Gene marker combination and system for predicting glioma risk level and application

Abstract

The invention provides a gene marker combination and system for predicting glioma risk and application, and belongs to the technical field of glioma prediction. The gene marker combination comprises an NTN4 gene, an RBP1 gene, a TWIST1 gene, a GADD45G gene, an NUAK2 gene, a GRIK2 gene, a WEE1 gene and an RRM2 gene. Eight senescence-related genes are screened by screening senescence-related genes and combining survival data of glioma patients. The research finds that the risk scoring model constructed by using the eight senescence-related genes can specifically and sensitively predict the glioma risk, and has good prediction efficiency. According to the method, the glioma risk assessment model constructed on the basis of combining the risk scoring model with the clinical pathological variable factors is established, and compared with the capability of predicting the glioma risk of the risk scoring model, the glioma risk assessment model is better in prediction efficiency.

Description

Technical Field

[0001] This invention belongs to the field of glioma prediction technology, and in particular relates to a combination of gene markers, a system, and applications for predicting the risk of glioma. Background Technology

[0002] Glioma is one of the most common malignant tumors of the central nervous system, especially glioblastoma (GBM), which is highly aggressive, has a very poor prognosis, and low survival rate. Currently, clinical practice mainly relies on the World Health Organization (WHO) tumor grading system and key molecular markers (such as IDH mutation, 1p / 19q co-deletion, and MGMT promoter methylation) for patient stratification and prognostic assessment. However, single molecular markers cannot fully reflect the biological heterogeneity of patients, resulting in limited guidance for individualized treatment plans. Therefore, constructing glioma prognostic prediction models based on multi-gene comprehensive bioinformatics analysis has become an important research direction in precision medicine.

[0003] In recent years, with the development of cancer genomics and high-throughput sequencing technologies, several multigene prognostic models have been developed for survival prediction in glioma patients. These models are typically based on public databases such as TCGA and CGGA, using bioinformatics analysis to screen for prognostic-related genes, and employing methods such as Cox regression, LASSO regression, and machine learning to construct risk scoring models. Current main research directions and existing problems include: (1) IDH mutation and 1p / 19q co-deletion model: IDH mutation and 1p / 19q co-deletion are key molecular subtypes in the WHO 2016 classification. Patients with these mutations have a good prognosis, but there are still significant differences in survival. A single molecule cannot accurately distinguish between high- and low-risk groups.

[0004] (2) Immune-related gene (IRG) model: Some studies have used immune-related genes to construct prognostic models. For example, the CIBERSORT algorithm was used to analyze immune microenvironment features such as T cell infiltration and macrophage polarization, and prognostic scores were constructed by combining immune gene expression. However, the dynamic changes in immune status limit the stability of these models.

[0005] (3) Stem-related gene model: Gliomas have strong tumor stemness, and some studies have stratified patients based on stem-related genes, such as SOX2 and OCT4. However, this type of model is only effective for specific subgroups (such as GBM) and cannot be universally applied to all patients.

[0006] (4) Senescence-Related Genes Model: Cellular senescence plays a crucial role in the tumor microenvironment; for example, the senescence-associated secretory phenotype (SASP) can promote tumor progression. Some studies have attempted to use senescence genes to screen for potential biomarkers, but their application is still immature and lacks large-scale data validation.

[0007] While the aforementioned models have improved the predictive ability of glioma prognosis to some extent, they still have the following limitations: I. Limitations of single biological features: Traditional molecular subtyping (such as IDH mutation, 1p / 19q co-deletion) cannot capture individual patient differences; immune, stem, or aging-related gene models are too one-sided and fail to incorporate multi-level molecular information. II. Limited model stability and generalization: Many studies have only built models on a single database (such as TCGA) without validating them in independent cohorts (such as CGGA), resulting in insufficient generalization; the use of only univariate Cox analysis or LASSO regression without incorporating more comprehensive biostatistical methods affects model stability. III. Lack of integrated tumor microenvironment analysis: Most existing studies have neglected the tumor immune microenvironment (TME), while immune cell infiltration plays a key role in glioma progression; traditional prognostic models do not consider single-cell sequencing (scRNA-seq) data and cannot analyze the interactions between different tumor cell populations. IV. Lack of guidance on drug sensitivity and individualized treatment: Most existing models are only used for prognostic assessment and are not combined with drug sensitivity data (such as the GDSC database), so they cannot directly guide clinical treatment decisions; the relationship between the expression of immune checkpoint genes (such as PD-1, PD-L1, CTLA4) and patient risk scores has not been fully studied, affecting their application value in immunotherapy. Summary of the Invention

