Tumor prognosis risk model, tumor prognosis risk determination method and electronic equipment

Abstract

The invention provides a tumor prognosis risk model determination method, a tumor prognosis risk determination method and electronic equipment. The method comprises the following steps: acquiring a data set of a tumor patient; determining gene expression information and progression-free lifetime of the tumor patient from the data set; combining the gene expression information and the progression-free lifetime as first input data, and inputting the first input data into a univariate proportional risk regression model to obtain a correlation value of each piece of gene expression information and a prognosis effect; selecting first gene expression information according to the correlation value, and determining a progress-free lifetime of the target from the data set according to the first gene expression information; determining second input data according to the first gene expression information and the target progression-free lifetime, and inputting the second input data into a multivariable proportional risk regression model to obtain a second expression gene and a target model coefficient; and constructing a tumor prognosis risk assessment model according to the second expression gene and the target model coefficient. The method can improve the accuracy of tumor prognosis risk assessment.

Description

Technical Field

[0001] The present application belongs to the biological field and relates to artificial intelligence technology, and in particular to a method for determining a tumor prognosis risk model, a method for determining a tumor prognosis risk, and an electronic device. Background Art

[0002] Diffuse large B-cell lymphoma (DLBCL) is a common lymphoma of intermediate to high malignancy with high invasiveness. Most patients can achieve long-term remission after treatment with standard R-CHOP (rituximab combined with cyclophosphamide, doxorubicin, vincristine and prednisone). However, approximately one-third of patients will experience refractory or relapsed disease. High-dose chemotherapy and autologous stem cell transplantation have been identified as second-line treatments for young patients with R / R disease, but only a few patients achieve long-term remission. Therefore, constructing a prognostic risk assessment model for lymphoma is of great significance for improving patient prognosis.

[0003] However, the current lymphoma prognostic risk assessment model has unclear training data on the time of recurrence and progression of patients, which leads to unclear prognostic criteria and poor accuracy of lymphoma prognostic risk assessment. Summary of the invention

[0004] In view of the above, it is necessary to propose a tumor prognosis risk model determination method, a tumor prognosis risk determination method and an electronic device, which can solve the problem of poor accuracy of tumor prognosis risk assessment.

[0005] The first aspect of an embodiment of the present application provides a method for determining a tumor prognosis risk model, the method comprising: obtaining a data set of tumor patients; determining gene expression information and progression-free survival of tumor patients from the data set; combining the gene expression information and the progression-free survival as first input data, and inputting them into a univariate proportional risk regression model to obtain a correlation value between each gene expression information and the prognostic effect; selecting first gene expression information according to the correlation value, and determining a target progression-free survival from the data set according to the first gene expression information; determining second input data according to the first gene expression information and the target progression-free survival, and inputting them into a multivariate proportional risk regression model to obtain a second expression gene and the target model coefficient; and constructing a tumor prognosis risk assessment model according to the second expression gene and the target model coefficient.

[0006] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application, the second expressed genes include: A4GALT gene, ATXN1 gene, DIAPH1 gene, EML1 gene, LMO2 gene, PBX4 gene, PLCL1 gene, SARDH gene, SGK1 gene, SLAMF1 gene and TNFRSF11B gene.

[0007] Further, in the above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application, the target model coefficient corresponding to the A4GALT gene is -0.116359642, the target model coefficient corresponding to the ATXN1 gene is -0.083380887, the target model coefficient corresponding to the DIAPH1 gene is -0.051781145, the target model coefficient corresponding to the EML1 gene is -0.055129472, and the target model coefficient corresponding to the LMO2 gene is -0.144639089. The target model coefficient corresponding to the PBX4 gene is -0.006071289, the target model coefficient corresponding to the PLCL1 gene is -0.283010482, the target model coefficient corresponding to the SARDH gene is -0.013092613, the target model coefficient corresponding to the SGK1 gene is -0.13304739, the target model coefficient corresponding to the SLAMF1 gene is -0.004109514, and the target model coefficient corresponding to the TNFRSF11B gene is -0.065225379.

[0008] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in an embodiment of the present application, the first gene expression information is selected according to the correlation value, and the target progression-free survival period is determined from the data set according to the first gene expression information, including: selecting gene expression information with a correlation value less than a preset correlation threshold as the first gene expression information; determining the tumor patient corresponding to the first gene expression information as the target tumor patient; and determining the progression-free survival period corresponding to the target tumor patient as the target progression-free survival period.

[0009] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in an embodiment of the present application, the tumor prognosis risk assessment model is constructed based on the second expressed gene and the target model coefficient, including: standardizing the second gene expression information corresponding to the second expressed gene to obtain third gene expression information; determining the product value of the third gene expression information and the corresponding target model coefficient; and obtaining the tumor prognosis risk assessment model based on the sum of multiple product values.

[0010] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in an embodiment of the present application, if the data set includes a first data set and a second data set, the first gene expression information is selected according to the correlation value, and the target progression-free survival period is determined from the data set according to the first gene expression information, including: gene differential analysis of the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set to obtain multiple first differentially expressed genes; selecting gene expression information with a correlation less than the preset correlation threshold as the first gene expression information; determining the first expression gene corresponding to the first gene expression information, and comparing the first expression gene with the first differentially expressed gene to obtain the second differentially expressed gene; and determining the target progression-free survival period corresponding to the second differentially expressed gene from the first data set and the second data set, respectively.

[0011] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in an embodiment of the present application, the second input data is determined based on the first gene expression information and the target progression-free survival period, including: determining the prognosis label corresponding to the second differentially expressed gene from the first data set and the second data set respectively; combining the second differential gene expression information corresponding to the second differentially expressed gene, the target progression-free survival period and the prognosis label as the second input data.

[0012] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application, the second expressed genes also include: ANXA1 gene, CD28 gene, CD3E gene, CPXM2 gene, CRTAP gene, CUEDC1 gene, DUSP10 gene, EGR1 gene, FAS gene, ITGAX gene, LMO2 gene, LY75 gene, MREG gene, PCDHGC3 gene, PLAU gene, RCAN3 gene, RFTN1 gene, RGS1 gene, SGK1 gene, SH3RF1 gene, SULF1 gene, TCF7 gene, TMEM163 gene and TNFRSF13C gene.

[0013] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application, the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set are separately genetically differentially analyzed to obtain multiple first differentially expressed genes, including: genetically differentially analyzing the gene expression information of each tumor patient in the first data set, and selecting first target expression genes that meet preset difference conditions; genetically differentially analyzing the gene expression information of each tumor patient in the second data set, and selecting second target expression genes that meet the preset difference conditions; comparing the first target expression genes with the second target expression genes, and selecting multiple identical expression genes as the multiple first differentially expressed genes.

[0014] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application, the method also includes: step 1, mixing and splitting the first data set and the second data set according to a preset ratio to obtain a model discovery group and a model verification group; step 2, randomly splitting the model discovery group to obtain a model training group and a model evaluation group; step 3, training a first penalized linear regression model according to the model training group to obtain a second penalized linear regression model; step 4, testing the second penalized linear regression model according to the model evaluation group to obtain a first prediction error, and adjusting the hyperparameters corresponding to the second penalized linear regression model according to the first prediction error; repeating the above steps 2-4 until a third penalized linear regression model and a target hyperparameter in which the first prediction error is less than a first preset error threshold are obtained, and the third penalized linear regression model is used as the multivariate proportional risk regression model.

[0015] Furthermore, in the above-mentioned tumor prognosis risk model determination method provided in an embodiment of the present application, after constructing the tumor prognosis risk assessment model according to the second expressed gene and the target model coefficient, the method also includes: testing the tumor prognosis risk assessment model according to the model validation group to obtain a second prediction error; if the second prediction error is less than or equal to a second preset error threshold, it is determined that the training of the tumor prognosis risk assessment model is completed.

[0016] The second aspect of an embodiment of the present application also provides a method for determining tumor prognosis risk, the method comprising: determining a preset tumor prognosis risk assessment model, the prognosis risk assessment model comprising multiple target expression genes and corresponding target model coefficients; determining initial gene expression information of a new tumor patient, and selecting target gene expression information corresponding to the target expression gene from the initial gene expression information; based on the target gene expression information, calling the tumor prognosis risk assessment model to determine a tumor prognosis risk score; and determining the prognosis risk status of the new patient based on the tumor prognosis risk score.

[0017] The third aspect of the embodiment of the present application also provides a device for determining a tumor prognosis risk model, the device comprising: a data set acquisition module, used to acquire a data set of a tumor patient; a data determination module, used to determine the gene expression information and progression-free survival of the tumor patient from the data set; a first input determination module, used to combine the gene expression information and the progression-free survival as the first input data, input them into a univariate proportional risk regression model, and obtain a correlation value between each gene expression information and the prognostic effect; a first gene expression determination module, used to select the first gene expression information according to the correlation value, and determine the target progression-free survival from the data set according to the first gene expression information; a second input determination module, used to determine the second input data according to the first gene expression information and the target progression-free survival, input them into a multivariate proportional risk regression model, and obtain the second expression gene and the target model coefficient; a risk assessment model construction module, used to construct a tumor prognosis risk assessment model according to the second expression gene and the target model coefficient.

[0018] The fourth aspect of an embodiment of the present application also provides an electronic device, which includes a processor and a memory, and the processor is used to implement any one of the above-mentioned methods for determining a tumor prognosis risk model or a method for determining a tumor prognosis risk when executing a computer program stored in the memory.

[0019] The fifth aspect of the embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a controller, it implements any one of the above-mentioned methods for determining a tumor prognosis risk model or methods for determining a tumor prognosis risk.

