Method for constructing prognosis model related to neoadjuvant chemotherapy pyroptosis of breast cancer

Abstract

The invention discloses a breast cancer neoadjuvant chemotherapy pyroptosis related prognosis model construction method which comprises the following steps: obtaining an original data set, and preprocessing to obtain an annotated whole-gene expression matrix and a survival clinical information matrix; screening out a gene database having significant differential expression on patient survival; obtaining a pyroptosis related gene database; taking an intersection of the pyroptosis related gene database and the gene database having significant differential expression on the survival of the patient to obtain a list of differential expression genes related to the survival of the pyroptosis; and carrying out disease survival prediction model batch construction on the differential expression gene list, screening out two optimal disease survival prediction models as sub-models, and weighting to obtain a prognosis model. According to the method, the problem that gene pathways of a prognosis model of the breast cancer neoadjuvant chemotherapy patient are deficient is solved, and the prognosis prediction capability of the breast cancer patient receiving the neoadjuvant chemotherapy is expected to be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and in particular to a method for constructing a pyroptosis-related prognosis model for adjuvant chemotherapy of breast cancer based on R language. Background Art

[0002] Breast cancer is a malignant tumor with a high incidence rate, threatening the health of women worldwide. At present, the main basis for the treatment and prognosis of breast cancer is the TNM staging system and the molecular classification of breast cancer. However, even among different breast cancer patients with the same TNM stage and the same molecular classification, their clinical outcomes or prognoses are still quite different, further proving that breast cancer is a highly heterogeneous disease. At present, neoadjuvant chemotherapy (NAC) has become an important part of the standard treatment of breast cancer and many other cancers. By downstaging the tumor, patients who cannot undergo surgery can meet the surgical conditions and improve the long-term prognosis of patients. However, due to the heterogeneity of patients' response to chemotherapy, even breast cancer patients with the same molecular classification still have large differences in the treatment results of neoadjuvant chemotherapy. Exploring new prognosis prediction methods and more advanced prognosis model construction methods has become a research hotspot. Therefore, there is an urgent need to build a more accurate prognosis prediction model and find new biomarkers to predict the treatment response and prognosis of breast cancer patients to neoadjuvant chemotherapy. In neoadjuvant chemotherapy for breast cancer, it has been found that many anticancer drugs limit the occurrence of malignant tumors by inducing cell pyroptosis. Studies have shown that chemotherapy drugs such as doxorubicin, paclitaxel and cisplatin can induce cell pyroptosis in breast cancer by activating the ROS / JNK signaling pathway dependent on the Caspase family and the downstream pathway mediated by the Gasdermin family protein, thereby exerting an anti-tumor effect. The above studies all suggest that the mechanism of some classic chemotherapy drugs in the anti-cancer process may have a strong correlation with cell pyroptosis. Predicting the effect of chemotherapy drugs through the cell pyroptosis pathway will help provide new strategies for the prognosis prediction and treatment plan selection of neoadjuvant chemotherapy for breast cancer.

[0003] In recent years, with the rapid development of high-throughput sequencing technology, the size of gene chip databases such as TCGA and GEO has been growing. Using bioinformatics technology, in-depth analysis of massive gene expression and sequencing information from laboratories around the world in the database is carried out. With the advent of the big data era, prediction of breast cancer treatment has become possible.

[0004] However, the current prediction models for neoadjuvant chemotherapy for breast cancer still face problems such as limited model gene pathway parameters and low predictive ability. Most of the genes in the existing models are concentrated in the ferroptosis pathway, the cell senescence pathway, or scattered in many other signaling pathways; in addition, there is no prognostic model for neoadjuvant chemotherapy for breast cancer constructed with genes related to the cell pyroptosis pathway. Summary of the invention

[0005] Based on the shortcomings in the research field of prediction models for neoadjuvant chemotherapy for breast cancer, the present invention proposes a prognosis prediction model constructed with genes related to the cell pyroptosis pathway, which makes up for the lack of gene pathways in the prognosis model of patients with neoadjuvant chemotherapy for breast cancer, and is expected to improve the clinical prognosis prediction ability for breast cancer patients who have received neoadjuvant chemotherapy.

[0006] The technical solution provided by the present invention is as follows:

[0007] In a first aspect, the present invention provides a method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer, comprising the following steps:

[0008] The original data set of breast cancer cases that received neoadjuvant chemotherapy was obtained, preprocessed, and the annotated full gene expression matrix and survival clinical information matrix were obtained through information collation and manual annotation;

[0009] Based on the annotated full gene expression matrix and survival clinical information matrix, we screened out the gene database with significant differential expression on patient survival;

[0010] Obtain the cell pyroptosis-related gene database;

[0011] Take the intersection of the pyroptosis-related gene database and the gene database with significant differential expression on patient survival to obtain a list of differentially expressed genes related to pyroptosis survival;

[0012] The disease survival prediction model was constructed in batches for the differentially expressed gene list through LASSO regression, and the risk score of each model was calculated in the annotated whole gene expression matrix, and the disease survival prediction model was divided into high and low groups; the high and low groups into which the risk scores of the disease survival prediction model were divided were regarded as the model's predictive indicators for the patient's prognosis, which were used for subsequent model screening;

[0013] By statistically analyzing whether the survival difference between the high and low groups is significant, and combining the evaluation of the model's predictive ability, two optimal disease survival prediction models were screened out as sub-models, and the prognostic model was obtained by weighting.