[0008] Therefore, the purpose of this invention is to provide a combination of gene markers, a system, and applications for predicting the risk of glioma.

[0009] To achieve the above-mentioned objectives, the present invention provides the following technical solution: This invention provides a combination of gene markers for predicting the risk of glioma, the combination of gene markers including NTN4 gene, RBP1 gene, TWIST1 gene, GADD45G gene, NUAK2 gene, GRIK2 gene, WEE1 gene and RRM2 gene.

[0010] This invention provides a reagent for detecting the expression level of the above-mentioned combination of gene markers for predicting glioma risk in the preparation of products for assessing glioma risk.

[0011] This invention provides an application of a reagent for detecting the expression level of the above-mentioned combination of gene markers, combined with a product for detecting clinicopathological variables, in the preparation of a product for assessing the risk of glioma, wherein the clinicopathological variables include tumor grade, age, and IDH status.

[0012] This invention provides a system for assessing the risk of glioma using the above-mentioned combination of gene markers, the system comprising a data detection unit, a data processing unit, and a result determination unit; The data detection unit detects the expression level of each gene in the above-mentioned gene marker combination in the sample to be tested. The data processing unit: constructs a risk scoring model using the above-mentioned gene marker combination, and substitutes the expression levels of each gene obtained by the data detection unit into the risk scoring model to calculate the risk score; The result determination unit receives the risk score obtained by the data processing unit and compares it with the median to determine the risk level of glioma in the tested sample.

[0013] Preferably, the risk scoring model is ∑(gene expression level × gene coefficient); the gene is the combination of the above-mentioned gene markers.

[0014] Preferably, the risk scoring model = RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748.

[0015] Preferably, the median is obtained by sorting all glioma samples in ascending order of risk scores obtained from the risk scoring model described above, and then obtaining the median.

[0016] Preferably, if the risk score of the sample to be tested is greater than or equal to the median, the sample to be tested is a high-risk patient for glioma; if the risk score of the sample to be tested is less than the median, the sample to be tested is a low-risk patient for glioma.

[0017] This invention provides a system for assessing the risk or survival probability of glioma using a combination of gene markers and clinicopathological variables. The system includes a data acquisition unit, a data processing unit, and a result determination unit. The data acquisition unit is used to acquire the tumor grade, age, IDH status, and aging-related risk score of the sample to be tested. The method for obtaining the aging-related risk score is as follows: a risk score model is constructed using the above-mentioned combination of gene markers, the expression level of each gene in the above-mentioned combination of gene markers in the sample to be tested is detected, and the obtained expression level of each gene is substituted into the risk score model to calculate the risk score. The data processing unit: establishes a glioma risk assessment model based on the obtained tumor grade, age, IDH status, and aging-related risk score; constructs a nomogram of the total score of the glioma risk assessment model versus the survival probability; and uses this nomogram to predict the glioma risk level or survival probability of glioma patients. The result determination unit receives the glioma risk score obtained by the data processing unit and uses a nomogram to determine the survival probability, thereby determining the glioma risk level or survival probability of the tested sample.

[0018] Preferably, the glioma risk assessment model = tumor grade score + age score + IDH status score + risk score. In the calculation of this glioma risk assessment model, tumor grade G2 is assigned a value of 0, tumor grade G3 is assigned a value of 52.5, and tumor grade G4 is assigned a value of 100; age ≤ 40 years is assigned a value of 0, age > 40 years and ≤ 60 years is assigned a value of 67, and age > 60 years is assigned a value of 98; IDH status IDH mutant is assigned a value of 0, and IDH status IDH wild-type is assigned a value of 53; the risk score is calculated from the risk score of the above risk score model to determine the level of risk, with low risk assigned a value of 0 and high risk assigned a value of 54. The survival probability is one or more of the following: 1-year survival probability, 3-year survival probability, and 5-year survival probability.

