2.1 Identification and Prognostic Relevance of Lactate-Associated Genes

We used data from 371 hepatocellular carcinoma (HCC) patients with complete survival information from The Cancer Genome Atlas (TCGA) to evaluate the prognostic significance of lactate-associated genes. Univariate Cox regression analysis initially identified 128 lactate-associated genes significantly correlated with HCC prognosis from a list of 332 genes linked to lactate processes according to the literature [16]. The top 10 genes significantly associated with HCC prognosis were G6PD, TCOF1, KIF2C, RAN, PPM1G, JPT1, CACYBP, CCT5, ENO1, and IK (Figure 1A). LASSO regression was then used to refine the selection of candidate prognostic genes, identifying nine potential genes from the list (Figure 1B,C). To confirm the independent prognostic value of the selected genes, we performed multivariate Cox regression analysis. This analysis identified three critical prognostic genes: CALML5, G6PD, and IK (Figure 1D). These genes were the strongest predictors of overall survival in HCC patients, adjusting for other clinical variables such as age, gender, and cancer stage. A prognostic model was constructed using these three genes (CALML5, G6PD, and IK), and risk scores were calculated based on their gene expression levels. Patients were stratified into high- and low-risk groups based on median risk scores (Figure 1E). The high-risk group showed significantly worse overall survival than the low-risk group, highlighting the prognostic relevance of the lactate-associated gene model. Finally, we performed univariate and multivariate Cox regression analyses to assess the impact of clinical variables, such as gene score and tumor stage, on prognosis. Both analyses showed that the gene score from the prognostic model and tumor stage were significantly correlated with survival outcomes in HCC patients (Figure 1F). Our analysis identified G6PD, CALML5, and IK as critical prognostic markers for HCC based on their association with lactate metabolism and survival. The prognostic model, incorporating these three genes, effectively stratifies patients into high- and low-risk groups, offering valuable insights for personalized treatment strategies in HCC.

2.2 Construction and Validation of Lactate-Associated Gene Prognostic Model

We constructed and validated a prognostic model based on lactate-associated genes (CALML5, G6PD, and IK), using the TCGA LIHC cohort, divided into training (Figure 2A–D) and internal validation (Figure 2E–H) sets, along with the external validation cohort (LIHC-JP) (Figure 2I–L). The model aimed to assess the predictive value of lactate metabolism-related genes in HCC and was evaluated across different datasets. In the training set, patients were stratified into high- and low-risk groups based on risk scores derived from the lactate-associated gene expression model. Kaplan–Meier survival analysis revealed a significant survival difference between the two groups. Patients in the high-risk group showed significantly poorer survival than those in the low-risk group, with increased mortality and decreased overall survival (OS) as risk scores increased (Figure 2A,E). The distribution of risk scores and survival status was visualized to highlight survival differences (Figure 2B,C,F,G). A heatmap of the three key genes in the prognostic model was generated to compare their expression patterns between high- and low-risk groups (Figure 2D,H). This analysis showed that the expression of these genes was significantly higher in the high-risk group than in the low-risk group, reinforcing the association between elevated gene expression and poor prognosis. The heatmap also illustrated the distinct molecular characteristics of each risk group, confirming the prognostic relevance of lactate-associated genes in HCC. The external validation cohort (LIHC-JP) was used to validate the predictive value of the lactate-associated gene model. Similar to the training set, the high-risk group in the validation set showed significantly worse prognosis than the low-risk group (Figure 2I), confirming the robustness of the model across datasets. The external validation heatmap (Figure 2L) showed consistent expression patterns of the three key genes, confirming their relevance in predicting patient outcomes in an independent cohort. The distribution of risk scores and survival analysis in the external validation cohort mirrored the findings in the training set, further validating the prognostic utility of the model (Figure 2J,K). These findings suggest that lactate metabolism-related genes may serve as valuable biomarkers for prognosis in HCC.

