Abstract Mitochondrial oxidative stress plays a critical role in cancer development and progression. However, there is limited research on the relationship between mitochondrial oxidative stress and liver hepatocellular carcinoma (LIHC). Mitochondrial oxidative stress-related genes were collected from Genecards Portal. Prognosis-linked genes (PLGs) were identified by univariate Cox regression analysis. A risk model was constructed based on the PLGs using least absolute shrinkage and selection operator (LASSO) analysis. Receiver operating characteristic (ROC) curves were used to determine the predictive ability of the model. The expression levels of the prognostic genes were verified in the cell lines. Cell proliferation, apoptosis, and invasion assays were conducted to investigate the functional role of the target gene. We constructed a novel risk model based on 9 prognostic genes (CYP2C19, CASQ2, LPL, TXNRD1, CACNA1S, SLC6A3, OXTR, BIRC5, and MMP1). Survival analysis showed that patients with a low-risk score had a much better overall survival (OS). Prognostic risk score was found to be an independent predictor of prognosis. Patients in the high-risk group had a less favorable tumor microenvironment characterized by a lower degree of immune cell infiltration. Among the nine prognostic genes, MMP1, identified as the most promising candidate, demonstrated the capacity to enhance tumor cell proliferation and invasion. Our investigation reveals the oncogenic role of mitochondrial oxidative stress in LIHC. For the first time, we established a risk prediction model for mitochondrial oxidative stress in patients with LIHC. MMP1 has the potential to function as a promising biomarker in LIHC. Supplementary Information The online version contains supplementary material available at 10.1038/s41598-025-10076-0. Keywords: Liver hepatocellular carcinoma, Mitochondrial oxidative stress, Risk model, Prognosis, Consensus clustering Subject terms: Tumour biomarkers, Cancer genetics Background Liver hepatocellular carcinoma (LIHC), commonly referred to as liver cancer, is a major global health concern and the most prevalent primary liver malignancy^[32]1. The prognosis of patients with HCC is generally poor, with a median survival of approximately 6–20 months depending on the stage of cancer^[33]2. The 5-year survival rate for LIHC is less than 20%, largely due to its highly aggressive nature and high recurrence rate even after treatment. Despite advances in treatment options, LIHC remains a significant challenge in the field of oncology, as it is often diagnosed at an advanced stage and is difficult to completely remove by surgery, leading to treatment failure and poor patient outcomes^[34]3. Oxidative stress refers to a series of physiological and pathological processes by which cells produce excessive reactive oxygen species during metabolic processes owing to various stresses and stimuli from internal and external environments^[35]4. Mitochondrial oxidative stress refers to a series of biological effects caused by an imbalance in the production of excessive amounts of reactive oxygen species^[36]5. This process can lead to various physiological and pathological processes. When mitochondria are exposed to various stress or stimuli such as hypoxia, toxins, heat shock, and inflammation, the redox balance within the mitochondria can be disrupted, leading to the production of an excessive amount of reactive oxygen species^[37]6,[38]7. These free radicals can cause oxidative stress such as lipid peroxidation, protein oxidation, and DNA damage^[39]8. They can also affect the structure and function of mitochondria, by causing a decrease in mitochondrial membrane potential and damage to the mitochondrial respiratory chain^[40]9. These effects not only affect the function of mitochondria but also affect the metabolism and function of the entire cell and can even lead to cell death or apoptosis. Mitochondrial oxidative stress plays an important role in various diseases such as cancer^[41]10cardiovascular^[42]11and diabetes^[43]12. It has been discovered that mitochondrial oxidative stress exists not only in LIHC^[44]13 but also in various other cancers such as breast cancer^[45]14lung cancer^[46]15and colorectal cancer^[47]16. In liver cancer cells, the level of mitochondrial oxidative stress is high, and the reactive oxygen species produced can cause cellular apoptosis, gene mutations, and other reactions, thereby promoting proliferation, metastasis, and drug resistance in liver cancer cells^[48]17. Fructokinase A (which is specifically expressed in liver cancer cells) responds to oxidative stress signals and acts as a protein kinase, activating the Nrf2 signaling pathway to promote the survival of liver cancer cells under oxidative stress^[49]18. Mitochondrial DNA is