Method for predicting immunotherapy benefit rate, computing device, and computer storage medium

Abstract

The invention relates to a method for predicting an immunotherapy benefit rate, a computing device and a computer storage medium. The method comprises the following steps: generating a plurality of input feature subsets and prediction index data sets based on acquired clinical data, blood routine information and tumor-related gene mutation data; based on each input feature subset in the multiple input feature subsets and the prediction index data set, a plurality of prediction models are trained respectively, so that a matched candidate prediction model is determined for each input feature subset; selecting a target prediction model from the candidate prediction models based on clinical data of the user, blood routine information before immunotherapy, sequencing data and input operation of the user, and constructing user input features; and generating a prediction result about the immunotherapy benefit rate of the user through the target prediction model based on the constructed user input features. The immunotherapy benefit rate can be objectively and accurately predicted according to different cancer species.

Description

Technical Field

[0001] The present invention generally relates to biological information detection processing, and in particular, to a method, a computing device and a computer storage medium for predicting the benefit rate of immunotherapy. Background Art

[0002] Immunotherapy refers to a disease treatment method that induces, enhances or suppresses immune responses. Cancer immunotherapy has been widely used in clinical practice, and many single-drug immunosuppressants have been approved. The approved target indications of immunosuppressants are: programmed death ligand-1 (i.e., PD-L1) inhibitors, high tumor mutation burden (i.e., TMB-H), high microsatellite instability (i.e., MSI-H), etc. In actual clinical practice, immunosuppressants can be used in various forms such as adjuvant, neoadjuvant, and advanced treatment, combined with chemotherapy, or targeted therapy, double-immunity therapy, etc. There is no clear regular basis for the overall strategy of how to choose the treatment time point and treatment form of specific immunosuppressants for different individual patients to maximize the benefit of patients. It depends more on the subjective experience and exploration of individual clinicians about specific cancers. There is a lack of effective means to objectively and accurately rank and predict the treatment benefit rate of different cancers, patients at different stages, and different immunotherapy combinations.

[0003] In summary, the shortcomings of existing methods for evaluating the therapeutic benefit rate under different combinations of immunotherapy regimens are that they rely on individual experience, lack objectivity and accuracy, and are difficult to apply to different types of cancer. Summary of the invention

[0004] The present invention provides a method, a computing device and a computer storage medium for predicting the benefit rate of immunotherapy, which can objectively and accurately predict the benefit rate of immunotherapy for different types of cancer.

[0005] According to the first aspect of the present invention, a method for predicting the benefit rate of immunotherapy is provided. The method includes: obtaining clinical data of a sample object, the clinical data at least including attribute information, tumor type, and tumor stage of the sample object; obtaining blood routine information of the sample object before immunotherapy, the blood routine information at least including: neutrophil-to-lymphocyte ratio; generating multiple input feature subsets and prediction index data sets based at least on the clinical data and blood routine information, the multiple input feature subsets including different feature numbers and feature categories; training multiple prediction models based on each input feature subset and prediction index data set in the multiple input feature subsets, so as to determine a matching candidate prediction model for each input feature subset; selecting a target prediction model from the candidate prediction models based on the user's clinical data, blood routine information before immunotherapy, sequencing data, and the user's input operation, and constructing user input features; and generating a prediction result about the user's immunotherapy benefit rate via the target prediction model based on the constructed user input features.

[0006] In some embodiments, generating multiple input feature subsets and predictive indicator data sets includes: generating multiple input feature subsets and predictive indicator data sets based on the clinical data, blood routine information, and tumor-related gene mutation data, wherein the tumor-related gene mutation data is obtained based on sequencing data of the sample object; the method also includes: sorting the immunotherapy benefit rate data for different treatment regimens for the same user from large to small; and presenting the sorting results.

[0007] In some embodiments, the method for predicting the immunotherapy benefit rate also includes: sorting the immunotherapy benefit rate data of different treatment regimens for users participating in Phase I clinical trials from large to small; selecting users whose scheduled immunotherapy is ranked first to form a Phase I immune benefit user subset; and generating an immunotherapy benefit rate filtering threshold condition based on the predicted results of the immunotherapy benefit rate of users in the Phase I immune benefit user subset, so as to screen the users enrolled in Phase II of the clinical trial based on the generated immunotherapy benefit rate filtering threshold condition.

[0008] In some embodiments, selecting a target prediction model from candidate prediction models, and constructing user input features includes: selecting a candidate prediction model from multiple candidate prediction models as a target prediction model based on user input operations on clinical data, blood routine information before immunotherapy, and sequencing data; presenting multiple user input windows associated with feature categories based on feature categories included in an input feature subset matching the selected candidate prediction model; and generating user input features based on user operation data and feature categories on the user input windows.

[0009] In some embodiments, determining a matching candidate prediction model for each input feature subset includes: generating a comprehensive weighted prediction model based on multiple prediction models and multiple weights corresponding to the multiple prediction models; for each input feature subset, respectively calculating the area curve value under the receiver operating characteristic of each prediction model and the area curve value under the receiver operating characteristic of the comprehensive weighted prediction model; comparing the area curve value under the receiver operating characteristic of each prediction model and the area curve value under the receiver operating characteristic of the comprehensive weighted prediction model; and determining a candidate prediction model that matches each input feature subset based on the comparison result.

[0010] In some embodiments, determining a matching candidate prediction model for each input feature subset includes: calculating the area curve value under the receiver operating characteristic of each prediction model for each input feature subset; and selecting the prediction model with the best area curve value under the receiver operating characteristic as the candidate prediction model.

[0011] In some embodiments, the routine blood information before immunotherapy also includes: platelet content data, hemoglobin content data, and albumin content data; the attribute information of the sample object includes: age, gender, height, and weight of the sample object.

[0012] In some embodiments, generating multiple input feature subsets and predictive indicator data sets includes: based on the clinical data, obtaining the sample subject's immunotherapy efficacy status data, overall survival data, progression-free survival months, endpoint event data related to patient survival time, and event data occurring within the time period after the patient receives treatment without disease progression or death from any cause, so as to generate a predictive indicator data set.

[0013] In some embodiments, obtaining the efficacy status data of immunotherapy for the sample object includes: obtaining the size information of all pathological lymph nodes of the sample object before and after treatment; in response to determining whether the short diameter of all pathological lymph nodes of the sample object after treatment is reduced to less than a first predetermined diameter threshold, determining that the efficacy status data of the sample object is "complete remission"; and in response to determining that the sum of the target lesion diameters of the sample object after treatment is reduced by a ratio exceeding a first predetermined ratio threshold compared with the baseline level, determining that the efficacy status data of the sample object is "partial remission".