2.3 Independent Prognostic Analysis and PCA

We performed Decision Curve Analysis (DCA) to evaluate the clinical utility of our lactate-associated gene prognostic model. DCA plots (Figure 3A–C) showed that the model provides significant net clinical benefits across a wide range of threshold probabilities. This indicates that the model offers valuable predictive information for clinicians, assisting in decision-making for overall survival (OS) predictions. The net benefit curve suggests that our model outperforms traditional strategies, including clinical variables alone, by identifying patients who would benefit most from early intervention based on risk stratification. The model's predictive accuracy was further assessed using Receiver Operating Characteristic (ROC) curves for 1-, 2-, and 3-year overall survival (OS) predictions. As shown in Figure 3D–F, the Area Under the Curve (AUC) values for both the TCGA cohort and external validation cohort were consistently above 0.65, with some exceeding 0.7 and even reaching 0.8 for certain time points. These findings highlight the strong performance of our model in accurately predicting OS, further supporting its potential utility in clinical practice for long-term survival predictions in HCC patients. To further validate the model's ability to stratify patient risk, we performed Principal Component Analysis (PCA) on the full TCGA dataset, focusing on lactate-associated genes and those identified in the risk model. PCA results, shown in Figure 3G–I, revealed clear separation between high-risk and low-risk groups based on gene expression profiles. The PCA plots show that the risk genes effectively distinguish patients with better prognosis (low-risk) from those with worse prognosis (high-risk), supporting their utility in risk stratification. Our analysis confirms that the lactate-associated gene prognostic model is a robust and clinically relevant tool for predicting overall survival in HCC patients.

2.4 Enrichment Analysis of High- and Low-Risk Groups in HCC

The circular plot illustrates gene ontology (GO) categories, highlighting key biological processes, such as cell cycle regulation and immune responses, which are more active in the high-risk group. This visualizes significant gene expression differences related to patient prognosis (Figure 4A). A dot plot details the top enriched Biological Processes, highlighting critical activities such as nuclear division and chromosome segregation, suggesting increased cellular proliferation in high-risk patients (Figure 4B). The bubble plot illustrates KEGG pathway analysis, identifying critical pathways, such as the cell cycle and cytokine–cytokine receptor interaction, associated with high-risk groups (Figure 4C). GSEA plots compare high-risk and low-risk groups, with high-risk patients showing enrichment in pathways related to cell adhesion and metastasis, while low-risk patients exhibit activity in metabolic pathways, potentially correlating with better outcomes (Figure 4D,E).

2.5 Immune Infiltration and Checkpoint Analysis

The tumor microenvironment (TME) plays a crucial role in tumor progression and response to immunotherapy. To explore the relationship between immune cell infiltration and lactate-associated gene expression, we analyzed the distribution of immune cells and immune checkpoint expression in high-risk and low-risk groups based on the lactate-associated gene model. Our analysis revealed that the low-risk group exhibited a significantly higher proportion of stromal cells, suggesting a more active stromal component in these patients (Figure 5A). This finding suggests that stromal cells in the tumor microenvironment may contribute to more favorable immune conditions in low-risk patients. We also observed a strong positive correlation between G6PD expression and immune scores (Figure 5B). This suggests that higher G6PD expression may shape the immune landscape, potentially promoting immune cell infiltration and influencing the immune response in HCC. This positive correlation with immune scores highlights the potential role of G6PD in modulating the TME, especially in immune activation. To further assess immune cell infiltration, we used single-sample gene set enrichment analysis (ssGSEA) to evaluate the levels of 22 immune cell types in the high- and low-risk groups. Our results indicated increased macrophage (M0) infiltration in high-risk groups, whereas CD4 memory resting cells were more prevalent in low-risk groups (Figure 5C–E). This suggests that high-risk tumors may exhibit a more immunosuppressive TME, with macrophages, often involved in promoting immune evasion, being more abundant in these patients. Conversely, the increased presence of CD4 memory resting cells in the low-risk group may reflect a more balanced immune environment, potentially enhancing immune surveillance and tumor control. We further analyzed the expression of key immune checkpoints, including CTLA-4, CD274 (PD-L1), PDCD1 (PD-1), LAG3, and HAVCR2. These checkpoints play a crucial role in immune evasion and tumor progression. Our findings revealed differential expression of these checkpoints between the high- and low-risk groups (Figure 5F,G). Notably, G6PD and IK expression levels were strongly correlated with immune checkpoint expression, especially in the high-risk group. This suggests that G6PD and IK may regulate immune evasion by modulating immune checkpoint pathways, further supporting their potential role in promoting immune suppression and escape in HCC.

