Liver cancer prognosis stratification construction method based on exosome ceRNA network-vascular infiltration characteristics

Abstract

The application provides a liver cancer prognosis hierarchical construction method based on an exosome ceRNA network-blood vessel infiltration feature, comprising the following steps: data collection; screening of HCC differential genes; construction of an lncRNA-miRNA-mRNA ceRNA regulation network and regulation axis identification; screening of Hub-RNA; construction of an exosome RNA score system and a prognosis model; and hierarchical mechanism research on HCC patient high-risk and low-risk groups. In the research, the lncRNA, mRNA and miRNA molecules different from normal in HCC patient exosomes are taken as cores to construct a ceRNA network, and the potential molecular regulation mechanism in the HCC exosome related ceRNA network is explored. In addition, based on the common gene features of the exosome ceRNA network and VI related RNA, an Exo-VI-RNA Score scoring system is established, a HCC prognosis model is constructed, and the risk stratification and mechanism research of patients are carried out, so that a new idea is provided for the precise treatment of HCC patients.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics technology, and in particular relates to a method for constructing a prognostic stratification of hepatocellular carcinoma based on exosome ceRNA network-vascular invasion characteristics. Background Technology

[0002] Hepatocellular carcinoma (HCC) is a malignant tumor of the digestive system with a high incidence and mortality rate, and a five-year survival rate of only 20%. HCC is characterized by insidious onset, rapid progression, high rates of metastasis and recurrence, and poor prognosis, and the mechanisms of its development and progression are not fully understood. Exosomes are extracellular vesicles with a lipid bilayer, which can originate from most cells in the body and carry bioactive molecules such as nucleic acids, proteins, and lipids. lncRNAs, miRNAs, and mRNAs in exosomes can form a competitive endogenous RNA (ceRNA) network and participate in HCC progression through multiple pathways by regulating the tumor microenvironment, such as promoting vascular invasion, angiogenesis, and extracellular matrix remodeling. Among these, vascular invasion (VI) is a major factor mediating tumor cell spread and metastasis and leading to poor prognosis. Therefore, this study is necessary to construct a ceRNA network based on lncRNAs, mRNAs, and miRNAs that are different from normal molecules in HCC patient exosomes and to explore the potential molecular regulatory mechanisms in the HCC exosome-related ceRNA network. Summary of the Invention

[0003] This invention overcomes the shortcomings of existing technologies and provides a method for constructing a prognostic stratification of hepatocellular carcinoma based on exosomal ceRNA network-vascular invasion characteristics. Based on the common gene characteristics of exosomal ceRNA network and VI-related RNA, an Exo-VI-RNA Score scoring system was established, an HCC prognostic model was constructed, and risk stratification and its mechanism were studied for patients, providing new ideas for the precision treatment of HCC patients.

[0004] The present invention achieves the above objectives through the following technical solutions:

[0005] A prognostic stratification method for hepatocellular carcinoma based on exosomal ceRNA network-vascular invasion characteristics includes:

[0006] Data collection;

[0007] Screening for differentially expressed genes in HCC;

[0008] Constructing a lncRNA-miRNA-mRNA ceRNA regulatory network and identifying its regulatory axes;

[0009] Screening for Hub-RNA;

[0010] Constructing an exosomal RNA scoring system and prognostic model;

[0011] A study was conducted to investigate the stratification mechanism of high- and low-risk HCC patients.

[0012] The data collection includes:

[0013] Data on lncRNA, miRNA, and mRNA expression and clinical data of TCGA-LIHC tumors and adjacent normal tissues were downloaded from the Cancer Genome Atlas (TCGA) database. Data on miRNA expression in liver tissues of normal individuals and HCC patients (GSE140370), exosomal lncRNA and mRNA expression in blood samples of normal individuals (GSE100206), and exosomal lncRNA and mRNA expression in blood samples of HCC patients (GSE100207) were downloaded from the Gene Expression Omnibus (GEO) database.

[0014] The screening for differentially expressed genes in HCC includes:

[0015] The "DESeq2" algorithm was used to screen differentially expressed genes (DEGs) between HCC and normal samples in the TCGA-LIHC, GSE140370, GSE100206, and GSE100207 datasets. The intersection of the two sets of tumor DE-miRNAs was taken to obtain Co-DE-miRNA. The intersection of TCGA-LIHC tumor and exosome DEGs was taken according to the two types of RNA molecules to obtain the target DE-lncRNA and DE-mRNA.

[0016] The construction of the lncRNA-miRNA-mRNA ceRNA regulatory network and identification of its regulatory axes include:

[0017] The lncRNA-miRNA interaction pairs were retrieved using the LncBase database; the miRNA-mRNA interaction pairs were retrieved using the TargetScan database; based on the interaction pair relationships obtained above, a ceRNA regulatory network was constructed using Cytoscape (V3.9.1); the interaction pairs in the ceRNA regulatory network were screened, and a comprehensive analysis of univariate Cox regression, binding sites, and subcellular localization was used to obtain the lncRNA-miRNA-mRNA regulatory axis related to prognosis.

[0018] The screening of Hub-RNA includes:

[0019] Weighted gene co-expression network analysis was performed on RNA in HCC samples from TCGA-LIHC. Hub-RNAs were obtained by intersecting genes in the feature modules related to VI with genes in the constructed ceRNA network. Prognostic-related Hub-RNAs were screened out by univariate and multivariate Cox regression analysis.

[0020] The construction of the exosomal RNA scoring system includes:

[0021] HCC samples from TCGA-LIHC were randomly divided into training and validation sets in a 7:3 ratio. The Exo-VI-RNA Score was calculated based on Hub-RNA obtained from the training set. HCC samples were then divided into high-risk and low-risk groups based on their Exo-VI-RNA Score, and survival analysis was performed on each risk group. Multivariate Cox regression analysis was conducted using Exo-VI-RNA Score, age, sex, TNM stage, pathological tissue grade, and VI. The survival differences between the risk groups and the multivariate Cox regression model were validated on the validation set.

[0022] The construction of the prognostic model includes:

[0023] Based on the Exo-VI-RNA Score and various clinical factors, a prognostic model was constructed to predict the 1-year, 3-year, and 5-year survival rates of HCC patients. The performance of the predictive model was evaluated using the consistency index, calibration curve, and decision curve analysis.

[0024] The study on the stratification mechanism of high- and low-risk HCC patients includes:

[0025] The ssGSEA algorithm was used to calculate the enrichment scores of 28 immune cells in different risk groups, and the differences in immune cell infiltration between high- and low-risk groups were compared. The correlation between VI and Exo-VI-RNA scores was calculated using the chi-square test, Wilcoxon rank-sum test, and KM survival analysis. The GSVA algorithm was used to analyze the differences in metabolic pathways, VI status, tumor pathways, and immune function among different risk groups. The GSEA algorithm was used to analyze the biological processes of patients in high- and low-risk groups. Sensitive small molecule compounds and drugs for high- and low-risk patients were screened using the ConnectivityMap (CMap) database and the tumor drug sensitivity multi-omics database.

[0026] The beneficial effects of this invention compared to the prior art are:

[0027] Screening based on exosome-related ceRNA networks revealed the AC007128.1 / hsa-miR-139-5p / GABRA3 regulatory axis, which is associated with prognosis and provides a potential molecular regulatory mechanism for the diagnosis and treatment of HCC. The Exo-VI-RNA Score can serve as an independent prognostic factor for HCC. A prognostic model integrating Exo-VI-RNA scores and clinicopathological factors provides a powerful tool for survival prediction and personalized treatment decisions for HCC patients, and offers potential candidate therapies for HCC patients with different Exo-VI-RNA Score scores. Attached Figure Description

[0028] Figure 1 This is a flowchart illustrating the construction and analysis of the exosomal ceRNA network.

[0029] Figure 2 This diagram illustrates the DEGs from tumor tissue and exosome data, showing Co-DE-miRNA (A), DE-lncRNA (B), and DE-mRNA (C). TCGA_UP represents upregulated RNAs from tumor tissue data, TCGA_DOWN represents downregulated RNAs from tumor tissue data, GSE140370_UP represents upregulated miRNAs from tumor tissue data, GSE140370_DOWN represents downregulated miRNAs from tumor tissue data, GEO_UP represents upregulated lncRNAs / mRNAs from exosome data, and GEO_DOWN represents upregulated lncRNAs / mRNAs from exosome data.

