Identification of prognostic genes for hepatocellular carcinoma based on hepatocyte and ketone body metabolism using integrated bulk and single cell RNA sequencing

Identification of prognostic genes for hepatocellular carcinoma based on hepatocyte and ketone body metabolism using integrated bulk and single cell RNA sequencing | 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 Identification of prognostic genes for hepatocellular carcinoma based on hepatocyte and ketone body metabolism using integrated bulk and single cell RNA sequencing Wei Yuan, Runyu Zhuang, Bailin Wang, Fan Wu, Benliang Mao, Shanfei Zhu, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9187321/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Background The precise role of ketone body (KB) metabolism and the underlying cellular mechanisms in hepatocellular carcinoma (HCC) remain unclear. This study aimed to identify and mechanistically explore KB metabolism-related prognostic genes in HCC. Methods All data were obtained from public databases. Single-cell RNA sequencing (scRNA-seq) analysis and high-dimensional weighted gene co-expression network analysis (hdWGCNA) were integrated to identify key module genes related to key cell type with high KB metabolism-related gene (KMRG) activity. After intersecting with differentially expressed genes (DEGs) in HCC and KMRGs, candidate genes were acquired. Subsequently, prognostic genes were recognized by univariate Cox and least absolute shrinkage and selection operator (LASSO) regression. A risk model for risk stratification was established and validated. Further risk group-based analyses mainly examined the impact of risk scores on function, tumor microenvironment (TME), immunotherapy response, and drug sensitivity. Prognostic gene expression dynamics during key cell type differentiation were analyzed by cell trajectory analysis. Eventually, prognostic gene expression was validated via reverse transcription-quantitative PCR (RT-qPCR) Results Hepatocytes were identified as the key cell type with high KMRG activity. Remarkably, PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were identified as prognostic genes. RT-qPCR demonstrated that OXCT1 was upregulated in HCC, while the other prognostic genes were downregulated. The constructed risk model exhibited robust predictive capacity, showing that high-risk patients had lower survival probabilities. Activities of pathways like "MYC targets V1", infiltration of immune cell types like activated CD4 T cells, response to immunotherapy, and sensitivities to drugs like camptothecin were altered by risk scores. Dynamic expression changes of PPARGC1A, PDK4, HSD17B6, and APOC3 were observed during hepatocyte differentiation. Conclusion Five prognostic genes related to hepatocytes and KB metabolism were identified in HCC, and a risk model with strong predictive utility was developed, offering novel insights into clinical prognostic prediction for HCC. Hepatocellular carcinoma Hepatocytes Ketone body metabolism Prognostic gene Tumor immunity Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 1. Background Hepatocellular carcinoma (HCC) is the most common type of primary liver tumor, accounting for 75–85% of all cases. With the widespread administration of the hepatitis B vaccine, the incidence of HCC caused by viral hepatitis-related risk factors has decreased[ 1 ]. Conversely, the number of HCC cases linked to metabolic disorders, such as nonalcoholic fatty liver disease (NAFLD) and diabetes, has steadily increased, paralleling global trends in obesity[ 2 ]. HCC is now recognized as the sixth most common cancer and the fourth leading cause of cancer-related deaths globally[ 3 ]. The persistently high mortality rate of HCC is attributed to its late diagnosis, high recurrence rate, ease of metastasis, and treatment resistance. Although immune checkpoint inhibitors (e.g., atezolizumab/bevacizumab) and targeted therapies (e.g., lenvatinib) have improved outcomes, their efficacy varies across patient subgroups, and resistance remains a challenge[ 4 ]. Current predictive models for long-term clinical outcomes in HCC patients are insufficient. Therefore, identifying new prognostic genes holds critical significance for elucidating the underlying mechanisms of HCC pathogenesis, improving patient prognosis management, and developing targeted therapeutic targets[ 5 ]. Ketone bodies (KBs), including acetoacetate (AcAc), D-β-hydroxybutyrate (BHB), and acetone, are alternative energy substrates during fasting or metabolic stress. In addition to their metabolic roles, KBs exhibit pleiotropic effects, such as neuroprotection and anti-inflammatory activity, that are mediated through receptor-dependent (e.g., HCAR2) and epigenetic mechanisms[ 4 , 6 ]. KBs act as anti-inflammatory signaling molecules, modulating inflammatory and immune functions through both receptor-dependent and receptor-independent pathways. BHB inhibits inflammatory responses and regulates immune cell functions by binding and activating hydroxy carboxylic acid receptor 2 (HCAR2) or directly modulating intracellular signaling pathways[ 7 – 9 ]. KBs regulate reactive oxygen species (ROS) metabolism and maintain redox homeostasis by activating cytoprotective regulators such as Nrf2, SIRT1/3, and AMPK[ 10 , 11 ]. Ketone metabolism can be divided into ketogenesis and ketolysis. HCC undergoes metabolic reprogramming, leading to abnormalities in both processes of ketone metabolism[ 12 – 14 ]. In HCC tissues, the expression of the rate-limiting enzyme for ketogenesis, 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2), is downregulated, resulting in reduced BHB production. This phenomenon has been associated with HCC pathogenesis and staging and impacts tumor proliferation and metastasis[ 15 , 16 ]. Conversely, the expression of the rate-limiting ketolysis enzyme 3-oxoacid CoA transferase 1 (OXCT1) is significantly upregulated in HCC, enhancing ketolysis. In HCC, increased ketolysis plays a critical role in suppressing AMPK activation, thereby preventing excessive autophagy in HCC cells and promoting tumor proliferation[ 17 – 19 ]. Single-cell RNA sequencing (scRNA-seq) resolves transcriptional profiles at individual-cell resolution, overcoming the limitations of bulk sequencing in dissecting cellular heterogeneity. Its workflow typically includes cell isolation, cDNA synthesis, and library construction[ 20 ]. By cataloging cell types and states across tissues, scRNA-seq has revealed novel subtypes in normal and diseased organs[ 21 ]. For example, in cancer research, the complexity of the tumor microenvironment (TME), such as immune cell infiltration and stromal interactions, is deciphered[ 22 ]. Integration of clinical and pathological data with scRNA-seq results enables the discovery of diagnostic and prognostic biomarkers and therapeutic-related cell states and supports precise molecular subtyping and personalized therapies[ 23 ]. Recent studies have increasingly used scRNA-seq to characterize tumor microenvironment changes in HCC, identify predictive biomarkers, and establish theoretical foundations for clinical interventions[ 24 – 26 ]. This study integrated bulk RNA sequencing and single-cell RNA sequencing data to systematically identify prognostic biomarker genes associated with key cell subpopulations and ketone metabolism in HCC via hierarchical dynamic weighted gene coexpression network analysis (hdWGCNA) and machine learning-based bioinformatics approaches. A survival prediction model was constructed to stratify HCC patient survival outcomes, and functional mechanism studies further revealed the synergistic roles of key cell type and metabolic pathways in tumor progression. This research aims to provide molecular evidence for precision-based prognosis assessment in HCC, deepen the understanding of ketone metabolism-tumor microenvironment crosstalk mechanisms, and offer a theoretical basis for developing metabolically reprogrammed personalized immunotherapy strategies. 2. Methods 2.1 Data sources Three HCC-related transcriptome datasets were included in this study. The TCGA-HCC dataset was obtained from The Cancer Genome Atlas (TCGA) database ( https://portal.gdc.cancer.gov/ ) on February 10, 2025, and included 368 HCC tissue samples (with survival and clinical information) and 50 normal tissue samples. Additionally, the GSE14520 (platform: GPL571) and GSE149614 (platform: GPL24676) datasets were obtained from the Gene Expression Omnibus (GEO) database ( https://www.ncbi.nlm.nih.gov/geo/ ). The GSE14520 dataset included 130 HCC tissue samples with survival information. GSE149614 is a single-cell RNA-seq (scRNA-seq) dataset comprising 10 primary tumor (HCC) tissue samples and 8 normal tissue samples. All the samples were collected from human subjects. Notably, the HCC and normal samples were defined as the HCC and control groups, respectively. Moreover, 225 ketone body (KB) metabolism-related genes (KMRGs) were identified from the "GOBP_CELLULAR_KETONE_METABOLIC_PROCESS" pathway (GO: 0042180) in the Molecular Signatures Database (MSigDB) ( https://www.gsea-msigdb.org/gsea/msigdb ) ( Additional file 1 ). 2.2 Exploration of cell type profiles and functions through scRNA-seq analysis An extensive investigation of the scRNA-seq data from the GSE149614 dataset was conducted to explore the underlying mechanisms of HCC progression. First, the cell type profiles in HCC were explored. Initially, the raw data were read via the CreateSeuratObject function in the Seurat package (v 4.1.0)[ 27 ]. To ensure data quality, rigorous quality control criteria were applied to retain high-quality cells (200 ≤ detected genes (nFeature) ≤ 6,000, mitochondrial gene content (percent.mt) < 10%, total gene count (nCount_RNA) < 20,000). The FindVariableFeatures function was subsequently employed to identify genes with high variation coefficients across cells via the variance-stabilizing transformation (VST) method. The top 2,000 most variable genes (highly variable genes) were retained for further analyses. NormalizeData and ScaleData functions were subsequently used to normalize the data, and RunPCA was applied to perform principal component analysis (PCA). An elbow plot was generated via the elbowplot function to visualize the proportion of variance explained by each principal component (PC). The statistical significance of each PC was assessed via the JackStraw function. PCs with p < 0.05 were retained for analyses. Afterwards, unsupervised clustering analysis of the cells was conducted via the FindNeighbors and FindClusters functions (resolution = 0.1). Through uniform manifold approximation and projection (UMAP) conducted via the RunUMAP function, distinct cell clusters were recognized. Cell cluster annotation was performed by referencing marker genes from the literature [ 28 ] and the CellMarker2.0 database ( http://biobbio-bigdata.hrbmu.edu.cn/CellMarker/ ), with marker gene expression illustrated for each cell type via the DotPlot function. The functional analysis of each cell type was performed via the analyze_sc_clusters function in the ReactomeGSA package (v 1.12.0) [ 29 ] to reveal the related biological functions. The result was extracted via pathway function. 2.3 Ascertainment of the key cell type through KMRG activity and proportion analyses The AUCell package (version 1.12.0) [ 30 ] was utilized to evaluate KMRG activity at single-cell resolution within the GSE149614 dataset. A predefined gene set comprising 225 KMRGs was used as input. For each cell, an area under the curve (AUC) value was computed by ranking genes on the basis of their expression levels, thereby estimating the activation level of the gene set. Higher AUC values indicate a greater number of highly expressed KMRGs in individual cells. To define the threshold for identifying cells with active expression of the gene set, the AUCell_exploreThresholds function was applied (AUC > 0.7). Remarkably, AUC values were projected onto a UMAP space via the FeaturePlot function in the Seurat package (v 4.1.0), enabling visualization of KMRG activity across different cell populations. Additionally, the proportions of different cell types were visualized via the ggplot2 package (v 3.5.1) [ 31 ], with differences in proportions between the HCC and control groups analyzed via the Wilcoxon rank-sum test (p < 0.05). The key cell type was ultimately determined on the basis of its notably high KMRG activity and significantly greater proportion in HCC. 2.4 High-dimensional weighted gene coexpression network analysis (hdWGCNA) Within the GSE149614 dataset, the key cell type and the gene expression profiles were extracted for further analyses. Following the previously described methodology, the key cell type was redimensionalized and reclustered (resolution = 0.1) via the Seurat package (v 4.1.0), which led to the identification of distinct cell subtypes. On the basis of all the samples from the GSE149614 dataset, hdWGCNA was subsequently conducted via the hdWGCNA package (v 0.4.5) [ 32 ] to identify genes associated with key cell subtypes. All analyses were conducted following the official standard workflow ( https://smorabit.github.io/hdWGCNA/articles/basic_tutorial.html ). To eliminate low-quality data, the MetacellsByGroups function was used to develop a metacell-level gene expression matrix. The TestSoftPowers function was subsequently applied to select the optimal soft power threshold on the basis of the scale-free topology criterion (scale-free R2 > 0.8) and mean connectivity (approaching 0). A coexpression network was subsequently established via the ConstructNetwork function, and genes were grouped into distinct modules (k = 25, min cells = 100, max shared = 15, target metacells = 1,000). Spearman correlation analysis was subsequently performed with the psych package (v 2.1.6) [ 33 ] to evaluate the associations between gene modules and key cell subtypes. Genes within modules that strongly correlated with one of the key cell subtypes were integrated and defined as key module genes (|correlation coefficient (cor)| > 0.3, p < 0.05). 2.5 Function and protein‒protein interaction (PPI) analyses of candidate genes Within the TCGA-HCC dataset, to identify differentially expressed genes (DEGs) in HCC, the DESeq2 package (v 1.42.0) [ 33 ] was used to perform differential expression analysis between the HCC and control groups (|log 2 -fold change (FC)| > 1, p < 0.05). Next, to identify genes related to key cell type and KB metabolism in HCC, the key module genes, DEGs, and KMRGs were intersected via the ggvenn package (v 0.1.10) (37), resulting in candidate genes. Potential roles of the candidate genes in HCC were subsequently explored with the clusterProfiler package (v 4.10.1) [ 34 ] to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses (p < 0.05). To explore candidate gene interactions at the protein level, the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database ( https://string-db.org/ ) was used to establish a protein–protein interaction (PPI) network (interaction score ≥ 0.4). Cytoscape software (v 3.9.1) [ 35 ] was used for visualization. 2.6 Risk model development and validation Within the TCGA-HCC dataset, the potential value of KMRGs for predicting overall survival (OS) was assessed on the basis of HCC samples with survival information. Specifically, the survival package (v 3.7.0) [ 36 ] was initially applied to conduct univariate Cox analysis (hazard ratio (HR) ≠ 1, p 0.05) conducted with the cox.zph function was applied to analyze the retained genes. The genes that passed the pH assumption test were defined as candidate prognostic genes. Thereafter, the glmnet package (v 4.1.8) [ 37 ] was applied to construct a 10-fold cross-validated least absolute shrinkage and selection operator (LASSO) model to identify prognostic genes whose coefficients were not penalized to 0. On the basis of prognostic genes related to key cell type and KB metabolism, a risk model was constructed according to this formula (the coef and expr represented the LASSO coefficient and expression of each prognostic gene, respectively): The HCC samples were subsequently classified into high-risk (HRG) and low-risk (LRG) groups according to the optimal risk score cutoff value calculated via the survminer package (v 0.4.9) [ 38 ]. Then, within the HRG and LRG, the ggplot2 package (v 3.5.1) was employed to determine the risk score and survival state distribution. The prognostic gene expression trends are shown via the pheatmap package (v 1.0.12) [ 39 ]. In addition, Kaplan‒Meier (KM) survival curves for HRGs and LRGs were generated through the survminer package (v 0.4.9), and the survival probability difference was subjected to the log-rank test (p < 0.05). The survivalROC package (v 1.0.3) [ 40 ] was utilized to assess the predictive ability of the risk model by generating receiver operating characteristic (ROC) curves at 1-, 2-, and 3-year time points (AUC value > 0.6). To investigate the model accuracy and generalizability, the risk model was verified with the GSE14520 dataset. 2.7 Clinical stratification and KM survival analyses To investigate the associations between clinical characteristics and risk scores related to key cell type and KB metabolism, HCC samples with survival and clinical information from the TCGA-HCC dataset were first classified into distinct clinical subgroups on the basis of age (≤ 61 and > 61), sex (female and male), clinical stage (stage 1/2 and stage 3/4), T stage (T1/2 and T3/4), N stage (N0 and N1), and M stage (M0 and M1). Thereafter, risk score differences across clinical subgroups were analyzed via the Wilcoxon rank-sum test (p < 0.05). Additionally, the ability of the risk score to predict survival was further explored on the basis of clinical characteristics (age, sex, race (white, Asian, and black), clinical stage, T stage, N stage, and M stage). Specifically, within different clinical subgroups, the ggsurvplot function in the survival package (v 3.7.0) was applied to plot KM survival curves for HRG and LRG, and the survival differences were evaluated via the log-rank test (p < 0.05). 2.8 Gene set variation analysis (GSVA) GSVA package (v 1.42.0) [ 41 ] was utilized to conduct GSVA between HRGs and LRGs in the TCGA-HCC dataset (p < 0.05). In detail, HALLMARK pathways from MSigDB were used as a reference gene set. The pathway scores for HRGs and LRGs were calculated, and the differences between groups were evaluated via the limma package (v 3.58.1) [ 42 ] (p 2). A t value > 2 indicated that the pathway was activated in the HRG, whereas a negative t value < 2 suggested that the pathway was suppressed. 2.9 Tumor microenvironment (TME) analysis Initially, the ssGSEA algorithm in the GSVA package (v 1.42.0) was applied to calculate the infiltration scores of 28 immune cell types[ 43 ] within samples in the HRG and LRG. The differential immune infiltrating cell types between the HRG and LRG groups were recognized via the Wilcoxon rank-sum test (p < 0.05). The psych package (v 2.1.6) was subsequently applied to perform Spearman correlation analysis, which correlated differential immune infiltrating cell types with each other and with metabolism-related prognostic genes (|cor| > 0.3, p < 0.05). 2.10 Immunotherapy response prediction The immunophenotype score (IPS) is a predictor of the checkpoint blocker response[ 44 ]. Within the TCGA-HCC dataset, the tumor immunogenicity of patients in the HRG and LRG was assessed via the IPSs. Specifically, IPS data were obtained from The Cancer Immunome Atlas (TCIA) database ( https://tcia.at/ ) on January 13, 2025. Differences in the IPSs between the HRG and LRG were assessed via the Wilcoxon rank-sum test (p < 0.05). IPS categorization considers key immune checkpoints, including CTLA-4 and PD-1. A total of 4 distinct IPS categories (IPS_CTLA4_neg_PD1_neg, IPS_CTLA4_neg_PD1_pos, IPS_CTLA4_pos_PD1_neg, and IPS_CTLA4_pos_PD1_pos) were included (pos: positive; neg: negative). High IPSs are related to high immunogenicity. 2.11 Drug sensitivity analysis To evaluate the effects of risk scores related to key cell type and KB metabolism on drug sensitivity, the half-maximal inhibitory concentration (IC50) values for 138 common drugs were estimated for HRG and LRG patients from the TCGA-HCC dataset via the OncoPredict package (v 1.2)[ 44 ]. Differences between groups were compared via the Wilcoxon rank-sum test (p < 0.05). The top 15 drugs with the most significant differences are shown via the ggplot2 package (v 3.5.1). Furthermore, the psych package (v 2.1.6) was used to conduct Spearman correlation analysis between risk scores and IC50 values for all 138 drugs. The top 10 drugs showing significant positive/negative correlations with risk scores are highlighted in a plot generated with the ggplot2 package (v 3.5.1) (|cor| > 0.3, p < 0.05). 