Hepatocellular carcinoma patient prognosis combined prediction model and construction method and application thereof

Abstract

The invention discloses a hepatocellular carcinoma patient prognosis combined prediction model and a construction method and application thereof. The construction method comprises the following steps: acquiring baseline clinical indexes and serum N-carbohydrate chain characteristic data of a patient; key clinical indexes and key glycomics markers are screened out through feature engineering; and fusing the screened features into a joint feature vector, training by using a Cox regression model, and constructing a joint prediction model capable of outputting a risk score and a corresponding risk hierarchical threshold. The invention further relates to an evaluation system based on the model, a storage medium and a kit applying the model. According to the method, the functional serum glycomics characteristics are introduced into prognosis prediction of the specific treatment scene for the first time, the defects that an existing prediction method is single in index and limited in precision are overcome through multi-dimensional information integration, more accurate individualized risk layering can be achieved, a reliable tool is provided for clinical treatment decision making, and the method has important clinical application value.

Description

Technical Field

[0001] This invention belongs to the field of biomedical information technology, specifically relating to a combined prognostic prediction model for hepatocellular carcinoma patients, its construction method, and its application. Background Technology

[0002] Hepatocellular carcinoma (HCC) is a malignant tumor with leading incidence and mortality rates worldwide. For patients with advanced, unresectable HCC, systemic therapy is the primary treatment. In recent years, combination therapy regimens (referred to as "dual-drug regimens") combining tyrosine kinase inhibitors (TKIs) (such as lenvatinib and sorafenib) with immune checkpoint inhibitors (hereinafter referred to as immunotherapy) (such as pembrolizumab and atezolizumab) have become an important first-line treatment option due to their synergistic anti-tumor effects, bringing significant survival benefits to some patients. However, this combination therapy regimen is extremely expensive and may cause a series of treatment-related adverse reactions. More importantly, clinical practice shows that this regimen is not effective for all patients, exhibiting significant heterogeneity in efficacy. Currently, there is a lack of effective tools in clinical practice that can accurately predict the extent to which patients will benefit from a specific combination therapy regimen (i.e., long-term survival prognosis) before or in the early stages of treatment. This leads to a certain degree of blindness in treatment decisions, which may cause some patients to bear unnecessary economic burdens and toxic side effects without achieving the desired extension of survival, and also results in a waste of medical resources.

[0003] Existing methods or models for predicting the prognosis of HCC patients have the following limitations: First, the predictive indicators are singular and have limited accuracy. Most existing models rely on traditional clinicopathological indicators, such as tumor size, alpha-fetoprotein (AFP) levels, vascular invasion status, and Child-Pugh classification of liver function. While these indicators have some predictive value, they fail to fully encompass and reflect the complex biological characteristics of tumors, the state of the immune microenvironment, and the deep molecular basis of individual responses to TKIs and immunotherapy, resulting in a bottleneck in predictive accuracy and making it difficult to achieve precise individualized risk stratification. Second, there is a lack of dynamic and prospective biomarker systems. Most assessment models are built based on static "snapshot" data before treatment, failing to integrate biomarkers that may dynamically change during treatment and can indicate treatment response or drug resistance trends earlier, thus limiting their predictive timeliness and prospective guidance value. Finally, there is a gap in the application of emerging omics technologies. Protein glycosylation modification plays a crucial role in tumorigenesis, development, immune escape, and drug response. Serum glycomics can systematically and sensitively detect changes in the glycan structure of serum proteins, thereby reflecting the overall disease status and treatment perturbations. However, there are currently no studies or products that utilize serum glycomic characteristics, especially based on pre-treatment baseline characteristics, to develop effective tools for predicting the survival of HCC patients receiving TKI combined with immunotherapy.

[0004] Therefore, there is an urgent need to develop a novel method, model, and system that can integrate multidimensional information and accurately predict the survival of HCC patients receiving TKI combined with immunotherapy. Such a tool will help clinicians identify the high-risk population most likely to benefit from this expensive combination therapy before treatment, as well as the high-risk population that may not benefit. This will provide objective and reliable decision support for developing individualized treatment plans, optimizing the allocation of medical resources, avoiding ineffective treatments, and reducing unnecessary suffering and financial burden on patients. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a combined prognostic prediction model for hepatocellular carcinoma (HCC) patients, its construction method, and its application. By integrating static clinical indicators with dynamic, high-dimensional serum glycomics characteristics, a combined prediction model is constructed to achieve non-invasive, early, and accurate prediction of overall survival (OS) for HCC patients receiving dual TKI and immunotherapy.

