Method and system for constructing renal clear cell carcinoma prognosis model based on metabolic gene set

Abstract

The invention belongs to the technical field of biomedicine, and particularly relates to a renal clear cell carcinoma prognosis model construction method and system based on a metabolic gene set. The system comprises: a metabolic gene set data acquisition module for acquiring metabolic gene set data of a patient with renal clear cell carcinoma, the metabolic gene set data including CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1; the survival time calculation module is used for calculating the survival time of the renal clear cell carcinoma patient based on the prediction model; the prediction model comprises a formula I for calculating the risk score of the patient and a formula II for calculating the survival probability of the patient at a certain time point; and the result output module is used for outputting a prognosis result of the patient based on the survival time of the renal clear cell carcinoma patient. The system can significantly improve the accuracy and reliability of prognosis of ccRCC patients, and has wide application prospects and commercial values.

Description

Technical Field

[0001] This invention belongs to the field of biomedical technology, specifically relating to a method and system for constructing a survival prediction model for clear cell renal cell carcinoma based on metabolic gene sets. Background Technology

[0002] Clear cell renal cell carcinoma (ccRCC) is the most common type of renal cell carcinoma, and its prognosis varies significantly among individual patients. Despite significant advances in molecular biology research in recent years, accurately predicting the survival time of ccRCC patients remains a challenge. Changes in metabolic pathways are closely related to the occurrence and progression of renal cell carcinoma; therefore, constructing predictive models based on the expression of metabolism-related genes can provide important insights into the prognosis of ccRCC patients. Summary of the Invention

[0003] This invention proposes a method and system for constructing a prognostic model for clear cell renal cell carcinoma based on a set of metabolic genes, in order to predict the survival time of patients with clear cell renal cell carcinoma. This invention analyzes the gene expression data of patients, selects key genes related to metabolism, and constructs a machine learning model to accurately predict the survival time of patients.

[0004] The present invention specifically adopts the following technical solution:

[0005] This invention provides a prognostic system for clear cell renal cell carcinoma based on a set of metabolic genes, the system comprising:

[0006] The metabolic gene set data acquisition module is used to acquire metabolic gene set data of patients with clear cell renal cell carcinoma. The metabolic gene set includes: CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1.

[0007] The survival time calculation module calculates the survival time of patients with clear cell renal cell carcinoma based on a predictive model.

[0008] The prediction model includes Equation I for calculating the patient's risk score and Equation II for calculating the patient's survival probability at a certain time point.

[0009] Risks core=-0.56814*e1+1.67719*e2+1.84934*e3-0.79149*e4-0.77888*e5-0.72417*e6-0.69921*e7-0.75221*e8+1.17177*e9-0.76067*e10 (Formula I);

[0010] In Formula I, e1 represents the gene expression level of CYP3A7;

[0011] e2: The gene expression level of ST3GAL5;

[0012] e3: Gene expression level of DBH;

[0013] e4: Gene expression level of UGT2B7;

[0014] e5: Gene expression level of GCNT4;

[0015] e6: Gene expression level of LIPA;

[0016] e7: ITPKB gene expression level;

[0017] e8: Gene expression level of HS3ST1;

[0018] e9: GYG2 gene expression level;

[0019] e10: Gene expression level of CYP51A1.

[0020] S(t) = exp(-Riskscore*365t) (Equation II), where t represents a point in time and the unit is years.

[0021] The results output module outputs the prognostic results of the patients based on their survival time in the clear cell renal cell carcinoma.

[0022] This invention also provides a method for constructing a prognostic model of clear cell renal cell carcinoma based on a set of metabolic genes, comprising the following steps:

[0023] Step 1: Data Collection and Preprocessing

[0024] Gene expression profile data and survival time data of patients with clear cell renal cell carcinoma were collected from publicly available databases. The gene expression profile data and survival time data were quality controlled and then standardized.

