Model for predicting non-small cell lung cancer immunotherapy applicability and application thereof

Abstract

The invention provides a model for predicting non-small cell lung cancer immunotherapy applicability and application thereof. The invention discloses a group of labeled molecules which can be effectively used for carrying out NSCLC immunotherapy applicability analysis. The invention also provides a prediction model established by taking the labeled molecule as a key factor. The invention provides a new analysis approach for formulating a clinical treatment scheme of NSCLC.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of disease prediction models, and more particularly, it relates to a model for predicting the suitability of immunotherapy for non-small cell lung cancer and applications thereof. BACKGROUND

[0002] Non-small cell lung cancer (NSCLC) is the most common and deadly malignancy worldwide, accounting for approximately 85% of all lung cancer cases. For patients with advanced NSCLC who are negative for driver genes such as EGFR, ALK, and ROS1, immunotherapy and its combination therapy have become one of the standard treatment options. Immunotherapy for NSCLC activates the patient's own immune system, overcomes the suppression of immune function by tumor cells, and allows immune cells to recognize and attack cancer cells again. For example, immune checkpoint inhibitors such as PD-1 / PD-L1 inhibitors can block the "immune escape" signal of tumor cells and enhance the anti-tumor activity of T cells. Studies have shown that immunotherapy significantly improves patient survival.

[0003] However, clinical drug use has also shown that not all patients can benefit from immunotherapy. The efficacy of immunotherapy can be influenced by a variety of factors, including tumor immunogenicity, tumor microenvironment, and patient individual differences. Therefore, how to accurately screen patient populations that can benefit from immunotherapy is an important challenge in current clinical research and practice.

[0004] Currently, there have been a number of studies confirming that the expression level of PD-L1 protein in tumor tissue, immune microenvironment, and peripheral blood circulating tumor DNA are correlated with the efficacy and prognosis of immunotherapy. However, these studies often have the problems of invasive sample acquisition, large volume of plasma use, small sample size, lack of verification, and other issues, limiting their application in clinical practice.

[0005] Therefore, there is an urgent need to develop new diagnostic protocols or strategies to improve the accuracy of predicting immunotherapy response, so as to provide more personalized and effective treatment options for patients.

[0006] Research based on molecules present in peripheral blood to detect the efficacy of immunotherapy has received increasing attention, and liquid biopsy technology is favored due to its minimal invasiveness and allowing repeated sampling. Future research in this field needs to focus on discovering and validating new marker molecules and exploring the predictive value of these markers in different patient populations to achieve accurate prediction of immunotherapy response in patients with non-small cell lung cancer. SUMMARY

[0007] The present application aims to provide a model for predicting the immunotherapy applicability of non-small cell lung cancer and the application thereof.

[0008] In the first aspect of the present application, the application of a marker molecule combination in establishing an immunotherapy applicability analysis system for non-small cell lung cancer (NSCLC) is provided; the marker molecule combination comprises: CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, and TNFRSF12A.

[0009] In one or more embodiments, the analysis can include (but not limited to): testing, predicting, evaluating, prognosticating, diagnosing, scoring, etc.

[0010] In one or more embodiments, the molecules include: proteins, genes, and / or transcripts; preferably, the molecules include proteins.

[0011] In one or more embodiments, the analysis system comprises: an analysis model; preferably, the analysis model is a survival support vector machine (Survival SVM) model established with the marker molecules in the marker molecule combination as modeling molecules (e.g., modeling proteins).

[0012] In one or more embodiments, the analysis system further comprises: a marker molecule expression level determination module; more preferably, the marker molecule expression level determination module comprises: detection reagents, kits, and / or detection devices.

[0013] In one or more embodiments, the analysis model can include (but not limited to): a testing model, a predicting model, an evaluating model, a prognosticating model, a diagnosing model, a scoring model, a predicting scoring model, etc.

[0014] In one or more embodiments, the detection reagents include: specific antibodies, PCR detection reagents, sequencing reagents; more preferably, the detection reagents include: antibodies specifically capturing the molecular markers, primers specifically amplifying the molecular marker genes, probes specifically recognizing the molecular marker genes.

[0015] In one or more embodiments, the detection reagents are included in the kits.

[0016] In one or more embodiments, the detection devices include: gene sequencing instruments, chips, probe sets, primer probe sets, or electrophoresis devices.

[0017] In one or more embodiments, the expression level of each marker in the marker molecule combination is substituted into the analysis model to calculate a prediction score, and the immunotherapy suitability of the sample to be tested is determined according to the prediction score (e.g., the sample is determined to be High-risk or Low-risk, or the patient corresponding to the sample is determined to be unsuitable for immunotherapy or suitable for immunotherapy).

[0018] In one or more embodiments, the sample is divided into High-risk or Low-risk according to the survival support vector machine model score; preferably, a cut-off value is determined to define the High-risk or Low-risk.

[0019] In one or more embodiments, the prediction scores are sorted from small to large, and different percentiles are used to calculate the log rank statistics of the total survival (OS) of the two groups at different percentiles (e.g., with an interval of 5% (10%, 15%...90%, 95%) between every two values), and the threshold value (optimal threshold value) is the percentile with the largest log rank statistics.

[0020] In one or more embodiments, when the prediction score is higher than the threshold value, the non-small cell lung cancer patient belongs to the Low-risk group and is analyzed to be suitable for immunotherapy; when the prediction score is lower than the threshold value, the non-small cell lung cancer patient belongs to the High-risk group and is analyzed to be unsuitable for immunotherapy.

[0021] In one or more embodiments, for a group of samples, the top 20% are determined to be the High-risk group with relatively poor prognosis and are not recommended for immunotherapy, and the bottom 80% are determined to be the Low-risk group with relatively good prognosis and are recommended for immunotherapy.

[0022] In one or more embodiments, the immunotherapy includes: immunological checkpoint inhibitor treatment, or combined treatment of immunological checkpoint inhibitor and chemotherapy drugs; more preferably, the immunological checkpoint inhibitor includes: anti-PD-L1 antibody, anti-PD-1 antibody, or anti-CTLA-4 inhibitor.