[0014] In some embodiments, obtaining the efficacy status data of immunotherapy for the sample object also includes: determining the minimum value of the sum of all measured target lesion diameters during the treatment of the sample object; in response to determining that the increase ratio of the sum of the target lesion diameters after treatment of the sample object relative to the minimum value exceeds a second predetermined ratio threshold, and the increase value of the sum of the target lesion diameters after treatment relative to the minimum value exceeds a first size increase threshold, determining that the efficacy status data of the sample object is "disease progression", and the second predetermined ratio threshold is less than the first predetermined ratio threshold; in response to determining that the sum of the target lesion diameters after treatment of the sample object is reduced by a ratio less than the first predetermined ratio threshold compared with the baseline level, or the increase ratio of the sum of the target lesion diameters after treatment of the sample object relative to the minimum value is less than the second predetermined ratio threshold, determining that the efficacy status data of the sample object is "stable disease".

[0015] In some embodiments, obtaining tumor-related gene mutation data based on sequencing data of the sample object includes: obtaining the following items based on the sequencing data of the sample object: tumor mutation load score, microsatellite instability data, and whether the target gene exists.

[0016] In some embodiments, the tumor mutation burden score is obtained by: calculating the number of gene mutations detected for somatic gene coding errors, base substitutions, gene insertions or deletions based on the sequencing data of the target object; and dividing the number of gene mutations by the length of the targeted sequencing coverage fragment to obtain the tumor mutation burden score.

[0017] In some embodiments, generating multiple input feature subsets and predictive indicator data sets includes: generating a first input feature subset, the feature categories included in the first input feature subset are: cancer type, tumor stage, neutrophil-to-lymphocyte ratio, and treatment regimen; generating a second input feature subset, the feature categories included in the second input feature subset are: cancer type, tumor stage, neutrophil-to-lymphocyte ratio, TMB, treatment regimen, and target gene; generating a third input feature subset, the feature categories included in the third input feature subset are: age, gender, pre-immune systemic treatment, cancer type, tumor stage, hemoglobin, albumin, neutrophil-to-lymphocyte ratio, platelets, treatment regimen, number of lines, and target gene; generating a fourth input feature subset, the feature categories included in the fourth input feature subset are: age, gender, pre-immune systemic treatment, cancer type, tumor stage, hemoglobin, albumin, neutrophil-to-lymphocyte ratio, platelets, TMB, treatment regimen, number of lines, and target gene.

[0018] According to a second aspect of the present invention, a computing device is also provided, comprising: a memory configured to store one or more computer programs; and a processor coupled to the memory and configured to execute the one or more programs so that the device performs the method of the first aspect of the present invention.

[0019] According to a third aspect of the present invention, a non-transitory computer-readable storage medium is further provided, wherein machine-executable instructions are stored on the non-transitory computer-readable storage medium, and when the machine-executable instructions are executed, the machine executes the method of the first aspect of the present invention.

[0020] This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the invention, nor is it intended to limit the scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A schematic diagram of a computing device for implementing a method for predicting the benefit rate of immunotherapy according to an embodiment of the present invention is shown.

[0022] Figure 2 A flow chart of a method for predicting the benefit rate of immunotherapy according to an embodiment of the present invention is shown.

[0023] Figure 3 A schematic diagram of an interface including multiple user input windows according to an embodiment of the present invention is shown.

[0024] Figure 4 A flowchart of a method for determining a candidate prediction model according to an embodiment of the present invention is shown.

[0025] Figure 5 A flow chart of a method for obtaining immunotherapy efficacy status data of a sample object according to an embodiment of the present invention is shown.

[0026] Figure 6 A flow chart of a method for screening users for enrollment in Phase II clinical trial according to an embodiment of the present invention is shown.

[0027] Figure 7 A schematic diagram of a ROC curve of an optimized target prediction model according to some embodiments of the present invention is shown.

[0028] Figure 8 A schematic diagram of evaluation indicators of an optimized target prediction model according to some embodiments of the present invention is shown.

[0029] Fig. 9A schematic diagram of a ROC curve of an optimized target prediction model according to some other embodiments of the present invention is shown.

[0030] Fig.10 A schematic diagram of evaluation indicators of an optimized target prediction model according to some other embodiments of the present invention is shown.

[0031] Fig.11 A block diagram of a computing device suitable for implementing an embodiment of the present invention is schematically shown.

[0032] In the various drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0033] The preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0034] As used herein, the term "including" and its variations mean open inclusion, i.e., "including but not limited to". Unless otherwise stated, the term "or" means "and / or". The term "based on" means "based at least in part on". The terms "an example embodiment" and "an embodiment" mean "at least one example embodiment". The term "another embodiment" means "at least one additional embodiment". The terms "first", "second", etc. may refer to different or the same objects.

[0035] As described above, the shortcomings of existing methods for evaluating the therapeutic benefit of different combinations of immunotherapy regimens are that they rely on individual experience and lack objectivity and accuracy.

[0036] In order to at least partially solve one or more of the above problems and other potential problems, an exemplary embodiment of the present invention proposes a method for predicting the benefit rate of immunotherapy. Based on the clinical data and blood routine information of the sample object, a plurality of input feature subsets and prediction index data sets with different feature categories and quantities are generated; based on each input feature subset and prediction index data set in the plurality of input feature subsets, a plurality of prediction models are trained respectively, so as to determine a matching candidate prediction model for each input feature subset. The present invention can predict the benefit rate of immunotherapy based on conventional, limited patient clinical and detection index data such as clinical and blood routine, which is conducive to improving the accuracy of the prediction results. In addition, the present invention can select matching candidate prediction models for input feature data of different feature categories and quantity levels, improves the adaptability of candidate prediction models to different individual feature differences, and makes the prediction results applicable to different types of cancer, and is not limited by the individual subjective experience of doctors in specific types of cancer. In addition, by selecting a target prediction model from candidate prediction models based on the user's clinical data, blood routine information before immunotherapy, sequencing data, and user input operations, and constructing user input features; and based on the constructed user input features, generating prediction results about the user's immunotherapy benefit rate through the target prediction model, the present invention can match the determination of the target prediction model with the user's existing, actual input features, and improve the accuracy and convenience of predicting the immunotherapy benefit rate. Therefore, the present invention can objectively and accurately predict the immunotherapy benefit rate for different types of cancer.