[0030] Figure 3 This is a ceRNA network diagram. Red indicates molecules upregulated in the HCC group compared to the normal group, while blue indicates downregulated molecules. The intensity of the color indicates the log2FC size. Diamond-shaped molecules represent 622 lncRNAs, squares represent 43 miRNAs, and circles represent 429 mRNAs.

[0031] Figure 4 The diagram shows the results related to the regulatory axis. In (A), the regulatory axis of AC007128.1 / hsa-miR-139-5p / GABRA3 is established; downregulated molecules are indicated by blue boxes, and upregulated molecules are indicated by red boxes. (B)

[0032] Binding sites of the lncRNA-miRNA-mRNA axis. (C) Biological processes involved by GABRA3.

[0033] Figure 5The results of constructing the WGCNA network are shown in the following figures: (A) Analyzing the scale-free fit index and average connectivity under different soft thresholds to determine the optimal soft threshold for lncRNA / mRNA. (B) Determining the optimal soft threshold for miRNA. (C) Cluster dendrogram of lncRNA / mRNA, where each color represents a module, and gray represents genes not included in any module. (D) Cluster dendrogram of miRNA. (E) Heatmap of the correlation between lncRNA / mRNA co-expression modules and clinical traits. Each column corresponds to a clinical trait, and each row corresponds to a co-expression module. Each box includes the corresponding correlation (first row) and p-value (second row). Red represents a positive correlation, and green represents a negative correlation. (F) Heatmap of the correlation between miRNA co-expression modules and clinical traits.

[0034] Figure 6 Construction of the Exo-VI-RNA Score risk score. Among them, (A) KM survival curves of the training set.

[0035] (B) KM survival curves for the validation set. (C) Risk score and survival status distribution plots for patients in the training set. (D) Risk score and survival status distribution plots for patients in the validation set. (E) Multivariate Cox regression forest plot for the training set. (F) Multivariate Cox regression forest plot for the validation set.

[0036] Figure 7 This is the Exo-VI-RNA Score prognostic model. (A) Nomogram. (B) Calibration curves of the hazard model at 1, 3, and 5 years. The diagonal line represents perfectly accurate predictions; the closer the prediction is to the diagonal line, the better the outcome.

[0037] (C) DCA curves for the three models.

[0038] Figure 8 The results show the correlation between VI and Exo-VI-RNA. (A) A bar graph showing the percentage of patients with VI distributed between high-risk and low-risk groups. (B) Risk scores comparing patients with large VI, micro VI, and no VI. (C) KM curves for patients with large VI, micro VI, and no VI. (D) KM curves for high- and low-risk patients in the large VI group. (E) KM curves for high- and low-risk patients in the micro VI group.

[0039] Figure 9 Results of ssGESA, GSVA, and GSEA analyses for high- and low-risk groups of Exo-VI-RNA Score. (A) Results of ssGSEA analysis, (B) Results of GSVA analysis, and (CD) Results of GSEA analysis.

[0040] Figure 10GDSC results for the low-risk group of Exo-VI-RNA Score; Detailed Implementation

[0041] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0042] Example:

[0043] Numerous studies have shown that aberrantly expressed exosomal RNAs can serve as HCC-specific biomarkers. Exosomal miR-638 expression levels are negatively correlated with tumor size, TNM stage, and vascular invasion status in HCC patients. Furthermore, miR-638 overexpression significantly inhibits HCC cell proliferation. Exosomal miR-638 levels are positively correlated with overall survival and are an independent benign prognostic factor for HCC. High expression of exosomal circUHRF1 in the plasma of HCC patients is closely associated with poor clinical prognosis. Exosomal circUHRF1 promotes immune evasion and resistance to anti-PD1 immunotherapy in HCC by regulating TIM-3 expression in NK cells. Serum exosomal-derived lncRNAs (EV-lncRNAs), EV-DLEU2, EV-HOTTIP, EV-MALAT1, and EV-SNHG1, can distinguish HCC patients from non-HCC patients and show a significant advantage over AFP in the diagnosis of very early-stage HCC.

[0044] The tumor microenvironment (TME) is composed of various components, including tumor-associated macrophages (TAMs), immune cells, extracellular matrix (ECM), and endothelial cells. Exosomal RNA crosstalk in HCC can mediate communication and interactions between tumor cells and other cell types within the TME, thereby influencing HCC progression and metastasis. lncRNAs (RP11-1100 L3.8) expressed in macrophages of HCC patients may mediate macrophage polarization towards M2 and promote HCC tumor proliferation by activating glucose metabolism in HCC cells. The exosomal lncRNA HMMR-AS1, associated with poor prognosis, can promote HCC growth. The mechanism involves a hypoxic environment promoting competitive binding of exosomal HMMR-AS1 to miR-147a, thereby regulating ARID3A-mediated macrophage polarization. Exosomal RNAs can also affect the function of some immune cells. HCC cell exosomes miR-92b inhibit the effects of the potential target CD69 on NK cells, leading to a reduction in the cytotoxic effects of both CD69 and NK cells, ultimately resulting in immune escape. CD90+ hepatocellular carcinoma cells release exosomal lncRNA H19, which is transported to human umbilical vein endothelial cells, inducing angiogenesis and affecting cell-cell adhesion, thereby influencing the tumor microenvironment. The mechanism of angiogenesis may be related to the increased mRNA expression levels of vascular endothelial growth factor (VEGF) and its receptor (VEGF-R1) in endothelial cells by CD90+ hepatocellular carcinoma cell exosomes.

[0045] Due to the extremely rich blood supply to the liver, vascular invasion (VI) is a common event in hepatocellular carcinoma (HCC), which can be divided into macrovascular invasion and microvascular invision (mVI). Macrovascular invasion refers to the pathological phenomenon of tumor cells invading the major blood vessels of the liver, which can be detected by imaging examinations, usually involving the portal vein or its branches. mVI refers to the presence of solid tumor cell clusters within the lumen of endothelial-lined veins. Both can lead to the metastasis of invasive tumor cells away from the primary tumor site, and are key risk factors for HCC recurrence and poor prognosis.

[0046] The above studies indicate that exosomal RNA plays a crucial regulatory role in the occurrence, progression, and prognosis of hepatocellular carcinoma (HCC). Microvascular and macrovascular invasion in HCC is a major contributing factor to tumor cell dissemination, leading to recurrence, metastasis, and low survival rates. Current research on HCC-derived exosomal RNA mainly focuses on its role in regulating biological processes and exploring its downstream target genes. Related bioinformatics analyses emphasize screening for differentially expressed exosomal RNAs after sequencing data to identify HCC-specific biomarkers, but have not delved into the regulatory networks involved in these biomarkers or the RNA regulatory axes. Therefore, exploring the mechanisms of exosomal RNA in HCC development from a global perspective of RNA regulatory networks, and investigating the impact of exosomal RNA regulatory mechanisms and VI-associated characteristics on the prognosis of HCC patients, can provide a theoretical basis and important reference for the clinical diagnosis and treatment of HCC patients.

[0047] The complex interactions between lncRNAs, miRNAs, and mRNAs constitute a ceRNA regulatory network. lncRNAs, acting as molecular sponges for miRNAs, regulate mRNA expression, influencing the translation of corresponding proteins and related cellular activities. Overexpression of lincSCRG1 in human HCC tissues and cell lines can induce HCC cell proliferation and migration; the mechanism is that lincSCRG1 binding to miR26a leads to upregulation of SKP2 expression. The lncRNA MCM3AP-AS1, associated with poor prognosis, acts as a miRNA-199p molecular sponge, promoting FOXA1 expression. The lncRNA MYLK-AS1, a ceRNA, promotes tumor angiogenesis by targeting miR-429p to upregulate E2F7 expression and activate VEGFR-2 signaling.

[0048] This demonstrates that aberrant expression of any component within the ceRNA regulatory network can disrupt complex regulatory pathways, ultimately leading to cancer development. The mechanisms of HCC are not yet fully understood, and further research is needed to identify the interactions of key RNAs in HCC, construct ceRNA regulatory networks, and screen for key RNAs from these networks. This will help identify potent biomarkers and therapeutic targets, leading to more effective cancer therapies and better patient prognosis and survival rates.