a genetic material within the mitochondria and its damage can trigger oxidative stress reactions, affecting the function of the mitochondrial membrane and leading to abnormalities in the mitochondrial respiratory chain. Recent studies have shown that mitochondrial DNA damage is closely associated with the development and occurrence of liver cancer. DNA damage may be related to impaired mitochondrial respiratory chain function, which in turn promotes the proliferation and invasion of liver cancer cells^[50]13. Studies have, furthermore, suggested that the dynamic balance of oxidative stress not only coordinates complex cellular signaling events in cancer cells but may also affect other components of the tumor microenvironment (TME) such as M2 macrophages, dendritic cells, and immunosuppressive cells such as Treg cells^[51]19. However, one study suggested that oxidative stress may exert an inhibitory effect on the growth and proliferation of liver cancer cells^[52]20. Hence, it is imperative to establish a risk prediction model based on mitochondrial oxidative stress that comprehensively reflects the impact of mitochondrial oxidative stress-related genes on LIHC prognosis. In the present study, we constructed a risk model based on mitochondrial oxidative stress-related genes in LIHC. Second, we thoroughly investigated immune infiltration and potential signaling pathways in patients with LIHC. Our analysis revealed that, among the modeled mitochondrial oxidative stress-related genes, MMP1 was significantly upregulated in patients with LIHC and may represent a promising therapeutic target. The findings of this study provide new insights and perspectives for personalized and precision treatment of LIHC. Materials and methods Data collection and processing The transcriptome and clinical data of LIHC patients were downloaded from TCGA ([53]https://portal.gdc.cancer.gov/; accessed on January 4, 2023). This study was based entirely on publicly available datasets and did not involve any direct human or animal experimentation. Therefore, ethical approval and informed consent were not required. After data sorting, 373 LIHC and 49 normal control samples were included in the subsequent analyses. The mitochondrial oxidative stress-related genes were downloaded from the GeneCards database ([54]https://www.genecards.org/; accessed on January 4, 2023). ID conversion and data integration were performed using Strawberry Perl software (version 5.30). Identification of mitochondrial oxidative Stress-Related genes First, the differentially expressed genes (DEGs) were identified depending on the difference between tumor and normal samples using R software’s “limma” package. The threshold was set at | log2FC | > 2.0 and FDR < 0.05. Further analysis of the DEGs was performed using univariate Cox regression analysis of the prognosis-linked genes (PLGs). The correlational analyses of PLGs were performed using “igraph,” “psych,” “reshape2,” and “RColorBrewer” packages. Copy number variation analysis The copy number variation data of patients with LIHC were downloaded from the UCSC website ([55]https://xena.ucsc.edu/). Data were collected and graphed using Perl and R software. Visualization of the chromosomal locations of the PLGs was performed using Perl. Consensus clustering Based on the differential expression of PLGs, consensus clustering was implemented in LIHC patients via k-means methods using the “ConsensusClusterPlus” package. An optional classification method was identified after numerous attempts. Patients with LIHC were grouped into either clusters A or B for further analysis. For validating the consistency between patient clusters the principal component analysis (PCA), T-distributed stochastic neighbor embedding (tSNE), and uniform manifold approximation and projection (UAMP) were performed using “ggplot2,” “Rtsne,” and “umap” packages. Survival analyses was performed within the two clusters via “survival” and “survminer” packages. Functional enrichment analysis The “c2. cp.kegg. symbols.gmt” and “c2. go. symbols.gmt” were downloaded from the MSigDB database to perform GSVA and GSEA for the two clusters. The GSVA analysis was carried out with “limma,” “GSEABase,” and “GSVA” packages. The GSEA analysis was performed using “limma,” “org.Hs.eg.db,” and “clusterProfiler” packages. Construction and validation of the risk score model PLGs were used to explore the prognostic value of LIHC. A total of 488 patients with LIHC were randomly divided into two sets (training and test) at a 1:1 ratio to construct and validate the risk score model. The least absolute shrinkage and selection operator (LASSO) algorithm was used to explore the PLGs which were correlated with the overall survival (OS) rates using the “glmnet” package. PLGs (which are independent risk factors for