[0006] This invention is achieved through the following technical solution:

[0007] A method for constructing a combined prognostic prediction model for hepatocellular carcinoma patients includes the following steps:

[0008] Step 1) Obtain a set of baseline data of hepatocellular carcinoma patients who received combination therapy with tyrosine kinase inhibitors and immune checkpoint inhibitors as a training sample set. The baseline data includes a clinical indicator dataset and a serum glycomics feature dataset.

[0009] Step 2) Perform feature filtering on the clinical indicator dataset and serum glycemic feature dataset in the training sample set to obtain key clinical indicator subsets and key glycemic feature subsets; fuse and standardize the feature values ​​in the key clinical indicator subsets and key glycemic feature subsets to construct a joint feature vector;

[0010] The key clinical indicators are: tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and number of tumors.

[0011] The key glycomic feature is six N-glycan chains with the following structures: NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb.

[0012] Step 3) Using the joint feature vector as input features and the patient's overall survival time as the endpoint event, a Cox proportional hazards regression model is used for training to obtain the final joint prediction model; the joint prediction model can output predicted values ​​based on the input joint feature vector.

[0013] A joint prognostic prediction model for hepatocellular carcinoma patients, the model being constructed using the method described above, is used to output a quantified risk score HCC_OS based on the input joint feature vector of the patient.

[0014] The risk score HCC_OS is calculated using the following formula:

[0015] HCC_OS score = 0.3645 × tumor size - 0.3213 × tumor burden score + 0.0415 × NA3 level - 0.0531 × NA3Fb level - 0.0237 × NA4 level + 0.001 × alpha-fetoprotein level - 0.0602 × NA4Fb level + 0.0406 × NA2FB level + 0.2208 × Child-Pugh score - 0.0503 × body mass index + 0.1472 × tumor number - 0.0598 × NGA2FB level.

[0016] Preferably, the model is configured with a risk stratification threshold of 0.045. When the risk score HCC_OS ≥ 0.045, the patient is determined to be in the high-risk group, and when the risk score HCC_OS < 0.045, the patient is determined to be in the low-risk group.

[0017] A prognostic assessment system for hepatocellular carcinoma patients includes:

[0018] The data receiving and preprocessing module is used to receive raw clinical indicator data and raw serum glycemic data of the target patient, and extract corresponding feature values ​​based on the key clinical indicator subset and key glycemic feature subset, and then perform standardized processing and fuse them into a joint feature vector.

[0019] The key clinical indicators are: tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and number of tumors; the key glycomic features are six N-glycans with the following structures: NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb.

[0020] The model calculation module integrates the aforementioned joint prediction model, which is used to receive the joint feature vector and calculate and output the risk score HCC_OS of the target patient.

[0021] The risk stratification module is used to compare the risk score HCC_OS output by the model calculation module with the preset risk stratification threshold to generate a high-risk or low-risk stratification result.

[0022] The visualization output module is used to display the risk score HCC_OS, risk stratification results, and corresponding survival probability curves in the form of graphs or charts.

[0023] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the functions of one or more modules as described above in the evaluation system.

[0024] A kit for assisting in predicting the prognosis of patients with hepatocellular carcinoma, comprising:

[0025] A reagent for detecting the abundance of at least six specific N-glycans in serum; the six specific N-glycans are NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb.

[0026] And the instruction manual;

[0027] The instruction manual describes the following: The abundance data of six specific N-glycans obtained by using this kit are combined with the subset data of key clinical indicators and input into the above-mentioned joint prediction model to obtain the patient's risk stratification results; the key clinical indicators are: tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and number of tumors.

[0028] A kit for assisting in predicting the prognosis of patients with hepatocellular carcinoma, comprising:

[0029] A reagent for detecting the abundance of at least six specific N-glycans in serum; the six specific N-glycans are NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb.

[0030] And the instruction manual;

[0031] The instruction manual states the following: The abundance data of six specific N-glycans obtained using this kit are combined with a subset of key clinical indicators and input into the aforementioned assessment system to obtain the patient's prognostic results; the key clinical indicators are: tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and tumor number.

[0032] The above-mentioned kit is used in the preparation of products for the adjuvant assessment of the prognosis of hepatocellular carcinoma patients receiving combination therapy with tyrosine kinase inhibitors and immune checkpoint inhibitors.

[0033] The aforementioned combined predictive model is applied in the development of products for assisting in the assessment of the prognosis of hepatocellular carcinoma patients receiving combination therapy with tyrosine kinase inhibitors and immune checkpoint inhibitors.