[0049] Currently, reliable biomarkers for the diagnosis and prognosis of hepatocellular carcinoma (HCC) remain lacking in clinical practice. Exosomal RNA is involved in multiple aspects of HCC development, including angiogenesis, vascular invasion (VI), and tumor drug resistance, and can play an important role as a novel diagnostic and prognostic biomarker in liquid biopsy. The presence and extent of vascular invasion in patients significantly influence HCC recurrence, metastasis, and prognosis. However, systematic research on the molecular regulatory mechanisms of exosomes and their association with vascular invasion is still lacking.

[0050] This study innovatively utilizes various bioinformatics methods to investigate the regulatory mechanisms of key RNAs and their regulatory networks in HCC exosomes. A ceRNA regulatory network of exosome-related lncRNA-miRNA-mRNA was constructed, and potential molecular regulatory mechanisms were discovered. Furthermore, leveraging the shared genetic characteristics of the constructed exosomal ceRNA network and VI-related RNAs, an Exo-VI-RNA Score risk scoring system and prognostic model were developed to stratify the risk of HCC patients. The mechanisms leading to different risks were explored in depth, and corresponding sensitive drug molecules were screened. This provides a powerful theoretical basis and tool for personalized clinical diagnosis and treatment decision-making systems from a novel perspective.

[0051] Workflow of this study

[0052] This study conducted risk stratification and mechanistic analysis of HCC patients based on exosome-related ceRNA networks and VI gene characteristics. The analysis process is as follows: Figure 1 As shown.

[0053] A method for constructing a prognostic stratification of hepatocellular carcinoma based on exosomal ceRNA network-vascular invasion characteristics includes the following steps:

[0054] The first step is to download the TCGA-LIHC tumor dataset and clinical data from the TCGA database, and the tumor dataset GSE140370 (miRNA), exosome dataset GSE100206 (lncRNA / mRNA), and exosome dataset GSE100207 (lncRNA / mRNA) from the GEO database.

[0055] The second step involves performing differential expression analysis on HCC and normal samples in the dataset, and taking the intersection of differentially expressed miRNAs in the TCGA and GSE140370 datasets to obtain common differentially expressed miRNAs; taking the intersection of DEGs in TCGA-LIHC tumors and GEO exosomes according to the two categories of lncRNA and mRNA respectively to obtain the target differentially expressed lncRNA and target differentially expressed mRNA.

[0056] The third step involves constructing an exosome-related ceRNA regulatory network by integrating the interaction pair results obtained from the database. Then, univariate Cox regression analysis is used to screen out the complete lncRNA-miRNA-mRNA regulatory axis from the ceRNA network, and its subcellular localization and binding sites are analyzed.

[0057] The fourth step involves screening gene modules associated with the VI characteristic based on WGCNA network analysis. The intersection of RNAs in the ceRNA network and RNAs in the VI-related modules of WGCNA is then used to obtain Hub-RNAs.

[0058] In the fifth step, HCC samples from TCGA-LIHC were randomly divided into a training set (262 cases) and a validation set (112 cases) at a ratio of 7:3. In the training set, survival-related molecules were screened, and the Exo-VI-RNA Score was calculated. The training set was then divided into high- and low-risk groups based on the Exo-VI-RNA Score, and survival analysis and multivariate Cox regression analysis were performed. The validation set was used to validate the survival differences between the different risk groups and the multivariate Cox regression model. Subsequently, an Exo-VI-RNA Score prognostic model was constructed and its performance was evaluated.

[0059] The sixth step is to conduct a prognostic stratification study on high- and low-risk patients, including differences in the tumor microenvironment and screening for sensitive drugs.

[0060] To obtain the necessary clinical data and complete expression profiles, this study processed the TCGA-LIHC, GSE100206, GSE100207, and GSE140370 datasets according to the following steps:

[0061] (1) TCGA-LIHC dataset

[0062] The TCGA official tool gdc-client was used to download RNA-seq, miRNA-seq, and clinical data from TCGA-LIHC. Expression matrices were then compiled, with behavioral genes listed as samples. The RNA-seq, miRNA-seq expression matrices, and clinical data were then integrated.

[0063] Since analyses other than differential expression require eliminating the influence of gene length and sequencing depth, and RNA-seq data is gene count data, this study normalized the expression matrix data of lncRNA and mRNA to Transcripts Per Kilobase of exon model per Million mapped reads (TPM) and saved it accordingly. Similarly, the normalized Reads / Counts of exon model per Million mapped reads (RPM) was obtained from the miRNA-seq data matrix.

[0064] (2) GSE140370 dataset

[0065] According to the ceRNA hypothesis, both lncRNAs and mRNAs can only competitively bind to miRNAs via MREs, thereby regulating each other's expression; therefore, lncRNAs and mRNAs have an indirect regulatory relationship. As a bridge connecting the two, miRNAs play a particularly important role in the ceRNA network. Therefore, in addition to the miRNA expression data from TCGA, this study also included miRNA expression data from tumors and normal tissues in the GSE140370 dataset. The intersection of the differentially expressed miRNAs from tumors relative to normal tissues in the two datasets was then used to obtain a common differentially expressed miRNA matrix.

[0066] (3) GSE100206 and GSE100207 datasets

[0067] The exosomal lncRNA and mRNA expression matrices were extracted from the blood of healthy individuals from the GSE100206 dataset, and from the blood of HCC patients from the GSE100207 dataset. The exosomal matrices related to healthy individuals and HCC were merged, and the SVA algorithm was used to eliminate batch effects. Because the GSE100206 and GSE100207 datasets are reads per kilobase of exon model per million mapped reads (FPKM), direct differential analysis is not possible. Therefore, the batch-effect-free exosomal matrices were normalized using Log2 and then rounded down for storage; the normalized exosomal lncRNA and mRNA values ​​are Log2FPKM.

[0068] Difference expression and analysis:

[0069] Differential analysis was performed on HCC group samples and normal group samples from the TCGA-LIHC, GSE100206, GSE100207, and GSE140370 datasets, and differentially expressed lncRNAs were identified using the "DESeq2" algorithm.

[0070] (Differentially expressed lncRNA, DE-lncRNA), mRNA (Differentially expressed mRNA, DE-mRNA), miRNA (Differentially expressed miRNA, DE-miRNA) thresholds are set to |log2FoldChange|>0, P adj <0.25. Statistics on upward and downward adjustments are provided with annotations.

[0071] The intersection of DE-miRNA from tumor tissue of TCGA-LIHC and DE-miRNA from tumor tissue of GSE140370 yielded Co-DE-miRNA. The intersection of DE-lncRNA and DE-mRNA from tumor tissue of TCGA-LIHC with DE-lncRNA and DE-mRNA from exosomal sites of GSE100206 and GSE100207 yielded the target DE-lncRNA and DE-mRNA.

[0072] Screening of RNA relation pairs

[0073] Constructing a ceRNA network requires identifying lncRNA-miRNA and miRNA-mRNA interaction pairs. This study will utilize the following methods to screen RNA relationship pairs for constructing the ceRNA network.

[0074] LncRNA-miRNA interaction pair screening: Co-DE-miRNA was predicted in the TarBase database with a threshold set to Score>0.5. The lncRNA targeting the miRNA was predicted and its intersection with the target DE-lncRNA was taken to construct lncRNA-miRNA interaction pairs.

[0075] miRNA-mRNA interaction pair screening: Co-DE-miRNAs are predicted in the TargetScan database to obtain the mRNAs targeted by the miRNAs. The intersection of these mRNAs with the target DE-mRNAs is then used to construct miRNA-mRNA interaction pairs.

[0076] Construction of ceRNA network

[0077] By integrating lncRNA-miRNA and miRNA-mRNA interaction pairs, a ceRNA regulatory network associated with exosomes, incorporating lncRNA-miRNA-mRNA interactions, was constructed. lncRNAs and miRNAs, as well as miRNAs and mRNAs, each have direct regulatory effects.

[0078] Using Cytoscape software (Version 3.9.1), the integrated exosomal ceRNA regulatory network files were combined with RNA differential expression data to create a network diagram.

[0079] Differentially expressed RNA between the tumor group and the normal group

[0080] This study analyzed the differential expression between the HCC group and the normal group using tumor tissue data from TCGA-LIHC and GSE140370, and exosome data from GSE100206 and GSE100207, with threshold values ​​set at 1.5.