[0020] The above-mentioned tumor prognosis risk model determination method provided in the embodiment of the present application obtains a data set of tumor patients; determines the gene expression information and progression-free survival of tumor patients according to the data set; combines the gene expression information and the progression-free survival as the first input data, inputs it into the univariate proportional risk regression model, and obtains the correlation value between each gene expression information and the prognostic effect; selects the first gene expression information according to the correlation value, and determines the target progression-free survival from the data set according to the first gene expression information; determines the second input data according to the first gene expression information and the target progression-free survival, inputs it into the multivariate proportional risk regression model, and obtains the second expression gene and the target model coefficient; constructs a tumor prognosis risk assessment model according to the second expression gene and the target model coefficient. The above method determines the progression-free survival of each tumor patient according to the data set, and then uses the gene expression information and the progression-free survival as input data for model processing. Model training is performed using progression-free survival, and the prognostic criteria are clearly defined to improve the accuracy of model processing, thereby improving the accuracy of tumor prognosis risk assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The following specific implementation manner will further illustrate the present invention in conjunction with the above-mentioned drawings.

[0022] Figure 1 It is a sample schematic diagram of the existing tumor prognosis risk assessment model provided in the embodiments of the present application.

[0023] Figure 2 This is a structural diagram of a device for tumor prognosis risk assessment provided in an embodiment of the present application.

[0024] Figure 3 It is a flow chart of a method for determining a tumor prognosis risk model provided in an embodiment of the present application.

[0025] Figure 4 It is a flow chart for determining input data provided in an embodiment of the present application.

[0026] Figure 5 This is a flowchart for constructing a tumor prognosis risk assessment model provided in an embodiment of the present application.

[0027] Fig. 6A It is a schematic diagram of the prognostic effect of the tumor prognosis risk assessment model corresponding to the discovery group provided in the embodiment of the present application.

[0028] Figure 6B It is a schematic diagram of the recurrence of the discovery group corresponding to the tumor prognosis risk assessment model provided in the embodiment of the present application.

[0029] Figure 6C It is a schematic diagram of the prognostic effect of the tumor prognosis risk assessment model provided in the embodiment of the present application corresponding to the validation group.

[0030] Fig.6D It is a schematic diagram of the recurrence of the validation group corresponding to the tumor prognosis risk assessment model provided in the embodiment of the present application.

[0031] Fig. 7A It is a schematic diagram of the independence of univariate and multivariate Cox regression analysis of the discovery group corresponding to the prognostic value of 11 gene expression information provided in the examples of the present application.

[0032] Figure 7B It is a schematic diagram of the independence of univariate and multivariate Cox regression analysis for evaluating the prognostic value of 11 gene expression information in the validation group provided in the examples of the present application.

[0033] Figure 8 This is a flow chart for determining a target progression-free survival period provided in an embodiment of the present application.

[0034] Fig. 9 This is a flow chart for determining second input data provided in an embodiment of the present application.

[0035] Fig.10 This is a flow chart for determining a first differentially expressed gene provided in an embodiment of the present application.

[0036] Fig.11 It is a training flow chart of a multivariate proportional hazard regression model provided in an embodiment of the present application.

[0037] Fig.12 It is a data schematic diagram of a second expression gene and a target model coefficient provided in an embodiment of the present application.

[0038] Fig.13 This is a verification flow chart of a tumor prognosis risk assessment model provided in an embodiment of the present application.

[0039] Fig.14 This is a flow chart for determining tumor prognostic risk provided in an embodiment of the present application.

[0040] Fig.15 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the first embodiment of the present application.

[0041] Fig.16 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the second embodiment of the present application.

[0042] Fig.17 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the third embodiment of the present application.

[0043] Fig.18 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the fourth embodiment of the present application.

[0044] Fig.19 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the fifth embodiment of the present application.

[0045] Fig. 20 This is a prognosis verification effect diagram corresponding to the model discovery group provided in the sixth embodiment of the present application.

[0046] Fig.21 This is a prognosis verification effect diagram corresponding to the model verification group provided in the first embodiment of the present application.

[0047] Fig. 22 This is a prognosis verification effect diagram corresponding to the model verification group provided in the second embodiment of the present application.

[0048] Fig.23 This is a prognosis verification effect diagram corresponding to the model verification group provided in the third embodiment of the present application.

[0049] Fig.24 This is a prognosis verification effect diagram corresponding to the model verification group provided in the fourth embodiment of the present application.

[0050] Fig.25 A schematic diagram comparing the tumor prognosis risk assessment model provided in this application with the prior art.

[0051] FIG. 26A to FIG. 26H This is a prognosis verification effect diagram provided by another embodiment of the present application.

[0052] FIG. 27A to FIG. 27C This is a prognosis verification effect diagram provided in yet another embodiment of the present application.

[0053] Fig.28 It is a structural schematic diagram of a tumor prognosis risk model determination device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. The embodiments described are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0057] Diffuse large B-cell lymphoma (DLBCL) is one of the common malignant tumors, and its mortality rate ranks among the top among malignant tumors. Epidemiological monitoring has found that the incidence and mortality of diffuse large B-cell lymphoma (DLBCL) are gradually increasing, and it has become one of the most deadly cancers. However, in clinical practice, some patients can achieve long-term remission after standard R-CHOP (rituximab combined with cyclophosphamide, doxorubicin, vincristine and prednisone) treatment; about one-third of patients will experience refractory or relapsed disease. In clinical practice, only a few patients still achieve long-term remission after high-dose chemotherapy and autologous stem cell transplantation, which greatly affects the clinical treatment effect of diffuse large B-cell lymphoma (DLBCL). Therefore, constructing a prognostic risk assessment system for lymphoma patients is of great significance for improving patient prognosis.

[0058] However, the current lymphoma prognostic risk assessment model has a poor training data record of the patient's recurrence and progression time, resulting in unclear prognostic criteria; and the current lymphoma prognostic risk assessment model has a small training cohort, see Figure 1 , Figure 1 This is a sample schematic diagram of the existing lymphoma prognosis risk assessment model provided in the embodiments of the present application. Some models have no validation cohort, and some models include the treatment results of both CHOP and R-CHOP patients in the model. However, including the treatment results of both CHOP and R-CHOP patients in the model will result in poor repeatability of the model. As a result, the current lymphoma prognosis risk assessment model has poor accuracy in corresponding lymphoma prognosis risk assessment.

[0059] In view of the above problems, the first embodiment of the present application provides a method for determining a tumor prognosis risk model and related equipment, which can improve the accuracy of tumor prognosis risk assessment.

[0060] Figure 2 is a structural diagram of a device for tumor prognosis risk assessment provided in an embodiment of the present application. Figure 2 As shown, the electronic device 3 includes a memory 31, at least one processor 32 and at least one communication bus 33. The processor 32 is used to implement the tumor prognosis risk model determination method and the tumor prognosis risk determination method provided in the embodiment of the present application when executing the computer program stored in the memory 31.

[0061] Understandably, Figure 2The structure of the electronic device shown does not constitute a limitation of the embodiments of the present application, and may be either a bus structure or a star structure. The electronic device 3 may also include more or less other hardware or software than shown in the figure, or a different component arrangement.

[0062] In some embodiments, the electronic device 3 is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors, and embedded devices.

[0063] In one embodiment, the electronic device 3 can also be connected to a user device, which includes but is not limited to any electronic product that can perform human-computer interaction with a user through a keyboard, mouse, remote control, touchpad, or voice control device, such as a personal computer, tablet computer, smart phone, etc. The user inputs the corresponding gene expression information through the user device, and the user device transmits the gene expression information to the electronic device 3, which processes the gene expression information accordingly to obtain the tumor prognosis risk of the user.

[0064] In one embodiment, the electronic device 3 may further include a display (not shown), which is used to display the user's tumor prognosis risk. In other embodiments, the electronic device 3 may also interact with the user through a keyboard, mouse, remote control, touchpad, or voice control device, and the user inputs the corresponding gene expression information to the electronic device 3, and the electronic device 3 processes the gene expression information accordingly to obtain the user's tumor prognosis risk and display the tumor prognosis risk on the display.

[0065] It should be noted that the electronic device 3 is only an example, and other existing or future electronic products that are suitable for the present application should also be included in the protection scope of the present application and included here by reference.

[0066] In some embodiments, at least one communication bus 33 is configured to implement connection and communication between the memory 31 and at least one processor 32, etc.

[0067] Although not shown, the electronic device 3 may also include a power source (such as a battery) for supplying power to each component. Preferably, the power source may be logically connected to at least one processor 32 through a power management device, so that the power management device can manage charging, discharging, and power consumption management. The power source may also include one or more DC or AC power sources, recharging devices, power failure detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device 3 may also include a variety of sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be repeated here.

[0068] Figure 3 1 is a flow chart of a method for determining a tumor prognosis risk model provided in an embodiment of the present application, and the method for determining a tumor prognosis risk model is applied to an electronic device. Figure 3 As shown, the following steps are included:

[0069] S11, obtain a dataset of tumor patients.

[0070] In one embodiment, tumor patients may include lymphoma patients, breast tumor patients, lung tumor patients, etc., which are not limited here. This application only takes the tumor patient as a lymphoma patient as an example for explanation. The data set of lymphoma patients includes tumor data of multiple populations. For example, the data set of lymphoma patients includes 108 lymphatic patient data corresponding to Swedish and Chinese populations, 219 lymphatic patient data corresponding to Singaporean populations (i.e., Schmitz cohort), and 1049 lymphatic patient data corresponding to British populations (including GSE117556 cohort (723 patients, British population) and GSE181063 cohort (326 patients, British population)).