[0034] The beneficial effects of this invention are as follows:

[0035] (1) This invention is the first to combine serum glycomics characteristics, which can profoundly reflect the biological characteristics and immune microenvironment of tumors, with traditional clinical indicators to construct a multi-dimensional joint prediction model. Glycomics characteristics provide deep information from the functional level of protein glycosylation regulation that traditional imaging and biochemical indicators cannot capture, forming a strong complement to clinical indicators. This innovative information fusion overcomes the limitations of existing models that rely on single-dimensional data, greatly improves the model's ability to distinguish and predict patient survival prognosis, and can more accurately classify patients into risk subgroups that benefit from combination therapy.

[0036] (2) The serum or plasma samples used in this invention can be obtained through routine venous blood collection, which is non-invasive and easy to repeat. This makes the scheme of this invention not only suitable for baseline assessment before treatment, but also easier to perform dynamic and continuous monitoring during the treatment cycle. By observing the dynamic changes of indicators such as glycomics characteristics, it is expected to indicate the trend of treatment response or drug resistance earlier, thereby providing clinicians with a prospective and objective basis for timely adjustment of treatment strategies, greatly enhancing its application value in clinical practice.

[0037] (3) Early risk stratification using the model of this invention can identify high-risk patients whose survival benefit may be very limited even if they receive the current expensive "two-drug regimen" at the beginning of treatment or before treatment. For these patients, clinicians can consider switching them to participate in new drug clinical trials, choosing other alternative combination regimens, or focusing on optimal supportive care earlier, thereby avoiding unnecessary economic burdens, drug side effects, and delays in treatment opportunities caused by ineffective treatment. At the same time, for low-risk patients with clear predicted benefits, it enhances the treatment confidence of both doctors and patients. This provides a key tool for achieving true "personalized precision medicine" and rationally allocating valuable medical resources.

[0038] (4) This invention not only provides a predictive model, but also encapsulates it into a complete assessment system or reagent kit product. The system can realize an automated analysis process from data input, feature processing, model calculation to result output. It is easy to operate and the results are intuitive (such as outputting risk scores, stratified results, and survival curves). This integrated design reduces the professional calculation requirements of users, making it easy to deploy and promote in medical institutions at different levels. It can be seamlessly integrated into the existing clinical diagnosis and treatment workflow, improving the overall diagnosis and treatment efficiency. Attached Figure Description

[0039] Figure 1 Serum N-glycan profiles of HCC patients in Example 1: A represents the high-risk group; B represents the low-risk group.

[0040] Figure 2 Forest plot for glycomics analysis and feature screening of clinical features in the training set of Example 1;

[0041] Figure 3 Survival curves for prognostic evaluation of the HCC_OS model training set in Example 1;

[0042] Figure 4 The survival curves are the prognostic evaluation curves of the HCC_OS model in Example 1. Detailed Implementation

[0043] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0044] Unless otherwise specified, the technical means used in the following embodiments are all conventional means well known to those skilled in the art, and the experimental methods without specific conditions are all conventional methods in the art.

[0045] Unless otherwise specified, all materials and reagents used in the following examples are commercially available.

[0046] Example 1

[0047] 1. Test Sample

[0048] This study collected serum samples from 616 patients with advanced HCC who received combined tyrosine kinase inhibitor and immune checkpoint inhibitor therapy. All samples were obtained from the 302nd Hospital of the Chinese People's Liberation Army. All liver cancer samples underwent the examinations recommended in the "Guidelines for the Diagnosis and Treatment of Primary Liver Cancer (2024 Edition)," and the test results met the clinical diagnostic criteria. The following experiments have been filed with and approved by the ethics committee.

[0049] The dataset was randomly divided into a training set and a validation set in an 8:2 ratio, with 492 cases in the training set and 124 cases in the validation set.

[0050] 2. Basic Clinical Features

[0051] Clinical information was collected from 492 HCC patients in the training set, including: age, sex, alpha-fetoprotein, abnormal prothrombin, tumor size, vascular invasion status, Child-Pugh liver function classification, tumor burden score, alkaline phosphatase, body mass index, and number of tumors.

[0052] 3. Detection of N-glycans in training set blood samples

[0053] (1) Instruments and equipment

[0054] Capillary gel electrophoresis analyzer, PCR, centrifuge.