[0081] |log2FoldChange|>0, P adj <0.25. Results showed that in the TCGA-LIHC tumor tissue data, compared with the normal group, HCC patients had 5956 DE-lncRNAs, of which 1190 were downregulated and 4766 were upregulated; 866 DE-miRNAs, of which 182 were downregulated and 684 were upregulated; and 12936 DE-mRNAs, of which 4493 were downregulated and 8443 were upregulated. In the GSE140370 tumor tissue data, there were 59 DE-miRNAs, of which 9 were downregulated and 50 were upregulated. In the exosome data, compared with the normal group, HCC patients had 2116 DE-lncRNAs, of which 628 were downregulated and 1488 were upregulated; and 1832 DE-mRNAs, of which 832 were downregulated and 1000 were upregulated. After intersecting the DE-miRNAs from the two sets of tumor tissues, 43 Co-DE-miRNAs were obtained, of which 7 were downregulated and 36 were upregulated. After intersecting the DEGs from TCGA tumor tissues and exosomes, 758 target DE-lncRNAs were found, of which 69 were downregulated and 689 were upregulated; and 661 target DE-mRNAs were found, of which 188 were downregulated and 473 were upregulated. To more intuitively illustrate the distribution of DEGs and the results after intersection, a Venn diagram was constructed, as shown below. Figure 2 As shown in AC.

[0082] Determination of RNA Relation Pairs

[0083] To obtain the exosomal ceRNA network, lncRNA-miRNA interaction pairs were screened from the database based on Co-DE-miRNA. The intersection of the lncRNAs in the obtained interaction pairs with the target DE-lncRNA in this study was then calculated. The target mRNAs were predicted based on the miRNAs in the lncRNA-miRNA interaction pairs, and the intersection of these predictions with the target DE-mRNA in this study was then calculated. Finally, a ceRNA network containing 622 lncRNAs, 34 miRNAs, and 429 mRNAs was determined.

[0084] Establishment of ceRNA network

[0085] After integrating lncRNA-miRNA and miRNA-mRNA interaction pairs, 622 lncRNAs, 34 miRNAs, and 429 mRNAs were incorporated into the construction of the exosomal ceRNA network, involving a total of 4430 interaction pairs (see Table 1). A ceRNA regulatory network diagram was then drawn in Cytoscape software, as shown below. Figure 3 As shown.

[0086] Table 1. lncRNA-miRNA / miRNA-mRNA interactions

[0087]

[0088]

[0089] Due to its insidious onset, hepatocellular carcinoma (HCC) lacks specific and sensitive biomarkers. Currently, pathological examination remains the "gold standard" for clinical diagnosis. With in-depth research on exosomes, the close relationship between exosomal RNA and tumor development is gradually being revealed, indicating that it can serve as a non-invasive and convenient biomarker, offering new possibilities for the diagnosis and treatment of HCC. The molecular mechanisms of HCC are highly complex and heterogeneous. Studying the molecular mechanisms of HCC from different perspectives is essential to providing better solutions for clinical diagnosis and treatment. Therefore, this study focuses on HCC-derived specific exosomal RNA to investigate the prognosis and molecular regulatory mechanisms of HCC.

[0090] The ceRNA network hypothesis provides a new direction for exploring the mechanisms of HCC and also helps in finding highly specific and sensitive biomarkers. LncRNAs can competitively bind to miRNAs to regulate the expression of target mRNAs. Since public databases lack exosomal miRNA data from HCC patients, and since there is material exchange between exosomes in the blood and tumor tissue, this study constructed a ceRNA network using tumor miRNAs, exosomal lncRNAs, and mRNAs. Exploring the molecular mechanisms between exosomal and tumor tissue RNAs helps to uncover gene functions and regulatory mechanisms at a deeper level, facilitating a more comprehensive understanding of the occurrence and development of HCC.

[0091] In summary, this study investigated the regulatory mechanism of hepatocellular carcinoma (HCC) using an exosome-associated ceRNA network. We successfully constructed an exosome-associated ceRNA network containing 622 lncRNAs, 34 miRNAs, and 429 mRNAs, which is one of the innovative aspects of this study and lays the foundation for future research on the prognosis and molecular mechanisms of HCC.

[0092] Univariate Cox regression analysis was performed on RNA in the exosome-associated ceRNA network.

[0093] In order to screen out RNA molecules related to the prognosis of HCC patients from the exosomal ceRNA network constructed in Chapter 2, this study conducted a univariate Cox regression analysis based on the RNA expression level of the ceRNA network and patient survival information. P < 0.05 was considered statistically significant.

[0094] Locating binding sites between RNA molecules and constructing complete regulatory axes.

[0095] Using the prognosis-related RNAs obtained through screening, the lncRNA-miRNA-mRNA regulatory axis was screened based on the correct regulatory relationships in the ceRNA network. The establishment of the regulatory axis requires the following two conditions to be met:

[0096] (1) lncRNA competitively binds to miRNA, weakening the inhibitory effect of miRNA on the expression of target mRNA, thereby indirectly promoting the expression of mRNA. Therefore, the differential expression directions of lncRNA and mRNA are consistent and opposite to the differential expression directions of miRNA.

[0097] (2) lncRNA and mRNA have a co-expression relationship (R Pearson >0.3, P<0.05).

[0098] After identifying the regulatory axis of lncRNA-miRNA-mRNA, its binding sites need to be validated. The lncRNA-miRNA binding sites are searched in the LncBase database; the miRNA-mRNA binding sites are searched in the TargetScan database, and a binding site map is plotted.

[0099] Subcellular localization and functional analysis of RNA molecules in the regulatory axis

[0100] Since different organelles often have different physicochemical environments, the subcellular localization of RNA determines whether it participates in potential regulatory mechanisms. This study combined the RNAlocate database, the lncATLAS database, and the iLoc-LncRNA database to perform subcellular localization of each molecule in the regulatory axis.

[0101] This study analyzed the function of mRNA in the inBioDiscover database and explored its role in the development and progression of cancer based on the biological processes involved by mRNA.

[0102] Prognostic RNAs in the ceRNA Network

[0103] Univariate Cox regression analysis was performed on each molecule in the exosomal ceRNA network, and the results are shown in Appendix Table 2. 56 lncRNAs, 3 miRNAs, and 53 mRNAs were statistically significant, indicating that these RNAs are independent factors affecting the survival of HCC patients.

[0104] Table 2. Major subcellular localizations of molecules in the regulatory axis.

[0105]

[0106] Establishment of AC007128.1 / hsa-miR-139-5p / GABRA3 control axis

[0107] Based on the results of univariate Cox regression analysis of RNA molecules in the ceRNA network and combined with screening criteria, a complete prognostic-related lncRNA-miRNA-mRNA regulatory axis was screened from the ceRNA network. Two upregulated DE-miRNAs (hsa-miR-10b-5p and hsa-miR-139-5p) and one downregulated DE-miRNA (hsa-miR-9-5p) were initially identified as candidate miRNAs for the regulatory axis. Subsequently, based on the negative regulatory relationship between lncRNA and miRNA, two lncRNA-miRNA interaction pairs (TMEM220-AS1 / hsa-miR-10b-5p, AC007128.1 / hsa-miR-139-5p) were obtained from the lncRNAs in the univariate Cox regression analysis results; and based on the negative regulatory relationship between miRNA and mRNA, two miRNA-mRNA interaction pairs (hsa-miR-139-5p / GABRA3, hsa-miR-139-5p / GDF10) were obtained from the mRNAs in the univariate Cox regression analysis results. Pearson correlation analysis was performed on the above lncRNA and mRNA expression levels, showing a correlation of 0.55 between AC007128.1 and GABRA3, and a correlation of 0.19 between AC007128.1 and GDF10. By integrating lncRNA-miRNA interaction pairs, miRNA-mRNA interaction pairs, and lncRNA-mRNA co-expression information, a potential lncRNA-miRNA-mRNA regulatory axis was finally obtained (AC007128.1 / hsa-miR-139-5p / GABRA3), as shown below. Figure 4 As shown in Figure A, down-regulating molecules are represented by blue boxes, and up-regulating molecules are represented by red boxes.

[0108] This study searched for the binding sites of AC007128.1 / hsa-miR-139-5p in the Tarbase database and the binding sites of hsa-miR-139-5p / GABRA3 in the TargetScan database. The resulting binding site map is shown below. Figure 4 As shown in Figure B, the red areas in the figure represent the specific binding sites of the regulatory axis.