[0071] In one embodiment, data on 108 lymphatic patients corresponding to Swedish and Chinese populations and data on 219 lymphatic patients corresponding to Singaporean populations are used as training data sets for the model, and data on 1049 lymphatic patients corresponding to British populations are used as validation data sets for the model for model processing. Since the model data involves a large number of patients, the robustness of the tumor prognosis risk model is improved; and the embodiment of the present application improves the reliability of the tumor prognosis risk model by adding the validation data set for model processing.

[0072] In one embodiment, the data set of lymphoma patients includes the pre-collected clinical content of patients, such as transcriptome data, progression-free survival (PFS), overall survival (OS), and prognostic labels of each lymphoma patient. Among them, transcriptome data refers to the collection of all transcription products in cells under specific physiological conditions. These transcription products include mRNA, lncRNA, small RNA, etc. By sequencing and analyzing these transcription products, we can obtain the expression information of different genes and further reveal the function and regulatory mechanism of genes in organisms; PFS refers to the time from the start of treatment to the occurrence of disease progression or death due to any cause; OS refers to the time from the start of treatment to the death of the patient or the end of the last follow-up; prognostic labels can include Good labels and Poor labels. Good labels usually refer to tumor samples with better prognosis. These samples tend to show lower heterogeneity, invasiveness and metastasis, better treatment response and longer survival; Poor labels usually refer to tumor samples with poor prognosis. These samples tend to show high heterogeneity, invasiveness and metastasis, poor treatment response and shorter survival.

[0073] S12, determining gene expression information and progression-free survival of tumor patients from the dataset.

[0074] In one embodiment, by determining the preset keywords corresponding to the transcriptome data, for example, the preset keywords are transcriptome data, and the data corresponding to the preset keywords are selected from the data set of the tumor patient as the transcriptome data of each tumor patient. In one embodiment, the gene expression information corresponding to the transcriptome data can be obtained by a preset gene expression calculation algorithm. The preset gene expression calculation algorithm may include RPKM (Reads Per Kilobase Million) and TPM (Transcripts Per Million). Taking the gene expression calculation algorithm as TPM as an example, the standardized read number of all transcripts is calculated, and then the standardized read number is converted into the number of transcripts according to the length of each transcript, and finally all transcript numbers are added to obtain the total number of transcripts, and the ratio of the number of transcripts of each gene to the total number of transcripts is calculated to obtain the gene expression information of each gene. The TPM method can eliminate the influence of the length of the transcripts between different samples, and can more objectively compare the expression levels of genes in different samples. Wherein, a transcript refers to one or more mature mRNAs for encoding proteins formed by transcription of a gene.

[0075] In one embodiment, the data set includes the transcriptome data and progression-free survival (PFS) of each tumor patient. In the data set, there is a corresponding relationship between the tumor patient and the progression-free survival, and the progression-free survival of each tumor patient is obtained according to the corresponding relationship.

[0076] S13, combining the gene expression information and the progression-free survival period as the first input data, inputting them into a univariate proportional hazard regression model, and obtaining a correlation value between each gene expression information and the prognostic effect.

[0077] In one embodiment, the gene expression information of multiple tumor patients and the progression-free survival period are combined as the first input data, which is input into the univariate proportional hazard regression model (also referred to as the univariate cox regression model in this application), and the correlation value between each gene expression information and the prognostic effect is obtained, and the correlation value represents the significance of the correlation. Wherein, the univariate proportional hazard regression model is an existing model, which is not described in detail here. According to the correlation value between gene expression information and the prognostic effect, it is possible to select multiple gene expression information as an effective indicator for assessing the prognosis of a patient from a large amount of gene expression information, which is helpful to provide a more accurate prognostic assessment for the patient. Wherein, the lower the correlation value, the higher the significance of the gene expression information and the prognostic correlation; the higher the correlation value, the lower the significance of the gene expression information and the prognostic correlation. The correlation value can be the p value in the analysis result of the univariate cox regression model.

[0078] S14, selecting first gene expression information according to the correlation value, and determining a target progression-free survival period from the data set according to the first gene expression information.

[0079] In one embodiment, gene expression information with a correlation value less than a preset correlation threshold is selected as the first gene expression information, wherein the preset correlation threshold refers to a preset value, for example, the preset correlation threshold is 0.05, and the gene expression information with a correlation value less than 0.05 is selected as the first gene expression information. The preset correlation threshold may also be 0.01.

[0080] In one embodiment, a tumor patient corresponding to the first gene expression information is determined as a target tumor patient, and the progression-free survival period corresponding to the target tumor patient is used as a target progression-free survival period.

[0081] In other embodiments, the first gene expression information may be screened, and the tumor patient corresponding to the screened first gene expression information may be used as the target tumor patient, and the progression-free survival period corresponding to the target tumor patient may be used as the target progression-free survival period.

[0082] S15, determining second input data according to the first gene expression information and the target progression-free survival period, and inputting the second input data into a multivariate proportional hazard regression model to obtain the second expression gene and the target model coefficient.

[0083] In one embodiment, a multivariate proportional hazard regression model (also referred to in this application as a multivariate cox regression model) is a statistical method for studying the effects of multiple risk factors on survival time, and the multivariate proportional hazard regression model is an existing model and will not be described in detail herein. In other embodiments, the multivariate proportional hazard regression model can also be replaced with a minimum absolute shrinkage and selection algorithm regression model, that is, the second input data is determined according to the first gene expression information and the corresponding target progression-free survival period, and the data is input into the minimum absolute shrinkage and selection algorithm regression model to obtain the second expression gene and the target model coefficient. The minimum absolute shrinkage and selection algorithm regression model is an existing model and will not be described in detail herein. The second expression gene refers to the risk expression gene of the tumor patient selected from the first expression gene corresponding to the first gene expression information, and the target model coefficient refers to the coefficient of the tumor prognosis risk assessment model. For each second expression gene, there is a corresponding target model coefficient.

[0084] In one embodiment, the first gene expression information and the target progression-free survival are used as the second input data and input into the multivariate proportional hazard regression model, and the number of second expressed genes obtained is 11, namely A4GALT gene, ATXN1 gene, DIAPH1 gene, EML1 gene, LMO2 gene, PBX4 gene, PLCL1 gene, SARDH gene, SGK1 gene, SLAMF1 gene and TNFRSF11B gene. The above 11 gene expression information can be used as an independent risk factor for tumor patients. Among them, the target model coefficient corresponding to the A4GALT gene is -0.116359642, the target model coefficient corresponding to the ATXN1 gene is -0.083380887, the target model coefficient corresponding to the DIAPH1 gene is -0.051781145, the target model coefficient corresponding to the EML1 gene is -0.055129472, the target model coefficient corresponding to the LMO2 gene is -0.144639089, and the target model coefficient corresponding to the PBX4 gene is -0.144639089. The target model coefficient is -0.006071289, the target model coefficient corresponding to the PLCL1 gene is -0.283010482, the target model coefficient corresponding to the SARDH gene is -0.013092613, the target model coefficient corresponding to the SGK1 gene is -0.13304739, the target model coefficient corresponding to the SLAMF1 gene is -0.004109514, and the target model coefficient corresponding to the TNFRSF11B gene is -0.065225379.

[0085] In other embodiments, the first gene expression information is screened, and the second input data is determined according to the screened first gene expression information and the corresponding target progression-free survival period, and input into the multivariate proportional hazard regression model, and the number of second expression genes obtained is 24, namely ANXA1 gene, CD28 gene, CD3E gene, CPXM2 gene, CRTAP gene, CUEDC1 gene, DUSP10 gene, EGR1 gene, FAS gene, ITGAX gene, LMO2 gene, LY75 gene, MREG gene, PCDHGC3 gene, PLAU gene, RCAN3 gene, RFTN1 gene, RGS1 gene, SGK1 gene, SH3RF1 gene, SULF1 gene, TCF7 gene, TMEM163 gene and TNFRSF13C gene. The above 24 gene expression information can be used as independent risk factors for tumor patients.Among them, the target model coefficient corresponding to the ANXA1 gene is -1.0011307168086931, the target model coefficient corresponding to the CD28 gene is -0.1945782514641136, the target model coefficient corresponding to the CD3E gene is -0.6650735281151826, the target model coefficient corresponding to the CPXM2 gene is -0.13920831290469785, the target model coefficient corresponding to the CRTAP gene is 0.6701675799140139, and the target model coefficient corresponding to the CUEDC1 gene is -0.2850483407837123. The target model coefficient corresponding to the DUSP10 gene is -0.2845139860972085, the target model coefficient corresponding to the EGR1 gene is -0.3823476393062679, the target model coefficient corresponding to the FAS gene is -0.215377353563582, the target model coefficient corresponding to the ITGAX gene is -0.7991976883264612, the target model coefficient corresponding to the LMO2 gene is -0.010821119725323255, the target model coefficient corresponding to the LY75 gene is -0.19214411248203492, and the target model coefficient corresponding to the MREG gene is -0.215377353563582. The corresponding target model coefficient is -1.0468851564236465, the target model coefficient corresponding to the PCDHGC3 gene is -1.0200031999566868, the target model coefficient corresponding to the PLAU gene is -0.6796629705373355, the target model coefficient corresponding to the RCAN3 gene is -0.40291379763947177, the target model coefficient corresponding to the RFTN1 gene is -1.4893840884042466, the target model coefficient corresponding to the RGS1 gene is -0.39605208210685383, and the target model coefficient corresponding to the SGK1 gene is -0.6796629705373355. The corresponding target model coefficient is -0.23199246559601805, the target model coefficient corresponding to the SH3RF1 gene is --0.4219802140882992, the target model coefficient corresponding to the SULF1 gene is -0.47676417514142877, the target model coefficient corresponding to the TCF7 gene is -0.21457198627540153, the target model coefficient corresponding to the TMEM163 gene is -0.16896583553105174, and the target model coefficient corresponding to the TNFRSF13C gene is ---0.5987386812735814.