[0019] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a combination of gene biomarkers, a system, and its application for predicting glioma risk. For the first time, this invention screens aging-related genes and, in conjunction with survival data from glioma patients, identifies eight aging-related genes. The study found that a risk scoring model constructed using these eight aging-related genes can specifically and sensitively predict glioma risk, demonstrating good predictive efficacy. This invention establishes a glioma risk assessment model based on a risk scoring model combined with clinicopathological variables. Compared to the risk scoring model's ability to predict glioma risk, this glioma risk assessment model exhibits superior predictive efficacy. Attached Figure Description

[0020] Figure 1A is a volcano plot of differentially expressed genes in the TCGA dataset, with red indicating significantly upregulated genes and blue indicating significantly downregulated genes; B is a volcano plot of differentially expressed genes in the CGGA dataset; C is a Venn plot showing 29 aging-related differentially expressed genes jointly identified in TCGA and CGGA; D is a hierarchical clustering heatmap based on these 29 genes, which can clearly distinguish between low-grade glioma (LGG) and glioblastoma (GBM) samples; E is a partial likelihood bias plot in LASSO regression analysis, showing that 12 key genes were selected at the optimal λ value; F is a regression coefficient trajectory plot of the 12 genes in LASSO regression, showing the trend of coefficient changes with λ value; G is a forest plot of multivariate Cox regression analysis, showing the hazard ratio (HR) and statistical significance (p-value) of the 8 independent prognostic-related genes finally included in the model. Figure 2 To illustrate the predictive performance of the risk scoring model on patient prognosis in the first cohort dataset, A-C represent Kaplan-Meier survival curves, showing that the overall survival of high-risk patients in the training set (A), test set (B), and the entire dataset (C) was significantly lower than that of the low-risk group (p<0.001). D-F represent the ROC curves of the risk model for 1-year, 3-year, and 5-year survival predictions in the training set (D), test set (E), and the entire dataset (F), respectively, with AUC values ​​indicating high predictive accuracy. G-I represent the risk score distribution plot and survival status plot, showing that high-risk patients in the training set (G), test set (H), and the entire dataset (I) had higher risk scores, shorter survival times, and higher mortality rates. Figure 3 The results show the independent validation of the risk scoring model in the second cohort dataset. A is the Kaplan-Meier survival curve, showing that the survival time of patients in the high-risk group was significantly shorter than that in the low-risk group (HR=3.72, p<0.001); B is the ROC curve of the risk scoring model for 1-year, 3-year, and 5-year survival in the second cohort dataset, with AUC values ​​of 0.750, 0.788, and 0.785, respectively, indicating that the model has good predictive performance; C is the risk score distribution plot and survival status plot, showing that patients in the high-risk group had higher risk scores, shorter survival times, and higher mortality rates. Figure 4The results of the construction and validation of the glioma risk assessment model are as follows: A shows the results of multivariate Cox regression analysis, which shows that tumor grade, age, IDH status, and risk score are independent prognostic factors; B shows that the constructed nomogram model integrates clinicopathological factors and risk scores, and can predict the survival probability of patients at 1 year, 3 years, and 5 years; C shows the calibration curve of the nomogram predicting survival rate, which shows that the model's prediction results are highly consistent with the actual observed values, and have good calibration ability; D shows the time-based ROC curve analysis, which shows that the AUC values ​​of the model in predicting 1-year, 3-year, and 5-year survival are 0.882, 0.934, and 0.902, respectively, indicating that the model has extremely high predictive accuracy. Detailed Implementation

[0021] This invention provides a combination of gene markers for predicting the risk of glioma, the combination of gene markers including NTN4 gene, RBP1 gene, TWIST1 gene, GADD45G gene, NUAK2 gene, GRIK2 gene, WEE1 gene and RRM2 gene.