2.12 Cell‒cell communication and cell trajectory analyses Within the GSE149614 dataset, cell‒cell interactions among various cell types were systematically explored across all samples via the CellChat package (v 1.6.1)[ 45 ] (p < 0.05, log2 mean expression of interacting molecules ≥ 0.1). In addition, ligand‒receptor (L‒R) pairs that mediate intercellular signaling were also examined (p < 0.01). Moreover, trajectory analysis was conducted via reduceDimension and plot_cell_trajectory functions in the Monocle package (v 2.30.1)[ 46 ], enabling the simulation of the differentiation process of the key cell type in the HCC and control groups. Additionally, the plot_genes_in_pseudotime function in the Monocle package (v 2.30.1) was applied to explore the dynamic expression patterns of prognostic genes related to key cell type and KB metabolism during the key cell type differentiation trajectory. 2.13 Reverse transcription‒quantitative PCR (RT‒qPCR) Prognostic gene expression was validated in clinical samples. Specifically, RNAs from 5 HCC samples and 5 control samples were isolated from individuals via TRIzol reagent (R401-01, Ambion, America). The collection was performed at Guangzhou Red Cross Hospital. The isolated RNAs were then used for cDNA synthesis via the Hifair® III 1st Strand cDNA Synthesis SuperMix for a qPCR kit (11141ES60, Yisheng, China). RT‒qPCR was subsequently performed using 2×Universal Blue SYBR Green qPCR Master Mix (G3326‒05, Servicebio, China). The primers used for the prognostic genes and the internal reference gene (GAPDH) are listed in Additional file 2. Following the RT‒qPCR procedure, the 2-ΔΔCт method was applied to identify expression profiles. A t test (p < 0.05) was employed to evaluate intergroup differences, and data visualization was conducted via GraphPad Prism 5 software (v 8.0)[ 47 ]. Ethical approval was granted by the Ethics Committee of Guangzhou Red Cross Hospital on February 16, 2019. (approval number: 020–34403034). All experiments were performed in accordance with relevant named guidelines and regulations and the authors complied with the ARRIVE guidelines. 2.14 Statistical analysis R software (v 4.2.3) was used to conduct the bioinformatics analyses. Notably, the Wilcoxon rank-sum test, the log-rank test, and the t test were employed in this study to assess differences between specific groups, with the significance threshold set at p < 0.05. 3. Results 3.1 Various functions of cell types in HCC Potential mechanisms underlying the progression of HCC were explored at the single-cell level within GSE149614 dataset. Cell type profiles in HCC were initially explored. Specifically, through data filtration, 52,471 cells and 25,712 genes were retained for further analyses ( Additional file 3a-b ). Additionally, a variation in these genes was depicted in Additional file 3c , highlighting 2,000 highly variable genes. After selecting the top 20 PCs (p < 0.0001) for UMAP dimensionality reduction according to elbow plot and PCA replacement test, 11 distinct cell clusters were determined (Fig. 1 a-c). Cell clusters were annotated into 9 types using marker genes ( Additional file 4 ), including T cells, myeloid cells, hepatocytes, natural killer (NK) cells, endothelial cells (ECs), fibroblasts, plasma cells, B cells, and epithelial cells (Fig. 1 d). The marker gene expression was illustrated in Fig. 1 e. Notably, these cell types were related to functional pathways like "biogenic amines are oxidatively deaminated to aldehydes by MAOA and MAOB" and "ethanol oxidation" (Fig. 1 f). The elucidation of these cell-related functions enhanced the understanding of the cellular processes involved and potentially offered new avenues for therapeutic strategies targeting HCC. The activities of KMRGs across various cell types were illustrated in Fig. 1 g. Hepatocytes demonstrated the highest KMRG activity. Furthermore, T cells represented the most abundant cell population across all samples (Fig. 1 h). Specifically, T cells and myeloid cells accounted for the highest proportions in the control and HCC groups, respectively. The proportion of hepatocytes was markedly increased in the HCC group (p < 0.01) (Fig. 1 i). Considering both the notably high KMRG activity and significantly higher proportion in HCC, hepatocytes were ultimately selected as the key cell type. 3.2 Relevant functional pathways and the PPI network of candidate genes In GSE149614 dataset, hepatocytes were extracted for further analyses. Through PCA, the top 20 PCs were selected (Fig. 2 a-b). Subsequent dimensionality reduction and reclustering revealed 3 hepatocyte subtypes (HC1, HC2, HC3) (Fig. 2 c). Next, hdWGCNA was performed based on these subtypes. Specifically, the optimal soft power threshold was confirmed to be 3, and a gene co-expression network was established, with genes clustered into 17 distinct modules (excluding the gray module) (Fig. 2 d-e). Then, 1,837 key module genes were acquired from the blue, greenyellow, green, yellow, and turquoise modules, which exhibited strong correlations with specific hepatocyte subtypes (|cor| > 0.3, p < 0.05) (Fig. 2 f). It was worth mentioning that the blue module demonstrated the strongest positive and negative associations with HC1 (cor = 0.49) and HC2 (cor = -0.52), respectively (p < 0.0001). Within TCGA-HCC dataset, 4,590 DEGs in the HCC group were acquired, including 3,383 upregulated and 1,207 downregulated genes (p < 0.05) (Fig. 2 g-h). Remarkably, after intersecting key module genes, DEGs, and KMRGs, 10 candidate genes related to hepatocytes and KB metabolism were identified, including PPARGC1A, PDK4, HSD17B6, APOC3, OXCT1, EGR1, SRD5A1, TDO2, FGF19, and FABP1 (Fig. 2 i). Candidate genes were significantly enriched in GO terms such as "cellular ketone metabolic process", as well as in KEGG pathways like "steroid hormone biosynthesis" (p < 0.05) (Fig. 2 j-k). Moreover, the constructed PPI network contained 4 interaction pairs (such as TIMP1-EZR and TIMP1-SFRP1) involving 7 proteins, elucidating the interplay of specific candidate genes at the protein level (Fig. 2 l). These results aided in understanding the multiple roles of hepatocytes and KB metabolism in HCC progression. 3.3 Strong predictive power of KMRGs for HCC prognosis demonstrated by a risk model Within TCGA-HCC dataset, 5 candidate genes linked to OS were retained through univariate Cox analysis (p 0.05) ( Additional file 5 ). Among them, PPARGC1A, PDK4, HSD17B6, and APOC3 were linked to better HCC prognosis (HR 1). Therefore, they were selected as candidate prognostic genes and subjected to LASSO regression analysis. PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were ultimately identified as prognostic genes as their regression coefficients were not penalized to 0 (optimal lambda = 0.00264) (Fig. 3 b-c). After screening prognostic genes, a risk model related to hepatocytes and KB metabolism (risk score = (-0.197860092) × PPARGC1A expression level + (-0.086179799) × PDK4 expression level + (-0.007066472) × HSD17B6 expression level + (-0.022773615) × APOC3 expression level + 0.208555505 × OXCT1 expression level) was constructed. Subsequently, HCC patients in TCGA-HCC dataset were divided into HRG and LRG (181 : 187) based on the optimal risk score cut-off value (-0.9058876). Risk score and survival state distribution within risk groups revealed that the number of dead patients increased with increasing risk scores (Fig. 3 d). Notably, PPARGC1A, PDK4, HSD17B6, and APOC3 exhibited higher expression in LRG, while OXCT1 demonstrated elevated expression in HRG (Fig. 3 e). Moreover, KM survival curves showed that HCC patients in LRG had markedly higher survival probabilities (p < 0.0001) (Fig. 3 f). Additionally, AUC values of ROC curves at 1-, 2-, and 3-year time po0ints all exceeded 0.6, reflecting notable predictive ability of this risk model (Fig. 3 g). Furthermore, to finalize assessment, the risk model was subjected to validation within GSE14520 dataset. Likewise, HCC patients in this dataset were divided into HRG and LRG (53 : 77) based on the optimal risk score cut-off value (-0.9715922). Results revealed by risk score and survival state distribution (Fig. 3 h), expression analysis (Fig. 3 i), KM survival curves (p < 0.05) (Fig. 3 j), and ROC curves (AUC values all exceeded 0.6) (Fig. 3 k) were largely consistent with those in TCGA-HCC dataset. The generalizability of this risk model was superior, and it might be a valuable tool for risk assessment in clinical practice for HCC. 3.4 Outstanding the clinical and survival relevance of risk scores In TCGA-HCC dataset, differences in risk scores related to hepatocytes and KB metabolism among various clinical subgroups were analyzed (Fig. 4 a). Risk scores were notably linked to clinical stage and T stage, with notably higher risk scores observed in patients with advanced stages (stage 3/4; T3/4) (p 61), gender (male), race (White and Asian), clinical stage (stage 1/2 and stage 3/4), T stage (T1/2 and T3/4), N stage (N0), and M stage (M0) (Fig. 4 b). The survival prediction ability of risk scores was excellent. Summarily, the robust clinical and prognostic significance of risk scores were found, emphasizing the potential roles of hepatocytes and ketone metabolism in the progression and outcome of HCC. 3.5 The prognostic genes were associated with metabolism Through GSEA, pathways associated with prognostic genes were analyzed in TCGA-HCC dataset. PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were significantly enriched in 58, 66, 72, 66, and 69 pathways, respectively (adjusted p < 0.05) ( Additional files 6–10 ). The top 5 most significant pathways of each prognostic gene were shown in Fig. 5 a-e. It was worth mentioning that they were co-enriched in 6 critical pathways, including "primary bile acid biosynthesis", "spliceosome", "cell cycle", "ubiquitin mediated proteolysis", "beta-alanine metabolism", and "homologous recombination". It was suggested that KB metabolism might influence HCC progression by altering these pathways. 3.6 Crucial functional pathways and TME profiles altered by risk score Within the TCGA-HCC dataset, biological pathways altered by risk scores related to hepatocytes and KB were explored through GSVA. In detail, the activities of pathways like "MYC targets V1" and "unfolded protein response" were activated in HRG, while the activities of pathways such as "fatty acid metabolism" and "bile acid metabolism" were suppressed in HRG (p < 0.05) (Fig. 6 a). Hepatocytes and KB metabolism might exert critical roles in HCC progression by modulating these pathways. Remarkably, TME profiles of HRG and LRG in the TCGA-HCC dataset were explored. The infiltration levels of central memory CD4 T cells and monocytes were relatively high across HCC patients. There were 14 differential immune infiltrating cell types between HRG and LRG (p < 0.05) (Fig. 6 b). Among them, 11 cell types exhibited increased infiltration levels in HRG, such as activated CD4 T cells, activated dendritic cells, and myeloid-derived suppressor cells (MDSCs). Furthermore, the associations of differential immune infiltrating cell types with each other and with prognostic genes were explored (Fig. 6 c). Most of the differential immune infiltrating cell types demonstrated significantly and positively strong correlations with each other. The strongest correlation was noted between immature B cells and activated B cells (cor = 0.86, p < 0.001). Regarding the relationships between differential immune infiltrating cell types and prognostic genes, central memory CD4 T cells exhibited the strongest positive correlation with OXCT1 (cor = 0.44, p < 0.001), while activated CD4 T cells showed the strongest inverse correlation with PDK4 (cor = -0.39, p < 0.001). These immune characteristics highlighted the intricate associations of TME in HCC with hepatocytes and KB metabolism. 3.7 Potential of the risk score as a predictor of immunotherapy and drug treatment outcomes In TCGA-HCC dataset, differences in IPS between the HRG and LRG were assessed under various immune checkpoint expression conditions. Notably, under the CTLA4_neg_PD1_neg condition, the LRG exhibited significantly higher IPSs than the HRG (p < 0.01), indicating a potentially greater tumor immunogenicity and enhanced responsiveness to immune checkpoint blockade therapy (Fig. 7 a). Regarding drug sensitivity, it was found that the IC50 values for drugs like BMS.708163 were lower in LRG, whereas drugs such as camptothecin exhibited lower IC50 values in HRG (p < 0.05) (Fig. 7 b). Lower IC50 values were indicative of higher drug efficacy. Notably, 4 drugs such as BMS.708163 exhibited significant positive correlations with risk scores, while 34 drugs like camptothecin were significantly and negatively correlated with risk scores (|cor| > 0.3, p < 0.0001) (Fig. 7 c). The differential drug sensitivities might be due to different drug metabolism mechanisms in HCC patients with varying risk scores. Taken together, risk scores related to hepatocytes and KB metabolism demonstrated superior potential in predicting immunotherapy and drug treatment outcomes for HCC patients. 3.8 Complex intercellular communication patterns Analysis of cell-cell communication in GSE149614 dataset revealed extensive intercellular interactions within microenvironment. Notably, fibroblasts, epithelial cells, hepatocytes, and ECs were involved in a greater number and intensity of interactions compared to other cell types ( Additional file 11a-d ). It was worth mentioning that hepatocytes exhibited interactions with all other cell types, with particularly frequent and relatively strong interactions observed between hepatocytes and myeloid cells. Compared to the control group, both the number and strength of interactions between hepatocytes and myeloid cells were reduced in the HCC group. Conversely, both the frequency and strength of the interactions between hepatocytes and other cell types were increased. These dynamic changes suggested a potential shift in the intercellular communication landscape during HCC progression. Furthermore, it was found that T cells participated in widespread intercellular signaling through numerous L-R interactions. In the control group, the GZMA-F2R pair demonstrated the highest communication probability in the T cell-to-EC interaction ( Additional file 11e ). In the HCC group, the MIF-(CD74 + CXCR4) pair exhibited the highest communication probability in the T cell-to-B cell interaction, indicating a potentially critical signaling axis in the HCC microenvironment ( Additional file 11f ). The T cell-to-hepatocyte interaction primarily involved the NAMPT-INSR pair. Overall, deciphering these intricate intercellular communication networks offered deeper insight into the cellular crosstalk underlying HCC progression. 3.9 Dynamic expression of prognostic genes during hepatocyte differentiation In GSE149614 dataset, differentiation trajectories of hepatocytes over time were tracked and presented in Additional file 12a , showing that hepatocytes differentiated into distinct subtypes at different stages. Three differentiation states were observed, with most cells in state 2. In both control and HCC groups, hepatocytes were predominantly in the intermediate and late stages of differentiation. Moreover, the expression dynamics of prognostic genes related to hepatocytes and KB metabolism were also illustrated during hepatocyte differentiation ( Additional file 12b-c ). Overall, the expression of PPARGC1A, PDK4, HSD17B6, and APOC3 exhibited an upward trend as differentiation progressed. Specifically, PPARGC1A expression remained relatively stable during the early and intermediate stages of differentiation, followed by an increase at the late stage before reaching a plateau. PDK4 expression initially decreased, then increased during the later stages, and eventually stabilized. HSD17B6 expression showed a continuous increase throughout the differentiation process. APOC3 expression increased during the early stages and stabilized during the late phase. In contrast, OXCT1 expression did not display significant changes across differentiation stages. In summary, the dynamic expression of these prognostic genes across hepatocyte differentiation stages might related to HCC progression. 3.10 RT‒qPCR validation of prognostic genes Expression of prognostic genes related to hepatocytes and KB metabolism was validated in clinical samples. Compared to the control group, expression of PPARGC1A, PDK4, HSD17B6, APOC3 was significantly decreased in HCC, while OXCT1 exhibited elevated expression (p < 0.05) ( Additional file 13a-e ). Significant differential expression of prognostic genes further confirmed their prognostic significance in HCC. 4. Discussion HCC is a highly heterogeneous malignant tumor, and its progression is closely associated with metabolic reprogramming. Disruptions in ketone body metabolism may affect the tumor microenvironment and patient prognosis[ 18 ]. In this study, multi-omics data from TCGA and GEO databases were integrated. Advanced bioinformatics methods were employed to identify five core prognostic genes (PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1) that are closely related to HCC function and ketone metabolism regulation. A specific HCC prognostic risk model was further constructed, while the dual clinical significance of hepatocyte-related KMRGs in prognostic assessment and tumor immune microenvironment regulation was systematically validated.Among the five core prognostic genes identified by the risk model, PPARGC1A, PDK4, HSD17B6, and APOC3 were significantly downregulated in HCC, whereas OXCT1 was markedly upregulated. Existing evidence indicates that these genes play critical roles in HCC cell proliferation, tumorigenicity, invasiveness, and drug resistance. The PPARGC1A gene encodes the transcriptional coactivator PGC-1α. Recent studies have identified PPARGC1A as a key mediator of metabolic reprogramming in cancer, enabling tumor cells to adapt to fluctuating metabolic demands[ 48 ]. Mechanistically, PGC-1α impedes HCC metastasis by suppressing the Warburg effect and aerobic glycolysis[ 49 ]. Conversely, SIRT1 promotes HCC metastasis by enhancing PGC-1α-mediated mitochondrial biogenesis[ 50 ]—a finding that reveals the dual regulatory role of PPARGC1A in HCC pathogenesis. As a mitochondrial matrix enzyme, PDK4 regulates glucose metabolism by phosphorylating the E1α subunit to inhibit pyruvate dehydrogenase complex (PDC) activity[ 51 ]. Loss of PDK4 exacerbates the proliferation, tumorigenicity, motility, and invasiveness of HCC[ 52 ]. Notably, this study found a significant negative correlation between activated CD4 + T cells and PDK4 expression, suggesting that PDK4 may be involved in tumor immune infiltration. HSD17B6 is an endoplasmic reticulum (ER)-localized oxidoreductase that regulates steroid hormone metabolism by catalyzing the reversible conversion between hydroxyl and ketone groups—particularly in androgen biosynthesis. Decreased HSD17B6 expression in HCC disrupts cell proliferation, migration, invasion, and androgen metabolism, while also affecting tumor-infiltrating immune cells[ 53 , 54 ]. While the role of HSD17B6 in HCC progression and immune checkpoint therapy response is well established, the mechanistic interaction between HSD17B6 and ketone metabolism remains to be further explored. As a key small apolipoprotein in lipid metabolism, APOC3 regulates triglyceride homeostasis by inhibiting lipoprotein lipase activity and delaying hepatic clearance of triglyceride-rich lipoproteins[ 55 ]. Recent studies have proposed APOC3 as a potential prognostic biomarker for hepatitis B virus (HBV)-associated HCC; its early upregulation is linked to steroid metabolism, PPAR signaling pathways, and fatty acid metabolism[ 56 , 57 ]. The expression pattern of APOC3 observed in this study differs from previous reports, which may be attributed to the temporal and microenvironmental heterogeneity of HCC and requires further validation. As a lysine succinyltransferase and a rate-limiting enzyme in ketone body metabolism, OXCT1 catalyzes the initiating and rate-limiting step of ketolysis. This enzyme converts extrahepatic ketone bodies into acetoacetyl-CoA, which then enters the tricarboxylic acid (TCA) cycle to participate in oxidative phosphorylation and ATP production[ 58 ]. Recent research evidence highlights the core role of this enzyme in metabolic reprogramming of malignant tumors[ 59 ]. In HCC, OXCT1 expression is significantly upregulated. It drives ketone body catabolism to provide energy for tumor cell proliferation, invasion, and metastasis[ 17 , 19 , 60 ]. These findings establish OXCT1 as a key regulator in HCC hepatocyte function and ketone body metabolism, though its specific mechanism of action remains to be fully elucidated.Notably, a study by the Zhu Chuxu team demonstrated that conditional knockout of OXCT1 in macrophages is significantly associated with increased OXCT1 expression in tumor-associated macrophages (TAMs) and poor prognosis in HCC patients. This suggests that OXCT1-mediated metabolic reprogramming of TAMs toward a pro-tumor phenotype accelerates tumor progression[ 17 ]. Our data reveal a positive correlation between OXCT1 levels and central memory CD4 + T cells, implying that this enzyme is involved in regulating tumor immune dynamics.Overall, these findings indicate that inhibiting OXCT1 is a potential therapeutic strategy to block HCC progression by simultaneously targeting metabolic and immune pathways. However, further mechanistic studies are required to clarify its context-dependent role in HCC pathogenesis and therapeutic resistance. Functional enrichment