[0014] In one possible implementation, the original data set is obtained through a public cancer database, including: a full gene expression matrix, labels of tumor samples or normal samples, survival time and survival status, breast cancer classification information, basic information, and rating information of multiple tumor staging systems.

[0015] Furthermore, the preprocessing includes: annotating the whole gene expression matrix file using the probe sequence annotation file to obtain an annotated whole gene expression matrix.

[0016] In one possible implementation, the method for screening out a gene database with significant differential expression on patient survival based on the annotated whole gene expression matrix and the survival clinical information matrix includes: calling R language, using the survival clinical information matrix to perform a survival test on the annotated whole gene expression matrix, setting a significance threshold, and screening out a gene database with significant differential expression on patient survival.

[0017] In one possible implementation, prediction models are built in batches for the differentially expressed gene list through LASSO regression, the risk score of each model is calculated in the annotated full gene expression matrix, and the prediction models are divided into high and low groups. The steps are as follows:

[0018] Call R language to construct a disease survival prediction model for the differentially expressed gene list through LASSO regression;

[0019] Use loop statements to adjust the random seed of the regression function and the gene groups involved in the construction of the prediction model one by one, and batch build the disease survival prediction model;

[0020] The risk scores of the genes selected in the disease survival prediction model were calculated, and the patients were divided into high-risk group and low-risk group according to the median risk score to test whether there was a survival difference between the high-risk group and the low-risk group.

[0021] In one possible implementation, the optimal disease survival prediction model meets the following conditions:

[0022] 1. The disease survival model has a very significant difference in the survival of patients in the experimental data set and all validation data sets, that is, P < 0.03;

[0023] 2. The area under the receiver operating curve (AUC) of the disease survival model for the 1-5-year predicted survival rate of patients in the experimental set and all validation data sets was > 0.5;

[0024] 3. The disease survival model has an AUC area in the validation set that is more than 50% higher than that of similar prediction models in the field of breast cancer;

[0025] 4. The selected sub-models should contain as many different gene parameters as possible to ensure the parameter volume and pathway volume of the overall model.

[0026] Furthermore, the screening method of the optimal disease survival prediction model comprises:

[0027] Eliminate disease survival models that have no significant difference in survival prediction between the experimental set and the validation set divided into high and low groups;

[0028] Disease survival models with an area under the receiver operating characteristic curve of <0.5 for the 1-5 year survival prediction rate of all patients were eliminated;

[0029] The disease survival model has an AUC area that is at least 50% higher than that of similar prediction models in the breast cancer field in the validation set.

[0030] Further, the expression of the prognostic model is as follows:

[0031] Riskscore=0.422338*BRCA1+0.227932*VEGFA+0.116192*IL6+0.827851*MYD88+(- 0.131573)*NEK9+(-0.435585)*TMMEM9B+(-0.012124)*BCL2+(-0.441609)*CASP6+( -0.177918)*PYCARD+0.325638*NOD1+0.123377*NOD2+0.041331*MALT1+(-0.296448 )*NAIP+(-0.030816)*CHMP4A+(-0.016860)*PER2+0.030703*BID+0.037198*FXYD5;

[0032] Among them, Riskscore is the risk score, and BRCA1, VEGFA, IL6, MYD, NEK9, TMMEM9B, BCL2, CASP6, PYCARD, NOD1, NOD2, MALT1, NAIP, CHMP4A, PER2, BID, and FXYD5 are the expression values ​​of gene fragments.

[0033] In a second aspect, the present invention provides a device for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer, comprising:

[0034] The first acquisition module is used to obtain the original data set of breast cancer cases that received neoadjuvant chemotherapy, perform preprocessing, and obtain an annotated full gene expression matrix and a survival clinical information matrix;

[0035] The screening module is used to screen out gene databases with significant differential expression on patient survival based on the annotated full gene expression matrix and survival clinical information matrix;

[0036] The second acquisition module obtains the cell pyroptosis-related gene database;

[0037] The intersection calculation module is used to obtain the intersection of the cell pyroptosis-related gene database and the gene database with significant differential expression for patient survival, and obtain a list of differentially expressed genes related to cell pyroptosis survival;

[0038] The disease survival prediction model batch construction module is used to batch construct disease survival prediction models for the differentially expressed gene list through LASSO regression, calculate the risk score of each model in the annotated full gene expression matrix, and divide the disease survival prediction models into high and low groups; the high and low groups into which the risk scores of the disease survival prediction models are divided are regarded as the model's predictive indicators for patient prognosis, which are used for subsequent model screening;

[0039] The prognostic model building module is used to statistically determine whether the survival difference between the high and low groups is significant, and combined with the evaluation of the model's predictive ability, to screen out two optimal disease survival prediction models as sub-models, and obtain the prognostic model through weighting.

