Metabonomics-based lung cancer patient immunotherapy prognosis prediction model

Abstract

The invention relates to bioinformatics and biomedicine, in particular to a lung cancer patient immunotherapy prognosis prediction model based on metabonomics. Metabonomics detection is carried out on peripheral blood of a lung cancer patient, and a PFS prognosis prediction model combining 17 differential metabolites is found. A prediction model training set AUC is 0.9497, the sensitivity is 0.9231, the specificity is 0.8846, the accuracy is 0.9077, the NPV is 0.8846, and the PPV is 0.9231; the test set AUC is 0.8810, the sensitivity is 0.9286, the specificity is 0.7333, the accuracy is 0.8276, the NPV is 0.9167, and the PPV is 0.7647. When the output cut off value is greater than or equal to 0.5, the prognosis of the lung cancer immunotherapy patient is good, and no progression lifetime gt exists; after 6 months, the immunotherapy prognosis of the lung cancer patient can be accurately predicted through the prediction model, and clinical guidance is provided. Wherein the 17 differential metabolites are as follows: Com56newg, Com513newg, Com327 newg, Com527newg, Com206newg, Com521newg, Com176newg, Com383newg, Com95newg, Com520newg, Com22newg, Com416newg, Com24newg, Com427newg, Com167newg, Com117newg and Com367newg. The invention also relates to a method for preparing the differential metabolite.

Description

Technical Field

[0001] This invention relates to bioinformatics and biomedicine, specifically to a prognostic prediction model for immunotherapy in lung cancer patients based on metabolomics. Background Technology

[0002] Lung cancer is the leading cause of cancer-related death. Tumor immunotherapy works by activating the body's immune system to fight tumors. In recent years, the use of immune checkpoint inhibitors (ICIs) has begun to change the status quo where chemotherapy was the primary first-line treatment for patients with driver gene-negative advanced lung cancer; however, only a small percentage of patients benefit from it. Most patients develop primary or acquired resistance to ICIs. Considering the high cost and side effects of immunotherapy, effective prediction of immunotherapy efficacy can help identify immunotherapy-sensitive and tolerant groups in a timely manner, reducing unnecessary medical expenses and adverse drug events. Currently, there is a lack of universally recognized and effective predictive indicators or tools for the prognosis of immunotherapy in clinical applications. Some studies suggest that PD-L1, neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR) may be prognostic indicators for lung cancer patients receiving ICI treatment, but their predictive efficacy and stability remain limited (Kiriu T, Yamamoto M, Nagano T, et al. The time-series behavior of neutrophil-to-lymphocyte ratio is useful as a predictive marker in non-small cell lung cancer[J]. PLoS One, 2018, 13(2)e0193018.). Therefore, finding biomarkers that can predict the prognosis of ICI treatment is of great significance for patient screening. At the same time, studies have found that the combined application of biomarkers is more helpful in predicting the effect of immunotherapy.

[0003] Metabolomics is an important component of systems biology, capable of detecting the state and changes of organisms at corresponding levels, similar to genomics, transcriptomics, and proteomics. As an emerging discipline in recent years, metabolomics fills the gap in other omics studies regarding small molecules at the final stage of biological reactions. Compared to changes at the gene, transcription, and protein levels, changes at the metabolic level are considered the final response of biological systems to biological events. Currently, metabolomics technology is widely used in various diseases, including eye diseases, endocrine diseases, and gastrointestinal diseases (Laíns I, M Gantner, S Murinello, et al. Metabolomics in the study of retinal health and disease. Prog Retin Eye Res, 2019, 69:57-79.). Especially in malignant tumors, a leading cause of death in clinical practice, metabolomics research in this field is constantly advancing and making breakthroughs, thanks to the continuous development of metabolomics, the gradual improvement of experimental equipment, the increasing maturity of experimental techniques, and its increasing application in clinical research. Currently, metabolomics has played a significant role in various malignant tumors, including colorectal cancer, bladder cancer, breast cancer, gastric cancer, esophageal cancer, and other malignant tumors (Pasikanti KK, KEsuvaranathan, YHong, et al. Urinary metabotyping of bladder cancer using two-dimensional gas chromatography time-of-flight mass spectrometry. J Proteome Res, 2013, 12(9): 3865-3873.; Qiu Y, G Cai, B Zhou, et al. A distinct metabolic signature of human colorectal cancer with prognostic potential. Clin Cancer Res, 2014, 20(8): 2136-2146.).