[0037] Figure 1 A schematic diagram of a computing device 100 for implementing a method for predicting the benefit rate of immunotherapy according to an embodiment of the present invention is schematically shown. Figure 1As shown, the computing device 100 may have one or more processing units, including special processing units such as a graphics processing unit (GPU), a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), or a general-purpose computing on graphics processing units (GPGPU), and a general-purpose processing unit such as a CPU. In some embodiments, the computing device 100 is also configured with at least: a clinical data acquisition unit 102, a blood routine information acquisition unit 104, a gene mutation data acquisition unit 106, an input feature subset and a prediction index data set generation unit 108, a candidate prediction model determination unit 110, a target prediction model selection unit 112, and an immunotherapy benefit rate prediction result generation unit 114. In some embodiments, the clinical data acquisition unit 102, the routine blood information acquisition unit 104, the gene mutation data acquisition unit 106, the input feature subset and prediction indicator data set generation unit 108, the candidate prediction model determination unit 110, the target prediction model selection unit 112, and the immunotherapy benefit rate prediction result generation unit 114 are, for example, program modules running on the computing device 100.

[0038] The clinical data acquisition unit 102 is used to acquire clinical data of the sample object, and the clinical data at least includes attribute information, tumor type, and tumor stage of the sample object.

[0039] The blood routine information acquisition unit 104 is used to acquire the blood routine information of the sample object before immunotherapy, and the blood routine information at least includes: the ratio of neutrophils to lymphocytes.

[0040] Regarding the gene mutation data acquisition unit 106, it is used to acquire tumor-related gene mutation data based on the sequencing data of the sample object, and the tumor-related gene mutation data at least includes a tumor mutation load score.

[0041] Regarding the input feature subset and prediction index data set generating unit 108, it is used to generate multiple input feature subsets and prediction index data sets based on the clinical data, blood routine information, and tumor-related gene mutation data, and the multiple input feature subsets include different numbers of features and feature categories.

[0042] The candidate prediction model determination unit 110 is used to train multiple prediction models based on each input feature subset and a prediction index data set in multiple input feature subsets, so as to determine a matching candidate prediction model for each input feature subset.

[0043] The target prediction model selection unit 112 is used to select a target prediction model from candidate prediction models based on the user's clinical data, blood routine information before immunotherapy, sequencing data and user input operations, and to construct user input features.

[0044] The prediction result generating unit 114 for immunotherapy benefit rate is used to generate a prediction result for the immunotherapy benefit rate of the user via a target prediction model based on the constructed user input features.

[0045] The following will be combined Figure 2 and Figure 3 A method for predicting the rate of benefit from immunotherapy according to an embodiment of the present invention is described. Figure 2 A flow chart of a method 200 for predicting the benefit rate of immunotherapy according to an embodiment of the present invention is shown. Figure 3 FIG. 2 is a schematic diagram of an interface 300 including multiple user input windows according to an embodiment of the present invention. It should be understood that the method 200 may be implemented in, for example, Fig.11 The described computing device 1100 may also be executed at Figure 1 The method 200 is executed at the described computing device 100. It should be understood that the method 200 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0046] At step 202 , the computing device 100 obtains clinical data of the sample object, where the clinical data at least includes attribute information, tumor type, and tumor stage of the sample object.

[0047] Regarding the attribute information of the sample object, in some embodiments, it also includes, for example: the age, gender, height, and weight of the sample object.

[0048] At step 204 , the computing device 100 obtains the routine blood information of the sample object before immunotherapy, and the routine blood information at least includes: a ratio of neutrophils to lymphocytes.

[0049] Regarding routine blood information, in some embodiments, it also includes, for example: platelet content, hemoglobin content, and albumin content.

[0050] In some embodiments, the computing device 100 obtains tumor-related gene mutation data based on sequencing data of the sample object, and the tumor-related gene mutation data at least includes a tumor mutation load score.

[0051] Regarding the method of obtaining tumor-related gene mutation data, it includes, for example: performing targeted amplification on the tissue sample and blood control of the sample subject, and sequencing the obtained genome sequence library on the machine. The sequencing read length will be mapped to the human genome reference sequence, and the variation search will be performed using a computer analysis program such as Varscan. After the candidate gene mutations found are filtered for reliability, the tumor-related gene mutation data are obtained.

[0052] Regarding tumor-related gene mutation data, in some embodiments, in addition to the tumor mutation load score, it also includes: tumor mutation load score, microsatellite instability data, and the presence or absence of the target gene.

[0053] Regarding the tumor mutation load score, it is obtained, for example, via the following method: the computing device 100 calculates the number of gene mutations detected such as somatic gene coding errors, base substitutions, gene insertion or deletion errors based on the sequencing data of the target object; and divides the number of gene mutations by the length of the targeted sequencing coverage fragment to obtain the tumor mutation load score (TMB).

[0054] Regarding the microsatellite instability data, the computing device 100 uses the calculation result of MSIsensor to make a judgment, for example. Specifically, if the MSI score is less than 3, the sample object is determined to be "microsatellite stable"; if the MSI score is between 3 and 10, the sample object is determined to be "intermediate"; if the MSI score is greater than 10, the sample object is determined to be "microsatellite unstable".

[0055] At step 206, the computing device 100 generates a plurality of input feature subsets and prediction indicator data sets based at least on the clinical data and routine blood test information, wherein the plurality of input feature subsets include different numbers of features and feature categories. In some embodiments, the computing device 100 generates a plurality of input feature subsets and prediction indicator data sets based at least on the clinical data, routine blood test information, and tumor-related gene mutation data.

[0056] Regarding the multiple input feature subsets, in some embodiments, they include, for example, 4 input feature subsets. Each input feature subset includes a different number of features and feature categories. Each input feature subset corresponds to a different matching candidate prediction model.

[0057] Regarding the method of separately generating multiple input feature subsets, for example, it includes: converting the acquired cancer type data, tumor stage data, neutrophil-to-lymphocyte ratio, and treatment regimen data about the sample object into corresponding feature values ​​to generate a first input feature subset; converting the cancer type data, tumor stage data, neutrophil-to-lymphocyte ratio, tumor mutation load score, treatment regimen data, and target gene data about the sample object into corresponding feature values ​​to generate a second input feature subset; converting the age, gender, pre-immune systemic treatment, and treatment regimen data about the sample object into corresponding feature values. The data, cancer type data, tumor stage data, hemoglobin, albumin, neutrophil-to-lymphocyte ratio, platelet, treatment regimen data, number of lines, and target gene data are converted into corresponding feature values ​​to generate a third input feature subset; and the age, gender, pre-immune system treatment data, cancer type data, tumor stage data, hemoglobin value, albumin value, neutrophil-to-lymphocyte ratio, platelet value, tumor mutation load score, treatment regimen, number of lines, and target gene data of the sample subject are converted into corresponding feature values ​​to generate a fourth input feature subset.

[0058] Regarding the method of converting into corresponding characteristic values, it includes, for example: determining the numerical values ​​indicated by the age (Age), hemoglobin value (HGB), albumin value (Albumin), neutrophil-to-lymphocyte ratio (NLR), and platelet value (Platelets) of the sample object as corresponding characteristic values.