[0022] In this invention, the NTN4 gene, RBP1 gene, TWIST1 gene, GADD45G gene, NUAK2 gene, GRIK2 gene, WEE1 gene and RRM2 gene are all known genes that can be found in the TCGA and CGGA databases.

[0023] This invention provides a reagent for detecting the expression level of the above-mentioned combination of gene markers for predicting glioma risk in the preparation of products for assessing glioma risk.

[0024] The reagents used in this invention to detect the expression level of the above-mentioned gene marker combinations include antibodies or primer pairs corresponding to each gene.

[0025] This invention provides an application of a reagent for detecting the expression level of the above-mentioned combination of gene markers, combined with a product for detecting clinicopathological variables, in the preparation of a product for assessing the risk of glioma, wherein the clinicopathological variables include tumor grade, age, and IDH status.

[0026] In this invention, the tumor grading is classified into G2, G3, and G4 according to the WHO method. The age is classified into ≤40 years, >40 years and ≤60 years, and >60 years; the IDH status is classified into IDH mutant type and IDH wild type.

[0027] This invention provides a system for assessing the risk of glioma using the above-mentioned combination of gene markers, the system comprising a data detection unit, a data processing unit, and a result determination unit; The data detection unit detects the expression level of each gene in the above-mentioned gene marker combination in the sample to be tested. The data processing unit: constructs a risk scoring model using the above-mentioned gene marker combination, and substitutes the expression levels of each gene obtained by the data detection unit into the risk scoring model to calculate the risk score; The result determination unit receives the risk score obtained by the data processing unit and compares it with the median to determine the risk level of glioma in the tested sample.

[0028] In this invention, the risk scoring model = ∑(gene expression level × gene coefficient); the gene is the combination of the gene markers mentioned above. As a preferred embodiment, the risk scoring model = RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748. The median is obtained by sorting all glioma samples according to the risk scores obtained from the above risk scoring model from smallest to largest.

[0029] In this invention, when the risk score of the sample to be tested is greater than or equal to the median, the sample to be tested is a high-risk patient for glioma; when the risk score of the sample to be tested is less than the median, the sample to be tested is a low-risk patient for glioma.

[0030] This invention provides a system for assessing the risk or survival probability of glioma using a combination of gene markers and clinicopathological variables. The system includes a data acquisition unit, a data processing unit, and a result determination unit. The data acquisition unit is used to acquire the tumor grade, age, IDH status, and aging-related risk score of the sample to be tested. The method for obtaining the aging-related risk score is as follows: a risk score model is constructed using the above-mentioned combination of gene markers, the expression level of each gene in the above-mentioned combination of gene markers in the sample to be tested is detected, and the obtained expression level of each gene is substituted into the risk score model to calculate the risk score. The data processing unit: establishes a glioma risk assessment model based on the obtained tumor grade, age, IDH status, and aging-related risk score; constructs a nomogram of the total score of the glioma risk assessment model versus the survival probability; and uses this nomogram to predict the glioma risk level or survival probability of glioma patients. The result determination unit receives the glioma risk score obtained by the data processing unit and uses a nomogram to determine the survival probability, thereby determining the glioma risk level or survival probability of the tested sample.

[0031] In this invention, the glioma risk assessment model = tumor grade score + age score + IDH status score + risk score. When calculating the glioma risk assessment model score, tumor grade G2 is assigned a value of 0, tumor grade G3 is assigned a value of 52.5, and tumor grade G4 is assigned a value of 100; age ≤ 40 years is assigned a value of 0, age > 40 years and ≤ 60 years is assigned a value of 67, and age > 60 years is assigned a value of 98; IDH status IDH mutant is assigned a value of 0, and IDH status IDH wild-type is assigned a value of 53; the risk score is calculated from the above risk score model to determine the level of risk, with low risk assigned a value of 0 and high risk assigned a value of 54. The survival probability is one or more of the following: 1-year survival probability, 3-year survival probability, and 5-year survival probability.