[0055] (2) Test reagents

[0056] Reagent A: 5 mM NH4HCO3 added to 1% SDS solution;

[0057] Reagent B: Add 2 U / μL of exoglycoside exonuclease solution to 1% NP-40;

[0058] Reagent C: Add 2 U / μL of sialidase solution to 100 mM, pH 5 NH4AC;

[0059] Reagent D: ddH2O;

[0060] Reagent E: A solution prepared by mixing 5 mM fluorescent labeling solution (trisodium 8-aminopyrene-1,3,6-trisulfonic acid) with DMSO solution (organic reducing agent NaBH3CN concentration of 1 M).

[0061] (3) N-glycan map detection and collection

[0062] 1) Release of N-glycan chains

[0063] Add 3 μL of reagent A to 5 μL of sample, heat at 95℃ for 5 min to denature, cool to room temperature, add 3 μL of reagent B and 4 μL of reagent C, react at 37℃ for 4 h, and add 80 μL of reagent D.

[0064] 2) Marking of N-glycan chains

[0065] Take 10 μL of the sample solution from step 1), dry it at 70℃ for 30 min, then add 3 μL of reagent E, react at 90℃ for 2 h, and finally add 80 μL of reagent D to terminate the reaction.

[0066] 3) Detection and chromatographic acquisition of N-glycan chains

[0067] Take 10 μL of the oligosaccharide chain sample prepared in step 2), place it in an ABI-specific 96-well plate, and detect it using an ABI 3500 sequencer to obtain the N-glycan map.

[0068] Blood samples from 492 HCC patients underwent protein denaturation, glycosidase cleavage, fluorescent labeling, and glycan mapping. Each sample ultimately yielded 12 different N-glycans, such as... Figure 1 As shown. These 12 N-glycans are: NGA2 (galactosylated biantennary N-glycan), NGA2F (galactosyl α-1, 6-core fucosylated biantennary N-glycan), NGA2FB (galactosyl α-1, 6-core fucosylated biantennary N-glycan), NG1A2F-1 (monobranched galactosyl α-1, 6-core fucosylated biantennary N-glycan), NG1A2F-2 (monobranched galactosyl α-1, 6-core fucosylated biantennary N-glycan), NA2 ( The product groups are galactosylated biantennary N-glycans, NA2F (galactosylated α-1,6-core fucosylated biantennary N-glycans), NA2FB (galactosylated α-1,6-core fucosylated bipartite N-glycans), NA3 (galactosylated triantennary N-glycans), NA3Fb (galactosylated α-1,3-branched fucosylated triantennary N-glycans), NA4 (galactosylated tetraantennary N-glycans), and NA4Fb (galactosylated α-1,3-branched fucosylated tetraantennary N-glycans). NG1A2F-1 and NG1A2F-2 are isomers.

[0069] 4. Feature screening of a combined prognostic prediction model for HCC patients receiving TKI combined with immunotherapy

[0070] Cox regression univariate analysis was performed on the collected clinical indicators and the detected N-glycan indicators. The results are as follows: Figure 2 As shown.

[0071] Depend on Figure 2It was found that six clinical indicators among the clinical characteristics—alpha-fetoprotein (AFP), tumor size, tumor burden score (TBS), tumor number (tumor_num), body mass index (BMI), and Child-Pugh score (Child-Pugh grade)—were associated with patient survival prognosis and all showed significant differences (P < 0.05).

[0072] Similarly, from Figure 2 As can be seen, six N-glycans, namely P9 (NA3), P11 (NA4), P10 (NA3Fb), P12 (NA4Fb), P8 (NA2FB), and P3 (NGA2FB), are associated with patient survival prognosis and all show significant differences (P < 0.05).

[0073] The above 6 clinical indicators and 6 N-glycan features are combined into a 12-dimensional joint feature vector as a feature for predicting patient prognosis.

[0074] 5. Construction of the HCC_OS Joint Prediction Model

[0075] Using 12 selected feature vectors (6 clinical indicators and 6 N-glycan indicators) as input, and patient survival (OS) as the label (0 = survival, 1 = death), the above 12 feature vectors are used as independent variables, and patient survival time and status are used as dependent variables. The data are then fitted to a multivariate Cox proportional hazards regression model to construct the HCC_OS model, the calculation formula of which is as follows:

[0076] HCC_OS score = 0.3645 × tumor_size - 0.3213 × TBS + 0.0415 × P9 level - 0.0531 × P10 level - 0.0237 × P11 level + 0.001 × AFP - 0.0602 × P12 level + 0.0406 × P8 level + 0.2208 × Child_Pugh_Score - 0.0503 × BMI + 0.1472 × tumor_num - 0.0598 × P3 level.