[0025] Step 2: Identify the differentially expressed gene set that shows significant changes compared to the control group (normal sample) in clear cell renal cell carcinoma samples (tumor group):

[0026] Using the DESeq2 software, gene expression profiles from the tumor group (535 cases) and the control group (72 cases) from the TCGA public database were compared. The differentially expressed genes with significant changes in the tumor group (|log2(Fold-change)>0.5|, adjust p-value<0.05, baseMean>50) were calculated and analyzed, resulting in a total of 7,118 differentially expressed genes in the tumor group.

[0027] Step 3: Identify the set of genes related to metabolism:

[0028] The gene set includes 1,582 metabolism-related genes from 85 KEGG metabolic pathway gene sets.

[0029] Step 4: Select the genes most valuable for survival time prediction through feature selection:

[0030] Using a feature selection algorithm, the most valuable genes for survival prediction are screened from the significantly changed metabolic-related candidate genes obtained in step 4 of the tumor group. Specific steps include:

[0031] Step 4.1: The intersection of the differentially expressed gene set from Step 2 and the metabolism-related gene set from Step 3 was used to obtain candidate metabolism-related genes that showed significant changes in the tumor group. This step identified a total of 768 abnormally expressed metabolic genes in tumors.

[0032] Step 4.2: Using Lasso regression, further screened the candidate genes to identify those significantly associated with survival time. This step identified 30 genes, including: ACHE, ADCY2, ASNS, B3GALT2, CHPT1, CYP21A2, CYP2C9, CYP3A7, CYP51A1, DBH, DMGDH, FUCA1, GCNT4, GYG2, HK3, HS3ST1, HS6ST2, ITPKA, ITPKB, LIPA, PANK1, PIGZ, POFUT2, PYCR1, RIMKLA, SCD5, SGMS1, ST3GAL5, UGT1A10, and UGT2B7.

[0033] Step 4.3: Univariate Cox regression analysis was used to further screen for genes significantly associated with survival time. This step identified 23 genes significantly associated with survival time, including: ACHE, ASNS, CYP2C9, CYP3A7, CYP51A1, DBH, DMGDH, GCNT4, GYG2, HK3, HS3ST1, ITPKA, ITPKB, LIPA, PANK1, PIGZ, POFUT2, PYCR1, SCD5, SGMS1, ST3GAL5, UGT1A10, and UGT2B7.

[0034] Step 5: Model Training

[0035] The model is trained using machine learning algorithms. The specific steps are as follows:

[0036] Step 5.1: Set the TCGA-KIRC dataset as the training set.

[0037] Step 5.2: Train the model using the training set. Input the 23 variables and their survival time and survival status obtained in the above steps as data. Use the My.stepwise.coxph function of the My.stepwise package to iterate the data multiple times to obtain the optimal multifactor Cox model combination. The variables of the optimal model combination include: CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1, and obtain the parameters of the multifactor Cox regression model.

[0038] Step 5.3: Build a multifactor Cox regression model using the variables of the selected optimal model combination.

[0039] Step 5.4: Evaluate the model performance on the training set. Commonly used evaluation metrics include C-index, AIC (Akaike information criterion), and AUC (Area under curve).

[0040] Step 6: Model Validation

[0041] The generalization ability of the model was verified using an independent validation dataset (E_MTAB_1980) to ensure the reliability and stability of the model in practical applications.

[0042] On the other hand, the present invention provides an electronic device comprising:

[0043] The system includes a memory and a processor, which communicate with each other via a bus. The memory stores program instructions that can be executed by the processor, which can then call these instructions to execute the aforementioned prognostic system for clear cell renal cell carcinoma based on a set of metabolic genes.

[0044] In another aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a prognostic system for clear cell renal cell carcinoma based on a set of metabolic genes.

[0045] The beneficial effects of this invention are as follows:

[0046] The survival prediction model for clear cell renal cell carcinoma based on metabolic gene sets constructed in this invention has the following advantages:

[0047] 1. High accuracy: By using expression data of metabolism-related genes, the model can accurately predict the survival time of ccRCC patients.