[0032] This invention constructs a nomogram based on tumor grade, age, IDH status, and aging-related risk scores. After assigning values ​​to each variable in the glioma risk assessment model, a total score for the glioma risk assessment model is obtained. Based on the total score of the glioma risk assessment model, the 1-year survival probability, the 3-year survival probability, and the 5-year survival probability, a nomogram is constructed. By drawing vertical lines on the total score of the glioma risk assessment model, the 1-year survival probability, the 3-year survival probability, and the 5-year survival probability can be obtained on the nomogram, thereby statistically analyzing the glioma risk level or survival probability of glioma patients.

[0033] In this invention, unless otherwise specified, all raw material components are commercially available products well known to those skilled in the art.

[0034] The technical solutions provided by the present invention will be described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0035] Example 1 Construction of prognostic prediction model 1.1.1 Data Acquisition and Preprocessing TCGA database (https: / / portal.gdc.cancer.gov): Download all glioma RNA-seq data and clinical information.

[0036] CGGA database (http: / / www.cgga.org.cn): Download mRNAseq_693 expression profile data and related clinical data.

[0037] HAGR database (https: / / genomics.senescence.info / ): Obtain all (279) aging-related genes through the HAGR database.

[0038] In R, the limma package was used for data standardization and background correction to remove samples with incomplete clinical information (such as missing survival time, survival status, and molecular characteristics) to ensure the accuracy of subsequent modeling and analysis.

[0039] 1.1.2 Differential Expression Analysis (DEGs Screening) Differential genes between LGG and GBM were calculated using the limma package. Screening criteria: |LogFC|>1; corrected p-value<0.05.

[0040] Results of significantly upregulated and significantly downregulated genes in differentially expressed genes in TCGA or CGGA datasets are shown below. Figure 1 A~B in the diagram.

[0041] Depend on Figure 1 The results from C indicate that 29 aging-related differentially expressed genes were jointly identified in TCGA and CGGA. Cross-linking the HAGR gene set yielded 29 aging-related differentially expressed genes (TACC3, PITX1, NTN4, GTSE1, PGR, ORC1, BIRC5, CLSPN, CDC45, CDC6, MYBL2, KIF20A, RBP1, TGFBI, TWIST1, GADD45G, ESPL1, DUSP6, TRPM8, TAGLN, BUB1B, CHAF1B, ATF3, NUAK2, GRIK2, WEE1, RRM2, SOX1, and KCNJ12).

[0042] Figure 1 The results of the D study showed that the hierarchical clustering heatmap based on the above 29 genes clearly distinguished between low-grade glioma (LGG) and glioblastoma (GBM) samples.

[0043] 1.1.3 Model Construction Using Cox Regression and LASSO Regression Univariate Cox regression analysis (p<0.001): Based on the 29 aging-related genes screened in section 1.1.2 above, univariate Cox regression analysis was performed to further screen genes significantly associated with survival. The `survival` package in R was used to perform regression analysis on each gene individually to determine survival time and survival status, and to assess whether its expression level was significantly associated with patient prognosis. The results showed that all 29 genes were significantly associated with survival.

[0044] LASSO Regression (λ-Optimal): Based on the 29 genes selected through univariate Cox regression, the optimal regularization parameter λ (lambda.min) was selected in LASSO regression using 10-fold cross-validation. Using the model corresponding to λ, only genes with non-zero regression coefficients were retained, ultimately resulting in 12 key genes (NTN4, PGR, BIRC5, RBP1, TWIST1, GADD45G, TRPM8, NUAK2, GRIK2, WEE1, RRM2, and SOX1). These 12 genes represent the subset of the 29 candidate genes that contribute most to survival prediction.

[0045] Multivariate Cox regression analysis: From the 12 key genes screened by LASSO regression, genes with independent predictive value for survival were further screened. The expression levels of the 12 LASSO-screened genes (as covariates), survival time, and survival status (as dependent variables) were calculated. The regression coefficient (coef), hazard ratio (HR), and p-value were calculated for each gene; genes with p < 0.05 and coef ≠ 0 were retained as independent prognostic factors, ultimately resulting in 8 independent prognostic genes (NTN4, RBP1, TWIST1, GADD45G, NUAK2, GRIK2, WEE1, and RRM2).