[0086] S16, constructing a tumor prognosis risk assessment model based on the second expressed gene and the target model coefficient.

[0087] In one embodiment, the input data of the tumor prognostic risk assessment model is the second gene expression information corresponding to one or more of the above-mentioned second expressed genes corresponding to the tumor patient, and the output data is the tumor prognostic risk score, wherein the higher the tumor prognostic risk score, the higher the prognostic risk; the lower the lymphoma prognostic risk score, the lower the prognostic risk.

[0088] The device adjustment method provided in the embodiment of the present application determines the progression-free survival of each tumor patient according to the data set, and then uses the gene expression information and the progression-free survival as input data for model processing. The progression-free survival is used for model training, the prognosis criteria are clearly defined, and the accuracy of model processing is improved, thereby improving the accuracy of tumor prognosis risk assessment.

[0089] In one embodiment, based on the correlation value between gene expression information and prognostic effect, multiple gene expression information for evaluating patient prognosis can be selected from a large amount of gene expression information as effective indicators, which helps to provide patients with more accurate prognostic evaluation. Figure 4 1 is a flow chart of determining input data provided by an embodiment of the present application, and the method for determining input data is applied to electronic equipment. Figure 4 As shown, the following steps are included:

[0090] S21, selecting gene expression information whose correlation value is less than a preset correlation threshold as the first gene expression information.

[0091] In one embodiment, the lower the correlation value, the higher the significance of the correlation between the gene expression information and the prognosis; the higher the correlation value, the lower the significance of the correlation between the gene expression information and the prognosis. The preset correlation threshold refers to a preset value, for example, the preset correlation threshold is 0.05, and the gene expression information with a correlation value less than 0.05 is taken as the first gene expression information.

[0092] S22, determining that the tumor patient corresponding to the first gene expression information is a target tumor patient.

[0093] In one embodiment, in the data set, there is a corresponding relationship between tumor patients and gene expression information, and by querying the corresponding relationship, the target tumor patient corresponding to the first gene expression information can be determined.

[0094] S23, determining the progression-free survival period corresponding to the target tumor patient as the target progression-free survival period.

[0095] In one embodiment, in a data set, there is a corresponding relationship between tumor patients and progression-free survival periods. By querying the corresponding relationship, the progression-free survival period corresponding to the target tumor patient can be obtained as the target progression-free survival period.

[0096] The embodiment of the present application selects gene expression information with a correlation value less than a preset correlation threshold as the first gene expression information based on the correlation value between gene expression information and prognostic effect, and uses the first gene expression information as an effective indicator for evaluating the patient's prognosis among multiple gene expression information, thereby improving the accuracy of model processing and thus improving the accuracy of tumor prognosis risk assessment.

[0097] Figure 5 is a flow chart of constructing a tumor prognosis risk assessment model provided in an embodiment of the present application, and a method for constructing a tumor prognosis risk assessment model is applied to an electronic device. Figure 5 As shown, the following steps are included:

[0098] S31, standardizing the second gene expression information corresponding to the second expressed gene to obtain third gene expression information.

[0099] In one embodiment, the second gene expression information corresponding to the second expressed gene is obtained, and the second gene expression information is standardized to obtain the third gene expression information. Among them, the standardization process can be Zscore standardization. Z-Score standardization is a commonly used data standardization method, which converts raw data into standard scores to eliminate the influence of dimensions and units between data. By converting data of different magnitudes into Z-Score scores of uniform measurement, the expression information of different genes can be uniformly compared.

[0100] In one embodiment, for each second gene expression information, there is a corresponding target model coefficient, and the third gene expression information is the value after standardized processing of the second gene expression information. In this way, according to the preset correspondence between the second gene expression information and the target model coefficient, the target model coefficient corresponding to the third gene expression information is obtained.

[0101] S32, determining a product value of the third gene expression information and the corresponding target model coefficient.

[0102] In one embodiment, the third gene expression information is recorded as x i , the target model coefficient is recorded as Coef, and the product of the third gene expression information and the target model coefficient is calculated, that is, Coef i *x i .

[0103] S33, obtaining a tumor prognosis risk assessment model according to the sum of multiple product values.

[0104] In one embodiment, according to the sum of multiple product values, a tumor prognosis risk assessment model is obtained as follows:

[0105]

[0106] Among them, the output data of the tumor prognostic risk assessment model is the tumor prognostic risk score, wherein the higher the tumor prognostic risk score, the higher the prognostic risk (that is, the higher the tumor prognostic risk score, the higher the risk of recurrence); the lower the tumor prognostic risk score, the lower the prognostic risk.

[0107] The embodiment of the present application constructs a tumor prognosis risk assessment model based on the second gene expression information and the target model coefficient. Since the second gene expression information and the target model coefficient are obtained by model training using progression-free survival, they have clearly defined prognosis criteria. This can improve the accuracy of constructing the tumor prognosis risk assessment model.

[0108] In one embodiment, a tumor prognosis risk assessment model is constructed according to the 11 second expression genes provided in the embodiments of the present application, see Fig. 6A , 6B , 6C and 6D, Fig. 6A It is a schematic diagram of the prognostic effect of the tumor prognosis risk assessment model provided in the embodiment of the present application corresponding to the discovery group; Figure 6B It is a schematic diagram of the recurrence of the discovery group corresponding to the tumor prognosis risk assessment model provided in the embodiment of the present application; Figure 6C It is a schematic diagram of the prognostic effect of the tumor prognosis risk assessment model provided in the embodiment of the present application corresponding to the validation group; Fig.6D Schematic diagram of the recurrence of the tumor prognosis risk assessment model corresponding to the validation group provided in the embodiment of the present application. Fig. 6A and Figure 6C As shown, the vertical axis represents progression-free survival (PFS), and the horizontal axis represents time. The progression-free survival (PFS) corresponding to the patient who is assessed as having a high risk of prognosis by the tumor prognostic risk assessment model of the embodiment of the present application is shown in FIG. Fig. 6A or the lower curve in 6C) is shorter than the progression-free survival of patients with a low prognostic risk (such as Fig. 6A or the upper curve in 6C); Figure 6B and Fig.6D As shown, the vertical axis represents the patient ratio, the horizontal axis represents recurrence and non-recurrence, and the proportion of patients with high prognosis risk corresponding to the recurrence of the tumor prognostic risk assessment model of the embodiment of the present application (such as Figure 6B or the darker part in 6D) is greater than the proportion of patients with low prognostic risk corresponding to recurrence (such as Figure 6B or the lighter part in 6D), the proportion of patients with high prognostic risk corresponding to patients without recurrence ( Figure 6Bor the darker part in 6D) is less than the proportion of patients without recurrence in patients with low prognostic risk ( Figure 6B or the lighter color portion in 6D). Fig. 6A , 6B , 6C, and 6D indicate that patients with a high prognostic risk have a significantly worse prognosis than patients with a low prognostic risk.

[0109] In one embodiment, a tumor prognosis risk assessment model is constructed according to the 11 second expression genes provided in the embodiments of the present application, see Fig. 7A and Figure 7B , Fig. 7A Schematic diagram of the independence of 11 gene expression information on prognostic value evaluated by univariate and multivariate cox regression analysis corresponding to the discovery group provided in the embodiment of the present application, Figure 7B The figure is a schematic diagram of the independence of 11 gene expression information on prognostic value evaluated by univariate and multivariate cox regression analysis corresponding to the validation group provided in the embodiment of the present application. In the figure, HR refers to the risk rate: >1 is a factor that promotes death, <1 is a factor that blocks death, =1 has no effect, CI is the 95% confidence interval, P is the significance in statistics, and P less than 0.05 is significant. Fig. 7A and 7B As shown, even taking into account other clinical factors such as age, gender (Male), tumor staging system (COO, including Ann Arbor staging, Boston staging and Common staging), stage and IPI (an indicator for evaluating tumor extent and prognosis), the 11 genes obtained in the embodiment of the present application (11-gene-high-risk in the figure) are still independent factors for predicting the two-year prognosis of patients treated with R-CHOP.

[0110] In one embodiment, in order to avoid the tumor prognosis risk assessment model being affected by the deviation of the patient population and the technical platform (for example, RNA-microarray technology and RNA-Seq technology), the accuracy of tumor prognosis risk assessment is further improved. The data set is divided into a first data set and a second data set, and the first data set and the second data set include tumor data of multiple populations and different technical platforms. For example, the first data set of lymphoma patients includes 108 lymphatic patient data corresponding to Swedish and Chinese populations and 219 lymphatic patient data corresponding to Singaporean populations (also referred to as Schmitz cohort in this application embodiment), and the first data set includes transcriptome data generated by RNA-Seq technology; the second data set includes 1049 lymphatic patient data corresponding to British populations (including GSE117556 cohort (723 patients, British population) and GSE181063 cohort (326 patients, British population)), and the second data set includes transcriptome data generated by RNA-microarray technology.

[0111] In one embodiment, the amount of first gene expression information selected according to the correlation value may be large, and further screening of the first gene expression information is required to improve the accuracy of the model input data and reduce the amount of input data. Figure 8 is a flow chart for determining a target progression-free survival period provided in an embodiment of the present application, and a method for determining a target progression-free survival period is applied to an electronic device. Figure 8 As shown, the following steps are included:

[0112] S41, performing gene differential analysis on the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set to obtain a plurality of first differentially expressed genes.