[0048] 2. High degree of automation: The model can process and analyze gene expression data of patients with clear cell renal cell carcinoma, reducing human error and subjective judgment.

[0049] 3. Clinical applicability: The model can be integrated into existing clinical decision support systems to provide doctors with real-time and reliable references.

[0050] In summary, this invention provides an innovative survival prediction system for clear cell renal cell carcinoma (ccRCC) based on metabolic gene sets, which can significantly improve the accuracy and reliability of prognosis for ccRCC patients and has broad application prospects and commercial value. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the construction of a prognostic model for clear cell renal cell carcinoma based on a set of metabolic genes.

[0052] Figure 2 A diagram showing the construction of a ccRCC risk prediction model for the metabolic gene set. A: Dynamic process diagram of LASSO selection variables; lambda selection variables were determined through 10-fold cross-validation, and λ with the smallest average error was selected as optimal. B: Forest plot of hazard ratios from the Cox multivariate regression model. C: AUC curves of 1-, 3-, 5-, and 10-year survival predicted by the multivariate Cox regression model. D: Calibration curves of 1-, 3-, 5-, and 10-year survival predicted by the multivariate Cox regression model. E: Kaplan-Meier survival curves for the high-risk and low-risk ccRCC groups.

[0053] Figure 3 The expression of model genes in the tumor group and the control group. Differences in the expression of AJ: CYP3A7(A), ST3GAL5(B), DBH(C), UGT2B7(D), GCNT4(E), LIPA(F), ITPKB(G), HS3ST1(H), GYG2(I), CYP51A1(J) between the tumor group and the control group.

[0054] Figure 4 The correlation matrix diagram among the model genes shows that there is no collinearity among the model features.

[0055] Figure 5 The nomograph, which is a model variable, can be scored based on the gene expression information of ccRCC patients, and the survival probability of patients in 1 year, 3 years, 5 years and 10 years can be predicted.

[0056] Figure 6Kaplan-Meier survival curves of model genes in ccRCC. Kaplan-Meier survival curves of AJ: CYP3A7(A), ST3GAL5(B), DBH(C), UGT2B7(D), GCNT4(E), LIPA(F), ITPKB(G), HS3ST1(H), GYG2(I), CYP51A1(J) in high-expression and low-expression groups.

[0057] Figure 7 The expression levels of model genes in the high-risk and low-risk groups of ccRCC are shown. AJ: Expression levels of CYP3A7(A), ST3GAL5(B), DBH(C), UGT2B7(D), GCNT4(E), LIPA(F), ITPKB(G), HS3ST1(H), GYG2(I), and CYP51A1(J) in the high-risk and low-risk groups.

[0058] Figure 8 A: AUC curves of the multivariate Cox regression model at 1, 3, 5, and 10 years on the validation dataset. B: Kaplan-Meier survival curves of the model on the validation dataset. C: Calibration curves of the model's predicted 1-year, 3-year, and 5-year survival on the validation dataset. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of the embodiments of this invention will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0060] Example 1: Establishment and validation of a survival prediction model for clear cell renal cell carcinoma based on metabolic gene sets. Figure 1 As shown, it includes:

[0061] 1. Data Collection and Preprocessing

[0062] The data collection and preprocessing process is as follows:

[0063] Step 1: Collect gene expression profiles and survival data from 535 ccRCC patients and 72 normal controls from the TCGA public database (https: / / xenabrowser.net / datapages / ).

[0064] Step 2: Extract mRNA genes from the gene expression profile data, and remove lncRNAs and miRNAs from the gene expression profile. Remove genes with low expression in the samples, and retain genes that are expressed in at least three samples and whose original expression readings are not less than 10.

[0065] Step 3: Standardize the gene expression profile data, specifically using the Variance Stabilization Transform (VST) function in the DESeq2 software.