[0046] Risk score calculation formula: Risk score = ∑(gene expression level × gene coefficient); Note: Risk score: a weighted total score used in tumor prognostic models to assess patient risk levels, integrating the influence of multiple genes into a single value to facilitate high / low risk stratification of patients. Gene expression level: represents the expression intensity of a gene in a sample. Gene coefficient: the regression coefficient of each gene obtained through Cox regression analysis, reflecting the strength and direction of the gene expression level's influence on prognosis (positive numbers represent risk factors, negative numbers represent protective factors). ∑: summation symbol, representing the sum of the products of all gene expression levels × regression coefficients, i.e., a "weighted sum" over all genes in the model.

[0047] Based on 612 glioma samples from the TCGA database, the total number of samples was randomly divided into a training set (n=428) and a test set (n=184). The training set was used for model training, hyperparameter tuning, and model performance monitoring. The test set was used for model performance evaluation in section 1.2. The specific risk score calculation formula based on the test set is as follows: Risk score = RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748.

[0048] The risk scoring model is now complete. Based on 612 glioma samples from the TCGA database, all patients were sorted by risk score from lowest to highest, and divided into high-risk and low-risk groups using the median score as the dividing line (median -0.420488607).

[0049] 1.2 Performance Evaluation of the Model The performance of the 8-gene risk scoring model constructed in section 1.1.3 was systematically evaluated in the first and second cohorts using Kaplan-Meier survival analysis and ROC curve evaluation. Statistical modeling and visualization were performed using the "survival" and "survminer" packages in R, with a significance threshold set at p < 0.05. The AUC value of the ROC curve was used to measure the model's predictive accuracy, where AUC > 0.9 indicates excellent predictive performance, 0.8–0.9 indicates good predictive ability, and 0.7–0.8 indicates acceptable discriminative power.

[0050] The first cohort consists of 612 glioma samples from the TCGA database, which are randomly divided into a training set (n=428) and a test set (n=184). The test set is used for model performance evaluation.

[0051] The second cohort consists of 657 glioma samples from the CGGA database.

[0052] Meanwhile, the risk scoring model constructed according to section 1.1.3 is as follows: Risk Score = RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748.

[0053] All glioma patients in the first and second cohorts were grouped into high-risk and low-risk groups: (1) Sort all glioma patients in the first cohort according to their risk scores from smallest to largest. Using the median as the boundary, if a patient's risk score is greater than or equal to the median, the patient is considered a high-risk patient. If a patient's risk score is less than the median, the patient is considered a low-risk patient. Thus, all glioma patients in the first cohort are divided into high-risk and low-risk groups. The calculated median for the first cohort is -0.420488607.

[0054] (2) Sort all glioma patients in the second cohort according to their risk scores from smallest to largest. Using the median as the boundary, if a patient's risk score is greater than or equal to the median, the patient is considered a high-risk patient. If a patient's risk score is less than the median, the patient is considered a low-risk patient. Thus, all glioma patients in the second cohort are divided into high-risk and low-risk groups. The calculated median for the second cohort is -0.111408085.

[0055] Figure 2 The results showed that in the first cohort, Kaplan-Meier survival curves indicated that patients in the high-risk group had significantly lower overall survival than those in the low-risk group across the training set, test set, and the entire dataset (p<0.001). The ROC curves and AUC values ​​of the risk model for 1-year, 3-year, and 5-year survival predictions across the training set, test set, and the entire dataset demonstrated high predictive accuracy. Specifically, the AUC values ​​for the 1-year, 3-year, and 5-year survival ROC curves across the entire dataset were 0.870, 0.904, and 0.853, respectively. Risk score distribution plots and survival status plots showed that patients in the high-risk group across the training set, test set, and the entire dataset had higher risk scores, shorter survival times, and higher mortality rates.

[0056] Figure 3 The results showed that in the second cohort, Kaplan-Meier survival curves indicated that the survival time of patients in the high-risk group was significantly shorter than that in the low-risk group (HR=3.72, p<0.001). The ROC curves of the risk scoring model for the 1-year, 3-year, and 5-year survival data in the second cohort dataset showed AUC values ​​of 0.750, 0.788, and 0.785, respectively, suggesting good predictive performance. Risk score distribution plots and survival status plots showed that patients in the high-risk group had higher risk scores, shorter survival times, and higher mortality rates.