analysis revealed that PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 are co-enriched in the ubiquitin-mediated proteolysis pathway. This suggests that dysregulation of these metabolic genes may affect HCC progression by disrupting protein homeostasis.The ubiquitin-proteasome system (UPS) is the most critical protein quality control mechanism in eukaryotic cells, responsible for eliminating misfolded, damaged, or regulatory proteins[ 61 ]. In HCC, UPS dysregulation is closely associated with tumorigenesis, invasiveness, and therapeutic resistance[ 62 ], and it regulates the stability of key tumor suppressors (e.g., p53, PTEN) and oncoproteins (e.g., β-catenin, c-Myc)[ 63 ]. Additionally, the UPS participates in the turnover of metabolic enzymes, regulates the metabolic plasticity of tumor cells, and mediates endoplasmic reticulum (ER)-associated degradation (ERAD) to maintain ER homeostasis[ 64 ]. HCC cells often exhibit enhanced proteasome activity and upregulation of specific E3 ubiquitin ligases; this adaptation enables them to sustain protein homeostasis under metabolic stress and support rapid proliferation. Among the five prognostic genes screened in this study, the PGC-1α protein itself is a key substrate of the UPS[ 65 ]. In HCC, the downregulation of PGC-1α may be partially attributed to increased ubiquitination-mediated degradation. Furthermore, PGC-1α-regulated mitochondrial biogenesis relies on the UPS to clear damaged mitochondrial proteins; its dysregulated expression may disrupt mitochondrial quality control (MQC), leading to the accumulation of dysfunctional mitochondria and exacerbated oxidative stress. HSD17B6 is localized to the ER, and its function as a steroid metabolic enzyme is closely linked to ER protein folding quality control[ 66 ]. Downregulation of HSD17B6 may result in the accumulation of androgen metabolic intermediates; these hydrophobic steroids may interfere with ER membrane integrity and the protein folding environment, triggering ER stress and activating the ERAD pathway—a key mechanism for maintaining ER homeostasis[ 67 ]. Our study found that the most significantly altered biological pathways in HCC, based on hepatocyte and KB metabolism-related risk scores, are "MYC Targets V1" and the "unfolded protein response (UPR)". Activation of the UPR pathway is closely associated with the aforementioned UPS dysregulation. When the ER’s protein-folding capacity is overwhelmed, the UPR not only initiates a transcriptional adaptation program but also recruits the UPS via the ERAD pathway to clear misfolded proteins[ 68 ]. In ERAD, misfolded ER proteins are retrotranslocated to the cytoplasm, labeled by E3 ubiquitin ligases (e.g., HRD1, gp78), and then degraded by the 26S proteasome[ 69 ]. Thus, the UPR and ubiquitin-mediated proteolysis pathway functionally form a tightly coupled quality control network, collectively maintaining protein homeostasis in HCC cells under metabolic stress.The c-Myc protein itself is a key substrate of the UPS, and its stability is regulated by multiple E3 ubiquitin ligases[ 70 ]. Under metabolic stress, UPR activation can affect c-Myc stability through multiple mechanisms: On one hand, enhanced proteasome activity mediated by ERAD may accelerate c-Myc turnover[ 71 ]; on the other hand, ER stress-induced translational inhibition (via PERK-eIF2α) may selectively upregulate the synthesis of oncoproteins such as c-Myc[ 72 ]. The activation of "MYC Targets V1" observed in this study may reflect the coordinated dysregulation of this ubiquitin system-metabolism-proliferation signaling axis.Studies have indicated functional crosstalk between the UPR and PGC-1α (encoded by the PPARGC1A gene). As a master regulator of mitochondrial biogenesis and energy homeostasis, PGC-1α transcription is regulated by CREB3—an ER-resident transcription factor that also coordinates UPR activation[ 73 , 74 ]. While existing studies have confirmed the association between UPR dysregulation and HCC development, the specific molecular mechanisms linking UPR signaling to HCC progression remain unclear.Based on the above evidence, we propose the following integrated mechanistic model: In HCC, dysregulation of metabolic genes such as PPARGC1A and HSD17B6 disrupts the UPS through multiple pathways, thereby triggering ER stress and UPR activation. The UPR further recruits the ubiquitin system via the ERAD pathway to clear misfolded proteins, while regulating the stability and transcriptional activity of key oncoproteins such as c-Myc. This cascade forms a loop that supports HCC progression. The complex triangular relationship among MYC Targets V1, the UPR, and PPARGC1A/PGC-1α reveals a molecular interaction network that may explain metabolic adaptation and therapeutic resistance in HCC, presenting broad research prospects. This study, for the first time, systematically reveals the key role of KMRGs in hepatocytes in the prognosis of HCC patients by integrating single-cell and transcriptome data. We successfully constructed a risk prediction model based on five key genes: APOC3, HSD17B6, PDK4, PPARGC1A, and OXCT1. This discovery not only provides new insights into the metabolic reprogramming mechanism of HCC but also lays an important foundation for clinical prognosis assessment and the development of personalized treatment strategies. However, this study has several limitations. First, the sample size of single-cell data is relatively limited (n = 18), which may not fully reflect the tumor heterogeneity of HCC. Second, although we identified candidate genes through bioinformatics methods, their specific regulatory mechanisms in ketone metabolism still need to be further verified through in vitro and in vivo experiments. Finally, the clinical application value of the model needs to be evaluated in larger-scale prospective cohorts. On the basis of these limitations, future research can focus on the following aspects. First, the single-cell sequencing sample size should be expanded, especially for HCC patients with different etiologies (such as HBV and HCV infection) and clinical stages, to improve the universality of the model; second, gene editing technology and animal models should be used to explore the molecular mechanisms of key genes in the ketone metabolism regulatory network, such as PDK4 and PPARGC1A; and third, metabolomics and drug sensitivity data should be integrated to develop precise treatment strategies for high-risk patients. These follow-up studies will help promote in-depth development in the field of HCC metabolic therapy. 5. Conclusion This study identified five key prognostic genes linked to hepatocyte-specific ketone body metabolism in HCC through integrated multi-omics analysis. The constructed risk model demonstrated robust predictive power for patient outcomes, correlating with altered pathways (e.g., MYC targets and UPR), immune infiltration, and drug sensitivity. Downregulation of PPARGC1A, PDK4, HSD17B6, and APOC3, along with OXCT1 upregulation, highlights metabolic reprogramming as a critical driver of HCC progression. These findings provide novel insights into HCC heterogeneity and therapeutic targeting, although further experimental validation and clinical cohort studies are warranted to translate these findings into precision medicine. Abbreviations Abbreviation full name KB ketone body HCC hepatocellular carcinoma scRNA-seq single-cell RNA sequencing hdWGCNA high-dimensional weighted gene coexpression network analysis DEGs differentially expressed genes KMRGs kb metabolism-related genes PH proportional hazard LASSO least absolute shrinkage and selection operator TME tumor microenvironment RT‒qPCR reverse transcription-quantitative PCR NAFLD nonalcoholic fatty liver disease AcAc acetoacetate BHB D-β-hydroxybutyrate ROS reactive oxygen species TCGA the cancer genome atlas GEO gene expression omnibus KMRGs ketone body metabolism-related genes MSigDB molecular signatures database VST variance-stabilizing transformation PCA principal component analysis PC principal component UMAP uniform manifold approximation and projection AUC area under the curve STRING the search tool for the retrieval of interacting genes/proteins ROC receiver operating characteristic GSVA gene set variation analysis IPS immunophenotype score TCIA The cancer immunome atlas Declarations Ethics approval and consent to participate The study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Guangzhou Red Cross Hospital on February 16, 2019 (approval number: 020--34403034). All subjects included in this study signed the "Informed Consent for the Use of Residual Samples" upon admission to our hospital, and all patients were notified prior to sample use. Due to the protection of patients' personal privacy, they were not provided individually. Clinical trial number not applicable. Consent for publication NA Availability of data and materials The datasets supporting this study are available from the following public repositories. The TCGA-HCC (LIHC) dataset was accessed via The Cancer Genome Atlas (https://portal.gdc.cancer.gov/) with the data version as of February 10, 2025. The GEO datasets GSE14520 (platform GPL571) and GSE149614 (platform GPL24676) were obtained from the Gene Expression Omnibus. The ketone body metabolism-related gene set “GOBP_CELLULAR_KETONE_METABOLIC_PROCESS” (GO:0042180) was sourced from the Molecular Signatures Database (MSigDB). CellMarker 2.0 (http://biobbio-bigdata.hrbmu.edu.cn/CellMarker/) was used for cell type marker annotation. The protein–protein interaction network was constructed using the STRING database (version 11.5, https://string-db.org/). Immunophenotype score (IPS) data were obtained from The Cancer Immunome Atlas (TCIA) with access date January 13, 2025. Competing interests The authors declare that they have no competing interests. Funding This research was funded by the [Guangdong Provincial Natural Science Foundation General Program] [2025A1515012455, PW) and the [Guangdong Provincial-Joint Funding Project of Jinan University and the Municipal Government] [2025A03J3595]. Authors' contributions Wei yuan and Runyu Zhuang: Conceptualization, Data curation, Validation, Visualization, Writing–original draft, Writing–review & editing. Benliang Mao: Data curation, Validation, Visualization, Writing–review & editing. Shanfei Zhu and Qi Cheng: Validation, Writing–review & editing. Bailin Wang and Fan Wu: Visualization, Writing–review & editing. Conceptualization, Supervision. Pengzhen Wang and Bo Ning: Conceptualization, project administration, supervision, writing–review & editing. This work currently described has not been published and is not being considered for publication elsewhere, and its publication was approved by all the authors. All the authors read and approved the final manuscript. Acknowledgments We would like to express our sincere gratitude to all the individuals and organizations who supported and assisted us throughout this research. Special thanks to the following authors: Bailin Wang, Fan Wu, Benliang Mao, Shanfei Zhu and Qi Cheng. In conclusion, we extend our thanks to everyone who has supported and assisted us along the way. Without your support, this research would not have been possible. Authors' information (optional) References Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024; 74:229-63. Lai SW. Risk factors for hepatocellular carcinoma. Cancer. 2019; 125:482. Shi T, Iwama H, Fujita K, Kobara H, Nishiyama N, Fujihara S, et al. Evaluating the Effect of Lenvatinib on Sorafenib-Resistant Hepatocellular Carcinoma Cells. Int J Mol Sci. 2021; 22. Ruan HB, Crawford PA. Ketone bodies as epigenetic modifiers. Curr Opin Clin Nutr Metab Care. 2018; 21:260-6. Lu Y, Han G, Zhang Y, Zhang L, Li Z, Wang Q, et al. M2 macrophage-secreted exosomes promote metastasis and increase vascular permeability in hepatocellular carcinoma. Cell Commun Signal. 2023; 21:299. Zhao C, Wang H, Liu Y, Cheng L, Wang B, Tian X, et al. Biased allosteric activation of ketone body receptor HCAR2 suppresses inflammation. Mol Cell. 2023; 83:3171-87.e7. Newman JC, Verdin E. β-Hydroxybutyrate: A Signaling Metabolite. Annu Rev Nutr. 2017; 37:51-76. Offermanns S. Free fatty acid (FFA) and hydroxy carboxylic acid (HCA) receptors. Annu Rev Pharmacol Toxicol. 2014; 54:407-34. Graff EC, Fang H, Wanders D, Judd RL. Anti-inflammatory effects of the hydroxycarboxylic acid receptor 2. Metabolism. 2016; 65:102-13. Shi X, Li X, Li D, Li Y, Song Y, Deng Q, et al. β-Hydroxybutyrate activates the NF-κB signaling pathway to promote the expression of pro-inflammatory factors in calf hepatocytes. Cell Physiol Biochem. 2014; 33:920-32. Miyauchi T, Uchida Y, Kadono K, Hirao H, Kawasoe J, Watanabe T, et al. Up-regulation of FOXO1 and reduced inflammation by β-hydroxybutyric acid are essential diet restriction benefits against liver injury. Proc Natl Acad Sci U S A. 2019; 116:13533-42. Kolb H, Kempf K, Röhling M, Lenzen-Schulte M, Schloot NC, Martin S. Ketone bodies: from enemy to friend and guardian angel. BMC Med. 2021; 19:313. Puchalska P, Crawford PA. Multi-dimensional Roles of Ketone Bodies in Fuel Metabolism, Signaling, and Therapeutics. Cell Metab. 2017; 25:262-84. Hwang CY, Choe W, Yoon KS, Ha J, Kim SS, Yeo EJ, et al. Molecular Mechanisms for Ketone Body Metabolism, Signaling Functions, and Therapeutic Potential in Cancer. Nutrients. 2022; 14. Wang YH, Liu CL, Chiu WC, Twu YC, Liao YJ. HMGCS2 Mediates Ketone Production and Regulates the Proliferation and Metastasis of Hepatocellular Carcinoma. Cancers (Basel). 2019; 11. Zhou X, Du J, Liu C, Zeng H, Chen Y, Liu L, et al. A Pan-Cancer Analysis of CD161, a Potential New Immune Checkpoint. Front Immunol. 2021; 12:688215. Zhu CX, Yan K, Chen L, Huang RR, Bian ZH, Wei HR, et al. Targeting OXCT1-mediated ketone metabolism reprograms macrophages to promote antitumor immunity via CD8(+) T cells in hepatocellular carcinoma. J Hepatol. 2024; 81:690-703. Huang D, Li T, Wang L, Zhang L, Yan R, Li K, et al. Hepatocellular carcinoma redirects to ketolysis for progression under nutrition deprivation stress. Cell Res. 2016; 26:1112-30. Guo D, Yu Q, Tong Y, Qian X, Meng Y, Ye F, et al. OXCT1 succinylation and activation by SUCLA2 promotes ketolysis and liver tumor growth. Mol Cell. 2025; 85:843-56.e6. Chen H, Ye F, Guo G. Revolutionizing immunology with single-cell RNA sequencing. Cell Mol Immunol. 2019; 16:242-9. Papalexi E, Satija R. Single-cell RNA sequencing to explore immune cell heterogeneity. Nat Rev Immunol. 2018; 18:35-45. Zhang Y, Wang D, Peng M, Tang L, Ouyang J, Xiong F, et al. Single-cell RNA sequencing in cancer research. J Exp Clin Cancer Res. 2021; 40:81. Cheng C, Chen W, Jin H, Chen X. A Review of Single-Cell RNA-Seq Annotation, Integration, and Cell-Cell Communication. Cells. 2023; 12. Lei Y, Tang R, Xu J, Wang W, Zhang B, Liu J, et al. Applications of single-cell sequencing in cancer research: progress and perspectives. J Hematol Oncol. 2021; 14:91. Deng H, Wang X, Jiang ZA, Xu J, Zhang Y, Zhou Y, et al. Clinical potential and experimental validation of prognostic genes in hepatocellular carcinoma revealed by risk modeling utilizing single cell and transcriptome constructs. Front Immunol. 2025; 16:1541252. Chen L, Yang H, Wei X, Liu J, Han X, Zhang C, et al. Integrated single-cell and bulk transcriptome analysis of R-loop score-based signature with regard to immune microenvironment, lipid metabolism and prognosis in HCC. Front Immunol. 2024; 15:1487372. Satija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015; 33:495-502. Lu Y, Yang A, Quan C, Pan Y, Zhang H, Li Y, et al. A single-cell atlas of the multicellular ecosystem of primary and metastatic hepatocellular carcinoma. Nat Commun. 2022; 13:4594. Griss J, Viteri G, Sidiropoulos K, Nguyen V, Fabregat A, Hermjakob H. ReactomeGSA - Efficient Multi-Omics Comparative Pathway Analysis. Mol Cell Proteomics. 2020; 19:2115-25. Lu Y, Li K, Hu Y, Wang X. Expression of Immune Related Genes and Possible Regulatory Mechanisms in Alzheimer's Disease. Front Immunol. 2021; 12:768966. Gustavsson EK, Zhang D, Reynolds RH, Garcia-Ruiz S, Ryten M. ggtranscript: an R package for the visualization and interpretation of transcript isoforms using ggplot2. Bioinformatics. 2022; 38:3844-6. Li R, Zhao M, Miao C, Shi X, Lu J. Identification and validation of key biomarkers associated with macrophages in nonalcoholic fatty liver disease based on hdWGCNA and machine learning. Aging (Albany NY). 2023; 15:15451-72. Orifjon S, Jammatov J, Sousa C, Barros R, Vasconcelos O, Rodrigues P. Translation and Adaptation of the Adult Developmental Coordination Disorder/Dyspraxia Checklist (ADC) into Asian Uzbekistan. Sports (Basel). 2023; 11. Zheng Y, Gao W, Zhang Q, Cheng X, Liu Y, Qi Z, et al. Ferroptosis and Autophagy-Related Genes in the Pathogenesis of Ischemic Cardiomyopathy. Front Cardiovasc Med. 2022; 9:906753. Wu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb). 2021; 2:100141. Liu P, Xu H, Shi Y, Deng L, Chen X. Potential Molecular Mechanisms of Plantain in the Treatment of Gout and Hyperuricemia Based on Network Pharmacology. Evid Based Complement Alternat Med. 2020; 2020:3023127. Lei J, Qu T, Cha L, Tian L, Qiu F, Guo W, et al. Clinicopathological characteristics of pheochromocytoma/paraganglioma and screening of prognostic markers. J Surg Oncol. 2023; 128:510-8. Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw. 2010; 33:1-22. Xie Y, He L, Lugano R, Zhang Y, Cao H, He Q, et al. Key molecular alterations in endothelial cells in human glioblastoma uncovered through single-cell RNA sequencing. JCI Insight. 2021; 6. Gu Z, Hübschmann D. Make Interactive Complex Heatmaps in R. Bioinformatics. 2022; 38:1460-2. Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics. 2000; 56:337-44. Hänzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013; 14:7. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015; 43:e47. Charoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D, et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell Rep. 2017; 18:248-62. 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. Azadian S, Doustmohammadi A, Naseri M, Khodarahmi M, Arab SS, Yazdanifar M, et al. Reconstructing the cell-cell interaction network among mouse immune cells. Biotechnol Bioeng. 2023; 120:2756-64. Joo EH, Kim S, Park D, Lee T, Park WY, Han KY, et al. Migratory Tumor Cells Cooperate with Cancer Associated Fibroblasts in Hormone Receptor-Positive and HER2-Negative Breast Cancer. Int J Mol Sci. 2024; 25. Varillas-Delgado D. Role of the PPARGC1A Gene and Its rs8192678 Polymorphism on Sport Performance, Aerobic Capacity, Muscle Adaptation and Metabolic Diseases: A Narrative Review. Genes (Basel). 2024; 15. Zuo Q, He J, Zhang S, Wang H, Jin G, Jin H, et al. PPARγ Coactivator-1α Suppresses Metastasis of Hepatocellular Carcinoma by Inhibiting Warburg Effect by PPARγ-Dependent WNT/β-Catenin/Pyruvate Dehydrogenase Kinase Isozyme 1 Axis. Hepatology. 2021; 73:644-60. Li Y, Xu S, Li J, Zheng L, Feng M, Wang X, et al. SIRT1 facilitates hepatocellular carcinoma metastasis by promoting PGC-1α-mediated mitochondrial biogenesis. Oncotarget. 2016; 7:29255-74. Dou X, Fu Q, Long Q, Liu S, Zou Y, Fu D, et al. PDK4-dependent hypercatabolism and lactate production of senescent cells promotes cancer malignancy. Nat Metab. 2023; 5:1887-910. Qin YJ, Lin TY, Lin XL, Liu Y, Zhao WT, Li XY, et al. Loss of PDK4 expression promotes proliferation, tumorigenicity, motility and invasion of hepatocellular carcinoma cells. J Cancer. 2020; 11:4397-405. Wu M, Jiang L. Hydroxysteroid 17-Beta Dehydrogenase 6 Is a Prognostic Biomarker and Correlates with Immune Infiltrates in Hepatocellular Carcinoma. Dig Dis Sci. 2022; 67:146-58. Lv L, Zhao Y, Wei Q, Zhao Y, Yi Q. Downexpression of HSD17B6 correlates with clinical prognosis and tumor immune infiltrates in hepatocellular carcinoma. Cancer Cell Int. 2020; 20:210. Chebli J, Larouche M, Gaudet D. APOC3 siRNA and ASO therapy for dyslipidemia. Curr Opin Endocrinol Diabetes Obes. 2024; 31:70-7. Wang X, Gong Y, Deng T, Zhang L, Liao X, Han C, et al. Diagnostic and prognostic significance of mRNA expressions of apolipoprotein A and C family genes in hepatitis B virus-related hepatocellular carcinoma. J Cell Biochem. 2019; 120:18246-65. Li X, Wang L, Wang L, Feng Z, Peng C. Single-Cell Sequencing of Hepatocellular Carcinoma Reveals Cell Interactions and Cell Heterogeneity in the Microenvironment. Int J Gen Med. 2021; 14:10141-53. Dong YN, Mesaros C, Xu P, Mercado-Ayón E, Halawani S, Ngaba LV, et al. Frataxin controls ketone body metabolism through regulation of OXCT1. PNAS Nexus. 2022; 1:pgac142. Zhang S, Xie C. The role of OXCT1 in the pathogenesis of cancer as a rate-limiting enzyme of ketone body metabolism. Life Sci. 2017; 183:110-5. Ma W, Sun Y, Yan R, Zhang P, Shen S, Lu H, et al. OXCT1 functions as a succinyltransferase, contributing to hepatocellular carcinoma via succinylating LACTB. Mol Cell. 2024; 84:538-51.e7. Pohl C, Dikic I. Cellular quality control by the ubiquitin-proteasome system and autophagy. Science. 2019; 366:818-22. Dawson SP. Hepatocellular carcinoma and the ubiquitin-proteasome system. Biochim Biophys Acta. 2008; 1782:775-84. Nakayama KI, Nakayama K. Ubiquitin ligases: cell-cycle control and cancer. Nat Rev Cancer. 