[0004] As a major subtype of cancer, lung cancer has consistently high incidence and mortality rates. Therefore, metabolomics-based methods are needed to predict the occurrence, development, and prognosis of lung cancer. Summary of the Invention

[0005] This application conducted metabolomics analysis on peripheral blood of lung cancer patients and found that a prognostic prediction model based on a combination of 17 differentially expressed metabolites could accurately predict the prognosis of lung cancer patients undergoing immunotherapy. Based on this, the present invention was completed.

[0006] In a first aspect, the present invention provides a set of biomarkers for predicting the prognostic effect of immunotherapy in lung cancer patients, the biomarker set comprising the following 17 differential metabolites: Com_56_neg, Com_513_pos, Com_327_pos, Com_527_pos, Com_206_neg, Com_521_pos, Com_176_neg, Com_383_pos, Com_95_neg, Com_520_pos, Com_22_neg, Com_416_pos, Com_241_neg, Com_427_pos, Com_167_neg, Com_117_neg, and Com_367_pos.

[0007] Secondly, the present invention provides a method for predicting whether the progression-free survival (PFS) of lung cancer patients treated with immunotherapy can exceed 6 months, the method comprising the following steps:

[0008] S01 Data Acquisition: Obtain raw data on the expression levels of metabolites in lung cancer patients as predictive indicators; the metabolites are as shown in the first aspect of the present invention;

[0009] S02 Data Preprocessing: The original data is used as the nodes of the input layer, standardized to obtain standardized data, activation function and optimizer are set, and batch training is performed;

[0010] S03 Model Construction: The standardized data mentioned above is input into the neural network model through the input layer, and the data is fitted through the hidden layer in the neural network model; a set of values ​​between 0 and 1 is output through the output layer; after training and optimization, a prognostic prediction model for PFS in cancer patients is formed.

[0011] S04 Evaluate model performance: Use the area under the receiver operating curve (AUC) to evaluate model performance; the model has predictive value only when the AUC is greater than 0.5, and the closer it is to 1, the stronger the predictive power.

[0012] Furthermore, in step S02, the standardization refers to subtracting the mean from the value of continuous data and dividing by the standard deviation to obtain standardized data.

[0013] Furthermore, in step S03, the data fitting is performed using batch training and batch standardized data.

[0014] Furthermore, the batch training preferably uses 10-100 data points per training session, and more preferably 65-85 data points per session.

[0015] Furthermore, in step S03, the data fitting is preferably verified using an early stopping function, and training is automatically terminated when the improvement in model performance is not significant.

[0016] Furthermore, in step S03, the neural network has a total of 8 layers. The initial input is a 17×8 node linear layer, followed by data fitting. Then, a dropout layer randomly silences 30% of neurons and passes them to the next 8×8 node linear layer. Data fitting and a dropout layer silence 30% of neurons again. Finally, it passes through an 8×1 node linear layer. After data fitting, the output is the predicted value (P) of whether the progression-free survival (PFS) of immunotherapy patients can be greater than 6 months.

[0017] Furthermore, the progression-free survival (PFS) of the immunotherapy patients can be greater than the predicted value P of 6 months, which is a value between 0 and 1; when the output P value is ≥0.5, the prognosis of lung cancer immunotherapy patients is good, and the progression-free survival is >6 months.

[0018] Furthermore, in step S03, the data fitting includes activation, normalization, weighting, transformation, and optimization.

[0019] Furthermore, the data activation is preferably performed using the ReLU function and / or the sigmoid function.

[0020] Furthermore, the data normalization refers to subtracting the mean of the data and dividing by the standard deviation.

[0021] Furthermore, the weighting includes linear layer weighting of 17×8 nodes, linear layer weighting of 8×8 nodes, and linear layer weighting of 8×1 nodes.

[0022] Preferably, the linear layer weighting of the 17×8 nodes can be selected from the following:

[0023] Table 1.17×8 node linear layer weighting values

[0024]

[0025]

[0026] Preferably, the linear layer weighting of the 8×8 nodes can be selected from the following:

[0027] Table 2.8×8 node linear layer weighting values

[0028]

[0029] Preferably, the linear layer weighting of the 8×1 nodes can be selected from the following:

[0030] Table 3.8×1 node linear layer weighting values

[0031]

[0032] Furthermore, the data conversion is preferably performed using the sigmoid function, which converts the output number into a value between 0 and 1.