[0066] Step 4: Use DESeq2 software to analyze differentially expressed genes that showed significant changes compared to the control group in ccRCC samples (|log2(Fold-change)|>0.5, adjust p-value<0.05, baseMean>50), resulting in a total of 7,118 tumor differentially expressed genes.

[0067] 2. Model inclusion gene set screening

[0068] Download the KEGG metabolic pathway gene set "KEGG_metabolism_nc.gmt"

[0069] (https: / / github.com / wu-yc / scMetabolism / tree / main / data) contains 85 sets of genes related to metabolic pathways, and the resulting gene set contains a total of 1,582 metabolic-related genes.

[0070] 3. Screening of candidate metabolic genes with significant changes in tumors

[0071] By taking the intersection of differentially expressed tumor genes obtained from ccRCC compared with the control group and the set of metabolic pathway genes, 768 metabolic-related candidate genes that showed significant changes in tumors were obtained.

[0072] 4. Feature selection is used to screen genes that are most valuable for predicting survival time.

[0073] Feature selection algorithms (including Lasso regression, univariate regression analysis, and multivariate regression analysis) are used to screen genes from the set of metabolic genes that are most valuable for predicting survival time. Specific steps include:

[0074] Step 4.1: As Figure 2Lasso regression was used to screen for genes significantly associated with survival time from the candidate genes. The goal of the regression was to minimize deviance, compressing the coefficients of unimportant variables to 0 and retaining a small number of important variables with non-zero coefficients. This step screened 30 genes, including: ACHE, ADCY2, ASNS, B3GALT2, CHPT1, CYP21A2, CYP2C9, CYP3A7, CYP51A1, DBH, DMGDH, FUCA1, GCNT4, GYG2, HK3, HS3ST1, HS6ST2, ITPKA, ITPKB, LIPA, PANK1, PIGZ, POFUT2, PYCR1, RIMKLA, SCD5, SGMS1, ST3GAL5, UGT1A10, and UGT2B7.

[0075] Step 4.2: Using univariate Cox regression modeling, calculate the hazard ratios of metabolic genes and test whether the variables meet the PH assumptions of the Cox proportional hazards model. Genes that meet the conditions of z_pvalue < 0.05 and the PH assumptions (cox.zph test ph_hypothesis_p > 0.05) are the genes screened that are significantly associated with survival time. This step screened 23 genes that are significantly associated with survival time, including: ACHE, ASNS, CYP2C9, CYP3A7, CYP51A1, DBH, DMGDH, GCNT4, GYG2, HK3, HS3ST1, ITPKA, ITPKB, LIPA, PANK1, PIGZ, POFUT2, PYCR1, SCD5, SGMS1, ST3GAL5, UGT1A10, and UGT2B7.

[0076] 5. Model Training:

[0077] The model was trained using a machine learning algorithm (Cox proportional hazards model). The specific steps are as follows:

[0078] Step 5.1: Set the TCGA-KIRC dataset (i.e., the data in Step 1) as the training set.

[0079] Step 5.2: Train the model using the training set. Input the 23 variables and their survival times and states obtained in the previous steps as data. Use the My.stepwise.coxph function of the My.stepwise package to iterate the model data multiple times to obtain the optimal multifactor Cox regression model combination. The optimal model combination includes 10 variables (i.e., 10 model genes): CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1. Obtain the parameters of the multifactor Cox regression model.

[0080] Step 5.3: Establish a multifactor Cox regression model using the selected optimal combination of model variables. Model the model using the `cox.ph()` function of the `survival` package, and use `cox.zph()` to test the PH hypothesis for each candidate gene. Calculate the VIF and correlation coefficient for each factor in the regression model to determine potential collinearity among factors. Variables that pass the PH hypothesis and collinearity test are then modeled a second time. All candidate genes that passed the `cox.zph()` test in this step are included in the second modeling, including: CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1. Furthermore, there is no collinearity among the model genes. Figure 3 , 4 And 7.

[0081] Step 5.4: Detect collinearity among variables. A correlation coefficient less than 0.5 and a VIF square root less than 2 are used as the test method. Figure 3 .