2006; 6:369-81. Ruggiano A, Foresti O, Carvalho P. Quality control: ER-associated degradation: protein quality control and beyond. J Cell Biol. 2014; 204:869-79. Trausch-Azar J, Leone TC, Kelly DP, Schwartz AL. Ubiquitin proteasome-dependent degradation of the transcriptional coactivator PGC-1{alpha} via the N-terminal pathway. J Biol Chem. 2010; 285:40192-200. Black VH, Sanjay A, van Leyen K, Möeller I, Lauring B, Kreibich G. Cholesterol and steroid synthesizing smooth endoplasmic reticulum of adrenocortical cells contains high levels of translocation apparatus proteins. Endocr Res. 2002; 28:425-30. Han J, Kaufman RJ. The role of ER stress in lipid metabolism and lipotoxicity. J Lipid Res. 2016; 57:1329-38. Ren C, Cui H, Bao X, Huang L, He S, Fong HKW, et al. Proteopathy Linked to Exon-Skipping Isoform of RGR-Opsin Contributes to the Pathogenesis of Age-Related Macular Degeneration. Invest Ophthalmol Vis Sci. 2023; 64:41. Wu X, Siggel M, Ovchinnikov S, Mi W, Svetlov V, Nudler E, et al. Structural basis of ER-associated protein degradation mediated by the Hrd1 ubiquitin ligase complex. Science. 2020; 368. Amati B. Myc degradation: dancing with ubiquitin ligases. Proc Natl Acad Sci U S A. 2004; 101:8843-4. Kropp EM, Matono S, Robida A, Kandarpa M, Wang O, Alt A, et al. Inhibition of Endoplasmic Reticulum Associated Degradation Is Cytotoxic to Relapsed Refractory Multiple Myeloma through Altered Proapoptotic Signaling. 2024; 144:3264. Holcik M. Could the eIF2α-Independent Translation Be the Achilles Heel of Cancer? Front Oncol. 2015; 5:264. Locke B, Campbell E, Lu R. CREB3 mediates the transcriptional regulation of PGC-1α, a master regulator of energy homeostasis and mitochondrial biogenesis. FEBS Lett. 2024; 598:1730-9. Kristensen CM, Dethlefsen MM, Tøndering AS, Lassen SB, Meldgaard JN, Ringholm S, et al. PGC-1α in hepatic UPR during high-fat high-fructose diet and exercise training in mice. Physiol Rep. 2018; 6:e13819. Additional Declarations No competing interests reported. Supplementary Files Additionalfile1.xlsx Additional files Additional file 1 225 Genes related to ketone body metabolism (KMRGs). Additionalfile2.xlsx Additional file 2 Primer tables of prognostic genes and internal reference genes (GAPDH). Additionalfile3.tif Additional file 3 Cell type profiles in HCC were initially explored. (a) A represent the filtered cell type. (b) B represent the filtered gene types. (c) C indicates the mutation status of these genes. Additionalfile4.xlsx Additional file 4 Cell cluster annotation table. Additionalfile5.xlsx Additional file 5 PH assumption test. Additionalfile6.xlsx Additional file 6 Through GSEA, pathways associated with PPARGC1A. Additionalfile7.xlsx Additional file 7 Through GSEA, pathways associated with PDK4. Additionalfile8.xlsx Additional file 8 Through GSEA, pathways associated with HSD17B6. Additionalfile9.xlsx Additional file 9 Through GSEA, pathways associated with APOC3. Additionalfile10.xlsx Additional file 10 Through GSEA, pathways associated with OXCT1. Additionalfile11.tif Additional file 11 Complex intercellular communication patterns. (a) Number of communication interactions between cells in the HCC group. (b) The strength of the communication interactions between cells in the HCC group. (c) The number of communication interactions between cells in the normal group. (d) The strength of the communication interactions between cells in the normal group. (e) Dot plot displaying key ligand‒receptor pairs involved in communication between T cells and various cell types in the control group. (f) Dot plot displaying key ligand‒receptor pairs involved in communication between T cells and various cell types in the HCC group. Additionalfile12.tif Additional file 12 Temporal trajectory analysis of the dynamic expression of prognostic genes during hepatocyte differentiation. (a) Differentiation trajectory map of hepatocytes over time. (b) Dynamic expression of prognostic genes during differentiation in the control group. (c) Dynamic expression of prognostic genes during differentiation in the HHC group. Additionalfile13.tif Additional file 13 Bar chart comparing biomarker expression levels between the control group and HCC group for (a) PPARGC1A, (b) PDK4, (c) HSD17B6, (d) APOC3, and (e) OXCT1. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 20 Apr, 2026 Reviewers agreed at journal 20 Apr, 2026 Reviews received at journal 14 Apr, 2026 Reviewers agreed at journal 14 Apr, 2026 Reviewers invited by journal 14 Apr, 2026 Editor invited by journal 06 Apr, 2026 Editor assigned by journal 02 Apr, 2026 Submission checks completed at journal 30 Mar, 2026 First submitted to journal 30 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9187321","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":624351831,"identity":"f1627e3b-e331-4c0a-b453-1a388a08235a","order_by":0,"name":"Wei Yuan","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Yuan","suffix":""},{"id":624351832,"identity":"b95b759c-591b-4f4f-ac37-86bb15709cb4","order_by":1,"name":"Runyu Zhuang","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Runyu","middleName":"","lastName":"Zhuang","suffix":""},{"id":624351833,"identity":"26f4c468-3adf-4bdf-a6b9-f8348eaa6fc5","order_by":2,"name":"Bailin Wang","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Bailin","middleName":"","lastName":"Wang","suffix":""},{"id":624351834,"identity":"fe034231-03b8-49bc-9901-0a7fb0f4a5ae","order_by":3,"name":"Fan Wu","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Fan","middleName":"","lastName":"Wu","suffix":""},{"id":624351835,"identity":"b9264333-9299-4b87-a5d9-f9e0271a256d","order_by":4,"name":"Benliang Mao","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Benliang","middleName":"","lastName":"Mao","suffix":""},{"id":624351836,"identity":"454eb948-22b6-4ea9-82f3-0ebf41d057ef","order_by":5,"name":"Shanfei Zhu","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Shanfei","middleName":"","lastName":"Zhu","suffix":""},{"id":624351837,"identity":"f9361e3f-b9ca-4818-ab91-55a53c7ccfb4","order_by":6,"name":"Qi Cheng","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital","correspondingAuthor":false,"prefix":"","firstName":"Qi","middleName":"","lastName":"Cheng","suffix":""},{"id":624351838,"identity":"23faa496-a498-4478-b0d7-ff5e31206d24","order_by":7,"name":"Pengzhen Wang","email":"","orcid":"","institution":"Guangzhou Red Cross Hospital，Jinan University","correspondingAuthor":false,"prefix":"","firstName":"Pengzhen","middleName":"","lastName":"Wang","suffix":""},{"id":624351839,"identity":"6e2cf44e-26ed-4fc0-8dc4-0b2f6ffc8f3f","order_by":8,"name":"Bo Ning","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYDACZjCSkLM/cPjAgQ8VxGuxMGY4eCzx4IwzxFtUkdhw+IzxYd4WIpSbs/Mefl1QIZHY2HbmwwHeBgZ5frED+LVYNvOlWc84I2HczHN2wwHJHQyGM2cn4NdicJjHzJi3TUK2TQKoxfAMQ4LBbSK1MPbIv3lwILGNOC3Gj4FaFGcwnGE4cJBILWbMIL8YMBwzONhwRoIIv5w/Y/y5oKJOzoDh8OPPfyps5PmlCWgBAjYJJI4ETmXIgPkDUcpGwSgYBaNg5AIA3L5JX+gyQh4AAAAASUVORK5CYII=","orcid":"","institution":"Guangzhou Red Cross Hospital，Jinan University","correspondingAuthor":true,"prefix":"","firstName":"Bo","middleName":"","lastName":"Ning","suffix":""}],"badges":[],"createdAt":"2026-03-21 17:23:30","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9187321/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9187321/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107869318,"identity":"a89ff6eb-dddb-49f3-a5e8-80c1acf8ed14","added_by":"auto","created_at":"2026-04-27 07:36:44","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":31686307,"visible":true,"origin":"","legend":"\u003cp\u003escRNA-seq delineates the cellular landscape of HCC and normal control samples. (a) Elbow plot of principal component analysis (PCA). (b) JackStraw significance test plot. (c) UMAP plot displaying original cell clusters. (d) Renamed cell clusters on the basis of marker gene annotation. (e) Dot plot showing the marker genes of each cell cluster. (f) Cell type‒pathway enrichment heatmap, with the x-axis representing nine major cell types (hepatocyte, T cell, fibroblast, etc.) and the y-axis showing 10 significant Reactome pathways screened by GSVA. The color indicates the enrichment score (red indicates high enrichment, and blue indicates low enrichment). (g) UMAP displaying the AUCell activity of the KRMG across cell subtypes. (h) Stacked bar chart illustrating differences in cell proportions between the HCC and normal control groups. (i) Box plot demonstrating differences in cell proportions between the HCC and normal control groups. ns P \u0026gt; 0.05, * P \u0026lt; 0.05, ** P \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/52f99de0d6eac2431d53938c.png"},{"id":107710205,"identity":"f7e04975-94d1-42f4-848f-dbb1e9190d7a","added_by":"auto","created_at":"2026-04-24 09:40:07","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":13108203,"visible":true,"origin":"","legend":"\u003cp\u003eKey genes for hepatocyte subpopulation identification. (a) Elbow plot of principal component analysis (PCA) for hepatocyte subpopulations. (b) JackStraw significance test plot. (c) UMAP plot displaying hepatocyte subtypes. (d) hdWGCNA plot. (e) Dendrogram showing hdWGCNA gene clustering modules, with each module represented by a distinct color. (f) Heatmap of correlations between gene clustering modules and hepatocyte subtypes. (g) Volcano plot of differentially expressed genes in the TCGA-HCC dataset. (h) Heatmap of the top 10 upregulated and top 10 downregulated genes among the DEGs in the TCGA-HCC dataset. (i) Venn diagram showing intersections among key module genes, differentially expressed genes, and metabolism-related genes. (j) GO enrichment analysis of the intersecting core genes. (k) KEGG enrichment analysis of the intersecting core genes. (l) PPI analysis illustrating the interaction relationships among the intersecting core genes.\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/f6e6cca84e14145036e5ba32.png"},{"id":107711285,"identity":"c6d06340-d58c-4c42-9022-2b544d850986","added_by":"auto","created_at":"2026-04-24 09:45:01","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":57509947,"visible":true,"origin":"","legend":"\u003cp\u003eConstruction and validation of the prognostic model. (a) Forest plot of the univariate Cox analysis. (b) Distribution plot of LASSO coefficients. (c) Distribution plot of partial likelihood deviation in LASSO regression, where 5 variables were retained when the partial likelihood deviation reached the minimum value (optimal lambda = 0.00264). (d) 1 Risk scores between the high-risk and low-risk groups in the TCGA-HCC dataset. 2 Survival status distribution. (e) Heatmap showing the expression levels of 5 prognosis-related genes in the high- and low-risk groups. (f) Kaplan‒Meier survival analysis of the high-risk and low-risk groups in the TCGA-HCC dataset. (g) ROC analysis demonstrating the ability of the prognostic model to predict 1-, 2-, and 3-year survival rates in HCC patients. (h1) (h2) (i) (j) (k) For the analysis of the validation set, same as before.\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/f1545d211bd0b9cdadc843c8.png"},{"id":107708307,"identity":"72bc599d-653b-4b6b-a176-b216f3e2ffa5","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":22466244,"visible":true,"origin":"","legend":"\u003cp\u003ePrognostic model and clinical correlation analysis. (a) Violin plot demonstrating the differences in hepatocellular and ketone body metabolism-related risk scores among different clinical subgroups in the TCGA-HCC cohort. (b) Kaplan‒Meier survival analysis across different clinical stratifications.\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/58dbed2117d9f45e1290fd01.png"},{"id":107708303,"identity":"8c1658b2-adb7-4672-9f8b-7abeb1bd69c2","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":5569470,"visible":true,"origin":"","legend":"\u003cp\u003eGSEA results of the three biomarkers, displaying the top 5 most significant pathways. (a) PPARGC1A. (b) PDK4. (c) HSD17B6. (d) APOC3. (e)\u003c/p\u003e\n\u003cp\u003eOXCT1.\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/7dc4da53f2b4545277246d6f.png"},{"id":107708926,"identity":"495d1976-8d27-4c2d-8489-47bb9c47e0a3","added_by":"auto","created_at":"2026-04-24 09:33:45","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":17921332,"visible":true,"origin":"","legend":"\u003cp\u003eCrucial functional pathways and TME profiles altered by risk score. (a) GSVA of the high-risk and low-risk groups in the TCGA-HCC dataset. (b) Violin plots showing the infiltration differences in 14 immune cell types between patients in the high-risk and low-risk groups. (c) Heatmap illustrating the correlations among differentially infiltrated immune cell types and their associations with prognostic genes.\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/6034da0482beb17288ff8ef3.png"},{"id":107708304,"identity":"ff9c833d-6177-43ba-badc-3f20f296dc3e","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":9080882,"visible":true,"origin":"","legend":"\u003cp\u003ePotential of the risk score as a predictor of immunotherapy and drug treatment outcomes. (a) IPS score boxplot. (b) Violin plot showing differences in drug sensitivity between the high- and low-risk groups. (c) Volcano plot related to drug sensitivity.\u003c/p\u003e","description":"","filename":"Figure7.png","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/d14510f2409e88aba44fac12.png"},{"id":107500098,"identity":"b17dafb2-7e89-4159-923d-a1fb015e6990","added_by":"auto","created_at":"2026-04-22 05:44:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":490168,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/b5b6323f-fbf9-45a0-a229-db43a6a976e5.pdf"},{"id":107708301,"identity":"afffd384-13ac-4f05-9a64-a627e738f2b0","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"xlsx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":12471,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional files\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional file 1 225 Genes related to ketone body metabolism (KMRGs).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile1.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/ac40bce6f33901e4b57bfdf4.xlsx"},{"id":107711409,"identity":"cc45433f-dc24-4dc8-8819-b106a808ce3c","added_by":"auto","created_at":"2026-04-24 09:45:24","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":10212,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 2 Primer tables of prognostic genes and internal reference genes (GAPDH).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile2.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/4ffdb93500c3c793b283f81d.xlsx"},{"id":107708305,"identity":"8f9aa178-2322-42a0-8afa-9645704ed0ce","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"tif","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":21302432,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 3 Cell type profiles in HCC were initially explored. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) A represent the filtered cell type. (\u003cstrong\u003eb\u003c/strong\u003e) B represent the filtered gene types. (\u003cstrong\u003ec\u003c/strong\u003e) C indicates the mutation status of these genes.\u003c/p\u003e","description":"","filename":"Additionalfile3.tif","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/57980ed687d8387f742d701a.tif"},{"id":107708984,"identity":"ca882266-19b3-4966-bea0-79c02767c048","added_by":"auto","created_at":"2026-04-24 09:33:59","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":10236,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 4 Cell cluster annotation table.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile4.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/f8e429154329d40ae34ea8a3.xlsx"},{"id":107708886,"identity":"a967f460-f0e1-49c6-8bd6-d0d0172e9e28","added_by":"auto","created_at":"2026-04-24 09:33:13","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":10128,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 5 PH assumption test.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile5.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/8d185246c4d92276e173d7f0.xlsx"},{"id":107708308,"identity":"dd45f570-2663-4d82-937d-53df42cfe62d","added_by":"auto","created_at":"2026-04-24 09:26:13","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":27701,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 6 Through GSEA, pathways associated with PPARGC1A.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile6.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/11281640dc87175a33f8eb5e.xlsx"},{"id":107711281,"identity":"e0dd9f80-88d5-460e-b424-3705af2e68fc","added_by":"auto","created_at":"2026-04-24 09:44:57","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":27068,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 7 Through GSEA, pathways associated with PDK4.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile7.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/045fda961c17953aa91e0542.xlsx"},{"id":107708288,"identity":"70993ec2-a6fe-4a2f-8ec3-75ca2c637daa","added_by":"auto","created_at":"2026-04-24 09:25:58","extension":"xlsx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":25483,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 8 Through GSEA, pathways associated with HSD17B6.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile8.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/52a103bba4ec9938198a58ba.xlsx"},{"id":107708933,"identity":"51521ce2-be00-467e-a01f-e962fcc63bfd","added_by":"auto","created_at":"2026-04-24 09:33:48","extension":"xlsx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":23611,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 9 Through GSEA, pathways associated with APOC3.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile9.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/fe721741faacf1a25d31887c.xlsx"},{"id":107708289,"identity":"f4ba9c33-4698-4ddb-bed1-6ead19231d3b","added_by":"auto","created_at":"2026-04-24 09:25:58","extension":"xlsx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":31455,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 10 Through GSEA, pathways associated with OXCT1.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"Additionalfile10.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/b70363575359966780651ce7.xlsx"},{"id":107708285,"identity":"c92ee794-7673-41f9-91fd-ca91939bcd37","added_by":"auto","created_at":"2026-04-24 09:25:54","extension":"tif","order_by":11,"title":"","display":"","copyAsset":false,"role":"supplement","size":11978452,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 11 Complex intercellular communication patterns. \u003c/strong\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Number of communication interactions between cells in the HCC group. (\u003cstrong\u003eb\u003c/strong\u003e) The strength of the communication interactions between cells in the HCC group. (\u003cstrong\u003ec\u003c/strong\u003e) The number of communication interactions between cells in the normal group. (\u003cstrong\u003ed\u003c/strong\u003e) The strength of the communication interactions between cells in the normal group. (\u003cstrong\u003ee\u003c/strong\u003e) Dot plot displaying key ligand‒receptor pairs involved in communication between T cells and various cell types in the control group. (\u003cstrong\u003ef\u003c/strong\u003e) Dot plot displaying key ligand‒receptor pairs involved in communication between T cells and various cell types in the HCC group.\u003c/p\u003e","description":"","filename":"Additionalfile11.tif","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/d8a4e0331886e2f81db418c9.tif"},{"id":107710215,"identity":"9fb271c0-31e5-41fb-99c7-c0482d8169ad","added_by":"auto","created_at":"2026-04-24 09:40:15","extension":"tif","order_by":12,"title":"","display":"","copyAsset":false,"role":"supplement","size":10633176,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 12 Temporal trajectory analysis of the dynamic expression of prognostic genes during hepatocyte differentiation.\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e) Differentiation trajectory map of hepatocytes over time. (\u003cstrong\u003eb\u003c/strong\u003e) Dynamic expression of prognostic genes during differentiation in the control group. (\u003cstrong\u003ec\u003c/strong\u003e) Dynamic expression of prognostic genes during differentiation in the HHC group.\u003c/p\u003e","description":"","filename":"Additionalfile12.tif","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/6c4b831c4501a888303048cf.tif"},{"id":107708287,"identity":"a6a3c3fe-4424-4d17-9f16-f263d3c52bf9","added_by":"auto","created_at":"2026-04-24 09:25:58","extension":"tif","order_by":13,"title":"","display":"","copyAsset":false,"role":"supplement","size":5645176,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAdditional file 13 Bar chart comparing biomarker expression levels between the control group and HCC group for\u003c/strong\u003e (\u003cstrong\u003ea\u003c/strong\u003e) PPARGC1A, (\u003cstrong\u003eb\u003c/strong\u003e) PDK4, (\u003cstrong\u003ec\u003c/strong\u003e) HSD17B6, (\u003cstrong\u003ed\u003c/strong\u003e) APOC3, and (\u003cstrong\u003ee\u003c/strong\u003e) OXCT1.