[0059] The sex (Sex) data of the sample object is converted into sex characteristic values ​​"1" / "0" according to the sex of the sample object, "male" / "female", respectively.

[0060] Determine whether the sample subject has undergone systemic treatment (Systemic_therapy_history) before immunotherapy. If it is determined that the sample subject has undergone systemic treatment, the converted immunotherapy feature value is "1". If it is determined that the sample subject has not undergone systemic treatment, the converted immunotherapy feature value is "0".

[0061] Based on the tumor stage data, the tumor stage characteristic values ​​are converted into numerical values ​​of "1", "2", "3", and "4" according to their "stage I", "stage II", "stage III", and "stage IV" respectively.

[0062] In response to determining that the calculated TMB value is greater than or equal to a predetermined tumor mutation burden score threshold, the TMB is converted to a TMB feature value "TMB-H", and in response to determining that the calculated TMB value is less than the predetermined tumor mutation burden score threshold, the TMB is converted to a TMB feature value "TMB-L".

[0063] Determine whether the target gene exists, and in response to determining that the target gene exists, determine the target gene characteristic value to be "1"; in response to determining that the target gene does not exist, determine the target gene characteristic value to be "0".

[0064] The number of treatment lines (Time) is converted into treatment line feature values ​​"0", "1", "2", and "3" according to whether it is neoadjuvant, 1st line, 2nd line, or ≥3rd line, respectively.

[0065] Regarding the efficacy status, if the efficacy status data is determined to be "complete remission" or "partial remission", the conversion efficacy status characteristic value is "1"; if the efficacy status data is determined to be "disease progression" or "stable disease", the conversion efficacy status characteristic value is "0".

[0066] Regarding the treatment plans, they are converted into treatment plan characteristic values ​​of "1", "2", "3", "4", "5" and "6" according to "chemotherapy plus immunotherapy", "immunotherapy alone", "dual immunotherapy", "targeted plus immunotherapy", "chemotherapy, targeted plus immunotherapy" and "others" respectively.

[0067] Regarding the immunization regimen, it is converted into the immunization regimen characteristic values ​​of "1" and "2" according to whether it is "single drug" or "combination chemotherapy / targeted" respectively.

[0068] Table 1 below illustrates the corresponding feature values ​​of each feature category included in the input feature subset of the sample object.

[0069] Table 1

[0070]

[0071] It should be understood that the above-mentioned cancer types, including various types, are not limited to the non-small cell lung cancer mentioned in the above examples.

[0072] Regarding the method for generating a predictive indicator data set, it includes, for example, the computing device 100 obtaining, based on the clinical data, the efficacy status data of immunotherapy of the sample object, the overall survival data, the number of progression-free survival months, the endpoint event data related to the patient's survival time, and the event data occurring during the time when the patient does not progress in the disease after receiving treatment or dies from any cause, so as to generate a predictive indicator data set.

[0073] At step 208 , the computing device 100 trains a plurality of prediction models based on each input feature subset of the plurality of input feature subsets and the prediction index dataset, respectively, so as to determine a matching candidate prediction model for each input feature subset.

[0074] Regarding multiple prediction models, for example, they are multiple different types of prediction models. In some embodiments, multiple prediction models are constructed based on machine learning frameworks such as neural network (NN_TORCH), gradient boosting machine (GBM), CATBoosting, extreme gradient boosting (XGB), fast learning neural network (FASTAI NN), random forest (RF), ExtraTrees classification, K-nearest neighbor algorithm (KNN) model, etc.

[0075] Regarding the method of training multiple prediction models separately, it includes, for example: first, the computing device 100 divides each input feature subset into a training set and a test set in a 4:1 ratio, the training set is used for the training of the aforementioned multiple prediction models, and the test set is used to evaluate the generalization ability of the multiple prediction models. For example, the computing device 100 calls the train_test_split function of the scikit-learn library to split the data set 4:1 so that it can be used for training and testing respectively. Then, the computing device 100 calls the TabularPredictor.fit function of autogluon to train the training set, and sets the function parameters (presets = 'best_quality', auto_stack = True, refit_full = True) to train multiple prediction models. The optimization of the parameters of the prediction model can be carried out by grid search, gradient descent, combinatorial optimization and other methods. In some embodiments, each prediction model is trained with replacement sampling, and the model hyperparameters are fine-tuned through 5-fold cross validation, and their various performance indicators are evaluated.

[0076] Regarding the method for determining the matching candidate prediction model, it includes, for example, multiple methods. For example, in some embodiments, the computing device 100 calculates the evaluation index of each prediction model in the multiple prediction models for each input feature subset in the multiple input feature subsets (the evaluation index is, for example, but not limited to, the area curve value under the receiver operating characteristic); and selects the prediction model with the best evaluation index (the evaluation index is, for example, but not limited to, the area curve value under the receiver operating characteristic) as the candidate prediction model.

[0077] Regarding evaluation indicators, they can also be accuracy, precision, recall, F-score, confusion matrix, log loss, etc.

[0078] For another example, in some other embodiments, the computing device 100 calculates the area curve value under the receiver operating characteristic of each prediction model for each input feature subset; and selects the prediction model with the best area curve value under the receiver operating characteristic as the candidate prediction model.

[0079] At step 210 , the computing device 100 trains a plurality of prediction models based on each input feature subset of the plurality of input feature subsets and the prediction index dataset, respectively, so as to determine a matching candidate prediction model for each input feature subset.

[0080] For example, the computing device 100 selects a candidate prediction model from multiple candidate prediction models as a target prediction model based on the user's input operations on clinical data, blood routine information before immunotherapy, and sequencing data; presents multiple user input windows associated with feature categories based on the feature categories included in the input feature subset matching the selected candidate prediction model; and generates user input features based on the user's operation data and feature categories on the user input windows.

[0081] like Figure 3 As shown, one candidate prediction model selected as the target prediction model from among the multiple candidate prediction models is, for example, the second candidate prediction model (for example, abbreviated as "Model_2"), as shown in FIG. Figure 3 Indicated by the mark 310. Figure 3 The feature categories of the input feature subset matching the selected second candidate prediction model are shown in Figure 1, including: cancer type, cancer stage, neutrophil-to-lymphocyte ratio (or abbreviated as "NLR"), tumor mutation burden (or abbreviated as "TMB"), and the presence of target genes (Driver_gene). Below each feature category, the user's operation data for the user input window associated with each feature category is shown.

[0082] At step 212 , the computing device 100 generates a prediction result regarding the immunotherapy benefit rate of the user based on the constructed user input features via the target prediction model.

[0083] The prediction result of the immunotherapy benefit rate of the user includes, for example, the ranking data of different treatment plans for the same user according to the immune effect, and the lowest immunotherapy benefit rate data. Figure 3 Indicated by the mark 320.