[0033] Furthermore, the data optimization preferably uses the Adam function as the optimizer, with a learning rate set to 0.01.

[0034] Thirdly, the present invention provides a system for predicting progression-free survival in patients undergoing immunotherapy for lung cancer, the system comprising a data input module, a data analysis and processing module, a progression-free survival prediction module, and an output module;

[0035] The data input module refers to acquiring and inputting data, and the data acquisition refers to the expression levels of 17 specific metabolites in lung cancer patients.

[0036] The data analysis and processing module is divided into data preprocessing, data standardization, and data fitting;

[0037] The survival prediction module refers to building a prediction model, validating it, and making predictions based on the acquired data.

[0038] The data output module outputs the progression-free survival prediction value P for lung cancer patients.

[0039] Furthermore, in the data analysis and processing module, standardization refers to subtracting the mean from the value of continuous data and dividing by the standard deviation to obtain standardized data.

[0040] Furthermore, in the data analysis and processing module, data fitting is performed using batch training and batch standardized data.

[0041] Furthermore, the batch training preferably uses 10-100 data points per training session, and more preferably 65-85 data points per session.

[0042] Furthermore, in the data analysis and processing module, the data fitting is preferably verified using an early stopping function, and training is automatically terminated when the model performance improvement is not significant.

[0043] Furthermore, the data analysis and processing module includes data fitting processes such as activation, normalization, weighting, transformation, and optimization.

[0044] Furthermore, the data activation is preferably performed using the ReLU function and / or the sigmoid function.

[0045] Furthermore, the data normalization refers to subtracting the mean of the data and dividing by the standard deviation.

[0046] Furthermore, the weighting includes linear layer weighting of 17×8 nodes, linear layer weighting of 8×8 nodes, and linear layer weighting of 8×1 nodes.

[0047] Preferably, the linear layer weighting of the 17×8 nodes can be selected from the following:

[0048] Table 1.17×8 node linear layer weighting values

[0049]

[0050]

[0051] Preferably, the linear layer weighting of the 8×8 nodes can be selected from the following:

[0052] Table 2.8×8 node linear layer weighting values

[0053]

[0054] Preferably, the linear layer weighting of the 8×1 nodes can be selected from the following:

[0055] Table 3.8×1 node linear layer weighting values

[0056]

[0057] Furthermore, the data transformation preferably uses the sigmoid function to convert the output numbers into values ​​between 0 and 1. Even further, the data optimization preferably uses the Adam function as the optimizer, with a learning rate set to 0.01.

[0058] Furthermore, the survival prediction module has a neural network with a total of 8 layers. The initial input is a 17×8 node linear layer, followed by data fitting. Then, a dropout layer randomly silences 30% of neurons and passes the data to the next 8×8 node linear layer. Data fitting and a dropout layer silence 30% of neurons again. Finally, the data passes through an 8×1 node linear layer. After data fitting, the output is a predicted value P for whether the progression-free survival (PFS) of immunotherapy patients can be greater than 6 months.

[0059] Furthermore, the progression-free survival (PFS) of the immunotherapy patients can be greater than the predicted value P of 6 months, which is a value between 0 and 1; when the output P value is ≥0.5, the prognosis of lung cancer immunotherapy patients is good, and the progression-free survival is >6 months.

[0060] Fourthly, the present invention provides the application of the system as described in the third aspect of the present invention in the preparation of a device for predicting progression-free survival in patients undergoing immunotherapy for lung cancer.

[0061] Furthermore, the application is used to predict whether a patient's progression-free survival (PFS) is greater than 6 months or less than 6 months in patients undergoing immunotherapy for lung cancer.

[0062] Furthermore, the lung cancer immunotherapy patient progression-free survival prediction device outputs a progression-free survival prediction value P for lung cancer immunotherapy patients. When the output P value is ≥0.5, the prognosis of lung cancer immunotherapy patients is good, and the progression-free survival is >6 months.

[0063] Fifthly, a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the above-described method.