2.6 Mutation Analysis and Drug Sensitivity in HCC Risk Groups

Somatic mutation analysis within the TCGA dataset revealed distinct mutational profiles between high-risk and low-risk groups (Figure 6A,B). High-risk patients exhibited a higher frequency of mutations in key oncogenes and tumor suppressor genes, such as TP53 and CTNNB1. Notably, high-risk patients showed a higher tumor mutational burden (TMB), suggesting a potential correlation with poor prognosis. Tumor Mutational Burden analysis (Figure 6C) quantified TMB, revealing that high-risk patients consistently exhibited higher TMB values compared to low-risk groups. This difference was statistically significant and supports the association between higher TMB and increased risk scores. Survival analysis showed that patients with higher TMB, particularly those in the high-risk group, had poorer overall survival rates (Figure 6D,E). These findings suggest that TMB could serve as an independent prognostic factor in risk stratification. Drug sensitivity testing using data from the GDSC database revealed variable responses to chemotherapy agents between the groups. High-risk patients generally exhibited reduced sensitivity to drugs like sorafenib and doxorubicin, as reflected by higher IC50 values. Conversely, some agents, such as cisplatin and paclitaxel, exhibited relatively lower IC50 values in high-risk groups, suggesting a nuanced response pattern that may inform therapeutic decisions (Figure 6F–O).

2.7 Gene Coexpression Network Analysis

Using weighted gene co-expression network analysis (WGCNA), we identified gene modules linked to tumor characteristics in HCC. The optimal soft-thresholding power for constructing a scale-free network was β = 13, with a scale-free fit index of 0.90, indicating a robust network topology (Figure 7A). Hierarchical clustering of genes revealed six distinct gene modules, each represented by a unique color (Figure 7B). The “Red” and “Green” modules showed significant correlations with clinical traits. The “Red” module had a strong positive correlation with tumor grade, indicating its association with more aggressive tumor features. In contrast, the “Green” module was most strongly correlated with tumor stage, suggesting its role in HCC progression (Figure 7C). Further analysis with Venn diagrams identified two genes, G6PD and IK, overlapping between our prognostic model's lactate-associated genes and the “Red” and “Green” modules, highlighting their relevance in HCC pathology (Figure 7D). Based on stepwise Cox regression analysis and LASSO regression, which identified G6PD and IK as the most robust independent prognostic markers, while CALML5 exhibited weaker prognostic significance (p > 0.05 in multivariate analysis). To enhance the model's predictive accuracy and clinical applicability, we refined it to focus on the two strongest predictors. Next, we performed immunohistochemical (IHC) analysis on the HCC tissue microarray (TMA), and H-score quantification was used to assess protein expression levels (Figure 7E,F). The analysis revealed that G6PD was significantly overexpressed in HCC tissues compared to adjacent normal tissues, suggesting its potential role in HCC progression. However, there was no significant difference in IK, which may be related to the quality of the antibody and the small number of samples. Furthermore, IHC staining data from The Human Protein Atlas database (https://www.proteinatlas.org) corroborated these findings, showing elevated expression of G6PD and IK in HCC tissues relative to normal liver tissues (Figure S1). This consistency across datasets strengthens the evidence supporting G6PD and IK as potential biomarkers for tumor progression and promising therapeutic targets.