[0084] In some embodiments, the computing device 100 calculates the ranking data of different treatment schemes according to immune effects for the same user for different tumor stages, respectively; then determines the correlation between the tumor stage and the ranking data of different treatment schemes according to immune effects, thereby predicting the corresponding treatment timing of different treatment schemes based on the correlation.

[0085] In the above scheme, the present invention generates a plurality of input feature subsets and prediction index data sets with different feature categories and quantities based on the clinical data and blood routine information of the sample object; based on each input feature subset and prediction index data set in the plurality of input feature subsets, multiple prediction models are trained respectively, so as to determine a matching candidate prediction model for each input feature subset. The present invention can predict the immunotherapy benefit rate based on conventional and limited patient clinical and test index data such as clinical and blood routine, which is conducive to improving the accuracy of the prediction results. In addition, the present invention can select matching candidate prediction models for input feature data of different feature categories and quantity levels, which improves the adaptability of candidate prediction models to the differences in different individual characteristics, and makes the prediction results applicable to different types of cancer, and is not limited by the individual subjective experience of doctors in specific types of cancer. In addition, by selecting a target prediction model from candidate prediction models based on the user's clinical data, blood routine information before immunotherapy, sequencing data, and user input operations, and constructing user input features; and based on the constructed user input features, generating prediction results about the user's immunotherapy benefit rate through the target prediction model, the present invention can match the determination of the target prediction model with the user's existing, actual input features, and improve the accuracy and convenience of predicting the immunotherapy benefit rate. Therefore, the present invention can objectively and accurately predict the immunotherapy benefit rate for different types of cancer.

[0086] In some embodiments, method 200 may also include: the computing device 100 sorts the immunotherapy benefit rate data for different treatment options for the same user from large to small; and presents the sorting results. It should be understood that the predicted results of the immunotherapy benefit rate data are not data used to directly identify, study and determine the cause or lesion status of a living human or animal body. Rather, it is a probability calculation of treatment benefit. In some embodiments, the predicted results of the immunotherapy benefit rate of a user generated by the present invention can also be used to determine the entry conditions for clinical trials. The following will be combined with Figure 6 The details are explained below and will not be repeated here.

[0087] The following will be combined Figure 4 A method 400 for determining a candidate prediction model according to an embodiment of the present invention is described. Figure 4FIG. 4 is a flowchart of a method 400 for determining a candidate prediction model according to an embodiment of the present invention. It should be understood that the method 400 may be implemented, for example, in Fig.11 The described computing device 1100 may also be executed at Figure 1 The method 400 is executed at the described computing device 100. It should be understood that the method 400 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0088] At step 402 , the computing device 100 generates a comprehensive weighted prediction model based on multiple prediction models and multiple weights corresponding to the multiple prediction models.

[0089] For example, the computing device 100 is configured with 13 prediction models that are respectively constructed based on different machine learning models. The computing device 100, for example, trains the 13 prediction models by the method of separately training multiple prediction models as exemplified above, so as to obtain 13 trained prediction models. Then, the computing device 100 constructs a comprehensive weighted prediction model based on the 13 trained prediction models and 13 weights according to the method shown in the following formula (1).

[0090] y=w1*x1+w2*x2+…w13*x13(1)

[0091] In the above formula (1), x1, x2...x13 represent the output data of 13 trained prediction models under data set A. w1, w2...w13 represent the weights corresponding to the 13 trained models. y represents the output data of the comprehensive weighted prediction model.

[0092] After constructing the comprehensive weighted prediction model, the computing device 100 uses the data set A (for example, based on each input feature subset and the prediction index data set in the multiple input feature subsets) to train the constructed comprehensive weighted prediction model, so as to obtain a trained comprehensive weighted prediction model.

[0093] At step 404, the computing device 100 calculates, for each input feature subset, the area curve value under the receiver operating characteristic of each prediction model and the area curve value under the receiver operating characteristic of the comprehensive weighted prediction model.

[0094] It should be understood that the receiver operating characteristic curve (ROC) means a probability curve, whose horizontal axis is the "False Positive Rate" (FPR) and the vertical axis is the "True Positive Rate" (TPR). AUC-ROC is usually used to describe the area under the ROC curve, which is an evaluation indicator to measure the quality of a binary classification model, indicating the probability that the predicted positive example is ranked before the negative example. That is, the degree to which the positive and negative classes can be correctly classified. Generally speaking, the higher the AUC-ROC value, the better the model's predictive performance. For example, an AUC-ROC value close to 1 means that it has good separability. A bad model has an AUC-ROC value close to 0, which means that it has the worst separability.

[0095] There may be multiple methods for calculating the area curve value (or simply referred to as "AUC_ROC value") under the receiver operating characteristic of each prediction model or comprehensive weighted prediction model. In some embodiments, the computing device 100 calculates the AUC_ROC value using, for example, the trapezoidal rule. That is, the area under the entire curve is approximated by the area of ​​the trapezoid formed by adjacent points on the ROC curve. In some embodiments, the computing device 100, for example in the scikit-learn library of Python, may use the roc_auc_score function to calculate the AUC_ROC value.

[0096] At step 406 , computing device 100 compares the area curve value under the receiver operating characteristic of each prediction model with the area curve value under the receiver operating characteristic of the comprehensive weighted prediction model.

[0097] At step 408 , the computing device 100 determines a candidate prediction model that matches each input feature subset based on the comparison result.

[0098] In some embodiments, if the computing device 100 determines that for the first to fourth input feature subsets mentioned above, the AUC_ROC values ​​of the comprehensive weighted prediction model are greater than the corresponding AUC_ROC values ​​of each of the 13 prediction models, then the comprehensive weighted prediction model is determined to be a candidate prediction model that matches each input feature subset.

[0099] By adopting the above means, the present invention can select the model with the best classification performance as the candidate prediction model for each input feature subset, thereby improving the accuracy of the prediction result of the present invention.

[0100] The following will be combined Figure 5A method 500 for obtaining efficacy status data of immunotherapy of a sample object according to an embodiment of the present invention is described. Figure 5 FIG. 5 is a flow chart showing a method 500 for obtaining immunotherapy efficacy status data of a sample object according to an embodiment of the present invention. It should be understood that the method 500 can be, for example, Fig.11 The described computing device 1100 may also be executed at Figure 1 The method 500 is executed at the described computing device 100. It should be understood that the method 500 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0101] At step 502 , the computing device 100 obtains size information of all pathological lymph nodes of the sample object before and after treatment.

[0102] At step 504 , if the computing device 100 determines whether the short diameters of all pathological lymph nodes of the sample subject after treatment are reduced to less than a first predetermined diameter threshold, the therapeutic effect status data of the sample subject is determined to be “complete remission”.