[0040] In a third aspect, the present invention provides a method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to the first aspect, which realizes the prediction of chemotherapy drug sensitivity and potential drug targets, comprising the following steps:

[0041] Obtain drug sensitivity data matrix and whole genome expression profile of all cell lines;

[0042] The pyroptosis gene set corresponding to breast cancer cell lines, breast cancer neoadjuvant chemotherapy drugs, and disease survival prediction models was screened to obtain a basic data set;

[0043] Based on the basic data set, the risk scores of breast cancer cell lines were calculated according to the disease survival prognosis prediction model and divided into high and low groups;

[0044] The gene set and risk score of the disease survival prediction model were used to test the correlation of each breast cancer chemotherapy drug, and a set of chemotherapy drugs closely related to gene expression differences was obtained;

[0045] Based on the whole genome sequencing results of a patient's pathological sample in clinical practice, the expression level and risk score level of the patient's chemotherapy drug-related genes are evaluated. Combined with the chemotherapy drug set that is closely related to gene expression differences, several chemotherapy drugs to which the patient is more sensitive are predicted.

[0046] The present invention provides a method for exploring the construction of a prediction model based on cell pyroptosis gene markers based on R language, and can further predict the survival expectancy and chemotherapy drug sensitivity of breast cancer patients who have received neoadjuvant chemotherapy. The present invention has the following beneficial effects:

[0047] 1. The method of the present invention makes up for the shortcomings of the lack of gene pathways in existing prognostic models. The existing prediction model for breast cancer patients undergoing neoadjuvant chemotherapy is only constructed through genes related to the ferroptosis pathway, and lacks the inclusion of genes related to other cell death modes, which to a certain extent affects the comprehensiveness and accuracy of the model. The method of the present invention is based on cell pyroptosis-related genes and proposes a prediction model based on this, which makes up for the lack of this gene pathway in the current research field.

[0048] 2. The method described in the present invention provides a new idea for model construction. First, several better sub-models are constructed through LASSO regression and other methods, and then the sub-models with different weights are merged into an overall model through MATLAB. The prediction ability of the overall model is visualized, and the optimal weights of the sub-models are selected to obtain a prediction overall model that comprehensively surpasses the sub-models, which has certain guiding significance.

[0049] 3. The present invention also provides a method based on a prognostic model that can predict a patient's sensitivity to chemotherapy drugs and potential drug targets, which can assist in the individualized use of drugs for patients in the clinic to a certain extent, make up for the defects of existing neoadjuvant chemotherapy prognosis prediction methods, and improve the accuracy of prognosis prediction.

[0050] 4. The prognostic model described in the present invention is expected to become a reference for clinically judging the prognosis of breast cancer patients receiving neoadjuvant chemotherapy, and has significant predictive ability for the survival expectation of breast cancer patients receiving neoadjuvant chemotherapy. Due to the small number of models in the field, the comparison interval is expanded to the ability to predict the survival of breast cancer patients. This model surpassed 80%, 80% and 86% of the mainstream LASSO regression and dimensionality reduction clustering algorithm models (including principal component dimensionality reduction analysis and hierarchical clustering, etc.) in GSE25055, GSE25066, and TCGA (n=1065), respectively, and can assist the clinic in predicting the outcome and drug sensitivity of patients' neoadjuvant chemotherapy to a certain extent, and then accurately adjust the treatment plan. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 The Venn diagram of the intersection of the survival significant genes and the pyroptosis gene database;

[0052] Figure 2 KM curve (top) and risk score (bottom) of the 9-gene submodel constructed for LASSO regression;

[0053] Figure 3 ROC curves of the 9-gene sub-model constructed for LASSO regression; the blue line is the ROC curve of the sub-model for 1-year survival prediction of the GSE25055 dataset, the red line is the ROC curve of the sub-model for 1-year survival prediction of the GSE25066 dataset, and the green line is the ROC curve of the sub-model for 1-year survival prediction of the TCGA dataset;

[0054] Figure 4 KM curve (top) and risk score (bottom) of the 11-gene submodel constructed for LASSO regression;

[0055] Figure 5 ROC curves of the 11-gene sub-model constructed for LASSO regression; the blue line is the ROC curve of the sub-model for 1-year survival prediction of the GSE25055 dataset, the red line is the ROC curve of the sub-model for 1-year survival prediction of the GSE25066 dataset, and the green line is the ROC curve of the sub-model for 1-year survival prediction of the TCGA dataset;

[0056] Figure 6 Display graph for MATLAB weights;

[0057] Figure 7 The KM curve (top) and risk score (bottom) of the final model;

[0058] Figure 8 is the ROC curve of the final model; the blue line, red line, and green line are the ROC curves of the final model for 1-, 3-, and 5-year survival prediction of the GSE25055 dataset, respectively;

[0059] Fig. 9 This is a graph showing the correlation between model scores and chemotherapy drug sensitivity;

[0060] Fig.10 This is the correlation diagram between the model parameter gene CASP6 and chemotherapy drug sensitivity;

[0061] Fig.11 This is the correlation diagram between the model parameter gene BID and chemotherapy drug sensitivity; DETAILED DESCRIPTION

[0062] The content of the present invention is further described below in conjunction with specific embodiments, but the content of the present invention is not limited thereto.

[0063] A method for constructing a prognostic model related to pyroptosis in neoadjuvant chemotherapy for breast cancer, comprising the following steps:

[0064] S100, obtained the original data set of breast cancer cases that received neoadjuvant chemotherapy from a public cancer database, was preprocessed, and the annotated full gene expression matrix and survival clinical information matrix were obtained through information extraction and annotation.