[0077] Using the above formula, a continuous HCC_OS risk score is calculated for each patient. To clearly categorize patients into "high-risk" and "low-risk" groups, a cutoff point is selected that maximizes the statistical significance of the survival difference between the two groups (minimum P-value in the Log-rank test) and ensures the most reasonable clinical difference (e.g., a sufficiently large hazard ratio (HR) while maintaining a reasonable sample size difference between the two groups). This cutoff point, in this model, is 0.045. That is, when the HCC_OS score is ≥0.045, the patient will be at high risk if receiving dual therapy with TKI and immunotherapy; when the HCC_OS score is <0.045, the patient is at low risk.

[0078] Based on the HCC_OS model described above, by inputting the desired prediction time range, the corresponding threshold can be evaluated, and a corresponding Kaplan-Meier curve can be generated based on the threshold. For example... Figure 3 As shown, the Kaplan-Meier survival curves visually demonstrate the significant association between the risk groups identified by the HCC_OS model and patient survival outcomes. Based on the optimal cutoff value calculated by the model, the training set patients were clearly distinguished into a low-risk group (blue curve) and a high-risk group (red curve). The analysis shows that the survival probability of patients in the high-risk group decreased sharply over time, significantly lower than that of the low-risk group.

[0079] Furthermore, this difference was highly statistically significant (Log-rank test, P < 0.0001), strongly validating the discriminative power of the HCC_OS model. Specifically, at several key survival time points (e.g., 20 months, 40 months), the survival rate of patients in the high-risk group was significantly worse than that of the low-risk group. Figure 3 As shown, the risk table below details the number of patients still under follow-up observation in both groups at each time point, providing data support for the curve above.

[0080] The above results demonstrate that the HCC_OS model possesses excellent prognostic predictive value. In clinical applications, when the model's output score for a specific HCC patient exceeds a threshold (0.045), the patient can be classified as a high-risk group. This suggests that the patient may have a poor response to TKI combined with immunotherapy or has developed treatment resistance, resulting in a relatively poor survival prognosis. This visualization provides important quantitative reference for individualized treatment decisions.

[0081] 6. Validation of the HCC_OS joint prediction model

[0082] Data from 124 independent validation patients were used to build the HCC_OS model from the training set. The expected prediction time range was input, and corresponding evaluation thresholds were derived. Kaplan-Meier curves were then generated based on these thresholds. The results are shown below. Figure 4 As shown, based on the same HCC_OS model and optimal cutoff value, the model still demonstrates excellent prognostic discrimination ability in an independent validation set cohort. The survival curves of the low-risk group (blue curve) and the high-risk group (red curve) also show significant and broad separation, with patients in the high-risk group having a significantly worse survival prognosis.

[0083] These results were strongly validated statistically in the validation set (Log-rank test, P < 0.0001), indicating that the predictive power of the HCC_OS model is not accidental and possesses good robustness and generalization ability. The survival probabilities of the two groups remained significantly different during long-term follow-up (e.g., 30 months, 40 months), further confirming the persistence of the model's predictions. The survival analysis of the validation set successfully reproduced the conclusions of the training set, strongly supporting the reliability and reproducibility of the HCC_OS model. This confirms that when the HCC_OS score is higher than a predetermined threshold (≥0.045), it can reliably identify high-risk patient groups who may not respond well to TKI combined with immunotherapy and have poor survival outcomes, providing crucial external validation evidence for the future clinical application of this scoring system.

[0084] The HCC_OS model and system constructed in this invention can be commercialized in the form of commercial detection kits (containing serum or plasma processing, glycan release and labeling reagents) and supporting analysis software, serving hospitals and testing institutions.

[0085] The embodiments described above are only some, not all, of the embodiments of the present invention. The detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments. The scope of protection of the present invention is determined by the scope claimed in the claims. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