[0103] Regarding all pathological lymph nodes, they include target nodes and non-target nodes.

[0104] Regarding the first predetermined diameter threshold, it is, for example but not limited to, 10 mm.

[0105] At step 506 , if the computing device 100 determines that the sum of the target lesion diameters of the sample subject after treatment is reduced by a ratio exceeding a first predetermined ratio threshold compared to the baseline level, the therapeutic efficacy status data of the sample subject is determined to be “partial remission”.

[0106] Regarding the first predetermined ratio threshold, it is, for example but not limited to, 30%.

[0107] At step 508, computing device 100 determines the minimum value of the sum of all measured target lesion diameters during treatment of the sample subject.

[0108] At step 510, if the computing device 100 determines that the increase ratio of the sum of the target lesion diameters of the sample object after treatment relative to the minimum value exceeds the second predetermined ratio threshold, and the increase value of the sum of the target lesion diameters after treatment relative to the minimum value exceeds the first size increase threshold, it is determined that the efficacy status data of the sample object is "disease progression", and the second predetermined ratio threshold is less than the first predetermined ratio threshold.

[0109] The second predetermined ratio threshold is, for example but not limited to, 20%.

[0110] Regarding the first size increase threshold, it is, for example but not limited to, 5 mm.

[0111] At step 512, if the computing device 100 determines that the ratio of the reduction of the sum of the target lesion diameters of the sample object after treatment compared with the baseline level is less than a first predetermined ratio threshold, or the ratio of the increase of the sum of the target lesion diameters of the sample object after treatment relative to the minimum value is less than a second predetermined ratio threshold, it is determined that the therapeutic efficacy status data of the sample object is "stable disease".

[0112] By adopting the above means, the present invention can accurately determine the efficacy status data of the immunotherapy of the sample object.

[0113] The following will be combined Figures 6 to 10 A method 600 for screening users for enrollment in a Phase II clinical trial according to an embodiment of the present invention is described. Figure 6 A flow chart of a method 600 for screening users for enrollment in a Phase II clinical trial according to an embodiment of the present invention is shown. Figure 7 A schematic diagram of a ROC curve 700 of an optimized target prediction model according to some embodiments of the present invention is shown. Figure 8 A schematic diagram of evaluation indicators of an optimized target prediction model according to some embodiments of the present invention is shown. Fig. 9 A schematic diagram of a ROC curve 900 of an optimized target prediction model according to some other embodiments of the present invention is shown. Fig.10 Schematic diagram of evaluating indicators of target prediction models through optimization according to some other embodiments of the present invention is shown. It should be understood that method 600 can be used, for example, in Fig.11 The described computing device 1100 may also be executed at Figure 1 The method 600 is executed at the described computing device 100. It should be understood that the method 600 may also include additional actions not shown and / or may omit the actions shown, and the scope of the present invention is not limited in this respect.

[0114] At step 602 , the computing device 100 sorts the immunotherapy benefit rate data of different treatment regimens for users participating in the Phase I clinical trial from large to small.

[0115] As for the user, it is, for example, a patient participating in Phase I of a clinical trial.

[0116] Regarding user input features, they are, for example, constructed based on clinical and test parameters of patients participating in Phase I clinical trials. In some embodiments, the user is, for example, based on patients participating in Phase I clinical trials and nearly a thousand pan-cancer patient data using Chowell MSK Train. For example, the characteristics of more than 250 patients participating in Phase I clinical trials include: TMB, tumor type, tumor stage, cancer type, age, whether there has been systemic treatment before or what specific systemic treatment has been added, blood cell biochemical test results nearly one month before enrollment (at least including albumin, neutrophil to lymphocyte ratio), tumor mutation load TMB. The nearly a thousand pan-cancer patient data of Chowell MSK Train include the following feature types: TMB, tumor type, cancer type, tumor stage, age, albumin before immunization and chemotherapy, and neutrophil to lymphocyte ratio.

[0117] Regarding the target prediction model, for example, the TMB, tumor type, cancer type, tumor stage, age, albumin, and neutrophil-to-lymphocyte ratio of existing enrolled patients in clinical trials and nearly a thousand pan-cancer patients in the Chowell MSK Train are used to construct an input feature subset, and then the constructed input feature subset is used to re-optimize the target prediction model.

[0118] like Figure 7 As shown, label 710 indicates the ROC curve of the target prediction model (e.g., Model_1) for the test set, and the area curve value under the receiver operating characteristic is 0.73. Label 712 indicates the ROC curve of the target prediction model (e.g., Model_1) for the training set, and the area curve value under the receiver operating characteristic is 0.80.

[0119] like Figure 8 As shown, mark 810 indicates the performance of the target prediction model for the test set on the corresponding evaluation index, and mark 812 indicates the performance of the target prediction model for the training set on the corresponding evaluation index. Figure 8 The evaluation indicators shown on the horizontal axis include, for example, accuracy, balanced accuracy, precision, recall, F1, MCC, and the area under the receiver operating characteristic curve value (ie, ROC_AUC).

[0120] For the target prediction model of the training set, the evaluation indicators are: ROC_AUC value is 0.796406064765707; accuracy is 0.7385892116182573; balanced accuracy is 0.7258168507310581; MCC is 0.424278038759879; F1 is 0.608695652173913; precision is 0.5414364640883977; recall rate is 0.6950354609929078.

[0121] For the target prediction model of the test set, the evaluation indicators are: ROC_AUC value is 0.7282855751278513; accuracy is 0.7126213592233009; balanced accuracy is 0.6689159022647941; MCC is 0.3101250551814306; F1 is 0.5; precision is 0.4378698224852071; recall rate is 0.5826771653543307.

[0122] The following example shows another implementation of constructing an input feature subset based on tumor type, cancer type, tumor stage, age, albumin, and neutrophil-to-lymphocyte ratio. Compared with the previous embodiment, the input feature subset is constructed without TMB. It should be understood that this embodiment is used, for example, in the case where pharmaceutical companies often do not perform NGS.

[0123] Among them, the characteristics of patients participating in the Phase I clinical trial include: age, tumor type, tumor stage, whether they have received systemic treatment or what specific systemic treatment they have received, and the results of blood cell biochemical tests one month before enrollment (including at least albumin and neutrophil-to-lymphocyte ratio). The characteristics of nearly 1,000 pan-cancer patient data from the Chowell MSK train include: tumor type, tumor stage, cancer type, age, albumin, and neutrophil-to-lymphocyte ratio.

[0124] like Fig. 9 As shown, label 910 indicates the ROC curve of the target prediction model for the test set, and the area curve value under the receiver operating characteristic is 0.71. Label 912 indicates the ROC curve of the target prediction model for the training set, and the area curve value under the receiver operating characteristic is 0.75.