[0109] Subcellular localization and functional analysis of AC007128.1 / hsa-miR-139-5p / GABRA3

[0110] This study integrated results from the lncATLAS, iLoc-LncRNA, and RNAlocate databases. Table 2 shows the subcellular localization of each molecule in the AC007128.1 / hsa-miR-139-5p / GABRA3 regulatory axis. AC007128.1 is mainly located in the cytoplasm and participates in the regulation of miRNAs; hsa-miR-139-5p is mainly located in microvesicles and exosomes; and GABRA3 is mainly located in exosomes and cytosol.

[0111] This study performed online predictions of biological processes involved in GABRA3, and the results are as follows: Figure 4 As shown in Figure C. iBioDiscover results indicate that GABRA3 is involved in the regulation of the ERBB4 signaling pathway, MHC class I complex, EGF / EGFR signaling pathway, and MHC complex, and plays a pro-cancer role in the development of HCC.

[0112] The exosome-related ceRNA network constructed in this study provides new insights into the molecular mechanism of HCC. However, the ceRNA network involves a large number of RNA molecules. In order to make clinical diagnosis and treatment more targeted, this study further screened key molecules in the network and obtained the prognostic regulatory axis AC007128.1 / hsa-miR-139-5p / GABRA3.

[0113] First, RNAs associated with HCC prognosis were screened from the molecules constituting the ceRNA network, yielding 56 lncRNAs, 3 miRNAs, and 53 mRNAs. These RNAs were independent factors predicting overall survival (OS) in HCC patients. Subsequently, based on the regulatory relationships of the ceRNA hypothesis and the results of univariate Cox regression screening, the AC007128.1 / hsa-miR-139-5p / GABRA3 regulatory axis was identified.

[0114] To further define this regulatory axis, the binding sites of lncRNA-miRNA and miRNA-mRNA interaction pairs were analyzed, confirming the regulatory relationships of AC007128.1 / hsa-miR-139-5p and hsa-miR-139-5p / GABRA3. Since the subcellular localization of each molecule determines its potential regulatory mechanism, the subcellular localization results showed that hsa-miR-139-5p screened from tumor tissue was mainly located in microvesicles and exosomes, AC007128.1 screened from both tumor tissue and exosomes was mainly located in the cytoplasm, and GABRA3 was mainly located in exosomes and cytosol, indicating that this regulatory axis may occur spatially. Functional enrichment of GABRA3 indicates that it participates in the ERBB4 signaling pathway and exerts a tumor-suppressive effect. At the same time, it participates in the regulation of the EGF / EGFR signaling pathway, MHC class I complex protein, and MHC complex protein, exerting a tumor-promoting effect. This shows that it can participate in the occurrence and development of HCC as both a tumor suppressor gene and a tumor-promoting gene, but under the combined effect, it promotes the growth of HCC cells.

[0115] Previous studies exploring the functions of molecules along the regulatory axis have shown that AC007128.1 is upregulated in both cells and tissues of esophageal squamous cell carcinoma and is associated with poor prognosis in patients; it has also been found to be associated with pyroptosis in bladder cancer. The low expression of hsa-miR-139-5p in HCC can increase HCC cell invasion and proliferation by increasing SLITRK4 expression. Hypomethylation of the promoter of the GABRA3 transcript expressed in gastric cancer induces increased miR-105-5p expression, leading to decreased PD-L1 expression. In this study, AC007128.1, hsa-miR-139-5p, and GABRA3 were all independent molecules affecting survival prognosis, with hsa-miR-139-5p acting as a protective miRNA, consistent with the findings of the previous studies.

[0116] Extensive research has been conducted on ceRNA molecular regulatory axes in cancer research, but there are currently no reports on exosomal ceRNA regulatory axes. This is another innovative aspect of our study, and it also provides a theoretical basis for molecular regulation research in HCC and precision medicine in clinical practice. However, due to the extremely limited availability of exosomal data in public databases, this study constructed a ceRNA network and regulatory axis based on signal interactions between blood and tumor tissue, utilizing tumor tissue miRNAs and blood exosomal lncRNAs and mRNAs. Further experimental validation is needed. Furthermore, the molecular regulatory mechanisms of HCC are complex, and the regulation of each molecule is not independent; multiple molecules may jointly regulate each other. Therefore, our research group will further validate the regulatory relationships within the regulatory axis through in vitro and in vivo experiments.

[0117] In summary, this study conducted an in-depth analysis of the exosome-related ceRNA network and proposed...

[0118] AC007128.1 / hsa-miR-139-5p / GABRA3 is a prognostic regulatory axis. Functional analysis suggests that this axis may exert a comprehensive pro-cancer effect through multiple signaling pathways, ultimately promoting the occurrence and development of HCC.

[0119] Building a WGCNA network

[0120] To identify modules associated with VI, cluster analysis was first performed on the whole transcriptome of HCC samples from TCGA-LIHC to identify outliers. Therefore, outliers needed to be removed before weighted correlation network analysis (WGCNA) was performed.

[0121] First, based on the expression data of the whole transcriptome in HCC samples, an optimal soft threshold was calculated to ensure the acquisition of a label-free scale network. Then, a co-expression matrix was constructed based on the soft threshold, and the adjacency degree between genes was calculated to obtain an adjacency matrix. After converting the adjacency matrix into a topological overlap matrix, the dissimilarity coefficient between genes was derived based on their similarity, resulting in a gene clustering tree. A dynamic tree slicing method was used to cluster co-expressed genes into different lncRNA / mRNA modules and miRNA modules. Finally, based on the grouped phenotypic information, the correlation between gene modules and phenotypes was calculated, identifying trait-related modules and obtaining gene modules associated with VI (lncRNA / mRNA modules, cor>0.15, P<0.05; miRNA modules, P<0.05).

[0122] Calculation of Exo-VI-RNA Score Risk Score

[0123] The RNAs in the co-expression modules related to VI that were screened and the RNAs used to construct the ceRNA network were intersected to obtain Hub-lncRNA, Hub-mRNA, and Hub-miRNA.

[0124] Thirty-seventy-seven HCC samples were randomly divided into training and validation sets (7:3 ratio). Univariate and multivariate Cox regression analyses were performed on Hub-lncRNA, Hub-mRNA, and Hub-miRNA in the training set to screen for risk factors that independently influenced the survival prognosis of HCC patients after excluding interference from clinical characteristics such as age, sex, histological grade, TNM stage, and VI. The selected RNAs were used to calculate the Exo-VI-RNA Score risk score. The product of the regression coefficient and expression level of each molecule was linearly summed, as shown in the following formula:

[0125]

[0126] In Formula 1, n represents the number of molecules used to construct the Exo-VI-RNA Score, i represents the i-th molecule, β represents the regression coefficient of the molecule in the multivariate Cox regression analysis, and X represents the expression level of the molecule.

[0127] KM Survival Prognosis Analysis and Validation

[0128] The Exo-VI-RNA Score risk score was calculated for each sample in the training set. HCC samples were then divided into a high-risk group and a low-risk group based on the optimal threshold of the risk score. Using the "survival" package, Kaplan-Meier (KM) survival analysis was performed on the two risk groups to calculate the survival difference between the high-risk and low-risk groups based on the survival status and duration of HCC patients. In this study, overall survival (OS) of HCC patients was used for calculation; a p-value < 0.05 was considered statistically significant.

[0129] To verify whether the Exo-VI-RNA Score can serve as an independent prognostic factor, this study integrated clinical characteristics and performed univariate and multivariate Cox regression analyses. P < 0.05 was considered statistically significant.

[0130] Construction of a prognostic model based on Exo-VI-RNA Score

[0131] Exo-VI-RNA Score and clinical information (age, sex, TNM stage, histological grade, VI) were incorporated, and a nomogram was constructed using the "rms" package. These indicators were integrated to construct a Cox proportional hazards regression model. Based on the contribution of each influencing factor to overall survival (OS), scores were assigned to each factor, and the scores were summed to obtain a total score. The 1-, 3-, and 5-year survival rates were then predicted based on the function transformation relationship. This study evaluated the prediction results by calculating the concordance index (C-index) and plotting calibration curves. Simultaneously, the "rmda" package was used to plot decision curve analysis (DCA) for the single Exo-VI-RNA Score model, the combined clinical information model, and the combined Exo-VI-RNA Score and clinical information model to evaluate the model's net clinical return.