\u003c/p\u003e","description":"","filename":"Additionalfile13.tif","url":"https://assets-eu.researchsquare.com/files/rs-9187321/v1/e31a2b2cc0e8c82ee93b6d1f.tif"}],"financialInterests":"No competing interests reported.","formattedTitle":"Identification of prognostic genes for hepatocellular carcinoma based on hepatocyte and ketone body metabolism using integrated bulk and single cell RNA sequencing","fulltext":[{"header":"1. Background","content":"\u003cp\u003eHepatocellular carcinoma (HCC) is the most common type of primary liver tumor, accounting for 75\u0026ndash;85% of all cases. With the widespread administration of the hepatitis B vaccine, the incidence of HCC caused by viral hepatitis-related risk factors has decreased[\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Conversely, the number of HCC cases linked to metabolic disorders, such as nonalcoholic fatty liver disease (NAFLD) and diabetes, has steadily increased, paralleling global trends in obesity[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. HCC is now recognized as the sixth most common cancer and the fourth leading cause of cancer-related deaths globally[\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. The persistently high mortality rate of HCC is attributed to its late diagnosis, high recurrence rate, ease of metastasis, and treatment resistance. Although immune checkpoint inhibitors (e.g., atezolizumab/bevacizumab) and targeted therapies (e.g., lenvatinib) have improved outcomes, their efficacy varies across patient subgroups, and resistance remains a challenge[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Current predictive models for long-term clinical outcomes in HCC patients are insufficient. Therefore, identifying new prognostic genes holds critical significance for elucidating the underlying mechanisms of HCC pathogenesis, improving patient prognosis management, and developing targeted therapeutic targets[\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eKetone bodies (KBs), including acetoacetate (AcAc), D-β-hydroxybutyrate (BHB), and acetone, are alternative energy substrates during fasting or metabolic stress. In addition to their metabolic roles, KBs exhibit pleiotropic effects, such as neuroprotection and anti-inflammatory activity, that are mediated through receptor-dependent (e.g., HCAR2) and epigenetic mechanisms[\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. KBs act as anti-inflammatory signaling molecules, modulating inflammatory and immune functions through both receptor-dependent and receptor-independent pathways. BHB inhibits inflammatory responses and regulates immune cell functions by binding and activating hydroxy carboxylic acid receptor 2 (HCAR2) or directly modulating intracellular signaling pathways[\u003cspan additionalcitationids=\"CR8\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. KBs regulate reactive oxygen species (ROS) metabolism and maintain redox homeostasis by activating cytoprotective regulators such as Nrf2, SIRT1/3, and AMPK[\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. Ketone metabolism can be divided into ketogenesis and ketolysis. HCC undergoes metabolic reprogramming, leading to abnormalities in both processes of ketone metabolism[\u003cspan additionalcitationids=\"CR13\" citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. In HCC tissues, the expression of the rate-limiting enzyme for ketogenesis, 3-hydroxy-3-methylglutaryl-CoA synthase 2 (HMGCS2), is downregulated, resulting in reduced BHB production. This phenomenon has been associated with HCC pathogenesis and staging and impacts tumor proliferation and metastasis[\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Conversely, the expression of the rate-limiting ketolysis enzyme 3-oxoacid CoA transferase 1 (OXCT1) is significantly upregulated in HCC, enhancing ketolysis. In HCC, increased ketolysis plays a critical role in suppressing AMPK activation, thereby preventing excessive autophagy in HCC cells and promoting tumor proliferation[\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSingle-cell RNA sequencing (scRNA-seq) resolves transcriptional profiles at individual-cell resolution, overcoming the limitations of bulk sequencing in dissecting cellular heterogeneity. Its workflow typically includes cell isolation, cDNA synthesis, and library construction[\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. By cataloging cell types and states across tissues, scRNA-seq has revealed novel subtypes in normal and diseased organs[\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. For example, in cancer research, the complexity of the tumor microenvironment (TME), such as immune cell infiltration and stromal interactions, is deciphered[\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Integration of clinical and pathological data with scRNA-seq results enables the discovery of diagnostic and prognostic biomarkers and therapeutic-related cell states and supports precise molecular subtyping and personalized therapies[\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Recent studies have increasingly used scRNA-seq to characterize tumor microenvironment changes in HCC, identify predictive biomarkers, and establish theoretical foundations for clinical interventions[\u003cspan additionalcitationids=\"CR25\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThis study integrated bulk RNA sequencing and single-cell RNA sequencing data to systematically identify prognostic biomarker genes associated with key cell subpopulations and ketone metabolism in HCC via hierarchical dynamic weighted gene coexpression network analysis (hdWGCNA) and machine learning-based bioinformatics approaches. A survival prediction model was constructed to stratify HCC patient survival outcomes, and functional mechanism studies further revealed the synergistic roles of key cell type and metabolic pathways in tumor progression. This research aims to provide molecular evidence for precision-based prognosis assessment in HCC, deepen the understanding of ketone metabolism-tumor microenvironment crosstalk mechanisms, and offer a theoretical basis for developing metabolically reprogrammed personalized immunotherapy strategies.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1 Data sources\u003c/h2\u003e\n \u003cp\u003eThree HCC-related transcriptome datasets were included in this study. The TCGA-HCC dataset was obtained from The Cancer Genome Atlas (TCGA) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://portal.gdc.cancer.gov/\u003c/span\u003e\u003c/span\u003e) on February 10, 2025, and included 368 HCC tissue samples (with survival and clinical information) and 50 normal tissue samples. Additionally, the GSE14520 (platform: GPL571) and GSE149614 (platform: GPL24676) datasets were obtained from the Gene Expression Omnibus (GEO) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ncbi.nlm.nih.gov/geo/\u003c/span\u003e\u003c/span\u003e). The GSE14520 dataset included 130 HCC tissue samples with survival information. GSE149614 is a single-cell RNA-seq (scRNA-seq) dataset comprising 10 primary tumor (HCC) tissue samples and 8 normal tissue samples. All the samples were collected from human subjects. Notably, the HCC and normal samples were defined as the HCC and control groups, respectively. Moreover, 225 ketone body (KB) metabolism-related genes (KMRGs) were identified from the \u0026quot;GOBP_CELLULAR_KETONE_METABOLIC_PROCESS\u0026quot; pathway (GO: 0042180) in the Molecular Signatures Database (MSigDB) (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.gsea-msigdb.org/gsea/msigdb\u003c/span\u003e\u003c/span\u003e) (\u003cstrong\u003eAdditional file 1\u003c/strong\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2 Exploration of cell type profiles and functions through scRNA-seq analysis\u003c/h2\u003e\n \u003cp\u003eAn extensive investigation of the scRNA-seq data from the GSE149614 dataset was conducted to explore the underlying mechanisms of HCC progression. First, the cell type profiles in HCC were explored. Initially, the raw data were read via the CreateSeuratObject function in the Seurat package (v 4.1.0)[\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. To ensure data quality, rigorous quality control criteria were applied to retain high-quality cells (200\u0026thinsp;\u0026le;\u0026thinsp;detected genes (nFeature)\u0026thinsp;\u0026le;\u0026thinsp;6,000, mitochondrial gene content (percent.mt)\u0026thinsp;\u0026lt;\u0026thinsp;10%, total gene count (nCount_RNA)\u0026thinsp;\u0026lt;\u0026thinsp;20,000). The FindVariableFeatures function was subsequently employed to identify genes with high variation coefficients across cells via the variance-stabilizing transformation (VST) method. The top 2,000 most variable genes (highly variable genes) were retained for further analyses. NormalizeData and ScaleData functions were subsequently used to normalize the data, and RunPCA was applied to perform principal component analysis (PCA). An elbow plot was generated via the elbowplot function to visualize the proportion of variance explained by each principal component (PC). The statistical significance of each PC was assessed via the JackStraw function. PCs with p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were retained for analyses. Afterwards, unsupervised clustering analysis of the cells was conducted via the FindNeighbors and FindClusters functions (resolution\u0026thinsp;=\u0026thinsp;0.1). Through uniform manifold approximation and projection (UMAP) conducted via the RunUMAP function, distinct cell clusters were recognized. Cell cluster annotation was performed by referencing marker genes from the literature [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e] and the CellMarker2.0 database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://biobbio-bigdata.hrbmu.edu.cn/CellMarker/\u003c/span\u003e\u003c/span\u003e), with marker gene expression illustrated for each cell type via the DotPlot function. The functional analysis of each cell type was performed via the analyze_sc_clusters function in the ReactomeGSA package (v 1.12.0) [\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e] to reveal the related biological functions. The result was extracted via pathway function.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3 Ascertainment of the key cell type through KMRG activity and proportion analyses\u003c/h2\u003e\n \u003cp\u003eThe AUCell package (version 1.12.0) [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e] was utilized to evaluate KMRG activity at single-cell resolution within the GSE149614 dataset. A predefined gene set comprising 225 KMRGs was used as input. For each cell, an area under the curve (AUC) value was computed by ranking genes on the basis of their expression levels, thereby estimating the activation level of the gene set. Higher AUC values indicate a greater number of highly expressed KMRGs in individual cells. To define the threshold for identifying cells with active expression of the gene set, the AUCell_exploreThresholds function was applied (AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.7). Remarkably, AUC values were projected onto a UMAP space via the FeaturePlot function in the Seurat package (v 4.1.0), enabling visualization of KMRG activity across different cell populations. Additionally, the proportions of different cell types were visualized via the ggplot2 package (v 3.5.1) [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e], with differences in proportions between the HCC and control groups analyzed via the Wilcoxon rank-sum test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The key cell type was ultimately determined on the basis of its notably high KMRG activity and significantly greater proportion in HCC.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e2.4 High-dimensional weighted gene coexpression network analysis (hdWGCNA)\u003c/h2\u003e\n \u003cp\u003eWithin the GSE149614 dataset, the key cell type and the gene expression profiles were extracted for further analyses. Following the previously described methodology, the key cell type was redimensionalized and reclustered (resolution\u0026thinsp;=\u0026thinsp;0.1) via the Seurat package (v 4.1.0), which led to the identification of distinct cell subtypes.\u003c/p\u003e\n \u003cp\u003eOn the basis of all the samples from the GSE149614 dataset, hdWGCNA was subsequently conducted via the hdWGCNA package (v 0.4.5) [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e] to identify genes associated with key cell subtypes. All analyses were conducted following the official standard workflow (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://smorabit.github.io/hdWGCNA/articles/basic_tutorial.html\u003c/span\u003e\u003c/span\u003e). To eliminate low-quality data, the MetacellsByGroups function was used to develop a metacell-level gene expression matrix. The TestSoftPowers function was subsequently applied to select the optimal soft power threshold on the basis of the scale-free topology criterion (scale-free R2\u0026thinsp;\u0026gt;\u0026thinsp;0.8) and mean connectivity (approaching 0). A coexpression network was subsequently established via the ConstructNetwork function, and genes were grouped into distinct modules (k\u0026thinsp;=\u0026thinsp;25, min cells\u0026thinsp;=\u0026thinsp;100, max shared\u0026thinsp;=\u0026thinsp;15, target metacells\u0026thinsp;=\u0026thinsp;1,000). Spearman correlation analysis was subsequently performed with the psych package (v 2.1.6) [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] to evaluate the associations between gene modules and key cell subtypes. Genes within modules that strongly correlated with one of the key cell subtypes were integrated and defined as key module genes (|correlation coefficient (cor)| \u0026gt; 0.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003e2.5 Function and protein‒protein interaction (PPI) analyses of candidate genes\u003c/h2\u003e\n \u003cp\u003eWithin the TCGA-HCC dataset, to identify differentially expressed genes (DEGs) in HCC, the DESeq2 package (v 1.42.0) [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e] was used to perform differential expression analysis between the HCC and control groups (|log\u003csub\u003e2\u003c/sub\u003e-fold change (FC)| \u0026gt; 1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Next, to identify genes related to key cell type and KB metabolism in HCC, the key module genes, DEGs, and KMRGs were intersected via the ggvenn package (v 0.1.10) (37), resulting in candidate genes. Potential roles of the candidate genes in HCC were subsequently explored with the clusterProfiler package (v 4.10.1) [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e] to conduct Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). To explore candidate gene interactions at the protein level, the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://string-db.org/\u003c/span\u003e\u003c/span\u003e) was used to establish a protein\u0026ndash;protein interaction (PPI) network (interaction score\u0026thinsp;\u0026ge;\u0026thinsp;0.4). Cytoscape software (v 3.9.1) [\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e] was used for visualization.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e2.6 Risk model development and validation\u003c/h2\u003e\n \u003cp\u003eWithin the TCGA-HCC dataset, the potential value of KMRGs for predicting overall survival (OS) was assessed on the basis of HCC samples with survival information. Specifically, the survival package (v 3.7.0) [\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e] was initially applied to conduct univariate Cox analysis (hazard ratio (HR)\u0026thinsp;\u0026ne;\u0026thinsp;1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) on candidate genes, and the proportional hazard (PH) assumption test (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05) conducted with the cox.zph function was applied to analyze the retained genes. The genes that passed the pH assumption test were defined as candidate prognostic genes. Thereafter, the glmnet package (v 4.1.8) [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e] was applied to construct a 10-fold cross-validated least absolute shrinkage and selection operator (LASSO) model to identify prognostic genes whose coefficients were not penalized to 0. On the basis of prognostic genes related to key cell type and KB metabolism, a risk model was constructed according to this formula (the coef and expr represented the LASSO coefficient and expression of each prognostic gene, respectively):\u003c/p\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"438\" height=\"100\"\u003e\u003c/p\u003e\n \u003cp\u003eThe HCC samples were subsequently classified into high-risk (HRG) and low-risk (LRG) groups according to the optimal risk score cutoff value calculated via the survminer package (v 0.4.9) [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. Then, within the HRG and LRG, the ggplot2 package (v 3.5.1) was employed to determine the risk score and survival state distribution. The prognostic gene expression trends are shown via the pheatmap package (v 1.0.12) [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. In addition, Kaplan‒Meier (KM) survival curves for HRGs and LRGs were generated through the survminer package (v 0.4.9), and the survival probability difference was subjected to the log-rank test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The survivalROC package (v 1.0.3) [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e] was utilized to assess the predictive ability of the risk model by generating receiver operating characteristic (ROC) curves at 1-, 2-, and 3-year time points (AUC value\u0026thinsp;\u0026gt;\u0026thinsp;0.6). To investigate the model accuracy and generalizability, the risk model was verified with the GSE14520 dataset.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e2.7 Clinical stratification and KM survival analyses\u003c/h2\u003e\n \u003cp\u003eTo investigate the associations between clinical characteristics and risk scores related to key cell type and KB metabolism, HCC samples with survival and clinical information from the TCGA-HCC dataset were first classified into distinct clinical subgroups on the basis of age (\u0026le;\u0026thinsp;61 and \u0026gt;\u0026thinsp;61), sex (female and male), clinical stage (stage 1/2 and stage 3/4), T stage (T1/2 and T3/4), N stage (N0 and N1), and M stage (M0 and M1). Thereafter, risk score differences across clinical subgroups were analyzed via the Wilcoxon rank-sum test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Additionally, the ability of the risk score to predict survival was further explored on the basis of clinical characteristics (age, sex, race (white, Asian, and black), clinical stage, T stage, N stage, and M stage). Specifically, within different clinical subgroups, the ggsurvplot function in the survival package (v 3.7.0) was applied to plot KM survival curves for HRG and LRG, and the survival differences were evaluated via the log-rank test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e2.8 Gene set variation analysis (GSVA)\u003c/h2\u003e\n \u003cp\u003eGSVA package (v 1.42.0) [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e] was utilized to conduct GSVA between HRGs and LRGs in the TCGA-HCC dataset (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In detail, HALLMARK pathways from MSigDB were used as a reference gene set. The pathway scores for HRGs and LRGs were calculated, and the differences between groups were evaluated via the limma package (v 3.58.1) [\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e] (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, |t| \u0026gt; 2). A t value\u0026thinsp;\u0026gt;\u0026thinsp;2 indicated that the pathway was activated in the HRG, whereas a negative t value\u0026thinsp;\u0026lt;\u0026thinsp;2 suggested that the pathway was suppressed.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003e2.9 Tumor microenvironment (TME) analysis\u003c/h2\u003e\n \u003cp\u003eInitially, the ssGSEA algorithm in the GSVA package (v 1.42.0) was applied to calculate the infiltration scores of 28 immune cell types[\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e] within samples in the HRG and LRG. The differential immune infiltrating cell types between the HRG and LRG groups were recognized via the Wilcoxon rank-sum test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The psych package (v 2.1.6) was subsequently applied to perform Spearman correlation analysis, which correlated differential immune infiltrating cell types with each other and with metabolism-related prognostic genes (|cor| \u0026gt; 0.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e2.10 Immunotherapy response prediction\u003c/h2\u003e\n \u003cp\u003eThe immunophenotype score (IPS) is a predictor of the checkpoint blocker response[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Within the TCGA-HCC dataset, the tumor immunogenicity of patients in the HRG and LRG was assessed via the IPSs. Specifically, IPS data were obtained from The Cancer Immunome Atlas (TCIA) database (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://tcia.at/\u003c/span\u003e\u003c/span\u003e) on January 13, 2025. Differences in the IPSs between the HRG and LRG were assessed via the Wilcoxon rank-sum test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). IPS categorization considers key immune checkpoints, including CTLA-4 and PD-1. A total of 4 distinct IPS categories (IPS_CTLA4_neg_PD1_neg, IPS_CTLA4_neg_PD1_pos, IPS_CTLA4_pos_PD1_neg, and IPS_CTLA4_pos_PD1_pos) were included (pos: positive; neg: negative). High IPSs are related to high immunogenicity.