[0113] In one embodiment, by determining a preset keyword corresponding to the transcriptome data, for example, the preset keyword is transcriptome data, data corresponding to the preset keyword is selected from the first data set and the second data set of the tumor patient as the transcriptome data of each tumor patient. In one embodiment, the gene expression information corresponding to the transcriptome data can be obtained by a preset gene expression amount calculation algorithm.

[0114] In one embodiment, when performing analysis of high-throughput sequencing technologies such as RNA-seq, since the experimental process may be affected by factors such as different batches of sample processing, library construction and sequencing depth, these factors may interfere with the analysis results. This variation introduced by technical factors is called batch effect. For example, samples in the same queue may produce batch effects due to different processing methods. Therefore, in order to avoid the impact of batch effects on the model construction results, before the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set are analyzed by gene differentiation, and a plurality of first differentially expressed genes are obtained, the method also includes: batch effect correction of the gene expression information in the first data set and the second data set. Among them, there are many methods for correcting batch effects, including ComBat method, proxy variable method, ratio-based method, mean center method and distance weighted discrimination method.

[0115] In one embodiment, gene differential analysis is performed on the gene expression information of each tumor patient in the first data set to obtain differentially expressed genes in the first data set; gene differential analysis is performed on the gene expression information of each tumor patient in the second data set to obtain differentially expressed genes in the second data set; the differentially expressed genes in the first data set are compared with the differentially expressed genes in the second data set to obtain the same differentially expressed genes as the first differentially expressed genes.

[0116] S42, selecting gene expression information whose correlation is less than a preset correlation threshold as the first gene expression information.

[0117] In one embodiment, the lower the correlation value, the higher the significance of the correlation between the expressed gene and the prognosis; the higher the correlation value, the lower the significance of the correlation between the expressed gene and the prognosis. The expressed gene with a high significance of the correlation with the prognosis is selected as the first gene expression information. The preset correlation threshold refers to a preset value, for example, the preset correlation threshold is 0.01, and the gene expression information with a correlation value less than 0.01 is used as the first gene expression information.

[0118] S43, determining a first expressed gene corresponding to the first gene expression information, and comparing the first expressed gene with the first differentially expressed gene to obtain a second differentially expressed gene.

[0119] In one embodiment, the first expressed gene corresponding to the first gene expression information is compared with the first differentially expressed gene, and the same expressed gene is selected as the second differentially expressed gene.

[0120] S44, determining a target progression-free survival period corresponding to the second differentially expressed gene from the first data set and the second data set respectively.

[0121] In one embodiment, a tumor patient containing a second differentially expressed gene is selected from the first data set and the second data set as a target tumor patient, and the progression-free survival period corresponding to the target tumor patient is determined from the first data set and the second data set.

[0122] The embodiment of the present application obtains a first differentially expressed gene by differentially analyzing the gene expression information of each tumor patient in the first data set and the second data set. After selecting the first gene expression information according to the correlation value, the first gene expression information is screened according to the first differentially expressed gene, and a tumor prognosis risk assessment model is constructed according to the screened gene expression information to improve the accuracy and efficiency of model construction.

[0123] Fig. 9 1 is a flow chart for determining the second input data provided by an embodiment of the present application, and the method for determining the second input data is applied to an electronic device. Fig. 9 As shown, the following steps are included:

[0124] S51, determining the prognostic signature corresponding to the second differentially expressed gene from the first data set and the second data set respectively.

[0125] S52, combining the second differential gene expression information corresponding to the second differentially expressed gene, the target progression-free survival period, and the prognostic signature as second input data.

[0126] In one embodiment, a tumor patient containing a second differentially expressed gene is selected from the first data set and the second data set as a target tumor patient, a preset label corresponding to the target tumor patient is determined from the first data set and the second data set, and the second differentially expressed gene, progression-free survival and prognostic label are combined as second input data and input into a multivariate proportional hazard regression model.

[0127] The embodiment of the present application screens the first gene expression information, and uses the screened second differentially expressed genes, the target progression-free survival period and the prognostic label as the second input data, thereby improving the accuracy of the multivariate proportional risk regression model corresponding to the input data and reducing the amount of input data, thereby improving the accuracy and rate of determining the second expressed genes and the target model coefficients, and improving the accuracy and rate of constructing a tumor prognosis risk assessment model.

[0128] Fig.10 1 is a flow chart of determining a first differentially expressed gene provided in an embodiment of the present application, and the method for determining a first differentially expressed gene is applied to an electronic device. Fig.10 As shown, the following steps are included:

[0129] S61, gene differential analysis of the gene expression information of each tumor patient in the first data set, and selecting a first target expression gene that meets a preset differential condition.

[0130] In one embodiment, gene differential analysis is performed on the gene expression information of each tumor patient in the first data set to obtain differentially expressed genes in the first data set. For example, there are 1545 differentially expressed genes in the first data set, that is, 1545 first target expression genes.

[0131] In one embodiment, gene expression information can be analyzed for gene differentiation by using preset gene differentiation software, and the preset difference condition refers to a preset threshold condition for screening differentially expressed genes from a large number of expressed genes. For example, the preset difference condition can refer to the genes with a P value <0.1 after FDR correction and a foldchange>1.2 between groups (tumor patients with two-year recurrence and two-year non-recurrence) as differentially expressed genes, i.e., the first target expression genes. Among them, FDR correction is a method for controlling the proportion of false positive results. FDR (false discovery rate) refers to the probability that at least one of all possible hypotheses is wrong. In FDR correction, the P value is used to measure the false positive rate of a test. Foldchange is usually used to compare the differences in gene expression between tumor patients with two-year recurrence and two-year non-recurrence. Specifically, the larger the value of the foldchange indicator, the greater the difference in gene expression between patients with recurrence and patients without recurrence, that is, there is a large difference in gene expression between patients with recurrence and patients without recurrence.

[0132] S62, gene differential analysis of the gene expression information of each tumor patient in the second data set, and selecting a second target expression gene that meets the preset differential condition.

[0133] In one embodiment, gene differential analysis is performed on the gene expression information of each tumor patient in the second data set to obtain differentially expressed genes in the second data set. For example, there are 959 differentially expressed genes in the second data set, that is, 959 second target expression genes.

[0134] S63, comparing the first target expressed gene with the second target expressed gene, and selecting a plurality of identically expressed genes as the plurality of first differentially expressed genes.

[0135] In one embodiment, the differentially expressed genes in the first data set are compared with the differentially expressed genes in the second data set to obtain the same differentially expressed genes as the first differentially expressed genes. For example, the number of the first differentially expressed genes is 372.

[0136] The embodiment of the present application performs gene differentiation analysis on the gene expression information in the first data set and the second data set, and can obtain the same differentially expressed genes in the above two data sets, so as to construct a subsequent tumor prognosis risk assessment model based on the differentially expressed genes, thereby improving the accuracy of model construction.

[0137] In one embodiment, a multivariate proportional hazard regression model is trained based on the mixed first data set and the second data set. Exemplarily, the first data set and the second data set are mixed and split into two groups, namely a model discovery group and a model validation group, and the multivariate proportional hazard regression model is trained based on the model discovery group. In other embodiments, the multivariate proportional hazard regression model can also be replaced with a minimum absolute shrinkage and selection algorithm regression model, that is, the second differentially expressed gene, progression-free survival and prognostic label are combined as input data to obtain input data, which is input into the minimum absolute shrinkage and selection algorithm regression model to obtain the second expression gene and the target model coefficient. The minimum absolute shrinkage and selection algorithm regression model is an existing model and will not be described in detail here. Fig.11 1 is a training flow chart of a multivariate proportional risk regression model provided in an embodiment of the present application, and a training method of a multivariate proportional risk regression model is applied to electronic devices. Fig.11 As shown, the following steps are included:

[0138] S71, mixing and splitting the first data set and the second data set according to a preset ratio to obtain a model discovery group and a model verification group.

[0139] In one embodiment, the first data set and the second data set are mixed and split into two groups, namely a model discovery group and a model validation group. Exemplarily, the model discovery group accounts for 70%, that is, the model discovery group includes 964 patients, and the model validation group accounts for 30%, that is, the model validation group includes 412 patients. In one embodiment, in the model discovery group and the model validation group, the data proportions of the above four cohorts are the same.

[0140] S72, randomly splitting the model discovery group to obtain a model training group and a model evaluation group.

[0141] In one embodiment, the hyperparameters in the penalized linear regression model are optimized by cross-validation, for example, a 10-fold cross-validation method is used to optimize the hyperparameters. The model discovery group is randomly divided into a model training group and a model evaluation group, one part is used as a training set to train the model, and the other part is used as a validation set to test the generalization performance of the model. Exemplarily, the model training group accounts for 80%, and the model training group includes 772 patients; the model evaluation group accounts for 20%, and the model evaluation group includes 192 patients.

[0142] S73: Train a first penalized linear regression model according to the model training group to obtain a second penalized linear regression model.

[0143] In one embodiment, a first penalized linear regression model is trained according to the model training group to obtain a second penalized linear regression model, and hyperparameters corresponding to the second penalized linear regression model are recorded.

[0144] S74, testing the second penalized linear regression model according to the model evaluation group to obtain a first prediction error, and adjusting the hyperparameters corresponding to the second penalized linear regression model according to the first prediction error.

[0145] In one embodiment, the hyperparameter of the second penalized linear regression model is adjusted according to the size of the first prediction error. For example, if the first prediction error is large, the value of the hyperparameter may be increased to increase the complexity of the model; conversely, if the first prediction error is small, the value of the hyperparameter may be decreased to reduce the complexity of the model.