[0132] Association analysis between vascular invasion and Exo-VI-RNA score

[0133] The Exo-VI-RNA Score risk score in this study was constructed based on the VI characteristic-exosomal ceRNA network. To investigate the impact of VI on the survival of HCC patients with different scores, this study statistically analyzed the frequency and percentage of VI status (no VI, micro VI, and large VI) in high- and low-risk groups; compared the Exo-VI-RNA Score risk scores of patients with the three VI statuses; and performed KM survival analysis on patients with the three VI statuses in the high- and low-risk groups.

[0134] ME cell infiltration assessment

[0135] The immune system plays a dual role in the development and progression of cancer. On one hand, it can continuously monitor tissues for signs of cancer through immune surveillance. Once cancerous cells are detected, T cells recognize and lyse them, thereby inhibiting tumor development. On the other hand, cancer cells may prevent or suppress the immune response by inhibiting T cell proliferation or recruiting immunosuppressive cells, thus promoting cancer development. The tumor microenvironment (TME) includes tumor cells and surrounding immune cells, stromal cells, and the extracellular matrix. Tumor cells can directly interact with other components of the TME. Immune infiltration within the TME plays a crucial role in tumor development and influences clinical outcomes for cancer patients.

[0136] This study employed the single-sample gene set enrichment analysis (ssGSEA) algorithm to investigate the infiltration of immune cells in the total mesothelial endothelial growth factor (TME) of HCC patients with different disease scores. Gene sets representing 28 immune cell types were obtained from the literature, with one gene set corresponding to each immune cell type. The enrichment score of each immune cell type in each sample was calculated using the "GSVA" R package (Version 1.40.1), representing the degree of infiltration of that immune cell type in that sample. Then, the statistical difference in enrichment scores for each immune cell type between the high- and low-risk groups was calculated, with a threshold of P < 0.05. Box plots were generated using the "ggplot2" algorithm based on the infiltration status of various cells in the high- and low-risk groups.

[0137] Gene set variation analysis (GSVA)

[0138] Cellular metabolism, VI status, tumor signaling pathways, and immune function are all closely related to tumorigenesis and development. Aerobic glycolysis is the main energy utilization method for cancer cells, and the reprogramming of energy metabolism is considered a "marker of cancer." The pentose phosphate pathway plays a key role in regulating cancer cell growth. Vascular invasion is a clinicopathological phenomenon closely related to recurrence and metastasis of hepatocellular carcinoma. Therefore, to explore the differences in these pathways among patients with different risk scores, this study collected gene sets related to metabolism, vascular invasion, tumor signaling pathways, and immune-related specific biological processes from the MSigDB database (http: / / www.gsea-msigdb.org) and published literature. Gene set variation analysis (GSVA) ​​was used to calculate the enrichment scores of these specific biological processes in patients with different risk scores. P < 0.05 was considered statistically significant.

[0139] Gene set enrichment analysis (GSEA)

[0140] To investigate the biological functional differences between the high-risk and low-risk groups, the "DESeq2" algorithm was first used to perform differential gene analysis on the two groups (high-risk group vs. low-risk group), with the threshold set to |log2FoldChange|>0, P adj <0.25. Then, the "clusterProfiler" algorithm was used to perform gene set enrichment analysis (GSEA) on upregulated and downregulated mRNAs, selecting all biological pathway (BP) gene sets for GO, with P<0.05 as the cutoff criterion. This study aimed to investigate the biological processes that DE-mRNA might be involved in in patients with different risk scores.

[0141] Screening for sensitive drugs in high- and low-risk patients

[0142] This study stratified HCC patients into high- and low-risk groups based on the Exo-VI-RNA Score and screened for sensitive drugs in both groups. This approach facilitates individualized treatment, avoids indiscriminate drug use, and provides a reference for clinical decision-making. The Connectivity Map (CMap) database and the Genomics of Drug Sensitivity in Cancer (GDSC) database were used to screen for sensitive molecular compounds and drugs for patients with different risk scores.

[0143] The specific methods are as follows: Differential analysis was performed between high-risk and low-risk groups and the normal group. The differentially expressed RNA was identified using the "DESeq2" algorithm (threshold setting: |log2FoldChange|>0, corrected P<0.25). Upregulated and downregulated genes were input into the CMap webpage, and the top 5 small molecule compounds in the results were selected using a CMap connection score <-90 as the screening criterion. The "oncoPredict" algorithm was used to predict the half-maximum inhibitory concentration (IC50) of all drugs in the GDSC database, and the top 5 drugs in the results were selected using a P<0.05 screening criterion.

[0144] Gene modules related to vascular invasion in the WGCNA network

[0145] A total of 377 tumor samples were extracted from TCGA-LIHC. After removing outliers, the lncRNA and mRNA expression data of 369 samples were included to construct a hierarchical clustering tree. Similarly, miRNA expression data of 375 tumor samples were extracted from TCGA-LIHC to construct a hierarchical clustering tree. To determine the adjacency matrix, a scale-free topology fitting index of 0.9 was used, and the optimal soft threshold for lncRNA / mRNA was found to be 10. Figure 5 A), the optimal soft threshold for miRNA is 3 ( Figure 5 B). After merging similar modules, a total of 26 lncRNA / mRNA co-expression modules were constructed. Figure 5 C), 7 miRNA co-expression modules ( Figure 5 D), where the gray module is used to collect genes not assigned to any module and is not included in subsequent analyses. Further analysis examined the correlation between clinical traits of HCC patients, such as pathological grade, TNM stage, VI, and liver fibrosis, and various gene modules. Figure 5 As shown in E and F. Based on the screening criteria, gene modules significantly associated with VI were obtained, including 13 lncRNA / mRNA co-expression modules (yellow, blue, and light orange) and a brown miRNA co-expression module.

[0146] The intersection of RNAs from the constructed ceRNA network with RNAs from the feature modules obtained by WGCNA analysis yielded 41 Hub-lncRNAs, 1 Hub-miRNA, and 24 Hub-mRNAs. Further univariate and multivariate Cox regression analyses of the Hub-RNAs revealed 3 Hub-lncRNAs (AL118511.1, LINC01060, AC122710.1) and 2 Hub-mRNAs (PDE6A, INSM1), as shown in Table 3.

[0147] The Exo-VI-RNA Score risk score was constructed using the above five RNA molecules. Based on the optimal threshold of the score, the training set was divided into high-risk and low-risk groups, resulting in 38 cases in the high-risk group and 194 cases in the low-risk group; the validation set yielded 46 cases in the high-risk group and 53 cases in the low-risk group.

[0148] Table 3. Molecular components for constructing the Exo-VI-RNA Score and the results of single-factor and multi-factor Cox analyses.

[0149]

[0150]

[0151] The results of KM survival analysis on the high- and low-risk groups of the training set are as follows: Figure 6 As shown in Figure A, patients in the low-risk group had significantly better survival rates than those in the high-risk group (P < 0.001), which was statistically significant. Subsequent KM survival analysis of the high- and low-risk groups, based on the optimal threshold of the Exo-VI-RNAScore for each sample in the validation set, yielded the same results. Figure 6 As shown in B, P = 0.0019, which is statistically significant. The risk scores and survival status of high-risk and low-risk patients in the training and validation sets are as follows: Figure 6 As shown in C and D.

[0152] To verify whether the Exo-VI-RNA Score can serve as an independent prognostic risk factor, this study combined clinical characteristics (including age, sex, histological grade, TNM stage, and VI), using survival time and survival status as dependent variables and various characteristics as independent variables, and conducted a multivariate Cox regression analysis. Figure 6 As shown in E, the HR value of the Exo-VI-RNA Score was 2.37, P < 0.001, indicating that the Exo-VI-RNA Score can serve as an independent risk prognostic factor for HCC patients, and patients with high-risk scores have a worse prognosis. The validation set is as follows: Figure 6 As shown in F, the HR value of the Exo-VI-RNA Score was 1.16, P = 0.727. Although the statistical P value exceeded the threshold, the HR > 1, indicating that the score is a risk factor affecting the patient's OS. The reason why the statistical difference was not significant may be due to the limited number of validation set samples.