2.8 Single-Cell Transcriptomic Analysis

We analyzed single-cell RNA sequencing (scRNA-seq) data to explore the cellular heterogeneity and expression profiles of lactate-associated genes in HCC. Using t-SNE dimensionality reduction, cells were grouped into distinct clusters representing major cell types, including T cells, hepatic cells, endothelial cells, myeloid cells, and mesenchymal cells (Figure 8A,B). Analysis of lactate-related gene (LRG) scores, based on key lactate-associated gene expression, showed significantly higher scores in tumor cells compared to normal cells (Figure 8C). Stratification by cell type revealed that mesenchymal and myeloid cells had the highest LRG scores, indicating their critical role in lactate metabolism within the tumor microenvironment (Figure 8D). Further investigation of specific lactate-associated genes revealed distinct expression patterns across cell types. G6PD, a key gene in lactate metabolism, was predominantly expressed in hepatic and mesenchymal cells (Figure 8E). Similarly, IK expression was enriched in myeloid and mesenchymal cells, highlighting its potential role in the metabolic reprogramming of these cell populations (Figure 8F). These findings highlight the significance of lactate-associated genes in driving metabolic alterations and tumor progression at the single-cell level. Given G6PD's prominent role in lactate metabolism and its expression in key cell types linked to tumor progression, we selected G6PD for further functional validation.

2.9 Validation of G6PD Expression in HCC

To explore the functional impact of G6PD on immune responses in HCC, we performed functional assays. The TCGA LIHC dataset revealed that G6PD is significantly overexpressed in tumor tissues compared to normal tissues (Figure S2A). Kaplan–Meier survival analysis indicated that high G6PD expression is associated with poor prognosis in HCC patients (HR = 2.15, log rank p = 8.6e-05; Figure S2B). Stratification by cancer stage revealed a progressive increase in G6PD expression with advancing stages (Figure S2C), and nodal metastasis (N1) correlated with higher G6PD levels compared to nonmetastatic cases (Figure S2D). Age-specific analysis further revealed that older patients had significantly higher G6PD expression than younger patients (Figure S2E). These findings underscore the association of G6PD with tumor progression and adverse clinical outcomes in HCC. Further analysis in liver cancer cell lines (HepG2, Huh7, and MHCC97H) revealed significantly higher G6PD mRNA and protein expression in cancer cells compared to the normal liver cell line L02 by western blot assays (Figure S2F–G), suggesting that G6PD overexpression contributes to the metabolic alterations characteristic of HCC cells.

2.10 G6PD and Immune Modulation in HCC

Our analysis of G6PD in HCC indicates its significant role in regulating immune responses, particularly in CD8⁺ T cell activation. Immune infiltration analysis (Figure S2) showed that G6PD expression positively correlated with CD8⁺ T cell infiltration in the tumor microenvironment (Figure S2A–D). G6PD expression was also associated with immune cell types involved in immune suppression, such as macrophages. This suggests that G6PD may modulate both immune activation and suppression within the tumor. Functional assays showed that G6PD knockdown in HepG2 cells enhanced CD8⁺ T cell proliferation and IFN-γ secretion, indicating increased T cell activation (Figure S3E). In contrast, G6PD overexpression suppressed T cell activation, supporting the hypothesis that G6PD may contribute to immune evasion. These results suggest that G6PD modulates the immune response by regulating the tumor's ability to influence CD8⁺ T cell activity. The effects of G6PD on immune cell activation and suppression markers, such as CD274 (PD-L1) in Figure 5, highlight its potential role in immune checkpoint regulation.