prognosis) were identified by multivariate Cox regression analysis and used to construct the model. The training and test sets were divided into high- and low-risk groups based on their risk scores. The OS rates of patients with LIHC in the two groups were calculated from the Kaplan-Meier survival curve using the “Kaplan-Meier survival” package. The predictive ability of the model was detected using the receiver operating characteristic (ROC) curve at 1, 3, and 5 years. Association between tumor risk score and immune environment landscape The CIBERSORT and ssGSEA R scripts were utilized to evaluate the proportion of immune cells infiltrating a particular tissue^[56]21. These tools allow for the identification and quantification of different immune cell types based on gene expression data. To investigate the differences in immune cell proportions between the low- and high-risk groups, we used the CIBERSORT software. The estimated score for each immune cell type in each sample was 1. Additionally, we conducted Spearman’s rank correlation analysis to examine the association between risk score values and immune cell infiltration. Gene expression validation in cell lines The expression profiles of nine hub genes were examined using real-time quantitative PCR (RT-qPCR) in four HCC cell lines and one normal human liver cell line. The five cell lines were obtained from the General Surgery Laboratory of Tianjin Medical University General Hospital. LO2, SNU398, and Li7 cell lines were cultured in complete RPMI1640 medium. HepG2 and LM3 cells were cultured in Dulbecco’s modified DMEM medium (DMEM). Total RNA was extracted using the RNA isolater Total RNA Extraction Reagent (Vazyme) and reverse-transcribed to cDNA. Nine hub genes were detected using SYBR Green Master Mix (Vazyme). The primer sequences used are listed in Table [57]S1. The relative gene expression levels were calculated using fold change (2^−ΔΔCt). Western blot Cells were harvested and thoroughly rinsed twice, followed by incubation on ice for 30 min in RIPA buffer. Protein concentration was quantified using the BCA assay (Thermo Fisher). The protein samples were combined with a loading buffer containing β-mercaptoethanol. 10 ug of protein samples were loaded onto a 10% SDS-polyacrylamide gel for electrophoresis (SDS-PAGE) and subsequently transferred onto PVDF membranes at 110 V for 120 min. Following a 1-hour blocking step with 5% milk, the membranes were incubated overnight with primary antibodies. Subsequently, the membranes were incubated with secondary antibodies for 1 h at 4 °C. Images were acquired using the ECL detection system. Cell proliferation experiment Cell proliferation was assessed using EdU assays (Ribobio). Briefly, cells were cultured to the logarithmic phase, harvested, and washed for subsequent use. Following fixation with 4% paraformaldehyde, samples were incubated with glycine for 5 min and Triton X-100 for 10 min. Subsequently, Apollo staining solution was employed for labeling, while Hoechst 33,342 was utilized for nuclear staining. Images were acquired using a fluorescence microscope. Cell invasion assays For invasion assays, 1 × 10^5 cells were cultured in the upper chamber coated with Matrigel and incubated in medium containing 1% FBS. Medium containing 10% FBS was added to the lower chamber. After a 24-hour incubation, the cells were fixed with 4% paraformaldehyde and subsequently stained with 0.1% crystal violet. Three randomly selected regions were counted, and the average was utilized for statistical analysis. Cell apoptosis analysis The apoptosis of tumor cells was detected using Annexin V/7AAD staining (Yeasen). Cells were digested with Trypsin without EDTA. The Annexin V and 7AAD were added into suspension after adjusting concentration of cells. Following a 15-minute incubation, the samples were analyzed by flow cytometry (FCM) within 1 h. Statistical analysis In our study, all statistical analyses were performed using R version 4.1.3. The differential function between the two groups was analyzed using the Wilcoxon rank-sum test, with a P-value < 0.05 considered to indicate a statistically significant difference. Kaplan-Meier survival analysis and the log-rank test were used to compare the survival of different subgroups of patients with LIHC in each dataset. The Kaplan-–Meier method was used to estimate the survival probability over time for each subgroup, and the log-rank test was used to compare the survival curves between the different subgroups. Multivariate Cox regression was used to select prognostic variables based on risk score and clinical characteristics. All analyses were considered statistically significant at a P-value < 0.05. Results Genetic variations of mitochondrial oxidative Stress-Related genes in LIHC The GeneCards database identified