[0065] In one possible implementation, the original data set includes: a full gene expression matrix, labels of tumor samples or normal samples, survival time and survival status, breast cancer classification information, basic information, and rating information of multiple tumor staging systems.

[0066] In a possible implementation, the preprocessing includes: annotating the whole gene expression matrix file using the probe sequence annotation file to obtain an annotated whole gene expression matrix.

[0067] S200, based on the annotated whole gene expression matrix and survival clinical information matrix, screens out gene databases with significant differential expression on patient survival.

[0068] In one possible implementation, S200 includes:

[0069] The R language was called to perform a survival test on the annotated full gene expression matrix using the survival clinical information matrix, and a significance threshold was set to screen out the gene database with significant differential expression on patient survival.

[0070] S300, obtain the cell pyroptosis-related gene database.

[0071] In one possible implementation, S300 includes: performing literature search by keywords, and looking up genes related to the pyroptosis pathway, considering multiple possible pyroptosis-related pathways, and constructing a pyroptosis-related gene database.

[0072] S400, taking the intersection of the pyroptosis-related gene database and the gene database with significant differential expression on patient survival, to obtain a list of differentially expressed genes related to pyroptosis survival.

[0073] In one possible implementation method, based on the R language "dplyr" package and related technologies, the intersection of the pyroptosis-related gene database and the gene database with significantly differential expression for patient survival is taken to screen for genes with significant differences in pyroptosis-related survival.

[0074] S500, batch constructs disease survival prediction models for the differentially expressed gene list through LASSO regression, calculates the risk score of each model in the annotated whole gene expression matrix, and divides the disease survival prediction models into high and low groups; the high and low groups into which the risk scores of the disease survival prediction models are divided are regarded as the model's predictive indicators for patient prognosis, which are used for subsequent model screening.

[0075] In a possible implementation, S500 includes the following sub-steps:

[0076] S510, calling R language to construct a disease survival prediction model for the differentially expressed gene list through LASSO regression;

[0077] S520, using loop statements to adjust the random seed of the regression function and the gene groups involved in the construction of the prediction model one by one, and batch build the disease survival prediction model.

[0078] Furthermore, in S510, based on R language packages such as "glmnet", "survival", "ggplot2", "survminer", and "timeROC" and related technologies, a patient disease survival prediction model is constructed in batches through the LASSO regression algorithm.

[0079] Specifically, in S510, the LASSO regression algorithm outputs different regression results, namely, sub-models including different coefficients and gene population subsets, through different random seeds and input gene populations.

[0080] Further, in S520, the loop statement is: by changing the random seed and the input gene population subset one by one, that is, eliminating the genes involved in building the model one by one, the input and output results of the LASSO regression algorithm are changed.

[0081] S600, by statistically analyzing whether the survival difference between the high and low groups is significant, and combining the evaluation of the model's predictive ability, two optimal disease survival prediction models are selected as sub-models, and the prognostic model is obtained by weighting.

[0082] In one possible implementation, the method of selecting two optimal disease survival prediction models in S600 includes:

[0083] Eliminate disease survival prediction models that have no significant difference in survival prediction between high and low groups in the experimental set and the validation set, that is, P>0.05;

[0084] All disease survival prediction models with an area under the receiver operating characteristic curve of 1-5 year survival prediction rate < 0.5 were eliminated;

[0085] The disease survival prediction model has an AUC area in the validation set that is at least 50% higher than that of similar prediction models in the field of breast cancer, that is, the disease survival prediction model has a prediction ability that is at least 50% higher than that of similar prediction models in the field of breast cancer;

[0086] In one possible implementation, the optimal disease survival prediction model in S600 meets the following conditions:

[0087] 1. The disease survival prediction model has a very significant difference in the survival of patients in the experimental data set and all validation data sets, that is, P<0.03;

[0088] 2. The area under the receiver operating curve (AUC) of the disease survival prediction model for the 1-5-year prediction rate of patients in the experimental set and all validation data sets was > 0.5;

[0089] 3. The disease survival prediction model has an AUC area in the validation set that is 50% higher than that of the same type of prediction models in the field of breast cancer, that is, the disease survival prediction model has a prediction ability that is at least 50% higher than that of the same type of prediction models in the field of breast cancer;

[0090] 4. The selected sub-models should contain as many different gene parameters as possible to ensure the parameter volume and pathway volume of the overall model.

[0091] In a possible implementation, in S600, the method of obtaining the prognosis model by weighting includes:

[0092] Through MATLAB, the sub-models are weighted and merged, and the multi-dimensional predictive ability of the overall model is visualized (that is, the predictive ability between different data sets). From the large number of predictive models mentioned above, several models with the highest predictive ability are selected as sub-models. The sub-models are adjusted with different weights. After determining the optimal sub-model weights, a linear multi-parameter overall model is constructed with predictive ability better than each sub-model.

[0093] Furthermore, the method of adjusting the parameters of the sub-models through different weights includes: in the process of weighted merging of two sub-models, the two sub-models are added with weights ranging from 0% to 100% respectively, and the prediction ability of the final model obtained by the addition is evaluated through the ROC curve in the experimental set and the verification set, the weights of the sub-models are adjusted step by step and the prediction ability is evaluated, and the weight point with the highest average AUC value in the experimental set and the verification set is selected as the optimal weight.

