Tumor treatment prognosis evaluation method and system based on machine learning

Abstract

The invention provides a tumor treatment prognosis evaluation method and system based on machine learning. The method comprises the following steps: S1, analyzing the influence of each immune cell subset of peripheral blood of a patient on tumor treatment responsiveness through a single-factor logistic regression model; s2, based on the total lifetime of the patient, constructing a multi-factor COX risk scale model by using the immune cell subpopulation, and calculating a peripheral blood immune pattern risk score of the patient; and S3, dividing the patients into a high-risk group and a low-risk group according to the median value of the risk score. According to the tumor treatment prognosis evaluation method and system based on machine learning, the peripheral blood immune pattern of the advanced tumor patient at different time points can be evaluated while noninvasive and dynamic material taking is achieved, then the influence of small-dose rhythm chemotherapy on prognosis of the advanced tumor patient is accurately and dynamically predicted, and the method and system are suitable for clinical transformation and popularization and have good application prospects. And a reference value is provided for treatment scheme selection of advanced tumor patients.

Description

Technical Field

[0001] The present invention belongs to the field of biotechnology, and particularly relates to a method and system for evaluating the prognosis of tumor treatment based on machine learning. Background Art

[0002] At present, the treatment methods for cancer mainly include comprehensive means such as surgical operation combined with radiotherapy, chemotherapy or targeted drug treatment. Among them, chemotherapy plays an extremely important role in the treatment plan for advanced tumors. Conventional chemotherapy uses the maximum tolerated dose (MTD), which can directly kill tumor cells by using cytotoxic drugs. However, chemotherapy based on MTD has a killing effect on all proliferating cells, with obvious adverse reactions and poor patient compliance. Metronomic chemotherapy refers to the tight and regular administration of relatively low doses (usually 1 / 10 - 1 / 3 of the conventional dose) of cytotoxic drugs (such as cyclophosphamide, vinblastine, topotecan, cisplatin, etc.). The main anti-tumor mechanism of metronomic chemotherapy is to inhibit tumor-related angiogenesis and stimulate the body's immunity, rather than the direct cytotoxic effect on tumor cells. It has been confirmed that metronomic chemotherapy has minor adverse reactions in various tumors. Currently, for advanced tumor patients, a comprehensive treatment plan mainly based on metronomic chemotherapy is usually adopted, that is, the "tumor treatment" described in this patent. The prognosis evaluation of tumor treatment is of great significance for improving the survival rate of patients, optimizing the medication plan and reducing unnecessary treatments. The existing technologies mainly focus on the benefit prediction or prognosis evaluation of neoadjuvant chemotherapy or adjuvant chemotherapy for a specific type of cancer, and the modeling data used are mostly transcriptome data of surgical tissue samples, which are relatively difficult to obtain. Therefore, in order to more accurately evaluate whether tumor treatment can extend the survival time of advanced tumor patients, it is urgent to develop a reliable tumor treatment prognosis evaluation model.

[0003] Based on the above problems, the present invention provides a method and system for evaluating the prognosis of tumor treatment, which can accurately and dynamically predict the impact of metronomic chemotherapy for pan-cancer on the prognosis of advanced tumor patients, and provide reference value for the selection of treatment plans for advanced tumors. Summary of the Invention

[0004] The present invention provides a method and system for evaluating the prognosis of tumor treatment based on machine learning, which can non-invasively and dynamically obtain samples while evaluating the peripheral blood immune pattern of advanced tumor patients at different time points, predict the impact of metronomic chemotherapy for pan-cancer on the prognosis of advanced tumor patients, and is suitable for clinical transformation and promotion.

[0005] The technical solution adopted by the present invention to achieve the above object is: a method for evaluating the prognosis of tumor treatment based on machine learning, the method comprising:

[0006] S1: Analyze the impact of each immune cell subset in the peripheral blood of patients on the responsiveness to tumor treatment through a univariate logistic regression model;

[0007] S2: Based on the overall survival of patients, construct a multivariate COX proportional hazards model using immune cell subsets and calculate the risk score of the peripheral blood immune pattern of patients;

[0008] S3: Divide patients into a high-risk group and a low-risk group according to the median value of the risk score.

[0009] Preferably, the specific operation of S1 is as follows:

[0010] S11: Perform multi-color flow cytometry analysis on peripheral blood samples of advanced tumor patients receiving low-dose rhythmic chemotherapy at different time points (i.e., the time points when patients come for chemotherapy during follow-up), that is, monitor the dynamic immune pattern of peripheral blood to obtain evaluation data, and then analyze the impact of each immune cell subset in the peripheral blood of patients on the responsiveness to tumor treatment through a univariate logistic regression model;

[0011] S12: Perform univariate survival curve analysis on each immune cell subset in the peripheral blood of patients, using the median infiltration ratio as the cut-off value. It is found that there is no significant difference in survival between the high and low infiltration groups of each immune cell subset in the peripheral blood except for neutrophils. Therefore, it is necessary to integrate immune cell subsets to construct a multivariate model.