[0146] In one embodiment, the number of the first preset errors may be one or more. For example, the first prediction error refers to the difference between the actual index value and the preset index value of the multiple indexes set for evaluating the performance of the model, and the preset index value is a preset index value that makes the model performance meet the actual needs. The index may include but is not limited to the ROC curve, the area under the curve (AUC), accuracy, sensitivity and specificity, wherein the ROC curve is a graphical tool for evaluating the performance of a binary classification model, which shows the relationship between the true positive rate (TPR) and the false positive rate (FPR) of the model at different thresholds; AUC is the area under the ROC curve, which measures the overall performance of the model at all possible thresholds; accuracy refers to the proportion of the number of correct predictions of the model to the total number of predictions, which is an indicator for measuring the overall performance of the model; sensitivity refers to the proportion of the number of positive examples correctly predicted by the model to the number of actual positive examples, which is an indicator for measuring the recognition ability of the model for positive examples; specificity refers to the proportion of the number of negative examples correctly predicted by the model to the number of actual negative examples, which is an indicator for measuring the recognition ability of the model for negative examples.

[0147] S75, repeat the above steps S72-S74 until a third penalized linear regression model and target hyperparameters are obtained in which the first prediction error is less than a first preset error threshold, and the third penalized linear regression model is used as the multivariate proportional risk regression model.

[0148] In one embodiment, for each first prediction error, there is a corresponding first preset error threshold, and the number of first element error thresholds can be set according to actual needs and is not limited here. Repeat the above steps 2-4, randomly split the model discovery group each time to obtain a model training group and a model evaluation group, and then calculate the first prediction error and adjust the hyperparameters until a third penalized linear regression model and a target hyperparameter are obtained in which all first prediction errors are less than the first preset error threshold. Among them, the first preset error threshold can be set according to actual needs and is not limited here.

[0149] The embodiment of the present application performs model training by mixing the first data set and the second data set of tumor patients from different populations and different transcriptome sequencing technologies, so that the trained tumor prognosis risk assessment model is less biased towards different technical platforms and populations, thereby improving the accuracy of model processing; in addition, the present application randomly splits the model discovery group to obtain a model training group and a model evaluation group, and performs a cross-validation method on the model training group and the model evaluation group to obtain a tumor prognosis risk assessment model, thereby further improving the accuracy of model processing.

[0150] In one embodiment, Fig.12 is a data schematic diagram of a second expression gene and a target model coefficient provided in an embodiment of the present application, such as Fig.12As shown, the number of second expressed genes provided in the present application can be 24, namely ANXA1 gene, CD28 gene, CD3E gene, CPXM2 gene, CRTAP gene, CUEDC1 gene, DUSP10 gene, EGR1 gene, FAS gene, ITGAX gene, LMO2 gene, LY75 gene, MREG gene, PCDHGC3 gene, PLAU gene, RCAN3 gene, RFTN1 gene, RGS1 gene, SGK1 gene, SH3RF1 gene, SULF1 gene, TCF7 gene, TMEM163 gene and TNFRSF13C gene. The above 24 gene expression information can be used as independent risk factors for tumor patients.Among them, the target model coefficient corresponding to the ANXA1 gene is -1.0011307168086931, the target model coefficient corresponding to the CD28 gene is -0.1945782514641136, the target model coefficient corresponding to the CD3E gene is -0.6650735281151826, the target model coefficient corresponding to the CPXM2 gene is -0.13920831290469785, the target model coefficient corresponding to the CRTAP gene is 0.6701675799140139, and the target model coefficient corresponding to the CUEDC1 gene is -0.2850483407837123. The target model coefficient corresponding to the DUSP10 gene is -0.2845139860972085, the target model coefficient corresponding to the EGR1 gene is -0.3823476393062679, the target model coefficient corresponding to the FAS gene is -0.215377353563582, the target model coefficient corresponding to the ITGAX gene is -0.7991976883264612, the target model coefficient corresponding to the LMO2 gene is -0.010821119725323255, the target model coefficient corresponding to the LY75 gene is -0.19214411248203492, and the target model coefficient corresponding to the MREG gene is -0.215377353563582. The corresponding target model coefficient is -1.0468851564236465, the target model coefficient corresponding to the PCDHGC3 gene is -1.0200031999566868, the target model coefficient corresponding to the PLAU gene is -0.6796629705373355, the target model coefficient corresponding to the RCAN3 gene is -0.40291379763947177, the target model coefficient corresponding to the RFTN1 gene is -1.4893840884042466, the target model coefficient corresponding to the RGS1 gene is -0.39605208210685383, and the target model coefficient corresponding to the SGK1 gene is -0.6796629705373355. The corresponding target model coefficient is -0.23199246559601805, the target model coefficient corresponding to the SH3RF1 gene is --0.4219802140882992, the target model coefficient corresponding to the SULF1 gene is -0.47676417514142877, the target model coefficient corresponding to the TCF7 gene is -0.21457198627540153, the target model coefficient corresponding to the TMEM163 gene is -0.16896583553105174, and the target model coefficient corresponding to the TNFRSF13C gene is ---0.5987386812735814.

[0151] In one embodiment, after constructing a tumor prognosis risk assessment model based on the 24 second expressed genes and the corresponding target model coefficients, the model validation group can also be used to validate the effectiveness of the above model. Fig.13 1 is a flow chart of verifying a tumor prognosis risk assessment model provided in an embodiment of the present application, and a verification method of the tumor prognosis risk assessment model is applied to an electronic device. Fig.13 As shown, the following steps are included:

[0152] S81, testing the tumor prognosis risk assessment model according to the model validation group to obtain a second prediction error.

[0153] In one embodiment, the number of the second prediction errors may be one or more, and illustratively, the second prediction error refers to the difference between the actual index value and the preset index value of multiple indexes set for evaluating the performance of the model, and the preset index value is a preset index value that makes the model performance meet the actual requirements. The index may include but is not limited to the ROC curve, the area under the curve (AUC), accuracy, sensitivity and specificity.

[0154] S82: If the second prediction error is less than or equal to a second preset error threshold, it is determined that the training of the tumor prognosis risk assessment model is completed.

[0155] In one embodiment, each second prediction error has a corresponding second preset error threshold. If each second prediction error is less than or equal to the second preset error threshold, it is determined that the training of the tumor prognosis risk assessment model is completed; if there is a second prediction error greater than the second preset error threshold, the training set is increased to continue training the multivariate proportional risk regression model.

[0156] The embodiment of the present application performs model validation based on a first data set and a second data set of tumor patients from mixed different populations and using different transcriptome sequencing technologies, so that the trained tumor prognosis risk assessment model is less biased towards different technology platforms and populations, thereby improving the accuracy of model processing and thus improving the accuracy of tumor prognosis risk assessment.

[0157] In one embodiment, for a new tumor patient, the prognostic risk score of the tumor patient can be determined by analyzing the initial gene expression information of the tumor patient and according to a preset tumor prognostic risk model. Fig.14 is a flow chart for determining tumor prognosis risk provided by an embodiment of the present application, and the method for determining tumor prognosis risk is applied to electronic equipment. Fig.14 As shown, the following steps are included:

[0158] S91, determining a preset tumor prognosis risk assessment model, wherein the prognosis risk assessment model includes a plurality of target expression genes and corresponding target model coefficients.

[0159] In one embodiment, the preset tumor prognosis risk assessment model is a model that is pre-trained to a convergence state and assesses the prognosis risk of tumor patients, the input data of the tumor prognosis risk assessment model is the gene expression information of tumor patients after analysis and screening, and the output data is the tumor prognosis risk score. The preset tumor prognosis risk assessment model includes the product value of multiple target expression genes and corresponding target model coefficients, wherein the target expression gene is a predetermined gene that has a greater correlation with the patient's prognosis effect.

[0160] In one embodiment, the number of target expression genes may be 11, namely A4GALT gene, ATXN1 gene, DIAPH1 gene, EML1 gene, LMO2 gene, PBX4 gene, PLCL1 gene, SARDH gene, SGK1 gene, SLAMF1 gene and TNFRSF11B gene. The above 11 gene expression information can be used as an independent risk factor for cancer patients. Among them, the target model coefficient corresponding to the A4GALT gene is -0.116359642, the target model coefficient corresponding to the ATXN1 gene is -0.083380887, the target model coefficient corresponding to the DIAPH1 gene is -0.051781145, the target model coefficient corresponding to the EML1 gene is -0.055129472, the target model coefficient corresponding to the LMO2 gene is -0.144639089, and the target model coefficient corresponding to the PBX4 gene is -0.116359642. The target model coefficient is -0.006071289, the target model coefficient corresponding to the PLCL1 gene is -0.283010482, the target model coefficient corresponding to the SARDH gene is -0.013092613, the target model coefficient corresponding to the SGK1 gene is -0.13304739, the target model coefficient corresponding to the SLAMF1 gene is -0.004109514, and the target model coefficient corresponding to the TNFRSF11B gene is -0.065225379.