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e2.11 Drug sensitivity analysis\u003c/h2\u003e\n \u003cp\u003eTo evaluate the effects of risk scores related to key cell type and KB metabolism on drug sensitivity, the half-maximal inhibitory concentration (IC50) values for 138 common drugs were estimated for HRG and LRG patients from the TCGA-HCC dataset via the OncoPredict package (v 1.2)[\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e]. Differences between groups were compared via the Wilcoxon rank-sum test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). The top 15 drugs with the most significant differences are shown via the ggplot2 package (v 3.5.1). Furthermore, the psych package (v 2.1.6) was used to conduct Spearman correlation analysis between risk scores and IC50 values for all 138 drugs. The top 10 drugs showing significant positive/negative correlations with risk scores are highlighted in a plot generated with the ggplot2 package (v 3.5.1) (|cor| \u0026gt; 0.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e2.12 Cell‒cell communication and cell trajectory analyses\u003c/h2\u003e\n \u003cp\u003eWithin the GSE149614 dataset, cell‒cell interactions among various cell types were systematically explored across all samples via the CellChat package (v 1.6.1)[\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e] (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, log2 mean expression of interacting molecules\u0026thinsp;\u0026ge;\u0026thinsp;0.1). In addition, ligand‒receptor (L‒R) pairs that mediate intercellular signaling were also examined (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01).\u003c/p\u003e\n \u003cp\u003eMoreover, trajectory analysis was conducted via reduceDimension and plot_cell_trajectory functions in the Monocle package (v 2.30.1)[\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e], enabling the simulation of the differentiation process of the key cell type in the HCC and control groups. Additionally, the plot_genes_in_pseudotime function in the Monocle package (v 2.30.1) was applied to explore the dynamic expression patterns of prognostic genes related to key cell type and KB metabolism during the key cell type differentiation trajectory.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e2.13 Reverse transcription‒quantitative PCR (RT‒qPCR)\u003c/h2\u003e\n \u003cp\u003ePrognostic gene expression was validated in clinical samples. Specifically, RNAs from 5 HCC samples and 5 control samples were isolated from individuals via TRIzol reagent (R401-01, Ambion, America). The collection was performed at Guangzhou Red Cross Hospital. The isolated RNAs were then used for cDNA synthesis via the Hifair\u0026reg; III 1st Strand cDNA Synthesis SuperMix for a qPCR kit (11141ES60, Yisheng, China). RT‒qPCR was subsequently performed using 2\u0026times;Universal Blue SYBR Green qPCR Master Mix (G3326‒05, Servicebio, China). The primers used for the prognostic genes and the internal reference gene (GAPDH) are listed in \u003cstrong\u003eAdditional file 2.\u003c/strong\u003e Following the RT‒qPCR procedure, the 2-\u0026Delta;\u0026Delta;Cт method was applied to identify expression profiles. A t test (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) was employed to evaluate intergroup differences, and data visualization was conducted via GraphPad Prism 5 software (v 8.0)[\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e]. Ethical approval was granted by the Ethics Committee of Guangzhou Red Cross Hospital on February 16, 2019. (approval number: 020\u0026ndash;34403034). All experiments were performed in accordance with relevant named guidelines and regulations and the authors complied with the ARRIVE guidelines.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e2.14 Statistical analysis\u003c/h2\u003e\n \u003cp\u003eR software (v 4.2.3) was used to conduct the bioinformatics analyses. Notably, the Wilcoxon rank-sum test, the log-rank test, and the t test were employed in this study to assess differences between specific groups, with the significance threshold set at p\u0026thinsp;\u0026lt;\u0026thinsp;0.05.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Various functions of cell types in HCC\u003c/h2\u003e \u003cp\u003ePotential mechanisms underlying the progression of HCC were explored at the single-cell level within GSE149614 dataset. Cell type profiles in HCC were initially explored. Specifically, through data filtration, 52,471 cells and 25,712 genes were retained for further analyses (\u003cb\u003eAdditional file 3a-b\u003c/b\u003e). Additionally, a variation in these genes was depicted in \u003cb\u003eAdditional file 3c\u003c/b\u003e, highlighting 2,000 highly variable genes. After selecting the top 20 PCs (p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) for UMAP dimensionality reduction according to elbow plot and PCA replacement test, 11 distinct cell clusters were determined (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea-c). Cell clusters were annotated into 9 types using marker genes (\u003cb\u003eAdditional file 4\u003c/b\u003e), including T cells, myeloid cells, hepatocytes, natural killer (NK) cells, endothelial cells (ECs), fibroblasts, plasma cells, B cells, and epithelial cells (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ed). The marker gene expression was illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ee. Notably, these cell types were related to functional pathways like \"biogenic amines are oxidatively deaminated to aldehydes by MAOA and MAOB\" and \"ethanol oxidation\" (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ef). The elucidation of these cell-related functions enhanced the understanding of the cellular processes involved and potentially offered new avenues for therapeutic strategies targeting HCC. The activities of KMRGs across various cell types were illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eg. Hepatocytes demonstrated the highest KMRG activity. Furthermore, T cells represented the most abundant cell population across all samples (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eh). Specifically, T cells and myeloid cells accounted for the highest proportions in the control and HCC groups, respectively. The proportion of hepatocytes was markedly increased in the HCC group (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ei). Considering both the notably high KMRG activity and significantly higher proportion in HCC, hepatocytes were ultimately selected as the key cell type.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Relevant functional pathways and the PPI network of candidate genes\u003c/h2\u003e \u003cp\u003eIn GSE149614 dataset, hepatocytes were extracted for further analyses. Through PCA, the top 20 PCs were selected (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea-b). Subsequent dimensionality reduction and reclustering revealed 3 hepatocyte subtypes (HC1, HC2, HC3) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). Next, hdWGCNA was performed based on these subtypes. Specifically, the optimal soft power threshold was confirmed to be 3, and a gene co-expression network was established, with genes clustered into 17 distinct modules (excluding the gray module) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed-e). Then, 1,837 key module genes were acquired from the blue, greenyellow, green, yellow, and turquoise modules, which exhibited strong correlations with specific hepatocyte subtypes (|cor| \u0026gt; 0.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef). It was worth mentioning that the blue module demonstrated the strongest positive and negative associations with HC1 (cor\u0026thinsp;=\u0026thinsp;0.49) and HC2 (cor = -0.52), respectively (p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWithin TCGA-HCC dataset, 4,590 DEGs in the HCC group were acquired, including 3,383 upregulated and 1,207 downregulated genes (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eg-h).\u003c/p\u003e \u003cp\u003eRemarkably, after intersecting key module genes, DEGs, and KMRGs, 10 candidate genes related to hepatocytes and KB metabolism were identified, including PPARGC1A, PDK4, HSD17B6, APOC3, OXCT1, EGR1, SRD5A1, TDO2, FGF19, and FABP1 (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ei). Candidate genes were significantly enriched in GO terms such as \"cellular ketone metabolic process\", as well as in KEGG pathways like \"steroid hormone biosynthesis\" (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ej-k). Moreover, the constructed PPI network contained 4 interaction pairs (such as TIMP1-EZR and TIMP1-SFRP1) involving 7 proteins, elucidating the interplay of specific candidate genes at the protein level (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003el). These results aided in understanding the multiple roles of hepatocytes and KB metabolism in HCC progression.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Strong predictive power of KMRGs for HCC prognosis demonstrated by a risk model\u003c/h2\u003e \u003cp\u003eWithin TCGA-HCC dataset, 5 candidate genes linked to OS were retained through univariate Cox analysis (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). Besides, PH assumption test revealed that they all satisfied assumption (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05) (\u003cb\u003eAdditional file 5\u003c/b\u003e). Among them, PPARGC1A, PDK4, HSD17B6, and APOC3 were linked to better HCC prognosis (HR\u0026thinsp;\u0026lt;\u0026thinsp;1). Conversely, OXCT1 was related to adverse HCC prognosis (HR\u0026thinsp;\u0026gt;\u0026thinsp;1). Therefore, they were selected as candidate prognostic genes and subjected to LASSO regression analysis. PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were ultimately identified as prognostic genes as their regression coefficients were not penalized to 0 (optimal lambda\u0026thinsp;=\u0026thinsp;0.00264) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb-c).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eAfter screening prognostic genes, a risk model related to hepatocytes and KB metabolism (risk score = (-0.197860092) \u0026times; PPARGC1A expression level + (-0.086179799) \u0026times; PDK4 expression level + (-0.007066472) \u0026times; HSD17B6 expression level + (-0.022773615) \u0026times; APOC3 expression level\u0026thinsp;+\u0026thinsp;0.208555505 \u0026times; OXCT1 expression level) was constructed. Subsequently, HCC patients in TCGA-HCC dataset were divided into HRG and LRG (181 : 187) based on the optimal risk score cut-off value (-0.9058876). Risk score and survival state distribution within risk groups revealed that the number of dead patients increased with increasing risk scores (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). Notably, PPARGC1A, PDK4, HSD17B6, and APOC3 exhibited higher expression in LRG, while OXCT1 demonstrated elevated expression in HRG (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee). Moreover, KM survival curves showed that HCC patients in LRG had markedly higher survival probabilities (p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef). Additionally, AUC values of ROC curves at 1-, 2-, and 3-year time po0ints all exceeded 0.6, reflecting notable predictive ability of this risk model (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eg). Furthermore, to finalize assessment, the risk model was subjected to validation within GSE14520 dataset. Likewise, HCC patients in this dataset were divided into HRG and LRG (53 : 77) based on the optimal risk score cut-off value (-0.9715922). Results revealed by risk score and survival state distribution (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eh), expression analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ei), KM survival curves (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ej), and ROC curves (AUC values all exceeded 0.6) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ek) were largely consistent with those in TCGA-HCC dataset. The generalizability of this risk model was superior, and it might be a valuable tool for risk assessment in clinical practice for HCC.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Outstanding the clinical and survival relevance of risk scores\u003c/h2\u003e \u003cp\u003eIn TCGA-HCC dataset, differences in risk scores related to hepatocytes and KB metabolism among various clinical subgroups were analyzed (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). Risk scores were notably linked to clinical stage and T stage, with notably higher risk scores observed in patients with advanced stages (stage 3/4; T3/4) (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01). Furthermore, significant survival differences between HRG and LRG were observed across multiple clinical strata, including age (\u0026le;\u0026thinsp;61 and \u0026gt;\u0026thinsp;61), gender (male), race (White and Asian), clinical stage (stage 1/2 and stage 3/4), T stage (T1/2 and T3/4), N stage (N0), and M stage (M0) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). The survival prediction ability of risk scores was excellent. Summarily, the robust clinical and prognostic significance of risk scores were found, emphasizing the potential roles of hepatocytes and ketone metabolism in the progression and outcome of HCC.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e3.5 The prognostic genes were associated with metabolism\u003c/h2\u003e \u003cp\u003eThrough GSEA, pathways associated with prognostic genes were analyzed in TCGA-HCC dataset. PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were significantly enriched in 58, 66, 72, 66, and 69 pathways, respectively (adjusted p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (\u003cb\u003eAdditional files 6\u0026ndash;10\u003c/b\u003e). The top 5 most significant pathways of each prognostic gene were shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea-e. It was worth mentioning that they were co-enriched in 6 critical pathways, including \"primary bile acid biosynthesis\", \"spliceosome\", \"cell cycle\", \"ubiquitin mediated proteolysis\", \"beta-alanine metabolism\", and \"homologous recombination\". It was suggested that KB metabolism might influence HCC progression by altering these pathways.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e3.6 Crucial functional pathways and TME profiles altered by risk score\u003c/h2\u003e \u003cp\u003eWithin the TCGA-HCC dataset, biological pathways altered by risk scores related to hepatocytes and KB were explored through GSVA. In detail, the activities of pathways like \"MYC targets V1\" and \"unfolded protein response\" were activated in HRG, while the activities of pathways such as \"fatty acid metabolism\" and \"bile acid metabolism\" were suppressed in HRG (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea). Hepatocytes and KB metabolism might exert critical roles in HCC progression by modulating these pathways.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eRemarkably, TME profiles of HRG and LRG in the TCGA-HCC dataset were explored. The infiltration levels of central memory CD4 T cells and monocytes were relatively high across HCC patients. There were 14 differential immune infiltrating cell types between HRG and LRG (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb). Among them, 11 cell types exhibited increased infiltration levels in HRG, such as activated CD4 T cells, activated dendritic cells, and myeloid-derived suppressor cells (MDSCs). Furthermore, the associations of differential immune infiltrating cell types with each other and with prognostic genes were explored (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Most of the differential immune infiltrating cell types demonstrated significantly and positively strong correlations with each other. The strongest correlation was noted between immature B cells and activated B cells (cor\u0026thinsp;=\u0026thinsp;0.86, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Regarding the relationships between differential immune infiltrating cell types and prognostic genes, central memory CD4 T cells exhibited the strongest positive correlation with OXCT1 (cor\u0026thinsp;=\u0026thinsp;0.44, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), while activated CD4 T cells showed the strongest inverse correlation with PDK4 (cor = -0.39, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). These immune characteristics highlighted the intricate associations of TME in HCC with hepatocytes and KB metabolism.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e3.7 Potential of the risk score as a predictor of immunotherapy and drug treatment outcomes\u003c/h2\u003e \u003cp\u003eIn TCGA-HCC dataset, differences in IPS between the HRG and LRG were assessed under various immune checkpoint expression conditions. Notably, under the CTLA4_neg_PD1_neg condition, the LRG exhibited significantly higher IPSs than the HRG (p\u0026thinsp;\u0026lt;\u0026thinsp;0.01), indicating a potentially greater tumor immunogenicity and enhanced responsiveness to immune checkpoint blockade therapy (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ea). Regarding drug sensitivity, it was found that the IC50 values for drugs like BMS.708163 were lower in LRG, whereas drugs such as camptothecin exhibited lower IC50 values in HRG (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003eb). Lower IC50 values were indicative of higher drug efficacy. Notably, 4 drugs such as BMS.708163 exhibited significant positive correlations with risk scores, while 34 drugs like camptothecin were significantly and negatively correlated with risk scores (|cor| \u0026gt; 0.3, p\u0026thinsp;\u0026lt;\u0026thinsp;0.0001) (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003ec). The differential drug sensitivities might be due to different drug metabolism mechanisms in HCC patients with varying risk scores. Taken together, risk scores related to hepatocytes and KB metabolism demonstrated superior potential in predicting immunotherapy and drug treatment outcomes for HCC patients.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e3.8 Complex intercellular communication patterns\u003c/h2\u003e \u003cp\u003eAnalysis of cell-cell communication in GSE149614 dataset revealed extensive intercellular interactions within microenvironment. Notably, fibroblasts, epithelial cells, hepatocytes, and ECs were involved in a greater number and intensity of interactions compared to other cell types (\u003cb\u003eAdditional file 11a-d\u003c/b\u003e). It was worth mentioning that hepatocytes exhibited interactions with all other cell types, with particularly frequent and relatively strong interactions observed between hepatocytes and myeloid cells. Compared to the control group, both the number and strength of interactions between hepatocytes and myeloid cells were reduced in the HCC group. Conversely, both the frequency and strength of the interactions between hepatocytes and other cell types were increased. These dynamic changes suggested a potential shift in the intercellular communication landscape during HCC progression. Furthermore, it was found that T cells participated in widespread intercellular signaling through numerous L-R interactions. In the control group, the GZMA-F2R pair demonstrated the highest communication probability in the T cell-to-EC interaction (\u003cb\u003eAdditional file 11e\u003c/b\u003e). In the HCC group, the MIF-(CD74\u0026thinsp;+\u0026thinsp;CXCR4) pair exhibited the highest communication probability in the T cell-to-B cell interaction, indicating a potentially critical signaling axis in the HCC microenvironment (\u003cb\u003eAdditional file 11f\u003c/b\u003e). The T cell-to-hepatocyte interaction primarily involved the NAMPT-INSR pair. Overall, deciphering these intricate intercellular communication networks offered deeper insight into the cellular crosstalk underlying HCC progression.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e3.9 Dynamic expression of prognostic genes during hepatocyte differentiation\u003c/h2\u003e \u003cp\u003eIn GSE149614 dataset, differentiation trajectories of hepatocytes over time were tracked and presented in \u003cb\u003eAdditional file 12a\u003c/b\u003e, showing that hepatocytes differentiated into distinct subtypes at different stages. Three differentiation states were observed, with most cells in state 2. In both control and HCC groups, hepatocytes were predominantly in the intermediate and late stages of differentiation.