[0012] Preferably, the impact in S11 is that, compared with patients with no response to treatment, the IFNγ + PD1 - CD8 + T cell subsets have a positive impact, while PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cell subsets have a negative impact.

[0013] Preferably, the immune cell subsets in S2 include IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells.

[0014] Preferably, the specific operation of S2 is as follows:

[0015] Based on the overall survival of the patient, using IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells, by calculating the HR value (Hazard Ratio) of each immune cell subset, taking the ln value and multiplying it by the infiltration proportion of the cell subset and then summing, calculate the peripheral blood immune pattern risk score of the patient.

[0016] Preferably, the peripheral blood immune pattern risk score in S2 = Σln(HRi)*proportioni, where i in HRi is IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T immune cell subsets.

[0017] Preferably, the grouping criteria for the high-risk group and the low-risk group in S3 are: the high-risk group is the collection of patients with a peripheral blood immune pattern risk score greater than the median value (i.e., the median), and the low-risk group is the collection of patients with a peripheral blood immune pattern risk score less than or equal to the median value.

[0018] The present invention also provides a tumor treatment prognosis evaluation system based on machine learning, including:

[0019] A data screening module, which analyzes the influence of each immune cell subset in the peripheral blood of the patient on the responsiveness to low-dose rhythmic chemotherapy through a univariate logistic regression model;

[0020] A model construction module, which constructs a multivariate COX proportional hazards model using immune cell subsets based on the overall survival of the patient,

[0021] and calculates the peripheral blood immune pattern risk score of the patient;

[0022] A risk assessment module, which divides the patients into a high-risk group and a low-risk group according to the median value of the risk score.

[0023] The present invention also provides a terminal device, which includes a processor and a computer-readable storage medium. The processor is used to implement various instructions, and the computer-readable storage medium is used to store multiple instructions, and the instructions are adapted to be loaded and executed by the processor to perform the above-mentioned tumor treatment prognosis evaluation method.

[0024] The present invention also provides a computer-readable storage medium, in which multiple instructions are stored, and the instructions are loaded and executed by the processor of the terminal device to perform the above-mentioned tumor treatment prognosis evaluation method.

[0025] Compared with the prior art, the beneficial effects that can be achieved by at least one of the above technical solutions adopted in the embodiments of the present specification at least include:

[0026] First, source innovation: The prediction data used in the prognosis evaluation model of the present invention is the cell subsets identified by multi-dimensional panel flow cytometry of peripheral blood samples of advanced cancer patients at different time points. It can non-invasively and dynamically obtain samples while evaluating the peripheral blood immune pattern of advanced cancer patients at different time points, and is suitable for clinical transformation and promotion.

[0027] Second, accurate prediction: Based on the prognosis model of the present invention, the AUC value of the predicted 2-year overall survival rate is 0.794, the AUC value of the 3-year overall survival rate is 0.779, and the AUC value of the 4-year overall survival rate is 0.966. The effect is significant and very accurate.

[0028] Third, applicable to multiple cancers: The prognosis evaluation model of the present invention can predict the impact of low-dose rhythmic chemotherapy on the prognosis of advanced cancer patients by evaluating the peripheral blood immune pattern of patients with multiple cancers. It has a wide range of applications and high clinical application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required to be used in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the drawings without creative efforts.

[0030] Figure 1 It is a result diagram of analyzing the influence of each immune cell subset in the peripheral blood of patients on the responsiveness to tumor treatment by a univariate logistic regression model;

[0031] Figure 2 It is the HR value of each cell subset;

[0032] Figure 3 It is the survival analysis result diagram of Example 4;

[0033] Figure 4 It is the survival rate analysis result diagram. Detailed implementation manners

[0034] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0035] The following uses specific examples to illustrate the implementation manners of the present application. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The present application can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts fall within the protection scope of the present application.

[0036] Embodiment 1

[0037] A method for evaluating the prognosis of tumor treatment based on machine learning, the method comprising:

[0038] S1: Analyze the influence of each immune cell subset in the peripheral blood of the patient on the responsiveness of tumor treatment through a univariate logistic regression model;

[0039] S2: Based on the overall survival of the patient, use IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells, by calculating the HR value of each immune cell subset, taking the ln value and multiplying it by the infiltration ratio of the cell subset and then summing (specific formula: peripheral blood immune pattern risk score = Σln(HRi)*proportion i, where i in HRi is IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T immune cell subset), calculate the peripheral blood immune pattern risk score of the patient.