a total of 1294 mitochondrial oxidative stress-related genes. We found 89 DEGs in the tumor tissues of the TCGA-LIHC cohort compared to normal tissues (Fig. [58]1A, B). We then performed Kaplan-Meier survival and univariate Cox analysis for DEGs using consolidated data from TCGA-LIHC. Finally, 37 PLGs with potential prognostic relevance were identified (Fig. [59]1C). Herein, CYP2C19, CYP1A2, CYP2C8, CYP2B6, and FOS were the genes that were highly expressed in the LIHC. Due to the uncertainty of chromosomal mutations in the development of LIHC^[60]22we further explored possible mutations (Fig. [61]1D) and their locations on the chromosomes (Fig. [62]S2A). MSTO1, NDRG1, PYCR1, TK1, and BIRC5 showed widespread CNV amplification, whereas CDKN2A, BRCA2, ARG2, SLC2A1, and DNMT1 showed CNV deletions. Except for CASQ2, all other genes were associated with poor prognosis. The landscape of the complex relationship between mitochondrial oxidative stress-related genes and their prognostic value is presented in a network plot in Fig. [63]1E. Fig. 1. [64]Fig. 1 [65]Open in a new tab Genetics characteristics and expression of mitochondrial oxidative stress-related genes in LIHC. (A) Differential genes involved in mitochondrial oxidative stress in LIHC. Red color denotes upregulation of gene expression in tumor tissues, while blue color indicates downregulation. (B) Volcano plot of differentially expressed genes. Red dots represent upregulated genes, while green dots represent downregulated genes. Differential genes were selected based on log FC > 2 and P-value < 0.05. (C) The effect of 37 prognosis-related ARGs on the OS of patients with LIHC. (D) Copy number variations of 37 ARGs in LIHC. (E) The interaction of 37 ARGs in LIHC. The size of each circle represents the impact of each gene on survival prognosis. Consistent clustering of 37 PLGs in LIHC To better understand the functions of mitochondrial oxidative stress-related genes in LIHC, we performed consensus clustering based on the 37 PLGs. Eight attempts were made, and the cohort could be effectively divided only when k = 2, as shown in Fig. [66]S1A-B. In other methods, confounding factors between the groups could be eliminated (Fig. [67]S2B-F). Similarly, the classification of the samples tended to be stable when k = 2 (Fig. [68]S1C). Based on the expression of the 37 PLGs, we divided LIHC patients into two groups: cluster A and cluster B. The results of survival analysis showed a significant difference between clusters A and B (Fig. [69]S1D). Patients with LIHC who were classified in cluster A might have a worse prognosis. UMPA, and tSNE methods were performed to further test the typing accuracy (Fig. [70]S2G-H). As shown in Fig. [71]S1E, the samples were divided into two groups. The expression levels of the PLGs in each cluster are shown in Fig. [72]S1F. The heatmap shows the detailed relationship between mitochondrial oxidative stress-related genes and clinical traits in the two clusters (Fig. [73]2A). In addition, the level of immune cell infiltration differed significantly between the two groups (Fig. [74]2B). With the exception of eosinophils, compared to cluster B, cluster A had a higher degree of tumor-infiltrating immune cells, including: activated CD4 + T cells, activated dendritic cells, and MDSC, suggesting a higher degree of immunoreactivity. Subsequently, we performed a KEGG pathway enrichment analysis between the two clusters (Fig. [75]2C-E). Cluster A was mainly enriched in the cell cycle pathway, while cluster B was mainly associated with the coagulation and complement pathways, drug metabolism pathways, and some other common metabolism-related pathways. Fig. 2. [76]Fig. 2 [77]Open in a new tab Correlation of mitochondrial oxidative stress pattern with clinical traits and tumor-infiltrating immune cells. (A) Correlation between two clusters and the sex, age, and staging of patients with LIHC in TCGA. (B) Comparison of immune cell composition in the TME between two subtypes. (C) KEGG analyses for mitochondrial oxidative stress-related genes of the two clusters. (D, E) Top 5 KEEG-enriched pathways for two clustering methods. Construction and validation of a mitochondrial oxidative Stress-Related prognosis signature To further explore the clinical value of mitochondrial oxidative stress-related genes, we constructed a prognostic model using 37 PLGs. Nine anoikis-related genes (ARGs) associated with survival rates were identified by univariate Cox regression analysis (Fig. [78]3A-B). The prognostic index (PI) was calculated as= (−0.44* expression level of CASQ2) + (0.19* expression level of TXNRD1) + (0.46* expression level of SLC6A3) + (0.14* expression level of CYP2C19) + (0.20* expression level of OXTR) + (0.15* expression level of BIRC5) + (0.26* expression level of MMP1) + (0.27* expression