[0161] In other embodiments, the number of target expression genes may be 24, namely ANXA1 gene, CD28 gene, CD3E gene, CPXM2 gene, CRTAP gene, CUEDC1 gene, DUSP10 gene, EGR1 gene, FAS gene, ITGAX gene, LMO2 gene, LY75 gene, MREG gene, PCDHGC3 gene, PLAU gene, RCAN3 gene, RFTN1 gene, RGS1 gene, SGK1 gene, SH3RF1 gene, SULF1 gene, TCF7 gene, TMEM163 gene and TNFRSF13C gene. The above 24 gene expression information can be used as independent risk factors for tumor patients.Among them, the target model coefficient corresponding to the ANXA1 gene is -1.0011307168086931, the target model coefficient corresponding to the CD28 gene is -0.1945782514641136, the target model coefficient corresponding to the CD3E gene is -0.6650735281151826, the target model coefficient corresponding to the CPXM2 gene is -0.13920831290469785, the target model coefficient corresponding to the CRTAP gene is 0.6701675799140139, and the target model coefficient corresponding to the CUEDC1 gene is -0.2850483407837123. The target model coefficient corresponding to the DUSP10 gene is -0.2845139860972085, the target model coefficient corresponding to the EGR1 gene is -0.3823476393062679, the target model coefficient corresponding to the FAS gene is -0.215377353563582, the target model coefficient corresponding to the ITGAX gene is -0.7991976883264612, the target model coefficient corresponding to the LMO2 gene is -0.010821119725323255, the target model coefficient corresponding to the LY75 gene is -0.19214411248203492, and the target model coefficient corresponding to the MREG gene is -0.215377353563582. The corresponding target model coefficient is -1.0468851564236465, the target model coefficient corresponding to the PCDHGC3 gene is -1.0200031999566868, the target model coefficient corresponding to the PLAU gene is -0.6796629705373355, the target model coefficient corresponding to the RCAN3 gene is -0.40291379763947177, the target model coefficient corresponding to the RFTN1 gene is -1.4893840884042466, the target model coefficient corresponding to the RGS1 gene is -0.39605208210685383, and the target model coefficient corresponding to the SGK1 gene is -0.6796629705373355. The corresponding target model coefficient is -0.23199246559601805, the target model coefficient corresponding to the SH3RF1 gene is --0.4219802140882992, the target model coefficient corresponding to the SULF1 gene is -0.47676417514142877, the target model coefficient corresponding to the TCF7 gene is -0.21457198627540153, the target model coefficient corresponding to the TMEM163 gene is -0.16896583553105174, and the target model coefficient corresponding to the TNFRSF13C gene is ---0.5987386812735814.

[0162] In one embodiment, a tumor prognosis risk assessment model is determined according to the product value of multiple target expression genes and the corresponding target model coefficients. For example, the tumor prognosis risk assessment model is:

[0163]

[0164] Among them, x i is the target expression gene, Coef is the target model coefficient, and risk score is the output data of the tumor prognosis risk assessment model, that is, the tumor prognosis risk score.

[0165] S92, determining the initial gene expression information of the new tumor patient, and selecting the target gene expression information corresponding to the target expression gene from the initial gene expression information.

[0166] In one embodiment, gene analysis is performed on a new tumor patient to obtain all gene expression information of the patient as initial gene expression information, and then target gene expression information identical to the target expression gene is screened from the initial gene expression information.

[0167] S93, calling the tumor prognosis risk assessment model according to the target gene expression information to determine the tumor prognosis risk score.

[0168] In one embodiment, the number of expressed genes identical to the target expressed genes may be 0, or may be 1, 2, or more, and is not limited here. When the number of expressed genes identical to the differentially expressed genes is 0, the tumor prognosis risk score corresponding to the patient is 0, indicating that the patient is in a low risk state of prognosis. The higher the tumor prognosis risk score, the higher the prognosis risk; the lower the tumor prognosis risk score, the lower the prognosis risk.

[0169] S94, determining the prognostic risk status of the new patient according to the tumor prognostic risk score.

[0170] In one embodiment, the prognostic risk state may include a high prognostic risk state and a low prognostic risk state. A prognostic risk value is set. If the tumor prognostic risk score is higher than the prognostic risk value, the prognostic risk state is determined to be a high prognostic risk state; if the tumor prognostic risk score is lower than the prognostic risk value, the prognostic risk state is determined to be a low prognostic risk state.

[0171] The embodiment of the present application analyzes the initial gene expression information of new tumor patients and determines the prognostic risk score of the tumor patients according to a preset tumor prognostic risk model, which can improve the accuracy of determining the prognostic risk status and thus improve the accuracy of tumor prognostic risk assessment.

[0172] In one embodiment, a tumor prognosis risk assessment model is constructed according to the 24 second expression genes provided in the embodiments of the present application, see Figures 15 to 24 , Figures 15 to 20 Used to describe the prognostic effect corresponding to the model discovery group, Figure 21 to Figure 24 Used to describe the prognostic effect corresponding to the model validation group. Fig.15 As shown, for the prognostic effect of the tumor prognostic risk assessment model corresponding to the model discovery group, the 24 gene combinations finally screened by the embodiment of the present application are still independent factors for predicting the two-year prognosis of patients receiving R-CHOP treatment, where HR refers to the risk rate: >1 is a factor that promotes death. Fig.16 As shown in the figure, the horizontal axis represents the number of patients who experienced R / R within 2 years, and the vertical axis represents the prognostic risk score. The patients are sorted from low to high according to their prognostic risk scores. The vertical dotted line is the threshold for distinguishing high risk from low risk (obtained according to the ROC curve), and the actual PFS of these patients is also displayed. Fewer patients in the low-risk group experienced R / R within 2 years, while more patients in the high-risk group experienced R / R within 2 years. Fig.17 As shown in the figure, the horizontal axis represents time and the vertical axis represents PFS. There is a significant difference in survival prognosis between the high-risk and low-risk groups. Fig.18 As shown in the figure, more patients in the high-risk group (Poor group) experienced R / R within 2 years (darker part), while more patients in the low-risk group (Good group) did not experience R / R within 2 years (lighter part). Fig.19 As shown, the high-risk group calculated by the model coefficients obtained under univariate and multivariate conditions is compared with the common high-risk factors in clinical practice (such as age>60, male, ABC classification, clinical stage III-IV, IPI score>3 patients). The method provided in the embodiment of the present application corresponds to a higher HR risk rate and a higher significance. Fig. 20 As shown, the horizontal axis represents specificity, and the vertical axis represents sensitivity. The ROC curve of the high-risk group calculated by the model coefficient obtained under multivariate and the common high-risk factors in clinical practice (such as age>60, male, ABC classification, clinical stage III-IV, high IPI score, TP53 positive, MYD88 positive, TP53 combined with MYD88, MYC, SPEN positive patients) is compared. The area under the curve (i.e., AUC value) is still larger for the method provided in the embodiment of the present application. Figure 21 to Figure 24 As shown in Figure 2, the prognostic effect of the tumor prognostic risk assessment model corresponding to the model validation group is Figures 15 to 20 The prognostic effects of the corresponding models in the tumor prognostic risk assessment model were similar and will not be elaborated here.

[0173] Fig.25A schematic diagram showing a comparison between the tumor prognosis risk assessment model provided in this application and the prior art (the prior art may include Panet al, Yanet al and Zamani-Ahm admahmudi), as shown in FIG. Fig.25 As shown, the tumor prognosis risk assessment model provided in the embodiment of the present application (as shown by the target curve in the figure) has greater advantages in predicting the early recurrence rate of patients after treatment.

[0174] In addition, in order to verify the non-bias between technology platforms, such as Figure 26A-26D As shown, the horizontal axis represents time and the vertical axis represents PFS. We used the model to make separate predictions for 4 cohorts (as mentioned above, our cohort (108 patients, Swedish and Chinese populations, RNA-Seq data, also recorded as the "first cohort" in the figure), Schmitz cohort (219 patients, Singaporean population, RNA-Seq data, also recorded as the "second cohort" in the figure), GSE117556 cohort (723 patients, British population, RNAmicroarray data, also recorded as the "third cohort" in the figure), and GSE181063 cohort (326 patients, British population, RNAmicroarray data), also recorded as the "fourth cohort" in the figure), and the effect was verified. The survival prognosis between the high-risk and low-risk groups was significantly different. As shown in the figure Figure 26E-26H As shown, the horizontal axis represents the high-risk group and the low-risk group, and the vertical axis represents the number of patients who developed R / R within 2 years. In the high-risk group (Poor group), the number of patients who developed R / R within 2 years was greater (lighter color part), and in the low-risk group (Good group), the number of patients who did not develop R / R within 2 years was greater (darker color part).

[0175] like Fig.27A , 27B As shown in 27C, this application also used two separate cohorts, CNP0001327 (49 patients, Chinese population, RNA-Seq data, also recorded as the "fifth cohort" in the figure) and GSE181063 (485 patients with OS information in the cohort, British population, RNAmicroarray data, also recorded as the "sixth cohort" in the figure) for risk prediction, and the effect was also verified. Fig.27A As shown in the figure, the horizontal axis represents time and the vertical axis represents PFS. There is a significant difference in survival prognosis between the high-risk and low-risk groups. Fig.27B As shown in the figure, the horizontal axis represents the high-risk group and the low-risk group, and the vertical axis represents the number of patients who developed R / R within 2 years. In the high-risk group (Poor group), the number of patients who developed R / R within 2 years was more (the darker part), and in the low-risk group (Good group), the number of patients who did not develop R / R within 2 years was more (the lighter part). Fig.27CAs shown, the horizontal axis represents time and the vertical axis represents OS. There is a significant difference in survival prognosis between the high-risk and low-risk groups.

[0176] See also Fig.28 , Fig.28 is a schematic diagram of the structure of the tumor prognosis risk model determination device provided in an embodiment of the present application. In some embodiments, the tumor prognosis risk model determination device 20 may include multiple functional modules composed of computer program segments. The computer program of each program segment in the tumor prognosis risk model determination device 20 may be stored in the memory of the electronic device 30 and executed by at least one processor to execute (see Figure 3 Description) The function of tumor prognostic risk assessment.

