV analysis, and Connectivity Map (CMap) https://clue.io/ was used for drug screening.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that no funds, grants, or other support were received during the preparation of this manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eYang Li and Sinan Cao contributed equally to this work and share first authorship.\u003c/p\u003e\n\u003cp\u003eYang Li performed the bioinformatics analyses, integrated multi-omics data, conducted statistical analyses, and drafted the manuscript.\u003c/p\u003e\n\u003cp\u003eSinan Cao performed the experimental validation, including cell culture, gene knockdown, and RT-qPCR assays, and contributed to data interpretation.\u003c/p\u003e\n\u003cp\u003eYamei Kuang assisted with data collection, data processing, and figure preparation.\u003c/p\u003e\n\u003cp\u003eDachuan Shen participated in result interpretation and manuscript revision.\u003c/p\u003e\n\u003cp\u003eLili Tian conceived and supervised the study, provided critical revisions to the manuscript, and approved the final version for submission.\u003c/p\u003e\n\u003cp\u003eAll authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank the contributors of the TCGA, GEO, and ICGC databases for sharing the sequencing data.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eSung H, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. Cancer J Clin. 2021;71:209\u0026ndash;49. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3322/caac.21660\u003c/span\u003e\u003cspan address=\"10.3322/caac.21660\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLlovet JM, et al. Hepatocellular carcinoma. Nat Rev Dis Primers. 2021;7:6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41572-020-00240-3\u003c/span\u003e\u003cspan address=\"10.1038/s41572-020-00240-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng R, et al. Liver cancer incidence and mortality in China: Temporal trends and projections to 2030. Chin J Cancer Res. 2018;30:571\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21147/j.issn.1000-9604.2018.06.01\u003c/span\u003e\u003cspan address=\"10.21147/j.issn.1000-9604.2018.06.01\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBeaufr\u0026egrave;re A, et al. Gene expression signature as a surrogate marker of microvascular invasion on routine hepatocellular carcinoma biopsies. J Hepatol. 2022;76:343\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jhep.2021.09.034\u003c/span\u003e\u003cspan address=\"10.1016/j.jhep.2021.09.034\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChew V, et al. Delineation of an immunosuppressive gradient in hepatocellular carcinoma using high-dimensional proteomic and transcriptomic analyses. Proc Natl Acad Sci U S A. 2017;114:E5900\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.1706559114\u003c/span\u003e\u003cspan address=\"10.1073/pnas.1706559114\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZheng Y, et al. Immune suppressive landscape in the human esophageal squamous cell carcinoma microenvironment. Nat Commun. 2020;11:6268. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-020-20019-0\u003c/span\u003e\u003cspan address=\"10.1038/s41467-020-20019-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLlovet JM, et al. Immunotherapies for hepatocellular carcinoma. Nat Rev Clin Oncol. 2022;19:151\u0026ndash;72. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41571-021-00573-2\u003c/span\u003e\u003cspan address=\"10.1038/s41571-021-00573-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKurebayashi Y, et al. Landscape of immune microenvironment in hepatocellular carcinoma and its additional impact on histological and molecular classification. Hepatology. 2018;68:1025\u0026ndash;41. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hep.29904\u003c/span\u003e\u003cspan address=\"10.1002/hep.29904\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSangro B, Sarobe P, Herv\u0026aacute;s-Stubbs S, Melero. I. Advances in immunotherapy for hepatocellular carcinoma. Nat Rev Gastroenterol Hepatol. 2021;18:525\u0026ndash;43. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41575-021-00438-0\u003c/span\u003e\u003cspan address=\"10.1038/s41575-021-00438-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCabrita R, et al. Author Correction: Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature. 2020;580:E1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-020-2155-6\u003c/span\u003e\u003cspan address=\"10.1038/s41586-020-2155-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHelmink BA, et al. B cells and tertiary lymphoid structures promote immunotherapy response. Nature. 2020;577:549\u0026ndash;55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41586-019-1922-8\u003c/span\u003e\u003cspan address=\"10.1038/s41586-019-1922-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarnelo M, et al. Interaction between tumour-infiltrating B cells and T cells controls the progression of hepatocellular carcinoma. Gut. 2017;66:342\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/gutjnl-2015-310814\u003c/span\u003e\u003cspan address=\"10.1136/gutjnl-2015-310814\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang Q, et al. Landscape and Dynamics of Single Immune Cells in Hepatocellular Carcinoma. Cell. 2019;179:829\u0026ndash;e845820. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cell.2019.10.003\u003c/span\u003e\u003cspan address=\"10.1016/j.cell.2019.10.003\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa L, et al. Tumor Cell Biodiversity Drives Microenvironmental Reprogramming in Liver Cancer. Cancer Cell. 2019;36:418\u0026ndash;e430416. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ccell.2019.08.007\u003c/span\u003e\u003cspan address=\"10.1016/j.ccell.2019.08.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNewman AM, et al. Determining cell type abundance and expression from bulk tissues with digital cytometry. Nat Biotechnol. 2019;37:773\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41587-019-0114-2\u003c/span\u003e\u003cspan address=\"10.1038/s41587-019-0114-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchumacher TN, Schreiber RD. Neoantigens in cancer immunotherapy. Science. 2015;348:69\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/science.aaa4971\u003c/span\u003e\u003cspan address=\"10.1126/science.aaa4971\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa Y et al. PDIA6 promotes pancreatic cancer progression and immune escape through CSN5-mediated deubiquitination of β-catenin and PD-L1. \u003cem\u003eNeoplasia\u003c/em\u003e 23, 912\u0026ndash;928 (2021). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.neo.2021.07.004\u003c/span\u003e\u003cspan address=\"10.1016/j.neo.2021.07.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBai Y, et al. PDIA6 modulates apoptosis and autophagy of non-small cell lung cancer cells via the MAP4K1/JNK signaling pathway. EBioMedicine. 2019;42:311\u0026ndash;25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ebiom.2019.03.045\u003c/span\u003e\u003cspan address=\"10.1016/j.ebiom.2019.03.045\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eButler A, Hoffman P, Smibert P, Papalexi E, Satija R. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nat Biotechnol. 2018;36:411\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/nbt.4096\u003c/span\u003e\u003cspan address=\"10.1038/nbt.4096\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKorsunsky I, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nat Methods. 2019;16:1289\u0026ndash;96. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41592-019-0619-0\u003c/span\u003e\u003cspan address=\"10.1038/s41592-019-0619-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAran D, et al. Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol. 2019;20:163\u0026ndash;72. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41590-018-0276-y\u003c/span\u003e\u003cspan address=\"10.1038/s41590-018-0276-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXu S, et al. Using clusterProfiler to characterize multiomics data. Nat Protoc. 2024. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41596-024-01020-z\u003c/span\u003e\u003cspan address=\"10.1038/s41596-024-01020-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJin S, et al. Inference and analysis of cell-cell communication using CellChat. Nat Commun. 2021;12:1088. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41467-021-21246-9\u003c/span\u003e\u003cspan address=\"10.1038/s41467-021-21246-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilkerson MD, Hayes DN. ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinf (Oxford England). 2010;26:1572\u0026ndash;3. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/bioinformatics/btq170\u003c/span\u003e\u003cspan address=\"10.1093/bioinformatics/btq170\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLangfelder P, Horvath S. WGCNA: an R package for weighted correlation network analysis. BMC Bioinformatics. 2008;9:559. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/1471-2105-9-559\u003c/span\u003e\u003cspan address=\"10.1186/1471-2105-9-559\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoshida Y. Nearest template prediction: a single-sample-based flexible class prediction with confidence assessment. PLoS ONE. 2010;5:e15543. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0015543\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0015543\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitchie ME et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. \u003cem\u003eNucleic acids research\u003c/em\u003e 43, e47 (2015). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/nar/gkv007\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkv007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayakonda A, Lin DC, Assenov Y, Plass C, Koeffler HP. Maftools: efficient and comprehensive analysis of somatic variants in cancer. Genome Res. 2018;28:1747\u0026ndash;56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1101/gr.239244.118\u003c/span\u003e\u003cspan address=\"10.1101/gr.239244.118\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGao J et al. Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal. \u003cem\u003eScience signaling\u003c/em\u003e 6, pl1 (2013). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/scisignal.2004088\u003c/span\u003e\u003cspan address=\"10.1126/scisignal.2004088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCharoentong P, et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell Rep. 2017;18:248\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.celrep.2016.12.019\u003c/span\u003e\u003cspan address=\"10.1016/j.celrep.2016.12.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eH\u0026auml;nzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013;14:7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/1471-2105-14-7\u003c/span\u003e\u003cspan address=\"10.1186/1471-2105-14-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShi JY, et al. Exploring prognostic indicators in the pathological images of hepatocellular carcinoma based on deep learning. Gut. 2021;70:951\u0026ndash;61. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1136/gutjnl-2020-320930\u003c/span\u003e\u003cspan address=\"10.1136/gutjnl-2020-320930\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGonzalez LC, et al. A coreceptor interaction between the CD28 and TNF receptor family members B and T lymphocyte attenuator and herpesvirus entry mediator. Proc Natl Acad Sci USA. 2005;102:1116\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1073/pnas.0409071102\u003c/span\u003e\u003cspan address=\"10.1073/pnas.0409071102\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu DH, Mou FF, An M, Xia P. Human leukocyte antigen and tumor immunotherapy (Review). Int J Oncol. 2023;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3892/ijo.2023.5516\u003c/span\u003e\u003cspan address=\"10.3892/ijo.2023.5516\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAyers M, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Investig. 2017;127:2930\u0026ndash;40. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1172/jci91190\u003c/span\u003e\u003cspan address=\"10.1172/jci91190\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoshida Y, Brunet JP, Tamayo P, Golub TR, Mesirov JP. Subclass mapping: identifying common subtypes in independent disease data sets. PLoS ONE. 2007;2:e1195. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1371/journal.pone.0001195\u003c/span\u003e\u003cspan address=\"10.1371/journal.pone.0001195\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZeng D, et al. Multi-Omics Immuno-Oncology Biological Research to Decode Tumor Microenvironment and Signatures. Front Immunol. 2021;12:687975. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fimmu.2021.687975\u003c/span\u003e\u003cspan address=\"10.3389/fimmu.2021.687975\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMaeser D, Gruener RF, Huang RS. oncoPredict: an R package for predicting in vivo or cancer patient drug response and biomarkers from cell line screening data. Brief Bioinform. 2021;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1093/bib/bbab260\u003c/span\u003e\u003cspan address=\"10.1093/bib/bbab260\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLamb J, et al. The Connectivity Map: using gene-expression signatures to connect small molecules, genes, and disease. Sci (New York N Y). 2006;313:1929\u0026ndash;35. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1126/science.1132939\u003c/span\u003e\u003cspan address=\"10.1126/science.1132939\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu H, et al. A flexible machine-learning framework to construct and visualize models for clinical characteristics prediction and feature selection. Comput Struct Biotechnol J. 2024;23:2798\u0026ndash;810. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.csbj.2024.06.035\u003c/span\u003e\u003cspan address=\"10.1016/j.csbj.2024.06.035\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Mime.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShaw R, Tian X, Xu J. Single-Cell Transcriptome Analysis in Plants: Advances and Challenges. Mol Plant. 2021;14:115\u0026ndash;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.molp.2020.10.012\u003c/span\u003e\u003cspan address=\"10.1016/j.molp.2020.10.012\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang Y, et al. Changing Technologies of RNA Sequencing and Their Applications in Clinical Oncology. Front Oncol. 2020;10:447. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fonc.2020.00447\u003c/span\u003e\u003cspan address=\"10.3389/fonc.2020.00447\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaskov H, Orhan A, Christensen JP, G\u0026ouml;genur I. Cytotoxic CD8\u0026thinsp;+\u0026thinsp;T cells in cancer and cancer immunotherapy. Br J Cancer. 2021;124:359\u0026ndash;67. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41416-020-01048-4\u003c/span\u003e\u003cspan address=\"10.1038/s41416-020-01048-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAxelrod ML, Cook RS, Johnson DB, Balko JM. Biological Consequences of MHC-II Expression by Tumor Cells in Cancer. Clin cancer research: official J Am Association Cancer Res. 2019;25:2392\u0026ndash;402. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1158/1078-0432.Ccr-18-3200\u003c/span\u003e\u003cspan address=\"10.1158/1078-0432.Ccr-18-3200\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKo KP, et al. Key Genetic Determinants Driving Esophageal Squamous Cell Carcinoma Initiation and Immune Evasion. Gastroenterology. 2023;165:613\u0026ndash;e628620. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1053/j.gastro.2023.05.030\u003c/span\u003e\u003cspan address=\"10.1053/j.gastro.2023.05.030\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMidorikawa Y, et al. Allelic imbalances and homozygous deletion on 8p23.2 for stepwise progression of hepatocarcinogenesis. Hepatology (Baltimore MD). 2009;49:513\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hep.22698\u003c/span\u003e\u003cspan address=\"10.1002/hep.22698\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang P, et al. PDIA6, which is regulated by TRPM2-AS/miR-424-5p axis, promotes endometrial cancer progression via TGF-beta pathway. Cell Death Dis. 2023;14:829. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41419-023-06297-8\u003c/span\u003e\u003cspan address=\"10.1038/s41419-023-06297-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXiang P, et al. circSETD3 confers radiotherapy resistance in nasopharyngeal carcinoma by attenuating ER stress-induced autophagy and apoptosis via PDIA6 upregulation. Oncogene. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1038/s41388-025-03652-1\u003c/span\u003e\u003cspan address=\"10.1038/s41388-025-03652-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 and 2 are available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Hepatocellular carcinoma, B cells, Molecular subtypes, Machine learning, Prognostic signature, PDIA6","lastPublishedDoi":"10.21203/rs.3.rs-9213778/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9213778/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eB cells are integral components of the tumor microenvironment (TME) and play a pivotal role in regulating anti-tumor immunity. However, the systematic classification of B-cell-associated molecular features in hepatocellular carcinoma (HCC) and their value in clinical prognosis and immunotherapy remain elusive.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe elucidated the molecular characteristics of B cells within the HCC TME based on single-cell RNA sequencing (scRNA-seq) data and constructed B-cell-associated molecular subtypes by integrating bulk RNA sequencing data. We systematically evaluated differences across subtypes regarding clinical prognosis, biological processes, genomic variations, the immune microenvironment, and immunotherapeutic responses. Furthermore, a B-cell-associated gene signature score (BCAGS) prognostic model was established using 117 combinations of machine learning algorithms. Finally, key genes were selected for in vitro functional validation.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eBased on B cell marker genes, we identified three molecular subtypes with significantly distinct clinical outcomes, validating their robustness across multiple external cohorts. The C2 subtype exhibited higher genomic instability and the poorest prognosis, whereas the C3 subtype presented an \"immune-hot\" phenotype with predicted sensitivity to immunotherapy. Additionally, the BCAGS was confirmed as an independent risk factor for HCC patients, demonstrating superior predictive performance compared to existing models. Functional experiments further revealed that silencing the key gene, \u003cem\u003ePDIA6\u003c/em\u003e, significantly inhibited the proliferation, migration, and invasion of HCC cells.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eThis study systematically unveils the heterogeneity of B-cell-associated molecular subtypes in HCC and their clinical significance, establishing a robust BCAGS prognostic model. Combined with in vitro validation, the results suggest that \u003cem\u003ePDIA6\u003c/em\u003e may play a critical role in promoting HCC progression, providing a novel theoretical basis for precision stratification and the development of potential therapeutic targets in HCC.\u003c/p\u003e","manuscriptTitle":"Breaking Therapeutic Resistance in Hepatocellular Carcinoma: Systems Biology Deconvolution of the B-cell Microenvironment Identifies PDIA6 as a Novel Target","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-26 10:49:50","doi":"10.21203/rs.3.rs-9213778/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a3e31f21-1bc2-4b8c-b33d-7862985a0b6d","owner":[],"postedDate":"March 26th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-03-31T17:10:23+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-26 10:49:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9213778","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9213778","identity":"rs-9213778","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}