[0064] Beneficial effects

[0065] This application utilizes metabolomics analysis of peripheral blood from patients with good prognosis (PFS > 6 months) and patients with poor prognosis (PSF < 6 months) after immunotherapy. Statistical and machine learning analyses were performed on 78 differentially expressed metabolites, and a PFS prognostic prediction model based on a combination of 17 of these differentially expressed metabolites was established. The prediction model's training set AUC was 0.9497, sensitivity was 0.9231, specificity was 0.8846, accuracy was 0.9077, NPV was 0.8846, and PPV was 0.9231; the validation set AUC was 0.8810, sensitivity was 0.9286, specificity was 0.7333, accuracy was 0.8276, NPV was 0.9167, and PPV was 0.7647. When the output cut-off value is ≥0.5, the prognosis of lung cancer immunotherapy patients is good, with progression-free survival >6 months. The predictive model can accurately predict the prognosis of lung cancer patients after immunotherapy, providing guidance for clinical practice. Attached Figure Description

[0066] Figure 1 To construct a predictive model for the efficacy of immunotherapy in lung cancer patients based on metabolomics and machine learning.

[0067] Figure 2 ROC curve for predicting the efficacy of immunotherapy by PD-L1 expression in patient tumor tissue.

[0068] Figure 3 The ROC curve is based on a metabolomics-based efficacy prediction model. Detailed Implementation

[0069] The specific embodiments of the present invention will be further described below. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the embodiments described below can be combined with each other as long as they do not conflict with each other.

[0070] Unless otherwise specified, the experimental methods used in the following embodiments are conventional methods, and the experimental materials used in the following embodiments are all available through conventional commercial channels.

[0071] the term

[0072] Progression-free survival (PFS): The time from the start of randomization to the occurrence of tumor progression (in any aspect) or death (from any cause). PFS includes death and better reflects the toxic effects of the drug.

[0073] Immune checkpoint inhibitors (ICIs) work by blocking the binding of immune checkpoints to their ligands, thereby relieving the immunosuppression caused by immune checkpoints and reactivating immune cells to exert their anti-tumor effects. Several ICIs have already achieved significant success in clinical applications for anti-tumor treatment, representing a breakthrough in the field.

[0074] Example 1 Case Selection

[0075] In this invention, a total of 94 lung cancer patients who regularly received immunotherapy were enrolled (as shown in Table 4). After follow-up, 53 patients had a better prognosis with a progression-free survival (PFS) > 6 months, while 41 patients had a poorer prognosis with a PFS < 6 months. Metabolomics analysis was performed on the peripheral blood of these 94 patients to understand the differences in metabolite expression characteristics among patients with different prognoses during immunotherapy for lung cancer, and the relationship between metabolite expression and the prognosis of lung cancer immunotherapy.

[0076] Table 4. Information on lung cancer patients included

[0077]

[0078]

[0079] Unbiased detection of small molecule metabolites in peripheral blood was performed using non-targeted metabolomics techniques such as LC-MS, GC-MS, and NMR. Differentially expressed metabolites were screened using bioinformatics analysis, and pathway analysis was conducted on these differentially expressed metabolites. In 94 samples, differentially expressed metabolites were screened based on a fold change of more than 1.5 times (upregulation greater than 1.5 times or downregulation less than 0.67 times) and a P-value < 0.05. Among these, 73 metabolites showed significant changes in expression levels between the PFS > 6 months and PFS < 6 months groups.

[0080] Further lasso regression statistical analysis was used to screen out 17 differentially expressed metabolites that were closely related to the efficacy outcome. These 17 differentially expressed metabolites are: Com_56_neg, Com_513_pos, Com_327_pos, Com_527_pos, Com_206_neg, Com_521_pos, Com_176_neg, Com_383_pos, Com_95_neg, Com_520_pos, Com_22_neg, Com_416_pos, Com_241_neg, Com_427_pos, Com_167_neg, Com_117_neg, and Com_367_pos.

[0081] Example 2: Constructing a predictive model for the efficacy of immunotherapy in lung cancer patients based on 17 differentially expressed metabolites.

[0082] Overfitting is avoided by using batch training (65 data points per batch) and batch standardization (automatically calculating the mean and standard deviation of the batch, subtracting the mean, and then dividing by the standard deviation). An early stopping function is used for checking, and training is automatically terminated when the model performance improvement is not significant. Dropout is also used to avoid overfitting; it randomly silences a portion of neurons during training to prevent some neurons from having excessively high prediction weights.