[0023] In one or more embodiments, the analysis of immunotherapy suitability includes: analyzing the sensitivity of the non-small cell lung cancer patient to immunotherapy, performing risk analysis or scoring of non-small cell lung cancer development; or developing a treatment / drug regimen; or analyzing the pathological remission and / or survival period of the non-small cell lung cancer patient after receiving immunotherapy.

[0024] In another aspect of the present application, a method for establishing an immunotherapy applicability analysis model of non-small cell lung cancer is provided, comprising: establishing a survival support vector machine (Survival SVM) model with a combination of marker molecules as modeling molecules (e.g., modeling proteins), wherein the combination of marker molecules comprises CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, and TNFRSF12A.

[0025] In one or more embodiments, the method for establishing a model (training model) comprises: obtaining the expression level of each marker molecule in the combination of marker molecules in the sample, and establishing a survival support vector machine model with the expression level of each marker molecule.

[0026] In one or more embodiments, the parameters for establishing the model comprise: kernel, type, gamma.mu, and opt.meth.

[0027] In one or more embodiments, the kernel is set as add_kernel.

[0028] In one or more embodiments, the type is set as regression.

[0029] In one or more embodiments, the gamma.mu is set as 0.25.

[0030] In one or more embodiments, the opt.meth is set as quadprog.

[0031] In one or more embodiments, the modeling molecules are obtained by a method comprising: (1) providing a plasma sample (e.g., a pre-treatment plasma sample) of a non-small cell lung cancer patient, and obtaining protein expression data in the sample; (2) using the expression data (or a part of the data) of (1) as a training set; in the training set, performing single-factor COX regression analysis on the overall survival period and the expression of candidate proteins to obtain candidate proteins significantly correlated with the overall survival period; and using Lasso COX regression combined with stepwise Cox regression to analyze and obtain the modeling molecules.

[0032] In one or more embodiments, after step (2) of the method, the method further comprises: after establishing a survival support vector machine (Survival SVM) model with the modeling molecules, performing internal performance verification on the model; preferably, 10-fold cross-validation is selected to verify the ROC AUC of the model.

[0033] In another aspect of the present application, a non-small cell lung cancer immunotherapy applicability analysis model is provided, which is an analysis model established by any of the above-mentioned methods.

[0034] In another aspect of the present application, there is provided an analysis system for immunotherapy suitability of non-small cell lung cancer, comprising:

[0035] (i) a marker molecule expression level determination module for the test sample; the module is used to determine the expression level of the marker molecules in the marker molecule combination, which comprises: CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, TNFRSF12A; preferably, the marker molecule expression level determination module comprises: detection reagents, kits and / or detection devices;

[0036] (ii) an analysis model for immunotherapy suitability of non-small cell lung cancer, or a hardware system (such as a computer) integrated with the analysis model; the analysis model is an analysis model established by any of the above-mentioned methods.

[0037] In another aspect of the present application, there is provided a method for analyzing the immunotherapy suitability of non-small cell lung cancer, comprising: analyzing with the analysis model; or, analyzing with the analysis system.

[0038] In one or more embodiments, the method comprises: analyzing the sensitivity of non-small cell lung cancer patients to immunotherapy according to the expression level of each marker molecule in the marker molecule combination; preferably further comprising: developing a treatment / drug regimen; preferably, the immunotherapy comprises: immune checkpoint inhibitor therapy, or combined therapy of immune checkpoint inhibitor and chemotherapy drugs.

[0039] In one or more embodiments, the method for analyzing the immunotherapy suitability of non-small cell lung cancer comprises an auxiliary disease analysis method, which is not directly aimed at obtaining a diagnosis result of the disease, but only provides auxiliary analysis / evaluation / score.

[0040] Other aspects of the present application will be apparent to those skilled in the art from consideration of the disclosure herein. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 , LASSO regression cross-validation curve, X axis is the logarithm of penalty coefficient logλ, Y axis is the likelihood deviation.

[0042] Figure 2 , 8 HR values, 95% CI and p values of 8 important key proteins.

[0043] Figure 3 , Model evaluation based on the area under the receiver operating characteristic curve (ROC) (AUC) constructed for 8 marker proteins. In the training set, overall survival (OS) was calculated.

[0044] Figure 4 The model constructed based on the area under the receiver operating characteristic curve (AUC) of 8 marker proteins was evaluated. In the training set, progression-free survival (PFS) was calculated.

[0045] Figure 5 、 8 The optimal threshold of the SurvivalSVM prediction model constructed with 1 marker protein.

[0046] Figure 6 A. Divide the 92 training sets into high-risk and low-risk groups based on the optimal threshold, and draw the KM curve of overall survival.

[0047] Figure 6 B. The 92 training sets were divided into high-risk and low-risk groups according to the optimal threshold, and the progression-free survival KM curve was drawn.

[0048] Figure 7 A. Univariate COX regression analysis of the Survival SVM prediction model and clinical characteristics (age, gender, smoking history, pathological subtype, and clinical stage).

[0049] Figure 7 B. Multivariate COX regression analysis of the SurvivalSVM prediction model and clinical characteristics (age, gender, smoking history, pathological subtype, and clinical stage).

[0050] Figure 8 KM curves of AF, prediction score grouping of 42 samples in the training set and clinical prognosis, and overall survival.

[0051] Figure 9 A. The receiver operating characteristic (ROC) curve was calculated and plotted using the pROC software package (version 1.18.0) in the validation set data matrix, and the AUC value for 6-month overall survival in the validation set was calculated.

[0052] Figure 9 B. The receiver operating characteristic (ROC) curve was calculated and plotted using the pROC software package (version 1.18.0) in the validation set data matrix, and the AUC value for 6-month progression-free survival was calculated in the validation set.

[0053] Figure 10 A. KM curve of the prediction score grouping and clinical overall survival of 102 samples in the validation set calculated by the model.

[0054] Figure 10 B. KM curve of the prediction score grouping and progression-free survival of 102 samples in the validation set calculated by the model.

[0055] Figure 11A. Univariate COX regression analysis of the SurvivalSVM prediction model with clinical features (age, gender, smoking history, pathological subtype and clinical stage) in the validation set.

[0056] Figure 11 B. Multivariate COX regression analysis of the SurvivalSVM prediction model with clinical features (age, gender, smoking history, pathological subtype and clinical stage) in the validation set.