level of LPL) + (1.10* expression level of CACNA1S). As shown in Fig. [79]3C, this prognostic prediction model performed well in predicting relative survival. The survival curves for the training and test sets are shown in Fig. S3A-B. We also validated the 1-, 3-, and 5-year survival rates using this model via ROC curve analysis. The results indicated excellent sensitivity and specificity of this model (Fig. [80]3D). The ROC curves for the training and test sets are shown in Fig. S3C-D. We performed multivariate analysis to validate the independent prognostic value of the risk score, which showed that disease stage and risk scores were prognostic factors (Fig. [81]3E). The prognostic model was further validated using the clustering method described above. The patients in cluster A had a higher risk score and worse prognosis, which was consistent with the above results (Fig. [82]3F-G). Therefore, our prognostic model is highly suitable for evaluating patient outcomes. Fig. 3. [83]Fig. 3 [84]Open in a new tab Risk model construction and validation in LIHC based on mitochondrial oxidative stress-related genes. (A) Coefficient profile plots of nine ARGs. (B) The LASSO method of nine ARGs associated with prognosis. (C) OS analysis of patients with LIHC based on the risk score. (D) The receiver operating characteristics (ROC) based on the risk score. (E) The multivariate analysis of risk score, age, sex, and stage for patient survival. (F) Risk scores for the two clusters. (G) Correlations between risk score, the two clusters, and survival in the Sankey diagram. Analysis of immune activity with the risk scores The inflammatory status of immune cells plays an important role in the natural evolution of tumor development. Mitochondrial oxidative stress is strongly associated with the immune response. To this end, we applied a risk grouping method to patients with LIHC to portray the TME landscape. The proportions of 22 immune cells were analyzed using the CIBERSORT algorithm (Fig. S4A). No significant differences were observed in the overall distribution of immune cells. In contrast, CD8 + T cells, follicular helper T cells, and neutrophils decreased in the high-risk group (Fig. S4B); whereas M0 and M2 macrophages increased. The interplay among immune cells in patients with LIHC can offer insights into the composition of the immune microenvironment for a more comprehensive understanding. Therefore, we analyzed the correlations among immune cells in liver cancer to gain a deeper understanding of their interrelationships and their potential impact on the disease (Fig. S4C). M0 macrophages and CD8 + T cells displayed a strong negative correlation (r = −0.7) in the context of LIHC, suggesting that as the presence of one cell type increased, the other decreased, potentially affecting the overall immune response within the TME. Subsequently, we investigated the associations between the nine gene signatures used to construct a prognostic risk model and immune cells within the TME, to better understand the potential interactions and their implications for patient outcomes in LIHC (Fig. S5A). Excluding CACNA1S, all other genes that played a role in the development of this model demonstrated significant associations with the function and regulation of immune cells. The overall risk score was positively correlated with M2 macrophages, indicating that an increase in the risk score may be associated with a higher number of M2 macrophages in the studied context (Fig. S5B). Expression status and prognostic signatures of nine ARGs The gene expression status of the nine ARGs was verified at the cellular level. Four hepatoma cell lines were analyzed for this study. The LO2 cell line was used as a normal control. Except for CASQ2, the other eight genes were highly expressed in almost carcinoma cell lines, consistent with the prognostic signatures (Fig. [85]4). We examined the relationship between nine ARGs with prognosis of LIHC. CASQ2, LPL, SLC6A3, OXTR, BIRC5, and MMP1 were identified as factors associated with the prognosis of LIHC (Fig. [86]5A). Furthermore, we investigated whether alterations in gene expression were mirrored at the protein level. The results suggested that, with the exception of SLC6A3, the protein expression levels of the remaining genes were consistent with RNA expression (Fig. [87]5B). Fig. 4. [88]Fig. 4 [89]Open in a new tab The RNA expression level of genes in prognostic prediction model detected by RT-qPCR. * P < 0.05; ** P < 0.01; *** P < 0.001. Fig. 5. [90]Fig. 5 [91]Open in a new tab The prognostic features and protein expression of ARGs in LIHC. (A) The relationship between nine ARGs and prognosis in LIHC. (B) Protein expression levels of six ARGs were detected in LIHC. MMP1 promotes the proliferation and migration of LIHC cells To