[0083] The model building process is as follows:

[0084] 1. The user inputs the expression levels of 17 differentially expressed metabolites of the patient, namely: Com_56_neg, Com_513_pos, Com_327_pos, Com_527_pos, Com_206_neg, Com_521_pos, Com_176_neg, Com_383_pos, Com_95_neg, Com_520_pos, Com_22_neg, Com_416_pos, Com_241_neg, Com_427_pos, Com_167_neg, Com_117_neg, and Com_367_pos;

[0085] 2. Metabolite expression levels are used as nodes in the input layer. The initial input is a 17x8 node linear layer. The prediction system's neural network has a total of 8 layers. The prediction system activates the input data using the ReLU function. Batch training, batch standardization, and a 30% dropout are performed to help improve model performance.

[0086] 3. After multiple rounds of training, the early stopping function automatically ends the training, and the training curve is saved in Supplement Figure 1.

[0087] The 17×8 node linear layer weighting can be selected from the following:

[0088] Table 1.17×8 node linear layer weighting values

[0089]

[0090]

[0091] Preferably, the linear layer weighting of the 8×8 nodes can be selected from the following:

[0092] Table 2.8×8 node linear layer weighting values

[0093]

[0094] Preferably, the linear layer weighting of the 8×1 nodes can be selected from the following:

[0095] Table 3.8×1 node linear layer weighting values

[0096]

[0097] Example 3: PD-L1 Expression Status and Survival Prognosis in Tumor Patients

[0098] Meanwhile, the expression status of PD-L1 in tumor tissues and the prognostic outcomes of survival were analyzed in 94 patients.

[0099] like Figure 2 As shown, the training set AUC for predicting the prognosis of immunotherapy patients using patient PD-L1 expression status was 0.7632, and the validation set AUC was 0.4687.

[0100] Example 4: Predictive Models for 17 Differential Metabolites in Cancer Patients and Survival Prognosis

[0101] like Figure 3 As shown, the prediction model training set AUC: 0.9497, sensitivity: 0.9231, specificity: 0.8846, accuracy: 0.9077, NPV: 0.8846, PPV: 0.9231;

[0102] Validation set AUC: 0.8810, sensitivity: 0.9286, specificity: 0.7333, accuracy: 0.8276, NPV: 0.9167, PPV: 0.7647.

[0103] The model outputs a probability value P, with 0.5 as the cut-off value. A probability value greater than 0.5 will output "yes" (expected progression-free survival greater than 6 months), and a probability value less than 0.5 will output "no" (expected progression-free survival less than 6 months). The closer the probability value is to 1, the greater the probability that the patient's predicted progression-free survival will exceed 6 months, and the closer the probability value is to 0, the smaller the probability that the patient's predicted progression-free survival will exceed 6 months.

[0104] It is evident that the immunotherapy prognosis prediction model constructed based on metabolomics results in this application is significantly more effective than the current reference index PD-L1 in clinical applications.

Claims

1. A set of marker combinations for predicting the prognosis of immunotherapy in lung cancer patients, the marker combinations comprising the following 17 differential metabolites: Com_56_neg, Com_513_pos, Com_327_pos, Com_527_pos, Com_206_neg, Com_521_pos, Com_176_neg, Com_383_pos, Com_95_neg, Com_520_pos, Com_22_neg, Com_416_pos, Com_241_neg, Com_427_pos, Com_167_neg, Com_117_neg and Com_367_pos. 2.A method for predicting whether the progression-free survival (PFS) of immunotherapy in lung cancer patients can be greater than 6 months, the method comprising the expression amount of the marker combination of claim 1 as raw data, the method comprising the following steps: S01 data acquisition: obtaining the expression amount of metabolites in lung cancer patients as raw data for predicting indicators; S02 data preprocessing: taking the raw data as the nodes of the input layer, standardizing to obtain standardized data, setting the activation function and the optimizer, and performing batch training; S03 model construction: inputting the above standardized data into the neural network model through the input layer, fitting the data through the hidden layer in the neural network model, outputting a set of values between 0 and 1 through the output layer, and forming a PFS prognosis prediction model for tumor patients after training and optimization; S04 model performance evaluation: using the area under the receiver operating curve (AUC) to evaluate the performance of the model; when the AUC is greater than 0.5, the model has predictive value, and the closer to 1, the stronger the prediction performance.