2.11 G6PD Regulates PD-L1 Expression via the mTOR Pathway

We further investigated the role of G6PD in regulating PD-L1 expression in HCC cells. Numerous studies have demonstrated that mTOR is a key upstream regulator of PD-L1 expression [17-19], playing a pivotal role in immune evasion by suppressing T cell activity [20]. G6PD exhibited the most significant prognostic value in our survival analysis and Cox regression model, indicating its crucial role in hepatocellular carcinoma (HCC) progression. As a key enzyme in the pentose phosphate pathway, G6PD is strongly implicated in tumor metabolism, oxidative stress regulation, and lactate metabolism. Based on evidence that G6PD regulates immune activation and suppression, we hypothesized that it modulates PD-L1 expression via the mTOR signaling pathway. To test this hypothesis, we performed Western blot analyses to assess how G6PD knockdown and overexpression influence PD-L1 and mTOR signaling in HepG2 cells. Results depicted in Figure 9A show that G6PD knockdown significantly reduced phosphorylated mTOR (p-mTOR) and PD-L1 levels, suggesting that G6PD modulates PD-L1 expression via mTOR signaling. Conversely, G6PD overexpression in HepG2 cells increased p-mTOR and PD-L1 levels, further supporting that G6PD promotes immune evasion by upregulating PD-L1 through the mTOR pathway (Figure 9B). Rescue experiments, involving G6PD reintroduction in knockdown cells, restored p-mTOR and PD-L1 levels, confirming that G6PD specifically mediates these effects (Figure 9C). Furthermore, treatment with an mTOR inhibitor (Figure 9D) in G6PD-overexpressing cells significantly reduced p-mTOR and PD-L1 levels, reinforcing that G6PD exerts immune-modulatory effects via mTOR signaling. These findings indicate that G6PD impacts CD8⁺ T cell activation, immune cell infiltration, and PD-L1 expression via the mTOR pathway, thereby facilitating immune evasion in HCC. By regulating the immune checkpoint PD-L1, G6PD likely suppresses T cell activity, thus facilitating tumor growth and progression.

2.12 Molecular Docking Analysis of G6PD With Candidate Drugs

To investigate potential interactions between G6PD and candidate therapeutic compounds, molecular docking was performed using four drugs with the highest predicted sensitivity scores: Staurosporine_1034, Vinorelbine_2048, Vinblastine_1004, and Docetaxel_1007. The docking results revealed strong binding affinities between G6PD and the selected drugs, with binding energies below −7 kcal/mol, suggesting stable interactions. Specifically, Staurosporine_1034 exhibited the highest binding affinity (−9.9 kcal/mol), forming one hydrogen bond within the active site. Vinorelbine_2048 showed a binding affinity of −7.6 kcal/mol, forming two hydrogen bonds that stabilized its interaction with G6PD. Vinblastine_1004 demonstrated a binding affinity of −8.8 kcal/mol, forming a single hydrogen bond. Docetaxel_1007 showed a binding affinity of −8.9 kcal/mol, forming three hydrogen bonds, indicating specific and robust binding. Visualization of the docking results revealed the structural alignment of each drug within the G6PD binding pocket, with key residues interacting through hydrogen bonds and hydrophobic interactions (Figure 10A–D) [21-24]. These findings suggest that the selected drugs may modulate G6PD activity by directly targeting its active site, potentially affecting lactate metabolism in HCC.

4.1 Data Collection and Preparation

Hepatocellular carcinoma (HCC) data, including RNA-sequencing and clinical information, were downloaded from The Cancer Genome Atlas (TCGA) portal (https://portal.gdc.cancer.gov/). A total of 374 HCC samples along with clinical data for 371 patients were included. Additionally, 332 lactate-associated genes were compiled from relevant literature sources, specifically from PMID: 38419085 [16]. An external validation dataset, LIHC-JP, was also used.

4.2 Prognostic Gene Selection and Model Construction

Initially, univariate Cox regression analysis was performed to assess the prognostic value of each lactate-associated gene. Genes with a p < 0.05 were considered significant and were further analyzed using Lasso regression to prevent overfitting and to refine the selection of variables. Multivariate Cox regression analysis was subsequently conducted to finalize the selection of variables, resulting in three lactate-associated prognostic genes. These genes were used to construct a prognostic model where each patient's risk score was calculated using the formula:

Risk score = ∑(Expression level of gene × coefficient).

The TCGA dataset was divided into training and validation sets, with LIHC-JP serving as an external validation set. Kaplan–Meier curves were used to analyze survival differences between high- and low-risk groups in these datasets.