[0153] Construction and evaluation of a prognostic model based on Exo-VI-RNA Score

[0154] A Cox regression model was constructed using the Exo-VI-RNA Score and clinical characteristics (including age, sex, histological grade, TNM stage, and VI) as a risk model for predicting HCC prognosis, and a nomogram was plotted. Figure 7 As shown in Figure A, age, sex, histological grade, TNM stage, and VI have relatively little impact on prognosis, while the Exo-VI-RNA Score has a very significant impact. Mapping patient information to a nomogram yields corresponding scores, and summing these scores gives a total score that can predict the patient's 1-, 3-, and 5-year survival rates. The calculated C-index is 0.738, indicating that this risk model can predict patient survival well. Subsequently, calibration curves for the model at 1, 3, and 5 years were plotted, as shown below. Figure 7 As shown in Figure B, the horizontal axis represents the patient survival rate predicted by the risk model, and the vertical axis represents the actual patient survival rate. The curves are close to the diagonal line of perfectly accurate predictions, indicating that the risk model has good predictive performance. DCA results are as follows... Figure 7 As shown in Figure C, the red curve represents the single Exo-VI-RNA Score model, the green curve represents the model that integrates clinical information, and the light green curve represents the integrated prognostic risk model that includes both Exo-VI-RNA Score and clinical features. It can be seen that the net clinical benefit rates of the single Exo-VI-RNA Score model and the combined model are similar, with little difference in effectiveness, but both are better than the combined clinical information model.

[0155] Vascular invasion was significantly associated with the Exo-VI-RNA score.

[0156] The percentage of various VIs occurring in patients in high- and low-risk groups, such as Figure 8 As shown in Figure A, significantly more patients in the high-risk group developed VI (large VI: 9.7%, small VI: 38.7%, no VI: 51.6%, P<0.01). In the low-risk group, 4.5% of patients developed large VI, 27.8% developed small VI, and 67.6% did not develop VI. Furthermore, the risk score of patients with large VI was higher than that of patients with small VI and those without VI. Although the result was not statistically significant due to the limited number of patients, it still showed a trend of difference among the three VI states. Figure 8 B). Survival analysis showed that the overall survival (OS) of patients with major VI was significantly worse than that of patients without VI (P<0.03). Figure 8 C). Subsequently, survival analyses were performed on high- and low-risk patients in different VI groups. The results showed that in the large VI group, patients with high-risk scores had extremely poor overall survival (OS), P < 0.01. Figure 8 D), similarly, in patients with micro VI, the overall survival (OS) of those with high-risk scores was worse than that of those with low-risk scores, P = 0.02 ( Figure 8E). And according to Figure 8 As shown in D and E, patients with high-risk scores and large VI had the worst overall survival (OS), indicating that the combined occurrence of these two factors is the main factor affecting the OS of HCC patients.

[0157] Differences in TME cell infiltration between high- and low-risk groups

[0158] ssGSEA results showed that the low-risk group had a higher degree of infiltration of adaptive immune-related immune cells (activated B cells, activated CD8 T cells, effector memory CD8 T cells, memory B cells, etc.), such as Figure 9 As shown in Figure A, the low-risk group patients had a stronger immune response compared to the high-risk group patients, which is consistent with the better overall survival (OS) of the low-risk group patients.

[0159] Differences in metabolic pathways, tumor pathways, VI status, and immune function between high- and low-risk groups

[0160] The results of GSVA analysis of high- and low-risk groups in metabolic, VI, tumor-related pathways, and immune-related pathways are as follows: Figure 9 As shown in Figure B, this study selected gene sets with significant differences (P<0.05) for presentation. The high-risk group showed significant enrichment in genes related to glycolysis, the pentose phosphate pathway, aerobic glycolysis, and VI / VEGF-mediated angiogenesis, compared to the lower-risk group. Furthermore, the high-risk group showed greater enrichment in tumor-related pathways such as MAPK, Notch, PI3K_AKT_MTOR, and PTEN, as well as immune-related pathways such as the NOD-like receptor signaling pathway. The low-risk group, on the other hand, showed upregulation in gene sets related to fatty acid oxidation, glutamine metabolism, and CTL-mediated immune responses.

[0161] To further explore the underlying biological processes in patients with different risk scores, this study performed GSEA enrichment analysis based on DE-mRNA in high- and low-risk groups, identifying upregulated DE-mRNA (...). Figure 9 C) is mainly enriched in biological processes such as cell development, ion transport, metabolic processes, metabolic regulation, and transfer; downregulated DE-mRNA ( Figure 9 D) is enriched in cellular macromolecular metabolism, the metabolism of nucleobase-containing compounds, the regulation of the MAPK cascade, the regulation of primary metabolism, and RNA metabolism.

[0162] Sensitive drugs for patients in high- and low-risk groups

[0163] The CMap analysis results are shown in Table 4, which screened out potential sensitive compounds in high-risk and low-risk groups. The connectivity fraction ranges from -100 to +100. A positive connectivity fraction indicates that the small molecule compound can induce the disease, while a negative connectivity fraction indicates that the small molecule compound can reverse the disease state and has potential therapeutic value.

[0164] Table 4. Screening results of CMap small molecule compounds in the high- and low-risk groups of Exo-VI-RNA Score.

[0165]

[0166]

[0167] The responses of approximately 200 anticancer drugs in over 1000 tumor cell types were obtained from the GDSC database. The "oncoPredict" algorithm was used to predict the sensitivity of patients with different risk scores to these 200 drugs. Based on the selection criteria, the results are as follows: Figure 10 As shown, among the treatments AZD8186, NU7441, rapamycin, taprazole, and topotecan, the predicted IC50 values ​​for the low-risk group were significantly lower than those for the high-risk group, indicating that patients with lower Exo-VI-RNA scores were more sensitive to these treatments. This study did not find any sensitive drugs for eligible high-risk patients from the GDSC database.

[0168] HCC, due to its rich blood supply, is highly susceptible to vascular invasion, leading to increased intrahepatic and extrahepatic metastasis and a poor prognosis. This chapter constructs a WGCNA network to identify genes associated with the VI trait, and then intersects these genes with those from an exosome ceRNA network to construct the Exo-VI-RNA Score risk score. KM analysis and multivariate Cox regression analysis show that this risk score is an independent risk factor affecting patient prognosis.

[0169] Subsequently, this study constructed a prognostic prediction model based on the Exo-VI-RNA Score, integrating multiple predictive indicators including risk scores, enabling prognostic risk assessment for individual patients. The model was evaluated from three dimensions: discriminative power, calibration, and clinical effectiveness. Results from C-indx, calibration curves, and DCA curves all indicated that the model had good predictive performance.

[0170] The incidence of vascular invasion (VI) differed between high- and low-risk groups, with the high-risk group being more prone to vascular invasion. Patients with both high VI scores and large vessel invasion had the worst overall survival (OS). This study analyzed the molecular mechanisms contributing to the survival differences between high- and low-risk groups from the perspectives of immunity, metabolism, tumor pathways, and VI status. ssGSEA results showed greater adaptive immune cell infiltration in the low-risk group, while GSVA results indicated a higher enrichment of CTL-mediated immune response signals in the low-risk group. These findings corroborate each other, suggesting a stronger adaptive immune response in the low-risk group with better OS. Conversely, the high-risk group showed significant enrichment in the NOD-like receptor signaling pathway, indicating a stronger innate immune response. Glycolysis, pentose phosphate pathway, fatty acid oxidation, and glutamine metabolism, among other metabolic pathways, promote tumor development and progression through different pathways. The enrichment of these pathways in the high- and low-risk groups provides individualized guidance for subsequent treatment targeting these metabolic pathways.

[0171] The high-risk group showed enrichment in the VI and angiogenesis pathways, further demonstrating the clinical significance of the Exo-VI-RNAScore established in this study. Simultaneously, the high-risk group showed significant enrichment in multiple tumor-related pathways, revealing the molecular mechanism underlying the poor survival prognosis in this group. Subsequently, GSEA results demonstrated the major biological processes involved by DE-mRNA in high-risk patients compared to low-risk patients. Upregulated DE-mRNA was mainly enriched in cell development, ion transport, metabolic processes, metabolic regulation, and metastasis, while downregulated DE-mRNA was enriched in cellular macromolecular metabolism, the metabolism of nucleobase-containing compounds, the regulation of the MAPK cascade, the regulation of primary metabolic processes, and RNA metabolism. This further reveals the potential molecular mechanisms underlying the survival differences among patients with different risk levels.