[0057] In summary, the risk scoring model, which categorizes individuals into high-risk and low-risk groups in the first or second cohort, aligns with actual measurement results. This invention demonstrates good sensitivity and specificity in predicting glioma risk and risk stratification using the risk scoring model. Furthermore, the risk scoring model exhibits stable survival prediction capabilities across multiple cohorts and time points, demonstrating significant clinical applicability and potential for wider application.

[0058] 1.3 Construction and Validation of Prognostic Nodal Chart Univariate Cox proportional hazards regression model was used to screen all variables. Variables with p < 0.10 in the univariate Cox regression analysis (the variables being tumor grade, age, gender, IDH status, and aging-related risk score) were used as candidate variables, while variables with p < 0.05 in the multivariate Cox regression analysis were ultimately used to establish the glioma risk assessment model.

[0059] Figure 4 The results in section A indicate that tumor grade, age, IDH status, and aging-related risk score were included in the multivariate Cox regression analysis. The aging-related risk score is the risk score model constructed in section 1.1.3: Risk score = RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748.

[0060] Based on a multivariate Cox proportional hazards regression model, this embodiment establishes a glioma risk assessment model by combining the risk scoring model constructed in section 1.1.3 above with clinicopathological variables. The score includes four variables: tumor grade, age, IDH status, and aging-related risk score (where tumor grade is divided into G2, G3, and G3 according to the WHO method; age is divided into ≤40 years, >40 years and ≤60 years, and >60 years; IDH status is divided into IDH mutant and IDH wild type; the aging-related risk score is sorted from smallest to largest according to the risk scores constructed in section 1.1.3 for all glioma patients, with the median as the boundary. When a patient's risk score is ≥ the median, the patient is a high-risk patient; when a patient's risk score is < the median, the patient is a low-risk patient. In this way, all glioma patients are divided into high-risk and low-risk groups). The glioma risk assessment model score is calculated as follows: Tumor grade score + Age score + IDH status score + Risk score. Specifically, in calculating this glioma risk assessment model score, a tumor grade of G2 is assigned a value of 0, a tumor grade of G3 is assigned a value of 52.5, and a tumor grade of G4 is assigned a value of 100; age ≤ 40 years is assigned a value of 0, age > 40 years ≤ 60 years is assigned a value of 67, and age > 60 years is assigned a value of 98; IDH status of IDH mutant is assigned a value of 0, and IDH status of IDH wild-type is assigned a value of 53; a low-risk risk score is assigned a value of 0, and a high-risk risk score is assigned a value of 54.

[0061] Based on a prognostic model derived from multivariate Cox regression, a nomogram tool was constructed to predict the 1-year, 3-year, and 5-year survival rates of glioma patients by integrating multiple clinicopathological variables (age, tumor grade, IDH status, and aging-related risk score). This model was implemented using the "rms" package in R and systematically validated using calibration curves and ROC curves.

[0062] according to Figure 4 The total score is vertically lined in Figure B to analyze the 1-year, 3-year, and 5-year survival probabilities of the glioma risk assessment model.

[0063] Figure 4 The results showed that multivariate Cox regression analysis revealed tumor grade, age, IDH status, and risk score as independent prognostic factors. Nomograms constructed from the Cox-screened model, integrating clinicopathological factors and risk scores, could predict patients' 1-year, 3-year, and 5-year survival probabilities. Calibration curves of the nomograms predicting survival showed a high degree of consistency between the predicted results and actual observations, demonstrating good calibration capability. Time-based ROC curve analysis showed AUC values ​​of 0.882, 0.934, and 0.902 for 1-year, 3-year, and 5-year survival predictions, respectively, indicating that the model has extremely high predictive accuracy.

[0064] In summary, the model's predicted values ​​are highly consistent with the actual observed values ​​(C-index of 0.853, ROC and AUC both > 0.88), demonstrating good discriminative power and practicality. This nomogram provides clinicians with a visual and individualized survival assessment tool, which can assist in precise treatment decisions.