[0125] like Fig.10 As shown, mark 1010 indicates the performance of the target prediction model for the test set on the corresponding evaluation index, and mark 1012 indicates the performance of the target prediction model for the training set on the corresponding evaluation index. Fig.10The evaluation indicators shown on the horizontal axis include, for example, accuracy, balanced accuracy, precision, recall, F1, MCC, and the area under the receiver operating characteristic curve value (ie, ROC_AUC).

[0126] For the target prediction model of the training set, the evaluation indicators are: ROC_AUC value is 0.7503015744264887; accuracy is 0.6493775933609959; balanced accuracy is 0.6690064682514922; MCC is 0.3078618532795775; F1 is 0.5444743935309975; precision is 0.4391304347826087; and recall is 0.7163120567375887.

[0127] For the target prediction model of the test set, the evaluation indicators are: ROC_AUC value is 0.7118272587060638; accuracy is 0.654368932038835; balanced accuracy is 0.6646846334929783; MCC is 0.2862269603753328; F1 is 0.4943181818181818; precision is 0.386666666666666666; recall rate is 0.6850393700787402.

[0128] It can be seen that for the two embodiments with and without TMB features, the evaluation indication shows that the performance of the target prediction model is good. It should be understood that the evaluation index of the target prediction model for the test set will be further improved as the number of actual applications increases.

[0129] At step 604 , the computing device 100 selects the user who is ranked first in the scheduled immunotherapy to form a first-stage immunotherapy benefit user subset.

[0130] For example, the current clinical trial phase I and phase II involve scheduled immunotherapy. The computing device 100 determines whether the first one in the ranking results of the predicted immunotherapy benefit rate data of the users participating in the clinical trial phase I is the scheduled immunotherapy. If it is determined that the first one in the ranking results of the predicted immunotherapy benefit rate data of the users participating in the clinical trial phase I is the scheduled immunotherapy, the user is screened as a phase I immune benefit user subset. For example, the phase I immune benefit user subset includes the identification data of the user.

[0131] At step 606, the computing device 100 generates an immunotherapy benefit rate filtering threshold condition based on the predicted results of the immunotherapy benefit rate of users in the Phase I immune benefit user subset, so as to screen the users enrolled in the Phase II clinical trial based on the generated immunotherapy benefit rate filtering threshold condition.

[0132] For example, the immunotherapy benefit rate filtering threshold condition is that the immunotherapy benefit rate is 0.27. A patient who is about to be enrolled in the Phase II clinical trial has the following characteristics, for example: tumor type = lung adenocarcinoma, age = 64, neutrophil-to-lymphocyte ratio = 6.2 / 1.05 = 5.9, albumin = 62.4, whether or not he received chemotherapy before = yes, and the model predicts that the output probability is 0.732376. The computing device 100 determines that the patient's immunotherapy benefit rate is greater than the immunotherapy benefit rate filtering threshold condition 0.27. The patient is then included in the population to be enrolled in the Phase II clinical trial.

[0133] It should be understood that the traditional clinical trial has a single immune-related index for inclusion and exclusion, the corresponding efficacy of TMB and PD-L1 has not achieved the ideal effect, and the overall index lacks overall regularity. The present invention can accurately indicate the indication / screen candidate patients for inclusion by adopting the above-mentioned means.

[0134] Fig.11 The step diagram of a computing device 1100 suitable for implementing an embodiment of the present invention is schematically shown. The device 1100 may be a computer system for implementing Figure 2 , 4 , 5, 6, and the method 200, 400, 500, 600 shown in the apparatus. Fig.11 As shown, the device 1100 includes a central processing unit (CPU) 1101, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 1102 or loaded from a storage unit 1108 into a random access memory (RAM) 1103. In the RAM 1103, various programs and data required for the operation of the device 1100 can also be stored. The CPU 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0135] Multiple components in the device 1100 are connected to the I / O interface 1105, including: an input unit 1106, an output unit 1107, and a storage unit 1108. The processing unit 1101 performs the various methods and processes described above, such as performing methods 200, 400, 500, and 600. For example, in some embodiments, the methods 200, 400, 500, and 600 may be implemented as a computer software program, which is stored in a machine-readable medium, such as the storage unit 1108. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1100 via the ROM 1102 and / or the communication unit 1109. When the computer program is loaded into the RAM 1103 and executed by the CPU 1101, one or more operations of the methods 200, 400, 500, and 600 described above may be performed. Alternatively, in other embodiments, the CPU 601 may be configured to perform one or more actions of the methods 200 , 400 , 500 , 600 in any other appropriate manner (eg, by means of firmware).

[0136] It should be further noted that the present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0137] Computer readable storage medium can be a tangible device that can hold and store instructions used by an instruction execution device. Computer readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination thereof. More specific examples (non-exhaustive list) of computer readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove on which instructions are stored, and any suitable combination thereof. The computer readable storage medium used here is not interpreted as a transient signal itself, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagated by a waveguide or other transmission medium (for example, a light pulse by an optical fiber cable), or an electrical signal transmitted by a wire.

[0138] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0139] The computer program instructions for performing the operation of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present invention.

[0140] Here, various aspects of the present invention are described with reference to the flow chart and / or step diagram of the method, device (system), and computer program product according to an embodiment of the present invention. It should be understood that each side step of the flow chart and / or step diagram and the combination of the side steps in the flow chart and / or step diagram can be implemented by computer-readable program instructions.

[0141] These computer-readable program instructions can be provided to a processor in a voice interaction device, a general-purpose computer, a special-purpose computer, or a processing unit of other programmable data processing devices, thereby producing a machine, so that when these instructions are executed by a processing unit of a computer or other programmable data processing device, a device that implements the functions / actions specified in one or more steps in a flowchart and / or a step diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions enable a computer, a programmable data processing device, and / or other devices to work in a specific manner, so that a computer-readable medium storing instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more steps in a flowchart and / or a step diagram.

[0142] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more steps in the flowchart and / or step diagram.

[0143] The flowchart and step diagram in the accompanying drawings show the possible architecture, functions and operations of the equipment, methods and computer program products according to multiple embodiments of the present invention. In this regard, each square step in the flowchart or step diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the square steps can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive square steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square step in the step diagram and / or the flowchart, and the combination of the square steps in the step diagram and / or the flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0144] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

[0145] The above are only optional embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the benefit rate of immunotherapy, characterized in that: include: Acquire clinical data of the sample object, where the clinical data at least includes attribute information, tumor type, and tumor stage of the sample object; Obtaining the blood routine information of the sample subject before immunotherapy, the blood routine information at least includes: the ratio of neutrophils to lymphocytes; Generate multiple input feature subsets and prediction indicator data sets based at least on the clinical data and routine blood test information, wherein the multiple input feature subsets include different numbers of features and different feature categories; Based on each input feature subset in the plurality of input feature subsets and the prediction indicator dataset, a plurality of prediction models are trained respectively so as to determine a matching candidate prediction model for each input feature subset; Based on the user's clinical data, blood routine information before immunotherapy, sequencing data, and user input operations, a target prediction model is selected from candidate prediction models, and user input features are constructed; and Based on the constructed user input features, a prediction result about the user's immunotherapy benefit rate is generated through a target prediction model.