[0094] In a possible implementation, the expression of the prognostic model is as follows:

[0095] Riskscore=0.422338*BRCA1+0.227932*VEGFA+0.116192*IL6+0.827851*MYD88+(- 0.131573)*NEK9+(-0.435585)*TMMEM9B+(-0.012124)*BCL2+(-0.441609)*CASP6+( -0.177918)*PYCARD+0.325638*NOD1+0.123377*NOD2+0.041331*MALT1+(-0.296448 )*NAIP+(-0.030816)*CHMP4A+(-0.016860)*PER2+0.030703*BID+0.037198*FXYD5;

[0096] Among them, Riskscore is the risk score, and BRCA1, VEGFA, IL6, MYD88, NEK9, TMMEM9B, BCL2, CASP6, PYCARD, NOD1, NOD2, MALT1, NAIP, CHMP4A, PER2, BID, and FXYD5 are the expression values ​​of gene fragments.

[0097] It should be noted that BRCA1 is the breast cancer 1 gene, VEGFA is the vascular endothelial growth factor gene, IL6 is the interleukin 6 gene, MYD88 is the myeloid differentiation factor 88 gene, NEK9 is the NIMA-associated kinase 9 gene, TMMEM9B is the transmembrane protein 9β gene, BCL2 is the B lymphocytoma-2 gene, CASP6 is the caspase 6 gene, PYCARD is the apoptosis-associated spot-like protein gene, NOD1 is the NOD-like receptor 1 gene, NOD2 is the NOD-like receptor 2 gene, MALT1 is the human mucosa-associated lymphoid tissue protein 1 gene, NAIP is the neuronal apoptosis inhibitory protein gene, CHMP4A is the recombinant human chromatin modifying protein 4α gene, PER2 is the menstrual circadian rhythm regulator 2 gene, BID is the BH3-interacting domain death agonist gene, and FXYD5 is the FXYD domain-containing ion transport regulator 5 gene.

[0098] The following is further described in conjunction with specific embodiments.

[0099] A method for constructing a prognostic model related to pyroptosis in neoadjuvant chemotherapy for breast cancer, comprising the following steps:

[0100] 1. Select "GEO Datasets" on the NCBI website (https: / / www.ncbi.nlm.nih.gov / ) to enter the GEO database, use "Breast cancer neoadjuvant chemotherapy" as the keyword to traverse all data sets, and select qualified data sets according to the following criteria to download raw data: ① All breast cancer patients included in the data set received neoadjuvant chemotherapy. ② The data set contains the full gene expression matrix of all cases included in the study. ③ All cases have labels for tumor samples or normal samples. ④ All cases have survival time and survival status. ⑤ All cases have breast cancer classification information ⑥ All cases have basic information such as age, and multiple tumor staging system rating information, such as TNM staging and AJCC staging.

[0101] Then, the probe sequence annotation file was downloaded from the experimental platform corresponding to the data set, and the gene expression matrix file in the original data was annotated. On the TCGA website (https: / / portal.gdc.cancer.gov / ), the keywords "Breastcancer", "TCGA-BRCA", and "Gene Expression Quantification" were selected to download all case information. File processing was performed in R language to obtain the annotated TCGA patient group full gene expression matrix and survival clinical information matrix.

[0102] 2. Call the R language "Survival" package, use the survival clinical information matrix to perform a survival test on the annotated TCGA patient group full gene expression matrix, set the significance threshold to adj.P.value = 0.05, and screen and derive genes with significant differential expression on patient survival.

[0103] 3. Use the keywords "Breast cancer", "neoadjuvant chemotherapy", "pyroptosis" and "prognosticmodel" to search the PubMed literature database (https: / / pubmed.ncbi.nlm.nih.gov / ), collect the list of pyroptosis-related genes article by article, and then collect pyroptosis genes from the pathway level in the Genecards database (https: / / www.genecards.org / ). Collect and remove duplicate genes one by one to obtain a database of pyroptosis-related genes.

[0104] 4. Call the R language "dplyr" package to take the intersection of the cell pyroptosis-related gene database and the genes with significant differential expression for patient survival, and obtain a list of differentially expressed genes related to cell pyroptosis survival.

[0105] like Figure 1 As shown, there are 3420 genes with significant survival differences, and the collected cell pyroptosis gene library has a total of 111 genes, and there are 45 pyroptosis genes with significant survival differences formed by the intersection.

[0106] 5. Call the R language "glmnet", "survival", "ggplot2", "survminer", "timeROC", "surMisc", and "tidyverse" packages to build a model for the above gene list through LASSO regression. Use loop statements to adjust the random seed of the regression function and the gene groups involved in the model construction one by one, and build a large number of disease survival prediction models (about 6,000, the usual research method only builds less than 50). The model accumulates and sums the selected gene expression values ​​by the LASSO regression coefficient to obtain the risk score, and divides the patients into high-risk groups and low-risk groups according to the median risk score of all cases. Test whether there is a survival difference between the high and low groups as one of the screening conditions for the model. The two best models are selected as sub-models for the subsequent construction of the total model. The two sub-models are evaluated for predictive ability and authenticity through ROC curves and KM curves in the GEO experimental set, GEO validation set, and TCGA validation set (including the full gene expression matrix and survival information matrix of the three).