\u003c/p\u003e \u003cp\u003eMoreover, the expression dynamics of prognostic genes related to hepatocytes and KB metabolism were also illustrated during hepatocyte differentiation (\u003cb\u003eAdditional file 12b-c\u003c/b\u003e). Overall, the expression of PPARGC1A, PDK4, HSD17B6, and APOC3 exhibited an upward trend as differentiation progressed. Specifically, PPARGC1A expression remained relatively stable during the early and intermediate stages of differentiation, followed by an increase at the late stage before reaching a plateau. PDK4 expression initially decreased, then increased during the later stages, and eventually stabilized. HSD17B6 expression showed a continuous increase throughout the differentiation process. APOC3 expression increased during the early stages and stabilized during the late phase. In contrast, OXCT1 expression did not display significant changes across differentiation stages. In summary, the dynamic expression of these prognostic genes across hepatocyte differentiation stages might related to HCC progression.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e3.10 RT‒qPCR validation of prognostic genes\u003c/h2\u003e \u003cp\u003eExpression of prognostic genes related to hepatocytes and KB metabolism was validated in clinical samples. Compared to the control group, expression of PPARGC1A, PDK4, HSD17B6, APOC3 was significantly decreased in HCC, while OXCT1 exhibited elevated expression (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05) (\u003cb\u003eAdditional file 13a-e\u003c/b\u003e). Significant differential expression of prognostic genes further confirmed their prognostic significance in HCC.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eHCC is a highly heterogeneous malignant tumor, and its progression is closely associated with metabolic reprogramming. Disruptions in ketone body metabolism may affect the tumor microenvironment and patient prognosis[\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. In this study, multi-omics data from TCGA and GEO databases were integrated. Advanced bioinformatics methods were employed to identify five core prognostic genes (PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1) that are closely related to HCC function and ketone metabolism regulation. A specific HCC prognostic risk model was further constructed, while the dual clinical significance of hepatocyte-related KMRGs in prognostic assessment and tumor immune microenvironment regulation was systematically validated.Among the five core prognostic genes identified by the risk model, PPARGC1A, PDK4, HSD17B6, and APOC3 were significantly downregulated in HCC, whereas OXCT1 was markedly upregulated. Existing evidence indicates that these genes play critical roles in HCC cell proliferation, tumorigenicity, invasiveness, and drug resistance.\u003c/p\u003e \u003cp\u003eThe PPARGC1A gene encodes the transcriptional coactivator PGC-1α. Recent studies have identified PPARGC1A as a key mediator of metabolic reprogramming in cancer, enabling tumor cells to adapt to fluctuating metabolic demands[\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e]. Mechanistically, PGC-1α impedes HCC metastasis by suppressing the Warburg effect and aerobic glycolysis[\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e]. Conversely, SIRT1 promotes HCC metastasis by enhancing PGC-1α-mediated mitochondrial biogenesis[\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e]\u0026mdash;a finding that reveals the dual regulatory role of PPARGC1A in HCC pathogenesis. As a mitochondrial matrix enzyme, PDK4 regulates glucose metabolism by phosphorylating the E1α subunit to inhibit pyruvate dehydrogenase complex (PDC) activity[\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e]. Loss of PDK4 exacerbates the proliferation, tumorigenicity, motility, and invasiveness of HCC[\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e]. Notably, this study found a significant negative correlation between activated CD4\u0026thinsp;+\u0026thinsp;T cells and PDK4 expression, suggesting that PDK4 may be involved in tumor immune infiltration. HSD17B6 is an endoplasmic reticulum (ER)-localized oxidoreductase that regulates steroid hormone metabolism by catalyzing the reversible conversion between hydroxyl and ketone groups\u0026mdash;particularly in androgen biosynthesis. Decreased HSD17B6 expression in HCC disrupts cell proliferation, migration, invasion, and androgen metabolism, while also affecting tumor-infiltrating immune cells[\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e, \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e]. While the role of HSD17B6 in HCC progression and immune checkpoint therapy response is well established, the mechanistic interaction between HSD17B6 and ketone metabolism remains to be further explored. As a key small apolipoprotein in lipid metabolism, APOC3 regulates triglyceride homeostasis by inhibiting lipoprotein lipase activity and delaying hepatic clearance of triglyceride-rich lipoproteins[\u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e]. Recent studies have proposed APOC3 as a potential prognostic biomarker for hepatitis B virus (HBV)-associated HCC; its early upregulation is linked to steroid metabolism, PPAR signaling pathways, and fatty acid metabolism[\u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e56\u003c/span\u003e, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e57\u003c/span\u003e]. The expression pattern of APOC3 observed in this study differs from previous reports, which may be attributed to the temporal and microenvironmental heterogeneity of HCC and requires further validation.\u003c/p\u003e \u003cp\u003eAs a lysine succinyltransferase and a rate-limiting enzyme in ketone body metabolism, OXCT1 catalyzes the initiating and rate-limiting step of ketolysis. This enzyme converts extrahepatic ketone bodies into acetoacetyl-CoA, which then enters the tricarboxylic acid (TCA) cycle to participate in oxidative phosphorylation and ATP production[\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e]. Recent research evidence highlights the core role of this enzyme in metabolic reprogramming of malignant tumors[\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e59\u003c/span\u003e]. In HCC, OXCT1 expression is significantly upregulated. It drives ketone body catabolism to provide energy for tumor cell proliferation, invasion, and metastasis[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e60\u003c/span\u003e]. These findings establish OXCT1 as a key regulator in HCC hepatocyte function and ketone body metabolism, though its specific mechanism of action remains to be fully elucidated.Notably, a study by the Zhu Chuxu team demonstrated that conditional knockout of OXCT1 in macrophages is significantly associated with increased OXCT1 expression in tumor-associated macrophages (TAMs) and poor prognosis in HCC patients. This suggests that OXCT1-mediated metabolic reprogramming of TAMs toward a pro-tumor phenotype accelerates tumor progression[\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Our data reveal a positive correlation between OXCT1 levels and central memory CD4\u0026thinsp;+\u0026thinsp;T cells, implying that this enzyme is involved in regulating tumor immune dynamics.Overall, these findings indicate that inhibiting OXCT1 is a potential therapeutic strategy to block HCC progression by simultaneously targeting metabolic and immune pathways. However, further mechanistic studies are required to clarify its context-dependent role in HCC pathogenesis and therapeutic resistance.\u003c/p\u003e \u003cp\u003eFunctional enrichment analysis revealed that PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 are co-enriched in the ubiquitin-mediated proteolysis pathway. This suggests that dysregulation of these metabolic genes may affect HCC progression by disrupting protein homeostasis.The ubiquitin-proteasome system (UPS) is the most critical protein quality control mechanism in eukaryotic cells, responsible for eliminating misfolded, damaged, or regulatory proteins[\u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e61\u003c/span\u003e]. In HCC, UPS dysregulation is closely associated with tumorigenesis, invasiveness, and therapeutic resistance[\u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e62\u003c/span\u003e], and it regulates the stability of key tumor suppressors (e.g., p53, PTEN) and oncoproteins (e.g., β-catenin, c-Myc)[\u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e63\u003c/span\u003e]. Additionally, the UPS participates in the turnover of metabolic enzymes, regulates the metabolic plasticity of tumor cells, and mediates endoplasmic reticulum (ER)-associated degradation (ERAD) to maintain ER homeostasis[\u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e64\u003c/span\u003e]. HCC cells often exhibit enhanced proteasome activity and upregulation of specific E3 ubiquitin ligases; this adaptation enables them to sustain protein homeostasis under metabolic stress and support rapid proliferation. Among the five prognostic genes screened in this study, the PGC-1α protein itself is a key substrate of the UPS[\u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e65\u003c/span\u003e]. In HCC, the downregulation of PGC-1α may be partially attributed to increased ubiquitination-mediated degradation. Furthermore, PGC-1α-regulated mitochondrial biogenesis relies on the UPS to clear damaged mitochondrial proteins; its dysregulated expression may disrupt mitochondrial quality control (MQC), leading to the accumulation of dysfunctional mitochondria and exacerbated oxidative stress. HSD17B6 is localized to the ER, and its function as a steroid metabolic enzyme is closely linked to ER protein folding quality control[\u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e66\u003c/span\u003e]. Downregulation of HSD17B6 may result in the accumulation of androgen metabolic intermediates; these hydrophobic steroids may interfere with ER membrane integrity and the protein folding environment, triggering ER stress and activating the ERAD pathway\u0026mdash;a key mechanism for maintaining ER homeostasis[\u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e67\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eOur study found that the most significantly altered biological pathways in HCC, based on hepatocyte and KB metabolism-related risk scores, are \"MYC Targets V1\" and the \"unfolded protein response (UPR)\". Activation of the UPR pathway is closely associated with the aforementioned UPS dysregulation. When the ER\u0026rsquo;s protein-folding capacity is overwhelmed, the UPR not only initiates a transcriptional adaptation program but also recruits the UPS via the ERAD pathway to clear misfolded proteins[\u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e68\u003c/span\u003e]. In ERAD, misfolded ER proteins are retrotranslocated to the cytoplasm, labeled by E3 ubiquitin ligases (e.g., HRD1, gp78), and then degraded by the 26S proteasome[\u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e69\u003c/span\u003e]. Thus, the UPR and ubiquitin-mediated proteolysis pathway functionally form a tightly coupled quality control network, collectively maintaining protein homeostasis in HCC cells under metabolic stress.The c-Myc protein itself is a key substrate of the UPS, and its stability is regulated by multiple E3 ubiquitin ligases[\u003cspan citationid=\"CR70\" class=\"CitationRef\"\u003e70\u003c/span\u003e]. Under metabolic stress, UPR activation can affect c-Myc stability through multiple mechanisms: On one hand, enhanced proteasome activity mediated by ERAD may accelerate c-Myc turnover[\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e71\u003c/span\u003e]; on the other hand, ER stress-induced translational inhibition (via PERK-eIF2α) may selectively upregulate the synthesis of oncoproteins such as c-Myc[\u003cspan citationid=\"CR72\" class=\"CitationRef\"\u003e72\u003c/span\u003e]. The activation of \"MYC Targets V1\" observed in this study may reflect the coordinated dysregulation of this ubiquitin system-metabolism-proliferation signaling axis.Studies have indicated functional crosstalk between the UPR and PGC-1α (encoded by the PPARGC1A gene). As a master regulator of mitochondrial biogenesis and energy homeostasis, PGC-1α transcription is regulated by CREB3\u0026mdash;an ER-resident transcription factor that also coordinates UPR activation[\u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e73\u003c/span\u003e, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e74\u003c/span\u003e]. While existing studies have confirmed the association between UPR dysregulation and HCC development, the specific molecular mechanisms linking UPR signaling to HCC progression remain unclear.Based on the above evidence, we propose the following integrated mechanistic model: In HCC, dysregulation of metabolic genes such as PPARGC1A and HSD17B6 disrupts the UPS through multiple pathways, thereby triggering ER stress and UPR activation. The UPR further recruits the ubiquitin system via the ERAD pathway to clear misfolded proteins, while regulating the stability and transcriptional activity of key oncoproteins such as c-Myc. This cascade forms a loop that supports HCC progression. The complex triangular relationship among MYC Targets V1, the UPR, and PPARGC1A/PGC-1α reveals a molecular interaction network that may explain metabolic adaptation and therapeutic resistance in HCC, presenting broad research prospects.\u003c/p\u003e \u003cp\u003eThis study, for the first time, systematically reveals the key role of KMRGs in hepatocytes in the prognosis of HCC patients by integrating single-cell and transcriptome data. We successfully constructed a risk prediction model based on five key genes: APOC3, HSD17B6, PDK4, PPARGC1A, and OXCT1. This discovery not only provides new insights into the metabolic reprogramming mechanism of HCC but also lays an important foundation for clinical prognosis assessment and the development of personalized treatment strategies. However, this study has several limitations. First, the sample size of single-cell data is relatively limited (n\u0026thinsp;=\u0026thinsp;18), which may not fully reflect the tumor heterogeneity of HCC. Second, although we identified candidate genes through bioinformatics methods, their specific regulatory mechanisms in ketone metabolism still need to be further verified through in vitro and in vivo experiments. Finally, the clinical application value of the model needs to be evaluated in larger-scale prospective cohorts. On the basis of these limitations, future research can focus on the following aspects. First, the single-cell sequencing sample size should be expanded, especially for HCC patients with different etiologies (such as HBV and HCV infection) and clinical stages, to improve the universality of the model; second, gene editing technology and animal models should be used to explore the molecular mechanisms of key genes in the ketone metabolism regulatory network, such as PDK4 and PPARGC1A; and third, metabolomics and drug sensitivity data should be integrated to develop precise treatment strategies for high-risk patients. These follow-up studies will help promote in-depth development in the field of HCC metabolic therapy.\u003c/p\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eThis study identified five key prognostic genes linked to hepatocyte-specific ketone body metabolism in HCC through integrated multi-omics analysis. The constructed risk model demonstrated robust predictive power for patient outcomes, correlating with altered pathways (e.g., MYC targets and UPR), immune infiltration, and drug sensitivity. Downregulation of PPARGC1A, PDK4, HSD17B6, and APOC3, along with OXCT1 upregulation, highlights metabolic reprogramming as a critical driver of HCC progression. These findings provide novel insights into HCC heterogeneity and therapeutic targeting, although further experimental validation and clinical cohort studies are warranted to translate these findings into precision medicine.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"99%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAbbreviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003e\u003cstrong\u003efull name\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eKB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eketone body\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eHCC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ehepatocellular carcinoma\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003escRNA-seq\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003esingle-cell RNA sequencing\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ehdWGCNA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ehigh-dimensional weighted gene coexpression network analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eDEGs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003edifferentially expressed genes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eKMRGs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ekb metabolism-related genes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ePH\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eproportional hazard\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eLASSO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eleast absolute shrinkage and selection operator\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eTME\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003etumor microenvironment\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eRT‒qPCR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ereverse transcription-quantitative PCR\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eNAFLD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003enonalcoholic fatty liver disease\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eAcAc\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eacetoacetate\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eBHB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eD-\u0026beta;-hydroxybutyrate\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eROS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ereactive oxygen species\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eTCGA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ethe cancer genome atlas\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eGEO\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003egene expression omnibus\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eKMRGs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eketone body metabolism-related genes\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eMSigDB\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003emolecular signatures database\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eVST\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003evariance-stabilizing transformation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ePCA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eprincipal component analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003ePC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eprincipal component\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eUMAP\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003euniform manifold approximation and projection\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eAUC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003earea under the curve\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eSTRING\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ethe search tool for the retrieval of interacting genes/proteins\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eROC\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003ereceiver operating characteristic\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eGSVA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003egene set variation analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eIPS\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eimmunophenotype score\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 17px;\"\u003e\n \u003cp\u003eTCIA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 82px;\"\u003e\n \u003cp\u003eThe cancer immunome atlas\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study was conducted in accordance with the Declaration of Helsinki, and approved by the Ethics Committee of Guangzhou Red Cross Hospital on February 16, 2019 (approval number: 020--34403034). All subjects included in this study signed the \u0026quot;Informed Consent for the Use of Residual Samples\u0026quot; upon admission to our hospital, and all patients were notified prior to sample use. Due to the protection of patients\u0026apos; personal privacy, they were not provided individually.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003enot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNA\u003c/p\u003e\n\u003cp\u003eAvailability of data and materials\u003c/p\u003e\n\u003cp\u003eThe datasets supporting this study are available from the following public repositories. The TCGA-HCC (LIHC) dataset was accessed via The Cancer Genome Atlas (https://portal.gdc.cancer.gov/) with the data version as of February 10, 2025. The GEO datasets GSE14520 (platform GPL571) and GSE149614 (platform GPL24676) were obtained from the Gene Expression Omnibus. The ketone body metabolism-related gene set \u0026ldquo;GOBP_CELLULAR_KETONE_METABOLIC_PROCESS\u0026rdquo; (GO:0042180) was sourced from the Molecular Signatures Database (MSigDB). CellMarker 2.0 (http://biobbio-bigdata.hrbmu.edu.cn/CellMarker/) was used for cell type marker annotation. The protein\u0026ndash;protein interaction network was constructed using the STRING database (version 11.5, https://string-db.org/). Immunophenotype score (IPS) data were obtained from The Cancer Immunome Atlas (TCIA) with access date January 13, 2025.