[0177] In this embodiment, the tumor prognosis risk model determination device 20 can be divided into multiple functional modules according to the functions performed. The functional modules may include: a data set acquisition module 201, a data determination module 202, a first input determination module 203, a first gene expression determination module 204, a second input determination module 205, and a risk assessment model construction module 206. The module referred to in this application refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0178] The data set acquisition module 201 can be used to acquire data sets of tumor patients.

[0179] The data determination module 202 may be used to determine gene expression information and progression-free survival of a tumor patient from the data set.

[0180] The first input determination module 203 can be used to combine the gene expression information and the progression-free survival period as the first input data, input them into a univariate proportional hazard regression model, and obtain the correlation value between each gene expression information and the prognostic effect.

[0181] The first gene expression determination module 204 may be configured to select first gene expression information according to the correlation value, and determine a target progression-free survival period from the data set according to the first gene expression information.

[0182] The second input determination module 205 can be used to determine second input data according to the first gene expression information and the target progression-free survival period, and input the second input data into a multivariate proportional risk regression model to obtain the second expression gene and the target model coefficient.

[0183] The risk assessment model construction module 206 can be used to construct a tumor prognosis risk assessment model according to the second expressed gene and the target model coefficient.

[0184] In the embodiment of the present application, the functional modules in the tumor prognosis risk model determination device 20 and the tumor prognosis risk model determination method in the above-mentioned multiple embodiments belong to the same application concept, and the specific implementation method of each module in the tumor prognosis risk model determination device 20 corresponds to the various steps of the tumor prognosis risk model determination method in the above-mentioned embodiment, and the present application will not repeat the description here.

[0185] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation. The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions to enable an electronic device (which can be a personal computer, an electronic device, or a network device, etc.) or a processor to execute parts of the methods of each embodiment of the present application.

[0186] Continuing from the above Figure 2 In some embodiments, a computer program is stored in the memory 31, and when the computer program is executed by at least one processor 32, all or part of the steps in the above-mentioned tumor prognosis risk model determination method are implemented. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0187] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store created data, etc.

[0188] In some embodiments, at least one processor 32 is the control core (Control Unit) of the electronic device 3, and uses various interfaces and lines to connect various components of the entire electronic device 3, and executes various functions and processes data of the electronic device 3 by running or executing programs or modules stored in the memory 31, and calling data stored in the memory 31. For example, when at least one processor 32 executes the computer program stored in the memory, it implements all or part of the steps of the tumor prognosis risk model determination method in the embodiment of the present application; or implements all or part of the functions of the tumor prognosis risk model determination device. At least one processor 32 can be composed of an integrated circuit, for example, it can be composed of a single packaged integrated circuit, or it can be composed of multiple integrated circuits with the same function or different functions, including one or more central processing units (CPU), microprocessors, digital processing chips, graphics processors, and various control chips. The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0189] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0190] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "including" does not exclude other units or, and the singular does not exclude the plural. Multiple units or devices stated in the specification can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

[0191] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present application.

Claims

1. A method for determining a tumor prognosis risk model, characterized in that: The method comprises: Obtain a dataset of tumor patients; Determining gene expression information and progression-free survival of tumor patients from the dataset; Combining the gene expression information and the progression-free survival period as the first input data, inputting them into a univariate proportional hazard regression model, and obtaining a correlation value between each gene expression information and the prognostic effect; selecting first gene expression information according to the correlation value, and determining a target progression-free survival period from the data set according to the first gene expression information; Determine second input data based on the first gene expression information and the target progression-free survival period, and input the second input data into a multivariate proportional hazard regression model to obtain a second expression gene and target model coefficient; A tumor prognosis risk assessment model is constructed based on the second expressed gene and the target model coefficient.

2. The method according to claim 1, characterized in that The second expressed genes include: A4GALT gene, ATXN1 gene, DIAPH1 gene, EML1 gene, LMO2 gene, PBX4 gene, PLCL1 gene, SARDH gene, SGK1 gene, SLAMF1 gene and TNFRSF11B gene.

3. The method according to claim 2, characterized in that The target model coefficient corresponding to the A4GALT gene is -0.116359642, the target model coefficient corresponding to the ATXN1 gene is -0.083380887, the target model coefficient corresponding to the DIAPH1 gene is -0.051781145, the target model coefficient corresponding to the EML1 gene is -0.055129472, the target model coefficient corresponding to the LMO2 gene is -0.144639089, and the target model coefficient corresponding to the PBX4 gene is -0.051781145. The target model coefficient corresponding to the PLCL1 gene is -0.283010482, the target model coefficient corresponding to the SARDH gene is -0.013092613, the target model coefficient corresponding to the SGK1 gene is -0.13304739, the target model coefficient corresponding to the SLAMF1 gene is -0.004109514, and the target model coefficient corresponding to the TNFRSF11B gene is -0.065225379.

4. The method according to claim 1, characterized in that The selecting first gene expression information according to the correlation value, and determining a target progression-free survival period from the data set according to the first gene expression information, comprises: Selecting gene expression information whose correlation value is less than a preset correlation threshold as the first gene expression information; Determining that the tumor patient corresponding to the first gene expression information is a target tumor patient; The progression-free survival period corresponding to the target tumor patient is determined as the target progression-free survival period.

5. The method according to claim 1, characterized in that If the data set includes a first data set and a second data set, selecting first gene expression information according to the correlation value, and determining a target progression-free survival period from the data set according to the first gene expression information, further includes: Gene differential analysis is performed on the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set to obtain a plurality of first differentially expressed genes; Selecting gene expression information whose correlation is less than a preset correlation threshold as the first gene expression information; Determining a first expressed gene corresponding to the first gene expression information, and comparing the first expressed gene with the first differentially expressed gene to obtain a second differentially expressed gene; A target progression-free survival period corresponding to the second differentially expressed gene is determined from the first data set and the second data set respectively.

6. The method according to claim 5, characterized in that The determining of the second input data according to the first gene expression information and the target progression-free survival period includes: Determining the prognostic signature corresponding to the second differentially expressed gene from the first data set and the second data set respectively; The second differential gene expression information corresponding to the second differentially expressed gene, the target progression-free survival period, and the prognostic signature are combined as second input data.

7. The method according to claim 1, characterized in that The second expressed genes also include: ANXA1 gene, CD28 gene, CD3E gene, CPXM2 gene, CRTAP gene, CUEDC1 gene, DUSP10 gene, EGR1 gene, FAS gene, ITGAX gene, LMO2 gene, LY75 gene, MREG gene, PCDHGC3 gene, PLAU gene, RCAN3 gene, RFTN1 gene, RGS1 gene, SGK1 gene, SH3RF1 gene, SULF1 gene, TCF7 gene, TMEM163 gene and TNFRSF13C gene.

8. The method according to claim 5, characterized in that The gene differential analysis of the gene expression information of each tumor patient in the first data set and the gene expression information of each tumor patient in the second data set to obtain a plurality of first differentially expressed genes includes: Gene differential analysis of the gene expression information of each tumor patient in the first data set, and selecting a first target expression gene that meets a preset differential condition; Gene differential analysis of the gene expression information of each tumor patient in the second data set, and selecting the second target expression gene that meets the preset differential condition; The first target expressed gene and the second target expressed gene are compared, and a plurality of identically expressed genes are selected as the plurality of first differentially expressed genes.

9. The method according to claim 5, characterized in that The method further comprises: Step 1, mixing and splitting the first data set and the second data set according to a preset ratio to obtain a model discovery group and a model verification group; Step 2, randomly splitting the model discovery group to obtain a model training group and a model evaluation group; Step 3, training the first penalized linear regression model according to the model training group to obtain a second penalized linear regression model; Step 4: testing the second penalized linear regression model according to the model evaluation group to obtain a first prediction error, and adjusting a hyperparameter corresponding to the second penalized linear regression model according to the first prediction error; Repeat steps 2-4 above until a third penalized linear regression model and target hyperparameters are obtained in which the first prediction error is less than a first preset error threshold, and the third penalized linear regression model is used as the multivariate proportional risk regression model.

10. The method according to claim 9, characterized in that After constructing the tumor prognosis risk assessment model according to the second expressed gene and the target model coefficient, the method further includes: Testing the tumor prognosis risk assessment model according to the model validation group to obtain a second prediction error; If the second prediction error is less than or equal to a second preset error threshold, it is determined that the training of the tumor prognosis risk assessment model is completed.

11. The method according to claim 1, characterized in that The step of constructing a tumor prognosis risk assessment model according to the second expressed gene and the target model coefficient includes: Standardizing the second gene expression information corresponding to the second expressed gene to obtain third gene expression information; Determining a product value of the third gene expression information and the corresponding target model coefficient; According to the sum of multiple product values, a tumor prognosis risk assessment model is obtained.

12. A method for determining tumor prognosis risk, characterized in that: The method comprises: Determining a preset tumor prognosis risk assessment model, wherein the prognosis risk assessment model includes a plurality of target expression genes and corresponding target model coefficients; Determining initial gene expression information of a new tumor patient, and selecting target gene expression information corresponding to the target expression gene from the initial gene expression information; According to the target gene expression information, calling the tumor prognosis risk assessment model to determine the tumor prognosis risk score; The prognostic risk status of the new patient is determined according to the tumor prognostic risk score.

13. An electronic device, characterized in that: The electronic device includes a processor and a memory, and the processor is used to implement the tumor prognosis risk model determination method described in any one of claims 1 to 11 or the tumor prognosis risk determination method described in claim 12 when executing the computer program stored in the memory.

Images