[0107] like Figure 2 , Figure 3 and Figure 4 , Figure 5 As shown in the figure, the KM curve and ROC curve of the two sub-models constructed by LASSO regression show that there are significant survival differences between the high and low groups of the two sub-models, and the area under the ROC curve of the sub-model is greater than 0.6 in both data sets and greater than 0.5 in the validation set.

[0108] Two sub-models were constructed, the first sub-model contained 9 gene expression value parameters, and the second sub-model contained 13 gene expression value parameters.

[0109] The expression of the first submodel is:

[0110] Riskscore1=0.24039371*VEGFA+(-0.55361342)*TMEM9B+(-0.03694425)*BCL2+(-0.56600686)*CASP6+ (-0.23467788)*CHMP4A+(-0.12839202)*PER2+0.23381691*BID+0.28327413*FXYD5+0.31475154*MALT1;

[0111] The relevant data of this model are Figure 2 , 3 shown.

[0112] The expression of the second submodel is:

[0113] Riskscore2=0.486179905*BRCA1+0.226049174*VEGFA+0.133755440*IL6+0.952991735*MYD88+(-0.151461768)*NEK9+(-0.417743396)*TMEM9B+(-0.0 08372395)*BCL2+(-0.422805359)*CASP6+(-0.204812693)*PYCARD+0.3748 62775*NOD1+0.142027053*NOD2+0.415963855*MALT1+(-0.341260296)*NAIP

[0114] The relevant data of the model are Figure 4 , 5 shown.

[0115] Among them, BRCA1 is the breast cancer 1 gene, VEGFA is the vascular endothelial growth factor gene, IL6 is the interleukin 6 gene, MYD88 is the myeloid differentiation factor 88 gene, NEK9 is the NIMA-associated kinase 9 gene, TMMEM9B is the transmembrane protein 9β gene, BCL2 is the B lymphocytoma-2 gene, CASP6 is the caspase 6 gene, PYCARD is the apoptosis-associated speck-like protein gene, NOD1 is the NOD-like receptor 1 gene, NOD2 is the NOD-like receptor 2 gene, MALT1 is the human mucosa-associated lymphoid tissue protein 1 gene, NAIP is the neuronal apoptosis inhibitory protein gene, CHMP4A is the recombinant human chromatin modifying protein 4αA gene, PER2 is the biological clock menstrual period circadian rhythm regulator 2 gene, BID is the BH3-interacting domain death agonist gene, and FXYD5 is the FXYD domain-containing ion transport regulator 5 gene.

[0116] 6. If Figure 6 As shown in the figure, the total model formed by adding different weights of two sub-models is visualized in the MATLAB software in a three-dimensional coordinate system constructed by the prediction accuracy of one experimental set and two validation sets, and the weight point with the highest average AUC value in the experimental set and the validation set is selected as the optimal weight. The optimal weight point is selected as the weight coefficient for merging the sub-models.

[0117] The best weight coefficient finally obtained is 0.1313. The final model score is:

[0118] Riskscore=0.422338*BRCA1+0.227932*VEGFA+0.116192*IL6+0.827851*MYD88+(- 0.131573)*NEK9+(-0.435585)*TMMEM9B+(-0.012124)*BCL2+(-0.441609)*CASP6+( -0.177918)*PYCARD+0.325638*NOD1+0.123377*NOD2+0.041331*MALT1+(-0.29644 8)*NAIP+(-0.030816)*CHMP4A+(-0.016860)*PER2+0.030703*BID+0.037198*FXYD5

[0119] At this point, the prognostic prediction model for breast cancer patients who received neoadjuvant chemotherapy has been constructed.

[0120] Figure 7 , Figure 8 They are the KM curve and ROC curve of the final model respectively; the results show that there is a significant survival difference between the high and low groups of the final integrated model, and according to the area under the ROC curve, the model has excellent predictive ability within 5 years of the experimental set.

[0121] The following is a description of the device for constructing a prognostic model related to pyroptosis of breast cancer undergoing neoadjuvant chemotherapy provided by the present invention. The device for constructing a prognostic model related to pyroptosis of breast cancer undergoing neoadjuvant chemotherapy described below and the device for constructing a prognostic model related to pyroptosis of breast cancer undergoing neoadjuvant chemotherapy described above can be referenced to each other.

[0122] The device for constructing a prognostic model related to pyroptosis in neoadjuvant chemotherapy for breast cancer includes:

[0123] The first acquisition module is used to obtain the original data set of breast cancer cases that received neoadjuvant chemotherapy, perform preprocessing, and obtain an annotated full gene expression matrix and a survival clinical information matrix;

[0124] The screening module is used to screen out gene databases with significant differential expression on patient survival based on the annotated full gene expression matrix and survival clinical information matrix;

[0125] The second acquisition module obtains the cell pyroptosis-related gene database;

[0126] The intersection calculation module is used to obtain the intersection of the cell pyroptosis-related gene database and the gene database with significant differential expression for patient survival, and obtain a list of differentially expressed genes related to cell pyroptosis survival;

[0127] The disease survival prediction model batch construction module is used to batch construct disease survival prediction models for the differentially expressed gene list through LASSO regression, calculate the risk score of each model in the annotated full gene expression matrix, and divide the disease survival prediction models into high and low groups; the high and low groups into which the risk scores of the disease survival prediction models are divided are regarded as the model's predictive indicators for patient prognosis, which are used for subsequent model screening;

[0128] The prognostic model building module is used to statistically determine whether the survival difference between the high and low groups is significant, and combined with the evaluation of the model's predictive ability, to screen out two optimal disease survival prediction models as sub-models, and obtain the prognostic model through weighting.