\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\u003eThis research was funded by the [Guangdong Provincial Natural Science Foundation General Program] [2025A1515012455, PW) and the [Guangdong Provincial-Joint Funding Project of Jinan University and the Municipal Government] [2025A03J3595].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWei yuan and Runyu Zhuang: Conceptualization, Data curation, Validation, Visualization, Writing\u0026ndash;original draft, Writing\u0026ndash;review \u0026amp; editing. Benliang Mao: Data curation, Validation, Visualization, Writing\u0026ndash;review \u0026amp; editing. Shanfei Zhu and Qi Cheng: Validation, Writing\u0026ndash;review \u0026amp; editing. Bailin Wang and Fan Wu: Visualization, Writing\u0026ndash;review \u0026amp; editing. Conceptualization, Supervision. Pengzhen Wang and Bo Ning: Conceptualization, project administration, supervision, writing\u0026ndash;review \u0026amp; editing. This work currently described has not been published and is not being considered for publication elsewhere, and its publication was approved by all the authors. All the authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe would like to express our sincere gratitude to all the individuals and organizations who supported and assisted us throughout this research. Special thanks to the following authors: Bailin Wang, Fan Wu, Benliang Mao, Shanfei Zhu and Qi Cheng. In conclusion, we extend our thanks to everyone who has supported and assisted us along the way. Without your support, this research would not have been possible.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; information (optional)\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2024; 74:229-63.\u003c/li\u003e\n\u003cli\u003eLai SW. Risk factors for hepatocellular carcinoma. Cancer. 2019; 125:482.\u003c/li\u003e\n\u003cli\u003eShi T, Iwama H, Fujita K, Kobara H, Nishiyama N, Fujihara S, et al. Evaluating the Effect of Lenvatinib on Sorafenib-Resistant Hepatocellular Carcinoma Cells. Int J Mol Sci. 2021; 22.\u003c/li\u003e\n\u003cli\u003eRuan HB, Crawford PA. Ketone bodies as epigenetic modifiers. Curr Opin Clin Nutr Metab Care. 2018; 21:260-6.\u003c/li\u003e\n\u003cli\u003eLu Y, Han G, Zhang Y, Zhang L, Li Z, Wang Q, et al. M2 macrophage-secreted exosomes promote metastasis and increase vascular permeability in hepatocellular carcinoma. Cell Commun Signal. 2023; 21:299.\u003c/li\u003e\n\u003cli\u003eZhao C, Wang H, Liu Y, Cheng L, Wang B, Tian X, et al. Biased allosteric activation of ketone body receptor HCAR2 suppresses inflammation. Mol Cell. 2023; 83:3171-87.e7.\u003c/li\u003e\n\u003cli\u003eNewman JC, Verdin E. \u0026beta;-Hydroxybutyrate: A Signaling Metabolite. Annu Rev Nutr. 2017; 37:51-76.\u003c/li\u003e\n\u003cli\u003eOffermanns S. Free fatty acid (FFA) and hydroxy carboxylic acid (HCA) receptors. Annu Rev Pharmacol Toxicol. 2014; 54:407-34.\u003c/li\u003e\n\u003cli\u003eGraff EC, Fang H, Wanders D, Judd RL. Anti-inflammatory effects of the hydroxycarboxylic acid receptor 2. Metabolism. 2016; 65:102-13.\u003c/li\u003e\n\u003cli\u003eShi X, Li X, Li D, Li Y, Song Y, Deng Q, et al. \u0026beta;-Hydroxybutyrate activates the NF-\u0026kappa;B signaling pathway to promote the expression of pro-inflammatory factors in calf hepatocytes. Cell Physiol Biochem. 2014; 33:920-32.\u003c/li\u003e\n\u003cli\u003eMiyauchi T, Uchida Y, Kadono K, Hirao H, Kawasoe J, Watanabe T, et al. Up-regulation of FOXO1 and reduced inflammation by \u0026beta;-hydroxybutyric acid are essential diet restriction benefits against liver injury. Proc Natl Acad Sci U S A. 2019; 116:13533-42.\u003c/li\u003e\n\u003cli\u003eKolb H, Kempf K, R\u0026ouml;hling M, Lenzen-Schulte M, Schloot NC, Martin S. Ketone bodies: from enemy to friend and guardian angel. BMC Med. 2021; 19:313.\u003c/li\u003e\n\u003cli\u003ePuchalska P, Crawford PA. Multi-dimensional Roles of Ketone Bodies in Fuel Metabolism, Signaling, and Therapeutics. Cell Metab. 2017; 25:262-84.\u003c/li\u003e\n\u003cli\u003eHwang CY, Choe W, Yoon KS, Ha J, Kim SS, Yeo EJ, et al. Molecular Mechanisms for Ketone Body Metabolism, Signaling Functions, and Therapeutic Potential in Cancer. Nutrients. 2022; 14.\u003c/li\u003e\n\u003cli\u003eWang YH, Liu CL, Chiu WC, Twu YC, Liao YJ. HMGCS2 Mediates Ketone Production and Regulates the Proliferation and Metastasis of Hepatocellular Carcinoma. Cancers (Basel). 2019; 11.\u003c/li\u003e\n\u003cli\u003eZhou X, Du J, Liu C, Zeng H, Chen Y, Liu L, et al. A Pan-Cancer Analysis of CD161, a Potential New Immune Checkpoint. Front Immunol. 2021; 12:688215.\u003c/li\u003e\n\u003cli\u003eZhu CX, Yan K, Chen L, Huang RR, Bian ZH, Wei HR, et al. Targeting OXCT1-mediated ketone metabolism reprograms macrophages to promote antitumor immunity via CD8(+) T cells in hepatocellular carcinoma. J Hepatol. 2024; 81:690-703.\u003c/li\u003e\n\u003cli\u003eHuang D, Li T, Wang L, Zhang L, Yan R, Li K, et al. Hepatocellular carcinoma redirects to ketolysis for progression under nutrition deprivation stress. Cell Res. 2016; 26:1112-30.\u003c/li\u003e\n\u003cli\u003eGuo D, Yu Q, Tong Y, Qian X, Meng Y, Ye F, et al. OXCT1 succinylation and activation by SUCLA2 promotes ketolysis and liver tumor growth. Mol Cell. 2025; 85:843-56.e6.\u003c/li\u003e\n\u003cli\u003eChen H, Ye F, Guo G. Revolutionizing immunology with single-cell RNA sequencing. Cell Mol Immunol. 2019; 16:242-9.\u003c/li\u003e\n\u003cli\u003ePapalexi E, Satija R. Single-cell RNA sequencing to explore immune cell heterogeneity. Nat Rev Immunol. 2018; 18:35-45.\u003c/li\u003e\n\u003cli\u003eZhang Y, Wang D, Peng M, Tang L, Ouyang J, Xiong F, et al. Single-cell RNA sequencing in cancer research. J Exp Clin Cancer Res. 2021; 40:81.\u003c/li\u003e\n\u003cli\u003eCheng C, Chen W, Jin H, Chen X. A Review of Single-Cell RNA-Seq Annotation, Integration, and Cell-Cell Communication. Cells. 2023; 12.\u003c/li\u003e\n\u003cli\u003eLei Y, Tang R, Xu J, Wang W, Zhang B, Liu J, et al. Applications of single-cell sequencing in cancer research: progress and perspectives. J Hematol Oncol. 2021; 14:91.\u003c/li\u003e\n\u003cli\u003eDeng H, Wang X, Jiang ZA, Xu J, Zhang Y, Zhou Y, et al. Clinical potential and experimental validation of prognostic genes in hepatocellular carcinoma revealed by risk modeling utilizing single cell and transcriptome constructs. Front Immunol. 2025; 16:1541252.\u003c/li\u003e\n\u003cli\u003eChen L, Yang H, Wei X, Liu J, Han X, Zhang C, et al. Integrated single-cell and bulk transcriptome analysis of R-loop score-based signature with regard to immune microenvironment, lipid metabolism and prognosis in HCC. Front Immunol. 2024; 15:1487372.\u003c/li\u003e\n\u003cli\u003eSatija R, Farrell JA, Gennert D, Schier AF, Regev A. Spatial reconstruction of single-cell gene expression data. Nat Biotechnol. 2015; 33:495-502.\u003c/li\u003e\n\u003cli\u003eLu Y, Yang A, Quan C, Pan Y, Zhang H, Li Y, et al. A single-cell atlas of the multicellular ecosystem of primary and metastatic hepatocellular carcinoma. Nat Commun. 2022; 13:4594.\u003c/li\u003e\n\u003cli\u003eGriss J, Viteri G, Sidiropoulos K, Nguyen V, Fabregat A, Hermjakob H. ReactomeGSA - Efficient Multi-Omics Comparative Pathway Analysis. Mol Cell Proteomics. 2020; 19:2115-25.\u003c/li\u003e\n\u003cli\u003eLu Y, Li K, Hu Y, Wang X. Expression of Immune Related Genes and Possible Regulatory Mechanisms in Alzheimer\u0026apos;s Disease. Front Immunol. 2021; 12:768966.\u003c/li\u003e\n\u003cli\u003eGustavsson EK, Zhang D, Reynolds RH, Garcia-Ruiz S, Ryten M. ggtranscript: an R package for the visualization and interpretation of transcript isoforms using ggplot2. Bioinformatics. 2022; 38:3844-6.\u003c/li\u003e\n\u003cli\u003eLi R, Zhao M, Miao C, Shi X, Lu J. Identification and validation of key biomarkers associated with macrophages in nonalcoholic fatty liver disease based on hdWGCNA and machine learning. Aging (Albany NY). 2023; 15:15451-72.\u003c/li\u003e\n\u003cli\u003eOrifjon S, Jammatov J, Sousa C, Barros R, Vasconcelos O, Rodrigues P. Translation and Adaptation of the Adult Developmental Coordination Disorder/Dyspraxia Checklist (ADC) into Asian Uzbekistan. Sports (Basel). 2023; 11.\u003c/li\u003e\n\u003cli\u003eZheng Y, Gao W, Zhang Q, Cheng X, Liu Y, Qi Z, et al. Ferroptosis and Autophagy-Related Genes in the Pathogenesis of Ischemic Cardiomyopathy. Front Cardiovasc Med. 2022; 9:906753.\u003c/li\u003e\n\u003cli\u003eWu T, Hu E, Xu S, Chen M, Guo P, Dai Z, et al. clusterProfiler 4.0: A universal enrichment tool for interpreting omics data. Innovation (Camb). 2021; 2:100141.\u003c/li\u003e\n\u003cli\u003eLiu P, Xu H, Shi Y, Deng L, Chen X. Potential Molecular Mechanisms of Plantain in the Treatment of Gout and Hyperuricemia Based on Network Pharmacology. Evid Based Complement Alternat Med. 2020; 2020:3023127.\u003c/li\u003e\n\u003cli\u003eLei J, Qu T, Cha L, Tian L, Qiu F, Guo W, et al. Clinicopathological characteristics of pheochromocytoma/paraganglioma and screening of prognostic markers. J Surg Oncol. 2023; 128:510-8.\u003c/li\u003e\n\u003cli\u003eFriedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw. 2010; 33:1-22.\u003c/li\u003e\n\u003cli\u003eXie Y, He L, Lugano R, Zhang Y, Cao H, He Q, et al. Key molecular alterations in endothelial cells in human glioblastoma uncovered through single-cell RNA sequencing. JCI Insight. 2021; 6.\u003c/li\u003e\n\u003cli\u003eGu Z, H\u0026uuml;bschmann D. Make Interactive Complex Heatmaps in R. Bioinformatics. 2022; 38:1460-2.\u003c/li\u003e\n\u003cli\u003eHeagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics. 2000; 56:337-44.\u003c/li\u003e\n\u003cli\u003eH\u0026auml;nzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics. 2013; 14:7.\u003c/li\u003e\n\u003cli\u003eRitchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015; 43:e47.\u003c/li\u003e\n\u003cli\u003eCharoentong P, Finotello F, Angelova M, Mayer C, Efremova M, Rieder D, et al. Pan-cancer Immunogenomic Analyses Reveal Genotype-Immunophenotype Relationships and Predictors of Response to Checkpoint Blockade. Cell Rep. 2017; 18:248-62.\u003c/li\u003e\n\u003cli\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.\u003c/li\u003e\n\u003cli\u003eAzadian S, Doustmohammadi A, Naseri M, Khodarahmi M, Arab SS, Yazdanifar M, et al. Reconstructing the cell-cell interaction network among mouse immune cells. Biotechnol Bioeng. 2023; 120:2756-64.\u003c/li\u003e\n\u003cli\u003eJoo EH, Kim S, Park D, Lee T, Park WY, Han KY, et al. Migratory Tumor Cells Cooperate with Cancer Associated Fibroblasts in Hormone Receptor-Positive and HER2-Negative Breast Cancer. Int J Mol Sci. 2024; 25.\u003c/li\u003e\n\u003cli\u003eVarillas-Delgado D. Role of the PPARGC1A Gene and Its rs8192678 Polymorphism on Sport Performance, Aerobic Capacity, Muscle Adaptation and Metabolic Diseases: A Narrative Review. Genes (Basel). 2024; 15.\u003c/li\u003e\n\u003cli\u003eZuo Q, He J, Zhang S, Wang H, Jin G, Jin H, et al. PPAR\u0026gamma; Coactivator-1\u0026alpha; Suppresses Metastasis of Hepatocellular Carcinoma by Inhibiting Warburg Effect by PPAR\u0026gamma;-Dependent WNT/\u0026beta;-Catenin/Pyruvate Dehydrogenase Kinase Isozyme 1 Axis. Hepatology. 2021; 73:644-60.\u003c/li\u003e\n\u003cli\u003eLi Y, Xu S, Li J, Zheng L, Feng M, Wang X, et al. SIRT1 facilitates hepatocellular carcinoma metastasis by promoting PGC-1\u0026alpha;-mediated mitochondrial biogenesis. Oncotarget. 2016; 7:29255-74.\u003c/li\u003e\n\u003cli\u003eDou X, Fu Q, Long Q, Liu S, Zou Y, Fu D, et al. PDK4-dependent hypercatabolism and lactate production of senescent cells promotes cancer malignancy. Nat Metab. 2023; 5:1887-910.\u003c/li\u003e\n\u003cli\u003eQin YJ, Lin TY, Lin XL, Liu Y, Zhao WT, Li XY, et al. Loss of PDK4 expression promotes proliferation, tumorigenicity, motility and invasion of hepatocellular carcinoma cells. J Cancer. 2020; 11:4397-405.\u003c/li\u003e\n\u003cli\u003eWu M, Jiang L. Hydroxysteroid 17-Beta Dehydrogenase 6 Is a Prognostic Biomarker and Correlates with Immune Infiltrates in Hepatocellular Carcinoma. Dig Dis Sci. 2022; 67:146-58.\u003c/li\u003e\n\u003cli\u003eLv L, Zhao Y, Wei Q, Zhao Y, Yi Q. Downexpression of HSD17B6 correlates with clinical prognosis and tumor immune infiltrates in hepatocellular carcinoma. Cancer Cell Int. 2020; 20:210.\u003c/li\u003e\n\u003cli\u003eChebli J, Larouche M, Gaudet D. APOC3 siRNA and ASO therapy for dyslipidemia. Curr Opin Endocrinol Diabetes Obes. 2024; 31:70-7.\u003c/li\u003e\n\u003cli\u003eWang X, Gong Y, Deng T, Zhang L, Liao X, Han C, et al. Diagnostic and prognostic significance of mRNA expressions of apolipoprotein A and C family genes in hepatitis B virus-related hepatocellular carcinoma. J Cell Biochem. 2019; 120:18246-65.\u003c/li\u003e\n\u003cli\u003eLi X, Wang L, Wang L, Feng Z, Peng C. Single-Cell Sequencing of Hepatocellular Carcinoma Reveals Cell Interactions and Cell Heterogeneity in the Microenvironment. Int J Gen Med. 2021; 14:10141-53.\u003c/li\u003e\n\u003cli\u003eDong YN, Mesaros C, Xu P, Mercado-Ay\u0026oacute;n E, Halawani S, Ngaba LV, et al. Frataxin controls ketone body metabolism through regulation of OXCT1. PNAS Nexus. 2022; 1:pgac142.\u003c/li\u003e\n\u003cli\u003eZhang S, Xie C. The role of OXCT1 in the pathogenesis of cancer as a rate-limiting enzyme of ketone body metabolism. Life Sci. 2017; 183:110-5.\u003c/li\u003e\n\u003cli\u003eMa W, Sun Y, Yan R, Zhang P, Shen S, Lu H, et al. OXCT1 functions as a succinyltransferase, contributing to hepatocellular carcinoma via succinylating LACTB. Mol Cell. 2024; 84:538-51.e7.\u003c/li\u003e\n\u003cli\u003ePohl C, Dikic I. Cellular quality control by the ubiquitin-proteasome system and autophagy. Science. 2019; 366:818-22.\u003c/li\u003e\n\u003cli\u003eDawson SP. Hepatocellular carcinoma and the ubiquitin-proteasome system. Biochim Biophys Acta. 2008; 1782:775-84.\u003c/li\u003e\n\u003cli\u003eNakayama KI, Nakayama K. Ubiquitin ligases: cell-cycle control and cancer. Nat Rev Cancer. 2006; 6:369-81.\u003c/li\u003e\n\u003cli\u003eRuggiano A, Foresti O, Carvalho P. Quality control: ER-associated degradation: protein quality control and beyond. J Cell Biol. 2014; 204:869-79.\u003c/li\u003e\n\u003cli\u003eTrausch-Azar J, Leone TC, Kelly DP, Schwartz AL. Ubiquitin proteasome-dependent degradation of the transcriptional coactivator PGC-1{alpha} via the N-terminal pathway. J Biol Chem. 2010; 285:40192-200.\u003c/li\u003e\n\u003cli\u003eBlack VH, Sanjay A, van Leyen K, M\u0026ouml;eller I, Lauring B, Kreibich G. Cholesterol and steroid synthesizing smooth endoplasmic reticulum of adrenocortical cells contains high levels of translocation apparatus proteins. Endocr Res. 2002; 28:425-30.\u003c/li\u003e\n\u003cli\u003eHan J, Kaufman RJ. The role of ER stress in lipid metabolism and lipotoxicity. J Lipid Res. 2016; 57:1329-38.\u003c/li\u003e\n\u003cli\u003eRen C, Cui H, Bao X, Huang L, He S, Fong HKW, et al. Proteopathy Linked to Exon-Skipping Isoform of RGR-Opsin Contributes to the Pathogenesis of Age-Related Macular Degeneration. Invest Ophthalmol Vis Sci. 2023; 64:41.\u003c/li\u003e\n\u003cli\u003eWu X, Siggel M, Ovchinnikov S, Mi W, Svetlov V, Nudler E, et al. Structural basis of ER-associated protein degradation mediated by the Hrd1 ubiquitin ligase complex. Science. 2020; 368.\u003c/li\u003e\n\u003cli\u003eAmati B. Myc degradation: dancing with ubiquitin ligases. Proc Natl Acad Sci U S A. 2004; 101:8843-4.\u003c/li\u003e\n\u003cli\u003eKropp EM, Matono S, Robida A, Kandarpa M, Wang O, Alt A, et al. Inhibition of Endoplasmic Reticulum Associated Degradation Is Cytotoxic to Relapsed Refractory Multiple Myeloma through Altered Proapoptotic Signaling. 2024; 144:3264.\u003c/li\u003e\n\u003cli\u003eHolcik M. Could the eIF2\u0026alpha;-Independent Translation Be the Achilles Heel of Cancer? Front Oncol. 2015; 5:264.\u003c/li\u003e\n\u003cli\u003eLocke B, Campbell E, Lu R. CREB3 mediates the transcriptional regulation of PGC-1\u0026alpha;, a master regulator of energy homeostasis and mitochondrial biogenesis. FEBS Lett. 2024; 598:1730-9.\u003c/li\u003e\n\u003cli\u003eKristensen CM, Dethlefsen MM, T\u0026oslash;ndering AS, Lassen SB, Meldgaard JN, Ringholm S, et al. PGC-1\u0026alpha; in hepatic UPR during high-fat high-fructose diet and exercise training in mice. Physiol Rep. 2018; 6:e13819.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"dion","sideBox":"Learn more about [Discover Oncology](https://www.springer.com/12672)","snPcode":"","submissionUrl":"","title":"Discover Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Hepatocellular carcinoma, Hepatocytes, Ketone body metabolism, Prognostic gene, Tumor immunity","lastPublishedDoi":"10.21203/rs.3.rs-9187321/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9187321/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eThe precise role of ketone body (KB) metabolism and the underlying cellular mechanisms in hepatocellular carcinoma (HCC) remain unclear. This study aimed to identify and mechanistically explore KB metabolism-related prognostic genes in HCC.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eAll data were obtained from public databases. Single-cell RNA sequencing (scRNA-seq) analysis and high-dimensional weighted gene co-expression network analysis (hdWGCNA) were integrated to identify key module genes related to key cell type with high KB metabolism-related gene (KMRG) activity. After intersecting with differentially expressed genes (DEGs) in HCC and KMRGs, candidate genes were acquired. Subsequently, prognostic genes were recognized by univariate Cox and least absolute shrinkage and selection operator (LASSO) regression. A risk model for risk stratification was established and validated. Further risk group-based analyses mainly examined the impact of risk scores on function, tumor microenvironment (TME), immunotherapy response, and drug sensitivity. Prognostic gene expression dynamics during key cell type differentiation were analyzed by cell trajectory analysis. Eventually, prognostic gene expression was validated via reverse transcription-quantitative PCR (RT-qPCR)\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eHepatocytes were identified as the key cell type with high KMRG activity. Remarkably, PPARGC1A, PDK4, HSD17B6, APOC3, and OXCT1 were identified as prognostic genes. RT-qPCR demonstrated that OXCT1 was upregulated in HCC, while the other prognostic genes were downregulated. The constructed risk model exhibited robust predictive capacity, showing that high-risk patients had lower survival probabilities. Activities of pathways like \"MYC targets V1\", infiltration of immune cell types like activated CD4 T cells, response to immunotherapy, and sensitivities to drugs like camptothecin were altered by risk scores. Dynamic expression changes of PPARGC1A, PDK4, HSD17B6, and APOC3 were observed during hepatocyte differentiation.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eFive prognostic genes related to hepatocytes and KB metabolism were identified in HCC, and a risk model with strong predictive utility was developed, offering novel insights into clinical prognostic prediction for HCC.\u003c/p\u003e","manuscriptTitle":"Identification of prognostic genes for hepatocellular carcinoma based on hepatocyte and ketone body metabolism using integrated bulk and single cell RNA sequencing","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-22 05:44:02","doi":"10.21203/rs.3.rs-9187321/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-04-21T02:55:50+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"145937109406246074054725649880151300007","date":"2026-04-21T01:58:21+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-14T14:47:58+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"160377134263341446269576000047265447265","date":"2026-04-14T14:16:41+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-14T06:40:17+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-06T08:32:38+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-04-02T05:45:57+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-30T07:36:43+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Oncology","date":"2026-03-30T07:29:11+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-oncology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"dion","sideBox":"Learn more about [Discover Oncology](https://www.springer.com/12672)","snPcode":"","submissionUrl":"","title":"Discover Oncology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e784b592-f2fa-4335-b8ac-bb70c62b273e","owner":[],"postedDate":"April 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-22T05:44:02+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-22 05:44:02","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9187321","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9187321","identity":"rs-9187321","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}