further investigate the function of MMP1 in liver hepatocellular carcinoma (LIHC), we assessed the expression levels of MMP1 in four LIHC cell lines (Fig. [92]6A). The raw data of western blots were provided in Fig S6. The protein expression levels of MMP1 were consistent with those of mRNA. The results of the EdU experiment demonstrated that the addition of MMP1 enhanced the proliferation of tumor cells. (Fig. [93]6B). Du et al., suggested that overexpression of MMP1 could exacerbate the proliferation ability of colorectal cancer cells^[94]23. Additionally, we found that MMP1 enhanced the invasive capacity of LIHC cells (Fig. [95]6C). We also examined the effects of the treatment MMP1 on apoptosis of tumor cells in vitro. As anticipated, we observed that after treatment with MMP1, the proportion of apoptotic cells decreased compared to the control group (Fig. [96]6D). Subsequent analysis revealed a significant correlation between MMP9 expression and immune cell infiltration in LIHC. Notably, MMP9 was positively associated with multiple immune cell populations, including macrophages, T lymphocytes, and dendritic cells (Fig. [97]6E). Finally, we examined the expression of mitochondrial oxidative stress–related markers following MMP9 treatment and found that the expression levels of NRF-1 and TFAM were significantly upregulated (Fig. [98]6F). Fig. 6. [99]Fig. 6 [100]Open in a new tab MMP1 enhances the proliferation and migration functions of LIHC cells. (A) The protein expression level of MMP1 was detected using Western blot analysis. (B) The EdU proliferation assay was conducted following the addition of MMP1 to LM3 cell lines. (C) The invasion experiment showed that the MMP1 protein MMP1 protein enhances the invasive capacity of LM3 cell lines. (D) The apoptosis assay was performed to analyze the potential role of MMP1 on cellular apoptosis (E) The relationship between MMP9 and immune cell infiltration in LIHC. (F) The mRNA expression level of mitochondrial oxidative stress-related genes. Discussion Liver cancer is a malignancy with a high mortality incidence worldwide and is currently the most frequent cause of cancer-related deaths. According to global cancer statistics, liver cancer ranked sixth in terms of incidence and third in terms of mortality among all cancers^[101]24. These statistics highlight the significant burden of liver cancer on the global population. One challenge in treating liver cancer is that patients are often diagnosed at an advanced stage, which contributes to poor prognosis^[102]25. Therefore, early detection and timely treatment are critical to improve the outcomes of patients with LIHC. Mitochondrial oxidative stress refers to the disruption of oxidative reactions within the mitochondria, leading to mitochondrial dysfunction and an unstable cellular environment^[103]5. The relationship between mitochondrial oxidative stress and cancer is complex and multifactorial. High ROS levels can cause DNA damage and genomic instability^[104]26leading to accumulation of mutations that drive cancer development. Additionally, ROS can activate signaling pathways that promote cell growth and survival and inhibit programmed cell death (apoptosis), a mechanism that normally helps prevent the formation of cancer cells^[105]27. In our study, a set of risk score features were identified including the following genes: CYP2C19, CASQ2, LPL, TXNRD1, CACNA1S, SLC6A3, OXTR, BIRC5, and MMP1. Our findings suggest that CASQ2, LPL, SLC6A3, OXTR, BIRC5, and MMP1 are independent risk factors for LIHC. Currently, research on the role of these genes in liver cancer and mitochondrial oxidative stress is limited. In a previous study^[106]28differential gene expression analysis identified OXTR as a key gene influencing mitochondrial dysfunction and oxidative stress. During the identification and characterization of immune-related molecular subtypes in LIHC, OXTR has been identified as a critical regulator. Notably, its potential involvement in mitochondrial dysfunction and oxidative stress has also garnered significant interest^[107]29. Recent evidence suggests that aberrant expression of SLC6A3 may exacerbate mitochondrial oxidative stress by increasing ROS production, thereby contributing to tumor progression and immune dysregulation in the tumor microenvironment^[108]30,[109]31. Overexpression of BIRC5 may enable tumor cells to resist oxidative stress–induced cell death by sustaining mitochondrial function and suppressing excessive ROS accumulation, thereby contributing to therapeutic resistance and immune evasion^[110]32. Moreover, BIRC5 has been shown to play a role in the construction of prognostic models for various types of LIHC^[111]33–[112]35suggesting its potential as a valuable biomarker in the clinical management of cancer. The relationship