[0129] Based on the aforementioned method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer, the present invention provides a method for predicting the sensitivity of chemotherapy drugs and potential drug targets, comprising the following steps:

[0130] 1. Download all cell line gene expression matrices, drug sensitivity (IC50) matrices, and drug annotation files from the GDSC database (https: / / www.cancerrxgene.org / ). Organize the files using Perl language-related techniques to obtain files that can be imported into R language for processing.

[0131] 2. By searching for drugs in the DrugBank database (https: / / go.drugbank.com / ), we obtained all possible drugs commonly used in neoadjuvant chemotherapy for breast cancer patients and constructed a chemotherapy drug database.

[0132] 3. Through the R language "dplyr" package, all cell lines were screened to retain only breast cancer cell lines, the gene expression matrix only retained the expression matrix of genes contained in the prediction model, and the drug sensitivity only retained the drug objects in the above-mentioned chemotherapy drug database.

[0133] 4. Using the R language "ggstatsplot" package, the model risk score and the genes in the model were used to test the correlation of the chemotherapy drug set, with a significance threshold of P = 0.05. Fig. 9 , Fig.10 and Fig.11 shown.

[0134] The test requirements are as follows: first, perform normality tests on the model score / gene level and drug sensitivity respectively. If both of the correlation tests conform to the normal distribution, perform a t-test and calculate the Pearson coefficient, otherwise calculate the Spearman coefficient.

[0135] 5. In clinical applications, whole genome sequencing is performed on pathological samples of breast cancer patients who need neoadjuvant chemotherapy to obtain the average expression value of the model gene, and the patient's prognostic risk score is calculated according to the model formula. If the score is higher than the average, the prognosis is judged to be poor, otherwise the prognosis is good, and neoadjuvant chemotherapy is recommended at the algorithm level. On the other hand, through the prognostic risk score and the relative expression value level of the genes in the model, it is predicted that the patient may be more sensitive to certain drugs. Combined with the patient's tumor size, other molecular marker levels and objective clinical conditions, an individualized medication combination plan is selected.

[0136] It was predicted that at the gene level, docetaxel sensitivity was positively correlated with the expression levels of BCL2 and TMEM9B, and negatively correlated with the expression level of VEGFA. Paclitaxel was positively correlated with the expression level of BCL2, and negatively correlated with the expression levels of BID and FXYD5. Epirubicin was positively correlated with the expression level of CASP6. Gemcitabine was negatively correlated with the expression level of MALT1. Cisplatin was negatively correlated with the expression level of NOD1. Lapatinib was negatively correlated with the expression level of NOD1. Methotrexate was negatively correlated with the expression level of NOD2. Fulvestrant was negatively correlated with the expression levels of PER2 and TMEM9B. Tamoxifen was negatively correlated with the expression level of VEGFA. At the model level, fulvestrant was more sensitive in the high-Riskscore group, while paclitaxel was more sensitive in the low-Riskscore group.

[0137] The above description is only a preferred specific implementation manner of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention should be included in the scope of protection of the invention.

Claims

1. A method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer, characterized in that: The following steps are involved: The original data set of breast cancer cases that received neoadjuvant chemotherapy was obtained, preprocessed, and the annotated full gene expression matrix and survival clinical information matrix were obtained through information collation and manual annotation; Based on the annotated full gene expression matrix and survival clinical information matrix, we screened out the gene database with significant differential expression on patient survival; Obtain the cell pyroptosis-related gene database; Take the intersection of the pyroptosis-related gene database and the gene database with significant differential expression on patient survival to obtain a list of differentially expressed genes related to pyroptosis survival; The disease survival prediction model was constructed in batches for the differentially expressed gene list through LASSO regression, and the risk score of each model was calculated in the annotated whole gene expression matrix, and the disease survival prediction model was divided into high and low groups; the high and low groups into which the risk scores of the disease survival prediction model were divided were regarded as the model's predictive indicators for the patient's prognosis, which were used for subsequent model screening; By statistically analyzing whether the survival difference between the high and low groups is significant, and combining the evaluation of the model's predictive ability, two optimal disease survival prediction models were screened out as sub-models, and the prognostic model was obtained by weighting.

2. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 1, characterized in that: The original data set is obtained through a public cancer database, including: a full gene expression matrix, labels of tumor samples or normal samples, survival time and survival status, breast cancer classification information, basic information, and rating information of various tumor staging systems.

3. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 2, characterized in that: The preprocessing includes: annotating the whole gene expression matrix file by using the probe sequence annotation file to obtain an annotated whole gene expression matrix.

4. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 1, characterized in that: The method for screening out a gene database with significant differential expression on patient survival based on an annotated whole gene expression matrix and a survival clinical information matrix comprises: calling R language, performing a survival test on the annotated whole gene expression matrix using the survival clinical information matrix, setting a significance threshold, and screening out a gene database with significant differential expression on patient survival.

5. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 1, characterized in that: The prediction models are batch constructed for the differentially expressed gene list through LASSO regression, the risk score of each model is calculated in the annotated full gene expression matrix, and the prediction models are divided into high and low groups. The steps are as follows: Call R language to construct a disease survival prediction model for the differentially expressed gene list through LASSO regression; Use loop statements to adjust the random seed of the regression function and the gene groups involved in the construction of the prediction model one by one, and batch build the disease survival prediction model; The risk scores of the genes selected in the disease survival prediction model were calculated, and the patients were divided into high-risk group and low-risk group according to the median risk score to test whether there was a survival difference between the high-risk group and the low-risk group.

6. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 1, characterized in that: The optimal disease survival prediction model meets the following conditions:

1. The disease survival model has a very significant difference in the survival of patients in the experimental data set and all validation data sets, that is, P < 0.03; 2. The area under the receiver operating curve (AUC) of the disease survival model for the 1-5-year predicted survival rate of patients in the experimental set and all validation data sets was > 0.5; 3. The disease survival model has an AUC area in the validation set that is more than 50% higher than that of similar prediction models in the field of breast cancer; 4. The selected sub-models should contain as many different gene parameters as possible to ensure the parameter volume and pathway volume of the overall model.

7. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 6, characterized in that: The screening method of the optimal disease survival prediction model comprises: Eliminate disease survival models that have no significant difference in survival prediction between the experimental set and the validation set divided into high and low groups; Disease survival models with an area under the receiver operating characteristic curve of <0.5 for the 1-5 year survival prediction rate of all patients were eliminated; The disease survival model has an AUC area that is at least 50% higher than that of similar prediction models in the breast cancer field in the validation set.

8. The method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer according to claim 1, characterized in that: The expression of the prognostic model is as follows: Riskscore=0.422338*BRCA1+0.227932*VEGFA+0.116192*IL6+0.827851*MYD88+(- 0.131573)*NEK9+(-0.435585)*TMMEM9B+(-0.012124)*BCL2+(-0.441609)*CASP6+( -0.177918)*PYCARD+0.325638*NOD1+0.123377*NOD2+0.041331*MALT1+(-0.296448 )*NAIP+(-0.030816)*CHMP4A+(-0.016860)*PER2+0.030703*BID+0.037198*FXYD5; Among them, Riskscore is the risk score, and BRCA1, VEGFA, IL6, MYD, NEK9, TMMEM9B, BCL2, CASP6, PYCARD, NOD1, NOD2, MALT1, NAIP, CHMP4A, PER2, BID, and FXYD5 are the expression values ​​of gene fragments.

9. A device for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer, characterized in that: include: The first acquisition module is used to obtain the original data set of breast cancer cases that received neoadjuvant chemotherapy, perform preprocessing, and obtain an annotated full gene expression matrix and a survival clinical information matrix; The screening module is used to screen out gene databases with significant differential expression on patient survival based on the annotated full gene expression matrix and survival clinical information matrix; The second acquisition module obtains the cell pyroptosis-related gene database; The intersection calculation module is used to obtain the intersection of the cell pyroptosis-related gene database and the gene database with significant differential expression for patient survival, and obtain a list of differentially expressed genes related to cell pyroptosis survival; The disease survival prediction model batch construction module is used to batch construct disease survival prediction models for the differentially expressed gene list through LASSO regression, calculate the risk score of each model in the annotated full gene expression matrix, and divide the disease survival prediction models into high and low groups; the high and low groups into which the risk scores of the disease survival prediction models are divided are regarded as the model's predictive indicators for patient prognosis, which are used for subsequent model screening; The prognostic model building module is used to statistically determine whether the survival difference between the high and low groups is significant, and combined with the evaluation of the model's predictive ability, to screen out two optimal disease survival prediction models as sub-models, and obtain the prognostic model through weighting.

10. A method for predicting chemotherapy drug sensitivity and potential drug targets using the method for constructing a prognostic model related to neoadjuvant chemotherapy for breast cancer as described in any one of claims 1 to 8, characterized in that: The following steps are involved: Obtain drug sensitivity data matrix and whole genome expression profile of all cell lines; The pyroptosis gene set corresponding to breast cancer cell lines, breast cancer neoadjuvant chemotherapy drugs, and disease survival prediction models was screened to obtain a basic data set; Based on the basic data set, the risk scores of breast cancer cell lines were calculated according to the disease survival prognosis prediction model and divided into high and low groups; The gene set and risk score of the disease survival prediction model were used to test the correlation of each breast cancer chemotherapy drug, and a set of chemotherapy drugs closely related to gene expression differences was obtained; Based on the whole genome sequencing results of a patient's pathological sample in clinical practice, the expression level and risk score level of the patient's chemotherapy drug-related genes are evaluated. Combined with the chemotherapy drug set that is closely related to gene expression differences, several chemotherapy drugs to which the patient is more sensitive are predicted.

Images