[0065] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A combination of gene markers for predicting the risk of glioma, characterized in that, The combination of gene markers includes the NTN4 gene, RBP1 gene, TWIST1 gene, GADD45G gene, NUAK2 gene, GRIK2 gene, WEE1 gene, and RRM2 gene.

2. The use of a reagent for detecting the expression level of the combination of gene markers for predicting glioma risk as described in claim 1 in the preparation of products for assessing glioma risk.

3. The application of a reagent for detecting the expression level of the gene marker combination described in claim 1, combined with a product for detecting clinicopathological variables, in the preparation of a product for assessing the risk of glioma, characterized in that... The clinicopathological variables include tumor grade, age, and IDH status.

4. A system for assessing the risk of glioma using the combination of gene markers described in claim 1, characterized in that, The system includes a data detection unit, a data processing unit, and a result determination unit; The data detection unit detects the expression level of each gene in the gene marker combination of claim 1 in the sample to be tested. The data processing unit: constructs a risk scoring model using the gene marker combination described in claim 1, and substitutes the expression levels of each gene obtained by the data detection unit into the risk scoring model to calculate the risk score; The result determination unit receives the risk score obtained by the data processing unit and compares it with the median to determine the risk level of glioma in the tested sample.

5. The system according to claim 4, characterized in that, The risk scoring model is defined as ∑(gene expression level × gene coefficient); the gene is the combination of gene markers as described in claim 1.

6. The system according to claim 5, characterized in that, The risk scoring model is calculated as follows: RBP1 gene expression level × 0.001932862 - NTN4 gene expression level × 0.010305105 + TWIST1 gene expression level × 0.009929055 - GADD45G gene expression level × 0.001968495 + NUAK2 gene expression level × 0.034256348 - GRIK2 gene expression level × 0.014378515 + WEE1 gene expression level × 0.013430473 + RRM2 gene expression level × 0.017691748.

7. The system according to claim 4, characterized in that, The median is obtained by sorting all glioma samples in ascending order of risk scores obtained from the risk scoring model described in claim 5 or 6.

8. The system according to claim 4, characterized in that, If the risk score of the sample to be tested is greater than or equal to the median, the sample is considered a high-risk patient for glioma; if the risk score of the sample to be tested is less than the median, the sample is considered a low-risk patient for glioma.

9. A system for assessing the risk or survival probability of glioma using a combination of gene markers and clinicopathological variables, characterized in that, The system includes a data acquisition unit, a data processing unit, and a result determination unit; The data acquisition unit is used to acquire the tumor grade, age, IDH status, and aging-related risk score of the sample to be tested. The method for obtaining the aging-related risk score is as follows: a risk score model is constructed using the gene marker combination described in claim 1, the expression level of each gene in the gene marker combination described in claim 1 is detected in the sample to be tested, and the obtained expression level of each gene is substituted into the risk score model to calculate the risk score. The data processing unit: establishes a glioma risk assessment model based on the obtained tumor grade, age, IDH status, and aging-related risk score; constructs a nomogram of the total score of the glioma risk assessment model versus the survival probability; and uses this nomogram to predict the glioma risk level or survival probability of glioma patients. The result determination unit receives the glioma risk score obtained by the data processing unit and uses a nomogram to determine the survival probability, thereby determining the glioma risk level or survival probability of the tested sample.

10. The system according to claim 9, characterized in that, The glioma risk assessment model is defined as follows: tumor grade score + age score + IDH status score + risk score. In calculating this risk assessment model, tumor grade G2 is assigned a value of 0, tumor grade G3 is assigned a value of 52.5, and tumor grade G4 is assigned a value of 100; age ≤ 40 years is assigned a value of 0, age > 40 years ≤ 60 years is assigned a value of 67, and age > 60 years is assigned a value of 98; IDH status IDH mutant is assigned a value of 0, and IDH status IDH wild-type is assigned a value of 53; the risk score is calculated using the risk scoring model described in any one of claims 5-8, with low risk assigned a value of 0 and high risk assigned a value of 54. The survival probability is one or more of the following: 1-year survival probability, 3-year survival probability, and 5-year survival probability.

Images