[0172] Finally, screening for sensitive drugs in high- and low-risk patients revealed that low-risk patients showed higher sensitivity to small molecule compounds such as DDOST, LGMN, CAY-10618, SERPINC1, and CDA, as well as therapeutic agents such as AZD8186, NU7441, rapamycin, taprazole, and topotecan, which can provide a reference for clinical drug treatment. High-risk patients, on the other hand, were more sensitive to small molecule compounds such as DUSP28, IFNB1, IFNG, TES, and SHB.

[0173] In summary, this study constructed the Exo-VI-RNA Score based on exosomal ceRNA networks and VI characteristics, and used this score to build a risk assessment model for predicting 1-, 3-, and 5-year survival rates in patients. Subsequent studies on related molecular mechanisms may reveal that differences in immune, metabolic, VI, and tumor pathways may contribute to the poorer survival prognosis in high-risk patients. More specific molecular mechanisms require validation through subsequent molecular biology experiments.

[0174] Based on transcriptomic data from HCC tumor tissue and exosomes, this study constructed an exosome-related ceRNA network and a prognostic-related AC007128.1 / hsa-miR-139-5p / GABRA3 molecular regulatory axis, providing a theoretical basis for the molecular mechanisms and potential targets of HCC. Based on this exosome-related ceRNA network and VI gene characteristics, the Exo-VI-RNAScore risk scoring system and prognostic prediction model were constructed, enabling individualized assessment of the survival prognosis of HCC patients. The molecular mechanism studies and sensitive drug screening for high- and low-risk patients explained the survival differences between the two groups at the molecular level and theoretically guided precise clinical treatment strategies.

[0175] (1) Using transcriptomic data from tumor tissues and exosomes, an exosome-related ceRNA regulatory network was constructed. Differential analysis was performed on samples from the HCC group and the normal group. The intersection of DE-miRNAs from the TCGA-LIHC and GSE140370 tumor tissues was obtained to obtain Co-Hub-DE-miRNA. The intersection of DEGs from the TCGA-LIHC tumor tissue and GSE100206 and GSE100207 exosomes was obtained to obtain the target DE-lncRNA and target DE-mRNA. By integrating database information, lncRNA-miRNA and miRNA-mRNA interaction pairs were obtained, thereby constructing an exosome-related ceRNA regulatory network.

[0176] (2) Establishment of the AC007128.1 / hsa-miR-139-5p / GABRA3 regulatory axis associated with HCC patient prognosis. Univariate Cox regression analysis was performed on the RNAs used to construct the ceRNA network, yielding 56 statistically significant lncRNAs, 3 miRNAs, and 53 mRNAs. A complete molecular regulatory axis was then established based on the regulatory relationships, binding sites, and subcellular localization of lncRNAs, miRNAs, and mRNAs. Functional analysis indicated that this regulatory axis may promote the occurrence and development of HCC through multiple cancer-related signaling pathways.

[0177] (3) An Exo-VI-RNA Score risk score and a prognostic risk model based on this score were constructed, and the molecular mechanisms underlying the prognostic differences between high- and low-risk groups were analyzed. Hub-RNAs were obtained by intersecting genes from the constructed ceRNA network with genes related to VI characteristics. Univariate and multivariate Cox regression analyses screened three lncRNAs and two mRNAs for constructing the Exo-VI-RNA Score risk score. High- and low-risk groups were divided using the optimal threshold of the score. KM survival analysis and multivariate Cox regression results showed that the Exo-VI-RNA Score was a prognostic risk factor for HCC independent of clinicopathological features. Integrating the Exo-VI-RNA Score and clinicopathological features, a prognostic prediction model was constructed, which could predict patients' 1-, 3-, and 5-year survival rates relatively well. Comprehensive analysis of the Exo-VI-RNA Score and VI revealed that patients with both high scores and large VI had extremely poor prognoses. ssGSEA and GSVA results showed that the low-risk group had higher abundance of adaptive immune cell infiltration and significant enrichment in adaptive immune signaling pathways. The high-risk group was significantly enriched in tumor-related pathways such as MAPK, Notch, PI3K_AKT_MTOR, and PTEN.

[0178] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a prognostic stratification of hepatocellular carcinoma based on exosomal ceRNA network-vascular invasion characteristics, characterized in that, include: Data collection; Screening for differentially expressed genes in HCC; Constructing a lncRNA-miRNA-mRNA ceRNA regulatory network and identifying its regulatory axes; Screening for Hub-RNA; Constructing an exosomal RNA scoring system and prognostic model; A study on the stratification mechanism of high- and low-risk HCC patients; The screening for differentially expressed genes in HCC includes: The "DESeq2" algorithm was used to screen differentially expressed genes (DEGs) between HCC and normal samples in the TCGA-LIHC, GSE140370, GSE100206, and GSE100207 datasets. The intersection of the two sets of tumor DE-miRNAs was taken to obtain Co-DE-miRNA. The intersection of TCGA-LIHC tumor and exosome DEGs was taken according to the two types of RNA molecules to obtain the target DE-lncRNA and DE-mRNA. The construction of the lncRNA-miRNA-mRNA ceRNA regulatory network and identification of its regulatory axes include: The lncRNA-miRNA interaction pairs were retrieved using the LncBase database; the miRNA-mRNA interaction pairs were retrieved using the TargetScan database; based on the interaction pair relationships obtained above, a ceRNA regulatory network was constructed using Cytoscape; the interaction pairs in the ceRNA regulatory network were screened, and a comprehensive analysis of univariate Cox regression, binding sites, and subcellular localization was used to obtain the lncRNA-miRNA-mRNA regulatory axis related to prognosis. The screening of Hub-RNA includes: Weighted gene co-expression network analysis was performed on RNA in HCC samples from TCGA-LIHC. Hub-RNAs were obtained by intersecting genes in the feature modules related to VI with genes in the constructed ceRNA network. Prognostic-related Hub-RNAs were screened out by univariate and multivariate Cox regression analysis. The construction of the exosomal RNA scoring system includes: HCC samples from TCGA-LIHC were randomly divided into training and validation sets in a 7:3 ratio. The Exo-VI-RNA Score was calculated based on Hub-RNA obtained from the training set. HCC samples were then divided into high-risk and low-risk groups based on their Exo-VI-RNA Score, and survival analysis was performed on each risk group. Multivariate Cox regression analysis was conducted using Exo-VI-RNA Score, age, sex, TNM stage, pathological tissue grade, and VI. The survival differences between the risk groups and the multivariate Cox regression model were validated on the validation set.

2. The method for constructing a prognostic stratification of hepatocellular carcinoma based on exosome ceRNA network-vascular invasion characteristics according to claim 1, characterized in that, The data collection includes: Data on lncRNA, miRNA, and mRNA expression and clinical data of tumors and adjacent normal tissues were downloaded from the TCGA-LIHC tumor genome atlas database. Data on miRNA expression in liver tissues of normal individuals and HCC patients (GSE140370), exosomal lncRNA and mRNA expression in blood samples of normal individuals (GSE100206), and exosomal lncRNA and mRNA expression in blood samples of HCC patients (GSE100207) were downloaded from the Gene Expression Comprehensive Database.

3. The method for constructing a prognostic stratification of hepatocellular carcinoma based on exosome ceRNA network-vascular invasion characteristics according to claim 1, characterized in that, The construction of the prognostic model includes: Based on the Exo-VI-RNA Score and various clinical factors, a prognostic model was constructed to predict the 1-year, 3-year, and 5-year survival rates of HCC patients. The performance of the predictive model was evaluated using the consistency index, calibration curve, and decision curve analysis.

4. The method for constructing a prognostic stratification of hepatocellular carcinoma based on exosome ceRNA network-vascular invasion characteristics according to claim 1, characterized in that, The study on the stratification mechanism of high- and low-risk HCC patients includes: The ssGSEA algorithm was used to calculate the enrichment scores of 28 immune cell types in different risk groups, and the differences in immune cell infiltration between high- and low-risk groups were compared. The correlation between VI and Exo-VI-RNA scores was calculated using the chi-square test, Wilcoxon rank-sum test, and KM survival analysis. The GSVA algorithm was used to analyze the differences in metabolic pathways, VI status, tumor pathways, and immune function among different risk groups. The GSEA algorithm was used to analyze the biological processes of patients in high- and low-risk groups. Sensitive small molecule compounds and drugs for high- and low-risk patients were screened using the ConnectivityMap database and the tumor drug sensitivity multi-omics database.

Images