[0040] S3: Classify patients into a high-risk group and a low-risk group according to the median value (i.e., the median) of the risk score (i.e., the high-risk group is the collection of patients with a peripheral blood immune pattern risk score greater than the median value, and the low-risk group is the collection of patients with a peripheral blood immune pattern risk score less than or equal to the median value).

[0041] Among them, the specific operation of S1 is as follows:

[0042] S11: Perform multi-color flow cytometry analysis on peripheral blood samples of advanced cancer patients receiving low-dose rhythmic chemotherapy at different time points (i.e., the time points when the patients come for chemotherapy each time during follow-up), that is, perform dynamic monitoring of the peripheral blood immune pattern to obtain evaluation data, and then analyze the impact of each immune cell subset in the peripheral blood of the patients on the response to tumor treatment through a univariate logistic regression model. It is found that compared with patients with no response to treatment, the IFNγ + PD1 - CD8 + T cell subset has a positive impact, while PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cell subsets have a negative impact;

[0043] S12: Perform univariate survival curve analysis on each immune cell subset IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells in the peripheral blood of patients respectively, with the median infiltration ratio as the cut-off value. It is found that there is no significant difference in survival between the high and low infiltration groups of each immune cell subset in the peripheral blood except for neutrophils, so it is necessary to integrate immune cell subsets to construct a multi-factor model.

[0044] Example 2

[0045] The purpose of this example is to provide a tumor treatment prognosis evaluation system based on machine learning, including:

[0046] A data screening module that analyzes the impact of each immune cell subset in the peripheral blood of patients on the response to low-dose rhythmic chemotherapy through a univariate logistic regression model;

[0047] A model construction module, based on the overall survival of patients, constructs a multi-factor COX proportional hazards model using immune cell subsets and calculates the risk score of the peripheral blood immune pattern of the patients;

[0048] A risk assessment module divides the patients into a high-risk group and a low-risk group according to the median value of the risk score.

[0049] Example 3

[0050] The purpose of this example is to provide a terminal device, which includes a processor and a computer-readable storage medium. The processor is used to implement various instructions, and the computer-readable storage medium is used to store multiple instructions. The instructions are adapted to be loaded and executed by the processor to perform the tumor treatment prognosis assessment method described in Example 1.

[0051] Example 4

[0052] The purpose of this example is to provide a computer-readable storage medium, in which multiple instructions are stored. The instructions are loaded and executed by the processor of the terminal device to perform the tumor treatment prognosis assessment method described in Example 1.

[0053] Example 5

[0054] The following gives the effect verification of the tumor treatment prognosis assessment method based on machine learning of the present invention. The specific operations are as follows:

[0055] S1: Perform multi-color flow cytometry analysis on the peripheral blood samples of 83 patients with pan-cancer (colon cancer, gastric cancer) who received tumor treatment at Huashan Hospital at different time points (that is, the time points when the patients came for chemotherapy each time during the follow-up). That is, monitor the dynamic immune pattern of the peripheral blood a total of 185 times to obtain the data for evaluation. Then, analyze the influence of each immune cell subset in the peripheral blood of the patients on the response to tumor treatment through a univariate logistic regression model: For patients with treatment response (PR+CR) compared to patients without treatment response (PD+SD), IFNγ + PD1 - CD8 + The T cell subset has a positive effect (OR>1), while PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + Tcell, IFNγ - PD1 + CD8 + The T cell subset has a negative effect (OR<1) ( Figure 1 ); Subsequently, for each immune cell subset in the peripheral blood of 83 patients, IFNγ + PD1 - CD8 +T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + Univariate survival curve analysis was performed on T cells, with the median infiltration ratio as the cutoff value. It was found that except for neutrophils, there was no significant difference in survival between the high and low infiltration groups of peripheral blood immune cell subsets. Therefore, it is necessary to integrate immune cell subsets to construct a multivariate model.

[0056] S2: Using IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells, a multi-factor COX risk proportional model was constructed, and the HR values of each cell subset were as follows Figure 2 As shown, IFNγ + PD1 - CD8 + The HR value of T was 1.58, the HR value of PMN-MDSC was 0.68, the HR value of WBC was 0.92, the HR value of Neutrophils was 1.21, and the HR value of PD1 + CD8 + The HR value of T was 4.07, IFNγ - PD1 + CD8 + The HR for T was 0.19. The peripheral blood immune pattern risk score for each patient was calculated by calculating Σln(HRi)*proportioni (i.e., the HR value for each immune cell subset, multiplied by the infiltration proportion of that cell subset, and then summed. Where i represents the six different immune cell subsets mentioned above).