[0057] Figure 12 A-F. The prediction scores of 43 samples in the validation set were grouped and analyzed with clinical outcomes, i.e. KM curves of overall survival and progression-free survival. DETAILED DESCRIPTION

[0058] There is still lack of technical means and tools for predicting the effect of immunotherapy for non-small cell lung cancer (NSCLC) in the prior art. In view of this, the present application discloses a set of marker molecules which can be effectively used for analyzing the suitability of immunotherapy for NSCLC. The marker molecules include core molecules CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, and TNFRSF12A. The present application also provides a prediction model established by taking the marker molecules as key factors. The present application provides a new analysis approach for formulating the clinical treatment plan for NSCLC.

[0059] TERMS

[0060] As used herein, the cancer is non-small cell lung cancer (NSCLC), unless otherwise specified.

[0061] As used herein, the "patient", "subject" or "individual" can include a population at risk of a disease or a population diagnosed with a disease, especially a population suffering from a disease, in need of evaluating the subsequent treatment plan according to the disease condition, deciding whether to perform immunotherapy. Unless otherwise specified, the disease in the present application refers to cancer (tumor), especially NSCLC.

[0062] As used herein, the "sample" and "specimen" can be used interchangeably, which includes a substance obtained from an individual or isolated tissue, cells or body fluids, suitable for detecting marker molecules.

[0063] As used herein, "marker molecule" refers to a biomolecule or a fragment of a biomolecule, the expression level of which is related to a specific physical condition or state. The terms "marker molecule", "marker / marker" or "biomarker / marker" can be used interchangeably in the present application. The term "marker molecule" includes "marker protein".

[0064] As used herein, the term “analysis” includes “evaluation”, “assessment”, “determination”, “detection”, “prediction”, “evaluation” or “assessment” also includes “scoring”.

[0065] As used herein, the term “analysis model” includes “test model”, “prediction model”, “evaluation model”, “prognosis model”, “diagnosis model”, “scoring model”, “prediction and scoring model”, and the like, which are used interchangeably.

[0066] As used herein, the term “analysis system” includes “test system”, “prediction system”, “evaluation system”, “prognosis system”, “diagnosis system”, “scoring system”, “prediction and scoring system”, and the like, which are used interchangeably.

[0067] As used herein, the terms “prediction”, “prognosis” or “prognostic” refer to the prediction of the consequences of a disease or a drug administration scheme / treatment strategy for the disease, and the indicators of prognosis include short-term and long-term efficacy indicators, including but not limited to OS, PFS, ORR, DCR. Wherein, “OS” refers to “Overall Survival”, total survival, which refers to the total time of survival of all patients in the study. Wherein, “PFS” refers to “Progression Free Survival”, progression-free survival, which generally refers to the time during which the patient's disease does not progress or is resistant to treatment after treatment. In general, the median OS or median PFS is used to represent, i.e., the survival time / progression-free survival time reached by 50% of patients.

[0068] As used herein, the term “expression level” can refer to the concentration or amount of protein / gene of the marker molecule / indicator of the present application in a sample.

[0069] As used herein, the definition of high risk or low risk can be based on the existing standard in the art.

[0070] As used herein, “ / ” can mean “and” or “or”.

[0071] Marker molecule

[0072] In the present application, by in-depth analysis and comparison of samples of NSCLC clinical patients, factors related to the progression of NSCLC are identified, and marker molecules CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, TNFRSF12A are screened. The combined use of them as marker molecules has significant diagnostic significance for the immunotherapy applicability of NSCLC.

[0073] The amino acid sequence of the CD27 protein can refer to Uniprot ID P26842.

[0074] The amino acid sequence of the TWEAK protein can be found in UniprotID 043508.

[0075] The amino acid sequence of the CD8A protein can be found in UniprotID P01732.

[0076] The amino acid sequence of the GZMH protein can be found in UniprotID P20718.

[0077] The amino acid sequence of the IL6 protein can be found in UniprotID P05231.

[0078] The amino acid sequence of the PGF protein can be found in UniprotID P49763.

[0079] The amino acid sequence of the TIE2 protein can be found in niprotID Q02763.

[0080] The amino acid sequence of the TNFRSF12A protein can be found in UniprotID Q9NP84.

[0081] The present invention also encompasses sequence variants or homologous proteins of the above-mentioned proteins in organisms.

[0082] The marker molecules disclosed in the present invention can be used as markers (marker molecules) to assess the progression of NSCLC and the efficacy of treatment. This can be used to understand the disease state of an individual, assess or predict prognosis of disease risk, and formulate treatment / medication plans.

[0083] Model building

[0084] SurvivalSVM is a machine learning method that applies support vector machines (SVMs) to survival analysis. Survival analysis is a branch of statistics used to analyze life expectancy or time to an event. It is particularly suitable for the analysis scenario of this invention to assess the impact of specific factors on survival time.

[0085] In some specific embodiments of the present invention, a method for constructing a prediction scoring model is provided, the method comprising the following steps:

[0086] (1) Obtain protein expression data from plasma samples of all NSCLC patients before treatment;

[0087] (2) grouping the protein expression data into a training set and a validation set;

[0088] (3) In the training set, the total survival time and the expression of the candidate marker molecules are subjected to single factor COX regression analysis, a series of candidate marker molecules significantly related to the total survival time are obtained, and further analysis by Lasso COX regression combined with stepwise Cox regression obtains the preferred combination of marker molecules.

[0089] (4) A Survival SVM prediction model is established based on the preferred marker molecules to predict the efficacy of immunotherapy for NSCLC patients.

[0090] In some specific examples of the present application, the sample is a plasma sample isolated from the peripheral blood of an NSCLC patient before treatment.

[0091] In some specific examples of the present application, the lasso regression analysis is performed using the glmnet package of the R language software.

[0092] In some specific examples of the present application, according to the expression information of the NSCLC plasma marker molecules, the HR value and p value of each marker molecule are calculated by COX test, and the differentially expressed marker molecules are screened according to the standard of p<0.05, the probability and 95% confidence interval between the protein expression of all differentially expressed proteins and the total survival time are calculated by logistic single factor logistic regression analysis, and the proteins with p<0.05 are selected as key marker molecules.