3. The method of claim 2, wherein in step S02, the standardization refers to subtracting the mean value from the value of the continuity data and dividing by the standard deviation to obtain standardized data; in step S03, the data fitting uses batch training and batch standardized data for fitting, and the batch training preferably involves 65-85 data for each training; the data fitting preferably uses earlystopping function for verification and automatically ends the training when the model performance improvement is not obvious; and the data fitting includes activation, normalization, weighting, conversion, and optimization, wherein, The data activation preferably uses relu function and / or sigmoid function; the data normalization refers to subtracting the mean value of the data and dividing by the standard deviation; the weighting includes 17x8 node linear layer weighting, 8x8 node linear layer weighting and 8x1 node linear layer weighting; the data conversion preferably uses sigmoid function to convert the output number to a value between 0 and 1; the data optimization preferably uses adam function as the optimizer, and the learning rate is set to 0.01; the 17x8 node linear layer weighting is selected from the following: Table 1. 17x8 node linear layer weighting values The 8x8 node linear layer weighting is selected from the following: Table 2. 8x8 node linear layer weighting values The 8x1 node linear layer weighting is selected from the following: Table 3. 8x1 node linear layer weighting values 4. The method of claim 2, in step S03, the neural network has a total of 8 layers, the initial input is a linear layer of 17x8 nodes, after data fitting, a dropout layer randomly silences 30% of the neurons, and passes to the next linear layer of 8x8 nodes, again data fitting and dropout layer silences 30% of the neurons, and finally passes through a linear layer of 8x1 nodes, after data fitting, the output is the prediction value P of whether the immune therapy patient's progression-free survival (PFS) can be greater than 6 months; the prediction value P of whether the immune therapy patient's progression-free survival (PFS) can be greater than 6 months is a value between 0 and 1; when the output P value is greater than or equal to 0.5, the prognosis of the lung cancer immune therapy patient is good, and the progression-free survival is greater than 6 months.

5. A system for predicting the progression-free survival of a lung cancer immune therapy patient, the system comprising a data input module, a data analysis processing module, a progression-free survival prediction module, and an output module; wherein, the data input module refers to obtaining and inputting data, and the data obtained is the expression amount of 17 specific metabolites of a lung cancer patient as claimed in claim 1; the data analysis processing module is divided into data preprocessing, data standardization, and data fitting; the survival prediction module refers to constructing a prediction model, verifying it, and predicting the obtained data; the data output module outputs the prediction value P of the progression-free survival of the lung cancer patient.

6. The system of claim 5, the data fitting in the data analysis processing module includes activation, normalization, weighting, transformation, optimization: wherein, The data activation preferably uses the relu function and / or the sigmoid function; the data normalization refers to subtracting the mean value of the data and dividing by the standard deviation; the weighting includes linear layer weighting of 17x8 nodes, linear layer weighting of 8x8 nodes, and linear layer weighting of 8x1 nodes; the data conversion preferably uses the sigmoid function, the output number is converted to a value between 0 and 1; the data optimization preferably uses the adam function as the optimizer, and the learning rate is set to 0.01; the linear layer weighting of 17x8 nodes is selected from the following: Table 1. Linear layer weighting values of 17x8 nodes The linear layer weighting of 8x8 nodes is selected from the following: Table 2. Linear layer weighting values of 8x8 nodes The linear layer weighting of 8x1 nodes is selected from the following: Table 3. Linear layer weighting values of 8x1 nodes 7. The system of claim 5, the survival prediction module has a neural network, the neural network has a total of 8 layers, the initial input is a linear layer of 17x8 nodes, after data fitting, a dropout layer randomly silences 30% of the neurons, and passes to the next linear layer of 8x8 nodes, again data fitting and dropout layer silences 30% of the neurons, and finally passes through a linear layer of 8x1 nodes, after data fitting, the output is the prediction value P of whether the immune therapy patient's progression-free survival (PFS) can be greater than 6 months; the prediction value P of whether the immune therapy patient's progression-free survival (PFS) can be greater than 6 months is a value between 0 and 1; when the output P value is greater than or equal to 0.5, the prognosis of the lung cancer immune therapy patient is good, and the progression-free survival is greater than 6 months.

8. The use of the system of claim 5 in preparing a lung cancer immunotherapy patient progression-free survival prediction device, wherein the use is to predict the case of lung cancer immunotherapy patient PFS greater than 6 months and PFS less than 6 months.

9. The use of claim 8, wherein the lung cancer immunotherapy patient progression-free survival prediction device outputs a lung cancer immunotherapy patient progression-free survival prediction value P, and when the output P value is greater than or equal to 0.5, the lung cancer immunotherapy patient has a good prognosis and a progression-free survival greater than 6 months.

10. A computer readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the above method.

Images