[0082] Step 5.5: Plot the Cox regression forest plot. The forest plot includes the hazard ratios (HRs) and confidence intervals for each variable in the model. The model AIC value is 1783.92, and the concordance index is 0.77. The results show that ST3GAL5, DBH, and GYG2 are independent significant risk factors, and CYP3A7, UGT2B7, GCNT4, LIPA, ITPKB, and HS3ST1 are independent significant protective factors. Furthermore, the entire Cox model is statistically significant (p = 1.3551e). -31 ),like Figure 2 B.

[0083] Step 5.6: Plot the model ROC curve and calculate the AUC. Use the timeROC software package to calculate the survival curves and their AUC values ​​for 1 year, 3 years, 5 years and 10 years, where the predictive power for 1 year, 3 years, 5 years and 10 years are 0.819, 0.801, 0.837 and 0.872, respectively.

[0084] Step 5.7: Plot a nomograph based on the constructed model parameters. Gene expression levels in ccRCC patients can be used to score the nomograph and predict their 1-year, 3-year, 5-year, and 10-year survival probabilities. Figure 5 .

[0085] Step 5.8: Kaplan-Meier Survival Analysis. Kaplan-Meier analysis was used to compare the survival status between the two groups. The final model was used to predict the hazard scores of the original data. Based on the median hazard scores, the samples were divided into high-risk and low-risk groups. Kaplan-Meier analysis was performed using the survminer package and visualized. Significant differences were observed between the high-risk and low-risk groups. Figure 6 5. Model Validation

[0086] The model's generalization ability was validated using an independent validation dataset (E_MTAB_1980, https: / / ega-archive.org / studies / EGAS00001000509) to ensure its reliability and stability in practical applications. Results show that the AUC for predicting the model's overall survival over 1, 3, 5, and 10 years reaches 0.848, 0.825, 0.787, and 0.788, respectively, demonstrating the model's good generalization ability and suitability for widespread application. Figure 8 .

[0087] Furthermore, comparing this model with existing prediction methods (application publication number CN 115798703 A, "Device and Computer-Readable Storage Medium for Predicting Prognosis of Renal Clear Cell Carcinoma Based on Novel Fatty Acid Metabolism-Related Genes," with 1-year, 3-year, and 5-year overall survival AUCs of 0.763, 0.736, and 0.749, respectively), this invention uses the same number of genes and exhibits higher predictive efficacy for 1-year, 3-year, and 5-year overall survival, demonstrating that the prediction model of this invention is more accurate and superior to fatty acid metabolism gene models in predicting 1-year, 3-year, and 5-year overall survival. This proves the superiority of the model of this invention.

[0088] 6. Model Application

[0089] By integrating the trained model into a clinical decision support system, patients' survival time can be predicted by inputting their gene expression data.

[0090] First, calculate the Riskscore using the following formula;

[0091] Risk score=-0.56814*e1+1.67719*e2+1.84934*e3-0.79149*e4-0.77888*e5-0.72417*e6-0.69921*e7-0.75221*e8+1.17177*e9-0.76067*e10.

[0092] e1: Gene expression level of CYP3A7;

[0093] e2: The gene expression level of ST3GAL5;

[0094] e3: Gene expression level of DBH;

[0095] e4: Gene expression level of UGT2B7;

[0096] e5: Gene expression level of GCNT4;

[0097] e6: Gene expression level of LIPA;

[0098] e7: ITPKB gene expression level;

[0099] e8: Gene expression level of HS3ST1;

[0100] e9: GYG2 gene expression level;

[0101] e10: Gene expression level of CYP51A1.