[0093] In some specific embodiments of the present application, further comprising: in the training set, the internal performance of the prediction model is verified, and 10-fold cross-validation is selected to verify the ROC AUC of the model.

[0094] In some specific examples of the present application, the predicted score (Predicted Score) of each sample is divided into Risk-High and Risk-Low, wherein Risk-Low is expected to have greater benefit from immunotherapy.

[0095] The prediction score model of the present application shows high accuracy and specificity through ROC curve analysis, and can significantly distinguish between high-risk and low-risk patients for immunotherapy. Compared with the prior art, the model of the present application performs more superiorly in predicting the survival of NSCLC patients receiving immunotherapy combined with chemotherapy.

[0096] The prediction score model of the present application is consistent with the immune infiltration state in the tumor immune microenvironment. Experimental results show that samples predicted by the model to be in the low-risk group show higher immune infiltration, presenting a hot tumor state; while samples in the high-risk group show lower immune infiltration, presenting a cold tumor state. This finding provides a reliable method for predicting the risk of NSCLC immunotherapy.

[0097] Application of the model

[0098] The marker molecules described in the present application can be applied as a combination as a modeling protein to establish an analysis model (a prediction scoring model).

[0099] As a preferred mode of the present application, a survival support vector machine (Survival SVM) prediction scoring model is established by using the marker molecules of the present application to effectively predict the response of NSCLC patients to immunotherapy.

[0100] In some embodiments of the present application, a method for applying the prediction model to predict the efficacy of immunotherapy for NSCLC is provided, which comprises the following steps:

[0101] (1) detecting the expression level of the preferred marker molecules in the sample of the NSCLC patient;

[0102] (2) substituting the expression level of each protein obtained in step (1) into the prediction scoring model to calculate the prediction score; ranking the prediction scores from small to large, and then using different percentiles to calculate the c-value, with the percentiles being 10%, 15%, and up to 90%, and the interval between each two c-values being 5%; then calculating the log rank statistical value of the two groups of OS under different c-values, and taking the c-value with the largest log rank statistical value as the optimal threshold value. When the prediction score is higher than the optimal threshold value, the NSCLC patient belongs to the low-risk group, indicating that the NSCLC patient is suitable for immunotherapy; when the prediction score is lower than the optimal threshold value, the NSCLC patient belongs to the high-risk group, indicating that the NSCLC patient is not suitable for immunotherapy.

[0103] The present application also provides an analysis system for analyzing the immunotherapy applicability of non-small cell lung cancer, which comprises a protein expression level determination module of a sample to be tested and the analysis model (prediction scoring model).

[0104] The protein expression level determination module is used to analyze the scalar level of the marker molecules in the sample to be tested, and can comprise some detection reagents or kits suitable for protein expression level determination.

[0105] The determination of the expression level of the marker molecule can be carried out according to an established standard procedure well known in the art. For example, according to the protein of the marker molecule, an antibody specifically binding to the protein can be used to achieve detection, which can be implemented for a clinical sample by immunohistochemical staining. Commonly used immunohistochemical methods include but are not limited to immunofluorescence, immunoenzyme labeling, immunocolloid gold labeling, etc. Immunofluorescence utilizes the principle of specific binding of antigen and antibody, and a known antibody is labeled with fluorescein as a probe to examine the corresponding antigen in cells or tissues, which is observed under a fluorescence microscope. When the fluorescein in the antigen-antibody complex is irradiated by excitation light, it will emit fluorescence of a certain wavelength, so that the localization of a certain antigen in the tissue can be determined, and quantitative analysis can also be carried out. In immunoenzyme labeling, an enzyme-labeled antibody is used to react with tissues or cells, and then a substrate of the enzyme is added to generate a colored insoluble product or a particle with certain electron density, and various antigen components on the cell surface and in the cell are located by light microscopy or electron microscopy. In immunocolloid gold labeling, colloidal gold (an aqueous colloidal solution of gold) is used as a special metal particle as a marker. Colloidal gold can quickly and stably adsorb proteins without significantly affecting the biological activity of the proteins. An antibody or other molecules capable of specifically binding to immunoglobulin labeled with colloidal gold is used as a probe to qualitatively, quantitatively and quantitatively study the antigens in tissues or cells.

[0106] In some specific examples of the present application, the marker molecule expression data is obtained by isolating plasma from the peripheral blood of NSCLC patients, quantifying proteins in the plasma sample, and obtaining a protein expression matrix after sorting.

[0107] The analysis system (prediction scoring system) of the present application is suitable for clinical detection and is suitable for various medical environments, including applications in hospitals, clinics and other medical institutions, and can provide convenient detection services for different types of NSCLC patients. It also includes applications related to scientific research experiments in related research institutions.

[0108] The analysis system of the present application is suitable for all patient populations that need to evaluate the suitability of immunotherapy, especially in clinical trials and actual treatment. The method simplifies the operation process and improves the detection efficiency.

[0109] The present application adopts Lasso regression combined with stepwise Cox regression to ensure that the selected marker molecules are significantly related to survival time. Its maximum application range covers all related fields of data analysis, including biological statistics and epidemiological research.

[0110] The technical scheme of the present application has the following positive effects:

[0111] The application provides a new marker molecule combination for predicting the efficacy of NSCLC immunotherapy, including the preferred marker molecules. Compared with the prior art, especially the traditional method relying only on PD-L1 characteristics, this combination has higher prediction accuracy and stability. Through experimental verification, these marker molecules can effectively distinguish between high-risk and low-risk patients, thereby improving the effectiveness of clinical decision-making.

[0112] The prediction score model of the application can be applied to evaluate all advanced NSCLC patients, especially those who have not received immunotherapy. The model can be widely used in clinical trials and daily medical practice to help clinicians make better treatment decisions. The technical solution of the application provides a new idea and method for individualized immunotherapy of NSCLC patients, and has important clinical application value.

[0113] The analysis model developed by the application and its application method simplify the clinical detection process, enabling clinicians to more conveniently evaluate the suitability of patients for immunotherapy. This innovation not only improves detection efficiency, but also reduces the complexity of clinical operations, laying the foundation for widespread application in medical institutions.