4.3 Enrichment Analysis

Differentially expressed genes (DEGs) between risk groups were identified based on |log2FC| > 1.0 and a p < 0.05. Gene ontology (GO) analysis for biological processes, cellular components, and molecular functions, along with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, was conducted to explore the biological significance of these genes. Gene set enrichment analysis (GSEA) was also performed to further understand pathway differences between high- and low-risk groups.

4.4 Immune Analysis

The immune and stromal scores were estimated using the ESTIMATE algorithm to assess the impact of the tumor microenvironment on prognosis. Differences in immune cell infiltration between risk groups were analyzed using CIBERSORT. Correlations between gene expression and immune cell levels were assessed using Spearman's rank correlation.

4.5 Mutation Analysis

Somatic mutation data from the TCGA dataset were analyzed using the “maftools” R package. Tumor Mutational Burden (TMB) was calculated, and HCC patients were classified into high and low TMB groups based on the median TMB value. Survival differences between these groups were examined.

4.6 Drug Sensitivity Analysis

Drug sensitivity was assessed using the “pRRophetic” R package, which predicts IC50 values for chemotherapeutic agents based on gene expression profiles. Drugs showing significant differences in IC50 values between high- and low-risk groups were identified, and their potential efficacy was evaluated.

4.7 Single-Cell RNA Sequencing Analysis

Single-cell RNA-seq data from GEO datasets GSE245906 and GSE149614 were processed using the “Seurat” R package. Data quality control, normalization, and batch effect removal were performed, followed by clustering and differential gene expression analysis. The “SingleR” package was used for cell type annotation, and “AUCell” was utilized to quantify lactate-related gene expression across cell types.

4.8 Validation of Gene Expression

The expression levels of key genes, G6PD and IK, were validated using immunohistochemistry data from the Human Protein Atlas (https://www.proteinatlas.org/). This helped confirm their overexpression in HCC tissues compared to normal liver tissue.

4.9 Cell Culture and Lentivirus-Infection

Human normal hepatocytes L02, human hepatocellular carcinoma cells Huh7, HepG2, and MHCC97H were cultured in complete Dulbecco's modified Eagle's medium (DMEM) with 10% fetal bovine serum (FBS) (HyClone Laboratories, Logan, UT, USA) at 37°C in a 5% CO₂ atmosphere.

To generate stable knockdown or overexpressing cell lines for G6PD, G6PD shRNA plasmids (LV-Si-G6PD), G6PD overexpressing plasmids (LV-G6PD), and empty controls (LV-Sc-G6PD/LV-empty) were purchased from Santa Cruz Biotechnology. After coinfection of shRNA or overexpressing plasmids with the packaging plasmids psPAX2 (Addgene, Cat# 12260) and pMD2.G (Addgene, Cat# 12259) into HEK-293 T cells, the lentivirus was harvested, titrated, and stored for infecting HepG2 cells. These two cell lines (G6PD-KD-HepG2 and G6PD-OE-HepG2) infected with lentivirus were screened with puromycin, and the stable cell lines were further confirmed with detection of WB.

Cells were grown to 50% confluence in six-well cell culture plates (Nest, China) and transfection was performed using Lipofectamine 2000 (Invitrogen, Cat# 11668030), according to the manufacturer's instructions. In all the co-transfection experiments, the corresponding empty vectors were used as negative controls to ensure similar DNA concentrations.

4.10 Immunohistochemistry (IHC)

The hepatocellular carcinoma (HCC) tissue microarray (TMA) utilized in this study was sourced from Shanghai Outdo Biotechnology (Project No. OD-CT-DgLiv03; Ethics Committee Review Opinion No: YB M-05-02), comprising 30 paired HCC samples and their corresponding adjacent normal tissues. Immunohistochemical (IHC) staining was conducted on liver tissue sections following standard protocols. Tissue sections were initially heated at 70°C for 30 min in an oven, followed by deparaffinization using xylene and ethanol for 10 min. To quench endogenous peroxidase activity, the sections were incubated in 0.3% hydrogen peroxide (H₂O₂) for 20 min. Antigen retrieval was performed with citrate buffer at 95°C for 10 min. The slides were then incubated with primary antibodies against G6PD (1:100 dilution, Cat#AMRe02038, Enkilife) and IK (1:100 dilution, Cat#31833-1-AP, Proteintech). Signal detection was achieved using DAKO EnVision horseradish peroxidase (HRP)-conjugated anti-rabbit secondary antibody.