Claims

1. A method for constructing a combined prognostic model for hepatocellular carcinoma patients, characterized in that, The method comprises the following steps: Step 1) obtaining baseline data of a group of hepatocellular carcinoma patients receiving tyrosine kinase inhibitor combined with immune checkpoint inhibitor treatment as a training sample set, wherein the baseline data comprises a clinical indicator data set and a serum glycomics feature data set; Step 2) performing feature screening on the clinical indicator data set and the serum glycomics feature data set in the training sample set respectively to obtain a key clinical indicator subset and a key glycomics feature subset; fusing and standardizing the feature values in the key clinical indicator subset and the key glycomics feature subset to construct a joint feature vector; The key clinical indicators are tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index and tumor number. The key glycomics features are six N-glycan chains, and their structures are NGA2FB, NA2FB, NA3, NA3Fb, NA4 and NA4Fb respectively. Step 3) using the joint feature vector as an input feature and using the overall survival time of the patient as an endpoint event, training a Cox proportional hazards regression model to obtain a final joint prediction model; the joint prediction model can output a prediction value according to the input joint feature vector. 2.A combined prognostic model for hepatocellular carcinoma, characterized in that, The model is constructed by the construction method of claim 1, and is used to output a quantitative risk score HCC_OS according to the input joint feature vector of the patient; The risk score HCC_OS is calculated by the following formula: HCC_OS score = 0.3645 x tumor size - 0.3213 x tumor burden score + 0.0415 x NA3 level - 0.0531 x NA3Fb level - 0.0237 x NA4 level + 0.001 x alpha-fetoprotein level - 0.0602 x NA4Fb level + 0.0406 x NA2FB level + 0.2208 x Child-Pugh score - 0.0503 x body mass index + 0.1472 x tumor number - 0.0598 x NGA2FB level. 3.The combined prognostic model for hepatocellular carcinoma according to claim 2, wherein, The model is configured with a risk stratification threshold of 0.045, when the risk score HCC_OS is greater than or equal to 0.045, the patient is determined to be in a high-risk group, and when the risk score HCC_OS is less than 0.045, the patient is determined to be in a low-risk group.

4. A system for evaluating prognosis of a patient with hepatocellular carcinoma, characterized by, The method comprises the following steps: A data receiving and preprocessing module is configured to receive clinical indicator raw data and serum glycomics raw data of a target patient, extract corresponding feature values according to a key clinical indicator subset and a key glycomics feature subset, and fuse the feature values into a joint feature vector after standardization; The key clinical indicators are tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index and tumor number; the key glycomics features are six N-glycan chains, and their structures are NGA2FB, NA2FB, NA3, NA3Fb, NA4 and NA4Fb respectively. a model calculation module, which integrates the joint prediction model as claimed in claim 2 or 3, for receiving the joint feature vector and calculating a risk score HCC_OS of the target patient; a risk stratification module for comparing the risk score HCC_OS output by the model calculation module with a preset risk stratification threshold to generate a stratification result of high risk or low risk; a visualization output module for displaying the risk score HCC_OS, the risk stratification result and the corresponding survival probability curve in the form of a graph or chart.

5. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the functions of one or more modules of the evaluation system as claimed in claim 4.

6. A kit for aiding in the prognosis of a hepatocellular carcinoma patient, characterized in that, comprise: reagents for detecting the abundance of at least 6 specific N-glycan chains in serum; the 6 specific N-glycan chains are NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb, respectively; and a user manual; the user manual records the following content: combining the abundance data of the 6 specific N-glycan chains detected by the kit with the data of a subset of key clinical indicators, inputting into the joint prediction model as claimed in claim 2 or 3 to obtain the risk stratification result of the patient; the key clinical indicators are tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and tumor number.

7. A kit for aiding in the prognosis of a hepatocellular carcinoma patient, characterized in that, comprise: reagents for detecting the abundance of at least 6 specific N-glycan chains in serum; the 6 specific N-glycan chains are NGA2FB, NA2FB, NA3, NA3Fb, NA4, and NA4Fb, respectively; and a user manual; the user manual records the following content: combining the abundance data of the 6 specific N-glycan chains detected by the kit with the data of a subset of key clinical indicators, inputting into the evaluation system as claimed in claim 4 to obtain the prognosis result of the patient; the key clinical indicators are tumor size, tumor burden score, alpha-fetoprotein, Child-Pugh score, body mass index, and tumor number.

8. Use of the joint prediction model as claimed in claim 2 or 3 in the preparation of a product for assisting in the evaluation of the prognosis of a patient with hepatocellular carcinoma receiving combined treatment of a tyrosine kinase inhibitor and an immune checkpoint inhibitor.

9. Use of the kit as claimed in claim 6 in the preparation of a product for assisting in the evaluation of the prognosis of a patient with hepatocellular carcinoma receiving combined treatment of a tyrosine kinase inhibitor and an immune checkpoint inhibitor.

10. Use of the kit as claimed in claim 7 in the preparation of a product for assisting in the evaluation of the prognosis of a patient with hepatocellular carcinoma receiving combined treatment of a tyrosine kinase inhibitor and an immune checkpoint inhibitor.

Images