[0114] The prior art often needs to rely on more markers or detection means when predicting the effect of immunotherapy, resulting in a complex and costly detection process. The accuracy and stability of existing methods are also insufficient, limiting their widespread use in clinical applications. Compared with existing prediction methods (for example, compared with the clinical standard detection of PD-L1 expression level), the application achieves stable and superior prediction results with fewer marker molecules, which helps to more accurately screen out patients who may benefit from immunotherapy, so that patients can achieve better treatment effects. Therefore, by simplifying the number of marker molecules, the application improves the practicality and operability of the prediction model, providing clinicians with a more efficient tool to make more accurate decisions in actual treatment.

[0115] Compared with existing important gene sets, the prediction score model of the application shows better prediction ability in multiple experiments, especially in terms of survival. Compared with traditional methods, the application provides a more scientific and systematic method to evaluate the immunotherapy response of patients, which embodies its creativity and practical value.

[0116] Immunotherapy is effective for a portion of people, and considering the high price of immunotherapy drugs and possible side effects, it is important to find marker molecules that can predict the efficacy of the drugs. The TPS scoring system is a scoring system for evaluating the expression level of PD-L1 in tumor samples, which is used to evaluate the effectiveness of immunotherapy, and has been practiced in clinical practice. The analysis results of the application show that the application of the analysis system of the application has more significant advantages.

[0117] In summary, the present application overcomes the defects in the prior art by innovative marker molecule combinations and predictive scoring models, and provides more accurate and personalized immunotherapy prediction tools for NSCLC patients, bringing positive progress to clinical practice.

[0118] The present application will be further described in conjunction with specific examples. It should be understood that these examples are only used to illustrate the present application and not used to limit the scope of the present application. The experimental methods in the following examples, if not otherwise specified, are generally carried out according to the conventional conditions described in J. Sambrook et al., Molecular Cloning: A Laboratory Manual, Science Press, or according to the conditions suggested by the manufacturer.

[0119] Collection of patient samples

[0120] 194 plasma samples isolated from peripheral blood of non-small cell lung cancer (NSCLC) patients before treatment were collected, and all the patients received anti-PD-1 monoclonal antibody immunotherapy (monoclonal antibody combined with chemotherapy treatment). Unless otherwise specified, the samples used for analysis in the examples are blood samples taken before treatment. In the examples, the accuracy of the model constructed using the expression of marker molecules in blood samples taken before treatment was evaluated according to the results of clinical immunotherapy of these patients.

[0121] Among them, the anti-PD-1 monoclonal antibody is camrelizumab, pembrolizumab or toripalimab.

[0122] Among them, the chemotherapy drug is paclitaxel, carboplatin or cisplatin.

[0123] Sample collection and grouping

[0124] According to the order of specimen detection, all patients were divided into training set and validation set 2 groups. 92 samples were used as the training set, and 102 samples were used for the validation set.

[0125] 92: 92 patients were treated with anti-PD-1 monoclonal antibody + chemotherapy drug.

[0126] 102: 102 patients were treated with anti-PD-1 monoclonal antibody + chemotherapy drug.

[0127] Chemotherapy drugs: paclitaxel + carboplatin, paclitaxel + cisplatin, each half.

[0128] Statistics of overall survival and progression-free survival

[0129] Overall survival (Overall Survival, OS): The time from the patient's confirmation of having the disease to death due to any cause was counted.

[0130] Progression Free Survival (PFS): The time from randomization to the first occurrence of disease progression or death.

[0131] Detection reagent

[0132] In the antibody probe for detecting the protein expression level of the marker protein (protein in the marker molecule combination) of the patient, the antibody used includes the following:

[0133]

[0134]

[0135] Example I, sample acquisition and screening of key marker molecules

[0136] 1. Sample processing steps

[0137] Plasma separation: 2 mL of whole blood was collected from each patient using an EDTA anticoagulant tube. After collection, centrifuge at 3000 rpm for 5 minutes within 2 hours, and take the supernatant as the plasma sample.

[0138] 2. Sample detection steps

[0139] (1) Sample preparation: The plasma sample is pretreated as required, usually including centrifugation to remove cells and impurities.

[0140] (2) Probe incubation: Each target protein corresponds to a pair of specific antibody probes. The antibody part of the probe binds to a specific protein, while the accompanying DNA sequence is used for subsequent detection. The sample is mixed with the probe and incubated to allow the probe to bind to the target protein.

[0141] (3) Proximity Extension Assay (PEA) reaction: When two probes successfully bind to the same protein, the DNA sequences on the probes will be extended to each other through proximity effect, forming a new DNA sequence. This step ensures that only probes that bind to the target protein at the same time will form detectable DNA.

[0142] (4) PCR amplification: The new DNA sequence generated is amplified by PCR to increase its detection sensitivity. This step can be detected using a real-time quantitative PCR (qPCR) platform. Unlike traditional protein detection methods, here the concentration of protein is indirectly reflected through DNA amplification.

[0143] (5) Data reading and analysis: The signal of qPCR detection is converted into the corresponding protein concentration.

[0144] 3. Marker protein quantification

[0145] After the experiment, the data will be strictly quality controlled to ensure the specificity of the probes and the accuracy of the reactions, and the NPX data (Normalized Protein eXpression, NPX) will be output, which has been standardized relative to internal controls and reference samples. According to the NPX data, downstream analysis is entered.

[0146] 4. Differential marker protein screening

[0147] For the plasma marker protein expression data of 194 non-small cell lung cancer patients in peripheral blood before treatment, according to the detection batch, it is divided into training set and verification set, and the sample in the training set is 92.

[0148] In the training set, the relationship between the expression of marker proteins and overall survival was analyzed by single factor COX regression analysis, and 22 candidate marker proteins significantly related to overall survival (p value less than 0.05) were screened.

[0149] 5. Key marker protein screening

[0150] In the training set, COX single factor logistic regression analysis was performed, the risk value (HR) and 95% confidence interval (CI) of the correlation between the expression of all 92 proteins and OS (overall survival) results were calculated, and proteins with p<0.05 were screened. A total of 22 key proteins were identified.