between CASQ2 and mitochondrial oxidative stress is not fully understood, and further research is needed. However, other studies have suggested that CASQ2 affects mitochondrial function^[113]36. From a clinical perspective, the implementation of our 9-gene prognostic signature for HCC holds promising potential, but several practical considerations must be addressed to facilitate its translation into routine practice. Compared to conventional biomarkers such as AFP (alpha-fetoprotein), our multi-gene panel offers improved prognostic accuracy by capturing diverse biological processes, particularly those related to oxidative stress and tumor progression. Technically, the 9-gene panel can be adapted for use with quantitative real-time PCR or NanoString platforms, both of which are cost-effective, require minimal RNA input, and are already established in many clinical laboratories. This enhances the feasibility of incorporating the model into diagnostic workflows without the need for high-throughput sequencing infrastructure. Among the nine mitochondrial oxidative stress-associated genes, MMP1 was the most likely candidate to influence tumor prognosis. Studies indicated that MMPs are directly or indirectly involved in the regulation of mitochondrial oxidative stress. Some MMPs directly degrade mitochondrial inner membrane proteins, leading to mitochondrial dysfunction and oxidative stress^[114]37. Studies have suggested that MMP1 upregulation induced by oxidative stress may contribute significantly to the development and progression of various mitochondria-related diseases such as neurodegenerative disorders and metabolic syndromes^[115]38. Our findings suggested that MMP1 may facilitate tumor proliferation and invasion and could represent a promising therapeutic target. Although MMP1 has been previously identified as a prognostic factor in hepatocellular carcinoma (HCC), our study offers a novel perspective by framing MMP1 within the context of mitochondrial oxidative stress—a key but under-explored mechanism in HCC progression. Unlike earlier studies that mainly emphasized MMP1’s role in extracellular matrix remodeling and metastasis, we demonstrate for the first time that MMP1 expression is significantly correlated with oxidative stress-related gene networks and mitochondrial dysfunction signatures. This connection suggests that MMP1 may not only act through classical invasion-related pathways, but also participate in intracellular redox imbalance and metabolic reprogramming. Evidence suggests that immunotherapy may have a positive impact effect in patients with liver cancer^[116]39. However, owing to our limited understanding of the TME and immune cell infiltration in this disease, patients receiving immunotherapy may not have achieved the expected therapeutic effects. The relationship between mitochondrial oxidative stress and tumor immunity is complex. On the one hand, mitochondrial oxidative stress could promote tumor immune responses^[117]40thereby inhibiting the growth and spread of tumors. Mitochondrial oxidative stress can enhance the autophagy and apoptosis of tumor cells^[118]41as well as increase the phagocytic activity and reactive oxygen species levels of immune cells, thereby enhancing the immune cell-killing effect of tumors. Mitochondrial oxidative stress can also inhibit tumor immunity to some extent. For example, excessive mitochondrial oxidative stress can lead to apoptosis and impairment of immune cell function^[119]42thereby weakening the immune system response to tumors. In addition, certain tumor cells could promote tumor development and metastasis by changing mitochondrial oxidative metabolism to evade immune surveillance^[120]13,[121]19,[122]43. Therefore, in this study, we explored the relationship between prognostic models and tumor immunity. We analyzed the proportion of immune cells in the different risk subtypes. The level of immune cell infiltration was generally higher in the low-risk group than in the high-risk group, indicating that mitochondrial oxidative stress-related genes may inhibit immune cell aggregation. A novel risk model based on mitochondrial oxidative stress was developed in this study, and its effectiveness in predicting the prognosis of patients with LIHC was assessed. We also explored the potential implications of this model for immune-related therapies and conducted a functional enrichment analysis to investigate the association between genes and the tumor immune microenvironment. Overall, this research provided valuable insights into potential therapeutic strategies and antitumor targets for LIHC. Electronic supplementary material Below is the link to the electronic supplementary material. [123]Supplementary Material 1^ (72.1KB, pdf) [124]Supplementary Material 2^ (512.2KB, pdf) Acknowledgements