[0102] Then, the survival probability of a patient at a certain time point can be calculated using the formula S(t)=exp(-Riskscore*365t), where t represents the time point and the unit is years.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A prognostic system for clear cell renal cell carcinoma based on metabolic gene sets, characterized in that, The system includes: The metabolic gene set data acquisition module is used to acquire metabolic gene set data of patients with clear cell renal cell carcinoma. The metabolic gene set includes: CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1. The survival time calculation module calculates the survival time of patients with clear cell renal cell carcinoma based on a predictive model; The prediction model includes Equation I for calculating the patient's risk score and Equation II for calculating the patient's survival probability at a certain time point. Riskscore=-0.56814*e1+1.67719*e2+1.84934*e3-0.79149*e4-0.77888*e5-0.72417*e6-0.69921*e7-0.75221*e8+1.17177*e9-0.76067*e10, Formula I, In Formula I, e1, e2, e3, e4, e5, e6, e7, e8, e9, and e10 represent the gene expression levels of CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1, respectively. S(t) = exp(-Riskscore*365t), Equation II, where t represents a point in time in years; The results output module outputs the prognostic results of the patients based on their survival time in the clear cell renal cell carcinoma.

2. A method for constructing a prognostic model for clear cell renal cell carcinoma based on a metabolic gene set, characterized in that, Includes the following steps: Step 1: Collect gene expression profile data and survival time data of patients with clear cell renal cell carcinoma from public databases, perform quality control on the gene expression profile data and survival time data, and then standardize the gene expression profile data; Step 2: Identify the differentially expressed gene set that shows significant changes compared to normal samples from clear cell renal cell carcinoma; Step 3: Identify the set of genes related to metabolism; Step 4: Select the genes most valuable for survival time prediction through feature selection; Step 5: Model training to obtain the optimal combination of multi-factor models and the parameters of the multi-factor regression model, and to establish the multi-factor regression model; Step 6: Model Validation.

3. The method for constructing a prognostic model of clear cell renal cell carcinoma based on a metabolic gene set according to claim 2, characterized in that, Step 3 is to obtain a set of metabolism-related genes from multiple KEGG metabolic pathway gene sets.

4. The method for constructing a prognostic model of clear cell renal cell carcinoma based on a metabolic gene set according to claim 2, characterized in that, Step 4 includes: Step 4.1: Take the intersection of the differentially expressed gene set from Step 2 and the metabolism-related gene set from Step 3 to obtain the metabolism-related candidate genes that show significant changes in clear cell renal cell carcinoma; Step 4.2: Using Lasso regression, further screen the candidate genes from Step 4.1 to identify genes that are significantly associated with survival time; Step 4.3: Use univariate Cox regression analysis to further screen for genes that are significantly associated with survival time.

5. The method for constructing a prognostic model of clear cell renal cell carcinoma based on a metabolic gene set according to claim 2, characterized in that, Step 5 involves training the model using machine learning algorithms, including: Step 5.1: Use the gene expression profile data of patients with clear cell renal cell carcinoma from the publicly available database TCGA in Step 1 as the training set; Step 5.2: Train the model using the training set. Take the gene variables, their survival time and survival status obtained in Step 4 as data input, perform multiple iterations on the data to obtain the optimal combination of multifactor Cox models, and obtain the parameters of the multifactor Cox regression model. Step 5.3: Build a multifactor Cox regression model using the variables of the selected optimal model combination; Step 5.4: Evaluate the model performance on the training set.

6. The method for constructing a prognostic model of clear cell renal cell carcinoma based on a metabolic gene set according to claim 5, characterized in that, In step 5.3, the variables of the optimal model combination include: CYP3A7, ST3GAL5, DBH, UGT2B7, GCNT4, LIPA, ITPKB, HS3ST1, GYG2, and CYP51A1.

7. The method for constructing a prognostic model of clear cell renal cell carcinoma based on a metabolic gene set according to claim 2, characterized in that, The model's generalization ability was verified using an independent validation dataset.

8. An electronic device, characterized in that, The electronic device includes: The system includes a memory and a processor, which communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor can execute the renal clear cell carcinoma prognostic system based on metabolic gene sets as described in claim 1 by calling the program instructions.

9. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the renal clear cell carcinoma prognostic system based on a metabolic gene set as described in claim 1.

Images