[0151] In the training set, lasso regression analysis and stepwise COX regression were performed, 10-fold cross-validation and leave-one-out cross-validation were further used to reduce the characteristic proteins, and important key proteins were selected as model candidate proteins. Figure 1 The LASSO regression cross-validation curve. The X axis is the logarithm logλ of the penalty coefficient, and the Y axis is the likelihood deviation. The smaller the Y axis, the better the fitting effect of the equation. The top number is the number of variables left in the equation at different λ. The two dotted lines on the graph represent two special lambda (λ) values. The left dotted line is lambda.min, which means the λ when the deviation is the smallest, representing the highest fitting effect of the model at this lambda value. The right dotted line is lambda.1se, which means 1 standard error to the right of the minimum λ. Under the condition of lambda.min, 14 proteins are screened out, and then stepwise cox regression is used, and finally 8 proteins are screened out, including CD27, CD8A, GZMH, IL6, PGF, TIE2, TNFRSF12A and TWEAK. The HR value, 95% CI and p value of the eight important key proteins are shown in Figure 2

[0152] Among these proteins, the HR values of CD27 and TWEAK are less than 1, and are related to longer survival period; the HR values of other proteins are greater than 1, and are related to shorter survival period.​

[0153] Example 2, prediction score model construction

[0154] Among the 22 important key candidate marker proteins, the following 8 important marker proteins were selected as modeling proteins: CD27 protein, TWEAK protein, CD8A protein, GZMH protein, IL6 protein, PGF protein, TIE2 protein and TNFRSF12A protein.

[0155] Survival SVM is an R package for building a support vector machine Q(SVM) model for survival analysis. The package provides a function survivalsvm() for building the SVM model, which can be used to predict survival time.

[0156] The Survival SVM model described in this example was built according to the literature of Van Belle et al. (Van Belle, V. et al. A. k. (2011a); Bioinformatics (Oxford, England) 27, 87-94; Van Belle, V. et al. (2011b). Artificial intelligence in medecine 53 107-118).

[0157] Specific parameters of the model:

[0158] 1. kernel is add_kernel. The kernel parameter specifies the type of kernel function used by the support vector machine. "add_kernel" means using an additive kernel function. The additive kernel function is a special kernel function that can add the results of multiple kernel functions together, thereby integrating the advantages of different kernel functions.

[0159] 2. type is regression. The type parameter specifies the type of the survivalsvm model. "regression" means using a regression type of survival support vector machine model. This model is usually used to predict continuous results such as survival time or risk score.

[0160] 3. gamma.mu is 0.25. gamma.mu is a parameter in the kernel function. In support vector machines, the kernel function is used to map the input data into a high-dimensional space, so that linear classification or regression can be performed in the high-dimensional space. gamma.mu generally controls the width or smoothness of the kernel function. Smaller values of gamma.mu make the kernel function more smooth, and the model tends to consider global data; larger values of gamma.mu make the kernel function more sharp, and the model pays more attention to local data. Here gamma.mu = 0.25 is to set this parameter to 0.25.

[0161] 4. opt.meth is set to quadprog, opt.meth specifies the method used to solve the support vector machine optimization problem. "quadprog" means using the quadratic programming algorithm in the quadprog package to solve the optimization problem. Quadratic programming is an optimization method used to solve quadratic objective functions with linear constraints, which is used in support vector machines to find the optimal hyperplane or decision boundary.

[0162] Other parameters are default parameters.

[0163] The above-mentioned 8 protein expression amounts are substituted into the trained Survival SVM (Survival Support Vector Machine) model. According to the score of the Survival Support Vector Machine model, each sample is divided into High-risk and Low-risk by the model.

[0164] The model constructed based on the 8 marker proteins is evaluated based on the area under the receiver operating characteristic curve (AUC). In the training set, the overall survival (OS) is calculated. The results show that the AUC value for 6-month overall survival is 0.883, and the AUC value for 12-month overall survival is 0.860, as shown in Figure 3 .

[0165] The model constructed based on the 8 marker proteins is evaluated based on the area under the receiver operating characteristic curve (AUC). In the training set, the progression-free survival (PFS) is calculated. The results show that the AUC value for 6-month progression-free survival is 0.849, and the AUC value for 12-month progression-free survival is 0.680, as shown in Figure 4 .

[0166] The best threshold value of the Survival SVM prediction model constructed based on the 8 marker proteins is shown in Figure 5 , that is, the top 20% of the score (Predicted Score) from low to high is determined as the high-risk group of immunotherapy, and the prognosis is worse (the group not recommended for immunotherapy).

[0167] The training set 92 was divided into high-risk and low-risk groups according to the optimal threshold value, and the total survival KM curve was drawn, wherein the median overall survival of the low-risk group was 35.3 months (95% CI: 28.7 vs not reached), and the median overall survival of the high-risk group was 8 months (95% CI: 5.5 vs 13.3), as shown in Figure 6 A.

[0168] The training set 92 was divided into high-risk and low-risk groups according to the optimal threshold value, and the progression-free survival KM curve was drawn, wherein the median progression-free survival of the low-risk group was 14.8 months (95% CI: 12.5 vs 31.9), and the median progression-free survival of the high-risk group was 5.1 months (95% CI: 4.2 vs 9.9), as shown in Figure 6 B.

[0169] The single-factor COX regression analysis of the SurvivalSVM prediction model and the clinical characteristics (age, gender, smoking history, pathological subtype, and clinical stage) is shown in Figure 7 A, and the correlation P value of the model with OS (overall survival) and PFS (progression-free survival) is less than 0.001.

[0170] The multi-factor COX regression analysis of the SurvivalSVM prediction model and the clinical characteristics (age, gender, smoking history, pathological subtype, and clinical stage) is shown in Figure 7 B, and the correlation P value of the model with OS and PFS is less than 0.001.

[0171] Example Three, Performance Evaluation / Validation of the Prediction Score Model (I)

[0172] The established prediction score model was validated in the validation set (102 samples) data: the receiver operating characteristic curve (ROC) was calculated and drawn using the pROC software package (version 1.18.0), and the area under the curve (AUC) value was calculated.

[0173] The receiver operating characteristic curve (ROC) calculated and drawn in the validation set data matrix using the pROC software package (version 1.18.0). According to this, the AUC value for 6-month overall survival in the validation set was 0.793, and the AUC value for 12-month overall survival was 0.779, as shown in Figure 9 A.