[0057] S3: Take the median value of the risk score (0.05685611), and divide the patients into high-risk group and low-risk group according to the median value. That is, the high-risk group is the collection of patients whose peripheral blood immune pattern risk score is greater than the median value 0.05685611, and the low-risk group is the collection of patients whose peripheral blood immune pattern risk score is less than or equal to the median value 0.05685611. Survival analysis ( Figure 3 ) showed that there was a significant difference in overall survival between the high-risk and low-risk groups, with p value = 0.005.

[0058] Based on the above prognostic model (the longest follow-up time of patients was 4 years), the predicted 2-year overall survival rate AUC value was 0.794, the 3-year overall survival rate AUC value was 0.779, and the 4-year overall survival rate AUC value was 0.966 ( Figure 4 ).

[0059] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A method for evaluating the prognosis of tumor treatment based on machine learning, characterized in that, The method includes: S1: Analyze the influence of each immune cell subset in the peripheral blood of patients on the responsiveness to low-dose rhythmic chemotherapy through a univariate logistic regression model; S2: Based on the overall survival of patients, construct a multivariate COX proportional hazards model using immune cell subsets, and calculate the peripheral blood immune pattern risk score of patients; S3: Divide patients into a high-risk group and a low-risk group according to the median value of the risk score.

2. The tumor treatment prognosis evaluation method according to claim 1, wherein The specific operation of S1 is as follows: S11: Perform multi-color flow cytometry analysis on peripheral blood samples of advanced cancer patients receiving low-dose rhythmic chemotherapy at different time points, that is, monitor the dynamic immune pattern of peripheral blood to obtain evaluation data, and then analyze the influence of each immune cell subset in the peripheral blood of patients on the responsiveness to tumor treatment through a univariate logistic regression model; S12: Perform univariate survival curve analysis on each immune cell subset in the peripheral blood of patients, with the median infiltration ratio as the cut-off value. It is found that there is no significant difference in survival between the high and low infiltration groups of each immune cell subset in the peripheral blood except neutrophils, so it is necessary to integrate immune cell subsets to construct a multivariate model.

3. The tumor treatment prognosis evaluation method according to claim 2, characterized in that, The influence in S11 is that, compared with patients with no treatment response, IFNγ in patients with treatment response + PD1 - CD8 + The T cell subset has a positive influence, while PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + The T cell subset has a negative influence.

4. The tumor treatment prognosis evaluation method according to claim 1, wherein The immune cell subsets in S2 include IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells.

5. The tumor treatment prognosis evaluation method according to claim 1, wherein The specific operation of S2 is as follows: Based on the overall survival of patients, using IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T cells. By calculating the HR value of each immune cell subset, taking the ln value and multiplying it by the infiltration proportion of the cell subset and then summing, the peripheral blood immune pattern risk score of the patient was calculated.

6. The tumor treatment prognosis evaluation method according to claim 5, characterized in that, The peripheral blood immune pattern risk score in S2 = Σln(HRi)*proportioni, where i in HRi is IFNγ + PD1 - CD8 + T, PMN-MDSC, WBC, Neutrophils, PD1 + CD8 + T and IFNγ - PD1 + CD8 + T immune cell subsets.

7. The method for evaluating the prognosis of tumor treatment according to claim 1, wherein The grouping criteria for the high-risk group and the low-risk group in S3 are: the high-risk group is the collection of patients with a peripheral blood immune pattern risk score greater than the median value, and the low-risk group is the collection of patients with a peripheral blood immune pattern risk score less than or equal to the median value.

8. A tumor treatment prognosis evaluation system based on machine learning, characterized in that Including: A data screening module, which analyzes the influence of each immune cell subset in the peripheral blood of patients on the responsiveness to low-dose rhythmic chemotherapy through a univariate logistic regression model; A model construction module, which constructs a multivariate COX proportional hazards model using immune cell subsets based on the overall survival of patients, and calculates the peripheral blood immune pattern risk score of patients; A risk assessment module, which divides patients into a high-risk group and a low-risk group according to the median value of the risk score.

9. A terminal device, which includes a processor and a computer-readable storage medium, where the processor is used to implement various instructions, and the computer-readable storage medium is used to store multiple instructions, characterized in that, The instructions are suitable for being loaded and executed by a processor to perform the tumor treatment prognosis evaluation method according to any one of claims 1-5.

10. A computer-readable storage medium storing multiple instructions, characterized in that, The instructions are loaded and executed by the processor of the terminal device to perform the tumor treatment prognosis evaluation method according to any one of claims 1-5.

Images