4.11 Western Blot Analysis

Western blotting was performed to validate the protein expression levels of G6PD and related signaling molecules such as mTOR and PD-L1. Cells were lysed using RIPA buffer (Cat#P0013D, Beyotime) containing protease and phosphatase inhibitors (Cat#P1045, Beyotime). Protein concentrations were measured using the BCA Protein Assay Kit (Cat#P0009, Beyotime). Equal amounts of protein were separated on SDS-PAGE gels and transferred to PVDF membranes. The membranes were blocked with 5% nonfat milk and incubated with primary antibodies against G6PD (Cat#sc-373,886, Santa Cruz Biotechnology, 1:2000 dilution), mTOR (Cat#28273-1-AP, Proteintech, 1:1000 dilution), phospho-mTOR (Cat#67778-1-Ig, Proteintech, 1:1000 dilution), PD-L1 (Cat#83600-2-PBS, Proteintech, 1:1000 dilution), and GAPDH(Cat#60004-1-Ig, Proteintech, 1:10000 dilution, used as a loading control). Appropriate HRP-conjugated secondary antibodies (Cat#7074/7076, Cell Signaling Technology, 1:10000 dilution) were used, and protein bands were visualized using enhanced chemiluminescence (ECL) (Cat#A38556, ThermoFisher). Quantification of band intensity was performed using ImageJ software, and relative expression levels were calculated by normalizing to GAPDH.

4.12 Flow Cytometry Analysis

Flow cytometry was employed to assess the functional effects of G6PD modulation on T cell proliferation and IFN-γ secretion. Human CD8⁺ T cells were labeled with CFSE (carboxyfluorescein succinimidyl ester, Cat#C34554, Invitrogen) and stimulated with anti-CD3 (Cat#341090, Biosciences) and anti-CD28 (Cat#348047, Biosciences) antibodies. These T cells were cocultured with HepG2 cells, either with knocked down (KD) or overexpressed [27] G6PD, for 3 days. Postincubation, cells were harvested and stained for CD8 (Cat#335805, Biosciences) and IFN-γ (Cat#562213, Biosciences). Flow cytometric analysis was conducted to determine T cell proliferation (by CFSE dilution) and IFN-γ production. The proliferation index and the percentage of IFN-γ+ cells were calculated using FlowJo software.

4.13 Structural Analysis of G6PD and Therapeutic Agents

From the “oncoppredict” results, we selected the four drugs with the highest predicted sensitivity for molecular docking analysis. The molecular structures were obtained from PubChem (https://pubchem.ncbi.nlm.nih.gov/), and protein structures were retrieved from the Protein Data Bank (PDB; http://www.rcsb.org/). Protein and ligand files were converted to PDBQT (Protein Data Bank, Partial Charge (Q), and Atom Type (T)) format, water molecules were removed, and polar hydrogens were added to improve docking accuracy. Molecular docking simulations were conducted using AutoDock Vina 1.2.2 (http://autodock.scripps.edu/) to evaluate drug–protein binding interactions. The therapeutic potential of each drug was assessed based on its docking score, which reflects the strength and affinity of the binding [27]. The G6PD and Staurosporine complex structure was visualized by PyMol.

4.14 Statistical Analysis

Statistical analyses were performed using R software (version 4.1.2). The Wilcoxon test was used for differential analysis, Spearman's rank correlation for assessing correlations, and the Kaplan–Meier method for survival analysis. The Benjamini-Hochberg method was applied for multiple-testing correction. Significant differences were considered at p < 0.05.