[0174] The receiver operating characteristic curve (ROC) calculated and drawn in the validation set data matrix using the pROC software package (version 1.18.0). According to this, the AUC value for 6-month progression-free survival in the validation set was 0.783, and the AUC value for 12-month progression-free survival was 0.631, as shown in Figure 9 B.

[0175] Example 4, Performance evaluation / validation of the prediction scoring model (two)

[0176] The size of the prediction score in the prediction scoring model of the present application represents the size of the probability of the risk of immunotherapy for the non-small cell lung cancer patient. By detecting the protein expression level of the marker protein of the patient (at the time of detection, each target protein corresponds to 2 antibodies, and the 2 antibodies carry / couple a pair of single-stranded nucleotide sequences, and when the 2 antibodies are combined with the target protein, the 2 single-stranded nucleotide sequences are complementary to form a double-stranded DNA sequence. Then, amplification is performed, and the strength of the amplification signal is detected, and then the target protein is quantified), and the protein expression value of the marker protein of the patient is input into the scoring model, the prediction score of the non-small cell lung cancer patient can be obtained. According to the best threshold value of the Survival SVM prediction model in the training set, the top 20% of patients with the prediction score from low to high are classified into the high-risk group (High Risk). Accordingly, the efficacy or risk of immunotherapy for non-small cell lung cancer patients can be effectively predicted.

[0177] The prediction scores of the 102 samples in the validation set will be calculated according to the prediction scoring model of the present application, and the samples in the validation set will be divided into the high-risk group (High Risk) and the low-risk group (Low Risk) according to the best threshold value of the model.

[0178] The grouping of the prediction scores of the 102 samples in the validation set calculated by the model and the KM curve graphs of the total survival time and the progression-free survival time of the clinical are shown in Figure 10 A and Figure 10 B.

[0179] According to the results of Figure 10 A, it can be seen that in the validation set, the prediction scoring model of the present application predicts 21 people in the high-risk group, and the prediction scoring model of the present application predicts 81 people in the low-risk group. It can be seen that in the low-risk group predicted by the prediction scoring model of the present application, the median total survival time is significantly higher than that in the high-risk group, which is not reached and 15.5 months (P value is less than 0.001).

[0180] According to the results of Figure 10 B, it can be seen that in the validation set, the prediction scoring model of the present application predicts 21 people in the high-risk group, and the prediction scoring model of the present application predicts 81 people in the low-risk group. It can be seen that in the low-risk group predicted by the prediction scoring model of the present application, the median progression-free survival time is significantly higher than that in the high-risk group, which is 15.1 months and 21.5 months (P value is equal to 0.006). This shows that the prediction scoring model of the present application can more accurately predict the prognosis of non-small cell lung cancer patients receiving immunotherapy combined with chemotherapy.

[0181] Figure 11A shows the univariate COX regression analysis of the SurvivalSVM predictive model and clinical characteristics (age, gender, smoking history, pathological subtype and clinical stage) in the validation set, the correlation of the model with OS and PFS P values are less than 0.05.

[0182] Figure 11 B shows the multivariate COX regression analysis of the SurvivalSVM predictive model and clinical characteristics (age, gender, smoking history, pathological subtype and clinical stage) in the validation set, the correlation of the model with OS and PFS P values are less than 0.05.

[0183] Example Five, Performance Evaluation / Validation of the Predictive Score Model (Three)

[0184] In the training set of 92 samples, 42 samples with PD-L1 TPS score were analyzed for immunotherapy (patient data were collected retrospectively); in the validation set of 102 samples, 43 samples with PD-L1 TPS score were analyzed for immunotherapy (patient data were collected retrospectively). The analysis scheme was the score model analysis of the application, or TPS analysis, and the accuracy of the two analysis methods was compared.

[0185] 1. Calculation of the predictive score model of the application

[0186] For the training set of 42 samples and the validation set of 43 samples, the predictive score of each sample was calculated according to the predictive score model of the application, and the samples were divided into a high-risk group (High-Risk) and a low-risk group (Low-Risk).

[0187] The predictive score grouping of the 42 samples calculated by the predictive score model of the application in the training set and the clinical prognosis, that is, the KM curve of overall survival and progression-free survival, is shown in Figure 8 A and Figure 8 B. The patients were divided into two groups according to the best threshold of the SurvivalSVM predictive model, and the overall survival of the two groups was significantly separated (P<0.001); the progression-free survival of the two groups was significantly separated (P<0.05).

[0188] The predictive score grouping of the 43 samples calculated by the predictive score model of the application in the validation set and the clinical prognosis, that is, the KM curve of overall survival and progression-free survival, is shown in Figure 12 A and Figure 12 B. The patients were divided into two groups according to the best threshold of the SurvivalSVM predictive model, and the overall survival and progression-free survival of the two groups could be distinguished.

[0189] 2. TPS score analysis

[0190] Tumor Proportion Score (TPS) is a kind of index in tumor immunotherapy, using percentage numerical value, that is, 0-100%, the greater the value, the higher the sensitivity of tumor to immunotherapy, and the greater the opportunity to achieve good efficacy of immunotherapy alone, and without combination chemotherapy, which can reduce the treatment toxicity and treatment cost of patients. The TPS score value is used to determine whether to use immunotherapy drugs in clinic, and to guide the use of anti-tumor immunotherapy drugs.

[0191] In the training set of 92 samples, 42 samples were detected by PD-L1 immunohistochemistry, and the TPS scores of 42 samples were obtained. In the validation set of 102 samples, 43 samples were detected by PD-L1 immunohistochemistry, and the TPS scores of 43 samples were obtained. In this embodiment, PD-L1 immunohistochemical detection uses PD-L1 detection kit (immunohistochemical method) (SK006), and the experimental steps and TPS scores are shown in the instruction manual.

[0192] In the training set of 42 samples, when TPS analysis was used, according to the threshold value of 1% of PD-L1 expression level, the samples were divided into two groups of PD-L1 positive and negative, and the overall survival time of the two groups could not be significantly separated, and the P value was equal to 0.5, as shown in Figure 8 C.

[0193] In the training set of 42 samples, when TPS analysis was used, according to the threshold value of 1% of PD-L1 expression level, the samples were divided into two groups of PD-L1 positive and negative, and the overall survival time of the two groups could not be significantly separated, and the P value was equal to 0.5, as shown in Figure 8 D.