2. The method according to claim 1, characterized in that Generating multiple input feature subsets and prediction index data sets includes: generating multiple input feature subsets and prediction index data sets based on the clinical data, blood routine information, and tumor-related gene mutation data, wherein the tumor-related gene mutation data is obtained based on sequencing data of the sample object; The method further comprises: Sort the data of immunotherapy benefit rates of different treatment regimens for the same user from highest to lowest; and Presents the sorting results.

3. The method according to claim 1, characterized in that Also includes: The immunotherapy benefit rate data of different treatment regimens for users participating in Phase I clinical trials are sorted from high to low; Select the users who are the first to order immunotherapy to form a subset of users who benefit from the first phase of immunotherapy; as well as Based on the predicted results of the immunotherapy benefit rate of users in the Phase I immune benefit user subset, an immunotherapy benefit rate filtering threshold condition is generated, so as to screen the users enrolled in the Phase II clinical trial based on the generated immunotherapy benefit rate filtering threshold condition.

4. The method according to claim 2, characterized in that: Selecting a target prediction model from candidate prediction models and constructing user input features include: Based on the user's input operation on clinical data, blood routine information before immunotherapy, and sequencing data, selecting a candidate prediction model from multiple candidate prediction models as a target prediction model; Based on the feature categories included in the input feature subset matching the selected candidate prediction model, presenting a plurality of user input windows associated with the feature categories; and A user input feature is generated based on the user operation data on the user input window and the feature category.

5. The method according to claim 1, characterized in that: The candidate prediction models that are matched for each subset of input features include: generating a comprehensive weighted prediction model based on the multiple prediction models and the multiple weights corresponding to the multiple prediction models; For each subset of input features, the area under the receiver operating characteristic of each prediction model and the area under the receiver operating characteristic of the comprehensive weighted prediction model are calculated respectively; comparing the area under the receiver operating characteristic curve value of each prediction model with the area under the receiver operating characteristic curve value of the comprehensive weighted prediction model; and Based on the comparison results, candidate prediction models matching each input feature subset are determined.

6. The method according to claim 1, characterized in that The candidate prediction models that are matched for each subset of input features include: For each subset of input features, calculate the area under the receiver operating characteristic curve value for each prediction model; and The prediction model with the best area curve value under the receiver operating characteristic is selected as the candidate prediction model.

7. The method according to claim 1, characterized in that The blood routine information before immunotherapy also includes: platelet content data, hemoglobin content data, albumin content data; the attribute information of the sample object includes: age, gender, height, and weight of the sample object.

8. The method according to claim 1, characterized in that: Generate various input feature subsets and prediction indicator datasets including: Based on the clinical data, the efficacy status data of immunotherapy, overall survival data, progression-free survival months, endpoint event data related to patient survival time, and event data occurring during the time when the patient's disease does not progress or die from any cause after treatment are obtained for the sample subjects, so as to generate a predictive indicator data set.

9. The method according to claim 8, characterized in that The data on the efficacy of immunotherapy for sample subjects include: Obtaining the size information of all pathological lymph nodes of the sample subjects before and after treatment; In response to determining whether the short diameters of all pathological lymph nodes of the sample subject after treatment are reduced to less than a first predetermined diameter threshold, determining that the therapeutic efficacy status data of the sample subject is "complete remission"; and In response to determining that the sum of the target lesion diameters of the sample subject after treatment is reduced by a ratio exceeding a first predetermined ratio threshold compared with the baseline level, the efficacy status data of the sample subject is determined to be "partial remission".

10. The method according to claim 9, characterized in that The data on the efficacy of immunotherapy for the sample subjects also include: Determine the minimum value of the sum of all target lesion diameters measured during treatment of the sample subject; In response to determining that the increase ratio of the sum of the target lesion diameters of the sample subject after treatment relative to the minimum value exceeds a second predetermined ratio threshold, and the increase value of the sum of the target lesion diameters after treatment relative to the minimum value exceeds the first size increase threshold, determining that the therapeutic effect status data of the sample subject is "disease progression", and the second predetermined ratio threshold is less than the first predetermined ratio threshold; In response to determining that the ratio of the decrease in the sum of the target lesion diameters of the sample subject after treatment compared with the baseline level is less than a first predetermined ratio threshold, or the ratio of the increase in the sum of the target lesion diameters of the sample subject after treatment relative to the minimum value is less than a second predetermined ratio threshold, the therapeutic efficacy status data of the sample subject is determined to be "stable disease".

11. The method according to claim 2, characterized in that Based on the sequencing data of the sample object, the tumor-related gene mutation data is obtained, including: Based on the sequencing data of the sample subjects, the following items are obtained: tumor mutation burden score, microsatellite instability data, and the presence or absence of the target gene.

12. The method according to claim 11, characterized in that The tumor mutation burden score is obtained by: Based on the sequencing data of the target object, the number of gene mutations detected, such as somatic gene coding errors, base substitutions, gene insertions or deletions, is calculated; and The number of gene mutations was divided by the length of targeted sequencing coverage to obtain the tumor mutation burden score.

13. The method according to claim 2, characterized in that: Generate various input feature subsets and prediction indicator datasets including: Generate a first input feature subset, wherein the feature categories included in the first input feature subset are: cancer type, tumor stage, neutrophil-to-lymphocyte ratio, and treatment plan; Generate a second input feature subset, the feature categories included in the second input feature subset are: cancer type, tumor stage, neutrophil-to-lymphocyte ratio, TMB, treatment regimen, and target gene; Generate a third input feature subset, the feature categories included in the third input feature subset are: age, gender, pre-immune system treatment, cancer type, tumor stage, hemoglobin, albumin, neutrophil-to-lymphocyte ratio, platelets, treatment regimen, line number, and target gene; Generate a fourth input feature subset, the feature categories included in the fourth input feature subset are: age, gender, pre-immune systemic treatment, cancer type, tumor stage, hemoglobin, albumin, neutrophil-to-lymphocyte ratio, platelets, TMB, treatment regimen, number of lines, and target genes.

14. A computing device comprising: at least one processing unit; At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the device to perform the steps of the method according to any one of claims 1 to 13.

15. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 13 when executed by a machine.

Images