[0194] In the training set of 42 samples, when TPS analysis was used, according to the threshold value of 50% of PD-L1 expression level, the samples were divided into two groups of PD-L1 positive and negative, and the overall survival time of the two groups could not be significantly separated, and the P value was equal to 0.14, as shown in Figure 8 E.

[0195] In the training set of 42 samples, when TPS analysis was used, according to the threshold value of 50% of PD-L1 expression level, the samples were divided into two groups of PD-L1 positive and negative, and the overall survival time of the two groups could not be significantly separated, and the P value was equal to 0.16, as shown in Figure 8 F.

[0196] In the validation set of 43 samples, when TPS analysis was used, according to the threshold value of 1% of PD-L1 expression level, the samples were divided into two groups of PD-L1 positive and negative, and the overall survival time of the two groups could not be significantly separated, and the P value was equal to 0.16, as shown in Figure 12 C.

[0197] In the validation set of 43 samples, when TPS analysis was used, the 2 groups of PD-L1 positive and negative were divided according to the threshold value of PD-L1 expression level of 1%, and the progression-free survival of the 2 groups could not be significantly separated, with a P value equal to 0.46, as shown in FIG. Figure 12 D.

[0198] In the validation set of 43 samples, when TPS analysis was used, the 2 groups of PD-L1 positive and negative were divided according to the threshold value of PD-L1 expression level of 50%, and the overall survival of the 2 groups could not be significantly separated, with a P value equal to 0.69, as shown in FIG. Figure 12 E.

[0199] In the validation set of 43 samples, when TPS analysis was used, the 2 groups of PD-L1 positive and negative were divided according to the threshold value of PD-L1 expression level of 50%, and the progression-free survival of the 2 groups could not be significantly separated, with a P value equal to 0.76, as shown in FIG. Figure 12 F.

[0200] Therefore, through sufficient analysis and demonstration, it can be seen that, compared with the predictive score of PD-L1, the predictive model of the present application can more accurately predict the survival and recurrence of patients with non-small cell lung cancer receiving immunotherapy combined with chemotherapy. Figure 8 A to Figure 8 F.

[0201] Compared with the predictive score of PD-L1, the predictive model of the present application can more accurately predict the survival and recurrence of patients with non-small cell lung cancer receiving immunotherapy combined with chemotherapy. Figure 12 A to Figure 12 F.

[0202] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit the protection scope of the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the essence and scope of the technical solutions of the present application.

Claims

1. Application of a marker molecule combination in establishing an immunotherapy suitability analysis system for non-small cell lung cancer; the marker molecule combination comprises: CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, TNFRSF12A.

2. The use according to claim 1, characterized in that The analysis system includes: an analysis model; preferably, the analysis model is a survival support vector machine model established using the marker molecules in the marker molecule combination as modeling molecules; Preferably, the analysis system further comprises: a marker molecule expression level determination module; more preferably, the marker molecule expression level determination module comprises: a detection reagent, a kit and / or a detection device.

3. The use according to claim 2, characterized in that The expression level of each marker molecule in the marker molecule combination is substituted into the analysis model to calculate a prediction score, and the immunotherapy suitability of the sample to be tested is determined based on the prediction score.

4. The use according to any one of claims 1 to 3, characterized in that: The immunotherapy includes: immune checkpoint inhibitor therapy, or combined therapy with immune checkpoint inhibitors and chemotherapy drugs; more preferably, the immune checkpoint inhibitors include: anti-PD-L1 antibodies, anti-PD-1 antibodies or anti-CTLA-4 inhibitors; or The analysis of the suitability of immunotherapy includes: analyzing the sensitivity of non-small cell lung cancer patients to immunotherapy, performing risk analysis or scoring for the development of non-small cell lung cancer; or formulating a treatment / medication plan; or analyzing the pathological remission and / or survival of non-small cell lung cancer patients after immunotherapy.

5. A method for establishing a model for analyzing the suitability of immunotherapy for non-small cell lung cancer, comprising: A survival support vector machine model was established using a marker molecule combination as a modeling molecule; the marker molecule combination included: CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, and TNFRSF12A.

6. The method according to claim 5, wherein The method for establishing the model includes: obtaining the expression level of each marker molecule in the marker molecule combination in the sample, and establishing a survival support vector machine model based on the expression level of each marker molecule.

7. The method according to claim 5, wherein The modeled molecule is obtained by a method comprising: (1) Provide plasma samples from patients with non-small cell lung cancer and obtain protein expression data in the samples; (2) The expression data of (1) are used as a training set; in the training set, a univariate COX regression analysis is performed using the overall survival and the expression level of the candidate protein to obtain the candidate protein significantly correlated with the overall survival; and the modeling molecule is obtained by combining Lasso COX regression with stepwise Cox regression.

8. An analysis model for the suitability of immunotherapy for non-small cell lung cancer, which is an analysis model established by the method according to any one of claims 5 to 7.

9. A system for analyzing the suitability of immunotherapy for non-small cell lung cancer, comprising: (i) a module for measuring the expression level of marker molecules of a sample to be tested; This module is used to determine the expression levels of marker molecules in a marker molecule combination, wherein the marker molecule combination includes: CD27, TWEAK, CD8A, GZMH, IL6, PGF, TIE2, and TNFRSF12A; preferably, the marker molecule expression level determination module includes: a detection reagent, a kit and / or a detection device; (ii) An analysis model for the suitability of immunotherapy for non-small cell lung cancer, or a hardware system incorporating the analysis model; the analysis model is an analysis model established by the method according to any one of claims 5 to 7.

10. A method for analyzing the suitability of immunotherapy for non-small cell lung cancer, comprising: Performing analysis using the analysis model described in claim 8; or Performing analysis using the analysis system according to claim 9; Preferably, the method includes: analyzing the sensitivity of non-small cell lung cancer patients to immunotherapy based on the expression level of each marker molecule in the marker molecule combination; preferably, it also includes: formulating a treatment / medication plan; preferably, the immunotherapy includes: immune checkpoint inhibitor therapy, or a combination of immune checkpoint inhibitors and chemotherapy drugs.

Images