WO2018001295A1 - Molecular marker, reference gene, and application and test kit thereof, and method for constructing testing model - Google Patents

Abstract

Provided are a molecular marker, reference gene, and application and test kit thereof, and a method for constructing a testing model. Using follow-up information for a comparison method, the accuracy of the provided test kit in predicting the risk, in patients initially diagnosed as ER or PR positive for breast cancer, of recurrence or death 3-10 years after surgery is 70%, and accuracy in predicting a low-risk group and a high-risk group is 81.1% and 54.4%, respectively. The corresponding accuracies in predicting the FFPE pathology test results are 71.9% and 56.8%, respectively. The risk prediction model supporting the test kit requires only the Ct value of the molecular marker, the patient's age, the pT stage, and the LN quantity, and does not need to rely on other clinical pathology information; the model provides cancer prognosis assessment that is better than the pathology prediction result alone, and reduces to a certain extent the occurence of improper treatment caused by erroneous pathology prediction, thus further improving the technical method for cancer prognosis assessment.

Description

Molecular marker, internal reference gene and application thereof, detection kit and construction method of detection model

This application claims to be submitted to the Chinese Patent Office on June 30, 2016, the application number is 201610509983.0, and the Chinese patent application titled “Molecular Markers, Internal Reference Genes and Their Applications, Detection Kits, and Detection Model Construction Methods” is preferred. The entire contents are hereby incorporated by reference.

Technical field

The invention relates to the field of biotechnology, in particular to a molecular marker, an internal reference gene and an application thereof, a detection kit and a construction method of the detection model.

Background technique

Breast cancer is one of the main reasons that threaten the health of women around the world. In 2013, the American Cancer Society released the nation's cancer statistics showing that the incidence of breast cancer ranked first among women's cancer, and the mortality rate ranked second. According to the latest National Cancer Center data, in 2013, 232,340 new breast cancers and 39,620 deaths were reported among women in the United States. In the United States, one in every eight women has breast cancer. Although China is a country with low incidence of breast cancer, its incidence and mortality have increased significantly in recent years. About 15% of the 1.3 million newly diagnosed breast cancer patients diagnosed every year in the world are from China. According to the statistics of China Breast Cancer Network, the annual increase of breast cancer in China is 3%-4%, which is more than 1%-2% of the world level. The incidence rate is the first in women. Urgent development of breast cancer prevention, diagnosis, prognosis, and individualized treatment technologies.

Breast cancer is a kind of highly heterogeneous tumor with many prognostic factors. Breast cancer patients with the same clinical stage, histological grade and hormone receptor expression can receive the same treatment plan, and their prognosis may be different. How to accurately determine the prognosis of breast cancer patients and formulate corresponding individualized treatment programs to avoid the harm and burden caused by over-treatment and improper treatment is an urgent problem to be solved in clinical practice.

With the rapid development of molecular biology technology, it is possible to detect and detect prognosis-related genes in breast cancer using molecular biology methods such as polymerase chain reaction (PCR), probe hybridization and gene chip. In 2002, Van't Veer and others screened 117 cases of breast cancer by DNA chip technology and found 70 genes related to breast cancer prognosis. In 2004, American scientists used RT-PCR to verify 675 breast cancer samples. 21 prognostic-related genes, Genomic Health developed breast cancer prognosis-related products based on this research

It is currently the only breast cancer prognostic test product recommended by the NCCN guidelines, ASCO clinical guidelines and StGallen's clinical consensus. In addition, Yasuto Naoi et al. used DNA chip technology to study cancer tissue samples from patients with ER-positive and node-negative breast cancer in the Japanese population, and found 95 prognostic-related genes. The Torsten O. Nielsen team found that 50 gene combinations provide more breast cancer prognostic predictions than clinical factors and immunohistochemical staining, and use FFPE samples instead of fresh or frozen samples for testing. Detect the sample range. Breast cancer prognosis related products from 2002 to 2013

Mammaprint, ProsIgna ^TM and MapQuant Dx ^TM have successively obtained FDA and CE certification. However, these products are currently based on research and development in Europe and the United States. These products are not only expensive after entering China, but whether the genes and their detection models are applicable to the Chinese population has yet to be verified. Therefore, it is of great significance to develop a cost-effective prognostic detection technology for Chinese breast cancer.

Summary of the invention

In view of this, the present invention provides molecular markers and their applications, detection kits, and methods for constructing detection models. The kit is superior to the clinical pathological evaluation results in the prognosis evaluation performance of breast cancer, which can reduce the over-treatment and improper treatment caused by pathological diagnosis errors to meet the needs of individualized and precise treatment of breast cancer patients. Further improved the technical methods for predicting the prognosis of breast cancer in China.

In order to achieve the above object, the present invention provides the following technical solutions:

The invention provides genetic compositions, including the molecular markers MAPT and/or MS4A1.

The present invention provides a genetic composition consisting of the molecular markers BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT and MS4A1.

In some embodiments of the invention, the genetic composition further comprises the internal reference genes ACTB, GAPDH, GUSB, NUP214, VCAN.

The invention also provides the use of the genetic composition for the detection of a 3-10 year postoperative recurrence and/or mortality risk prediction for breast cancer.

In some specific embodiments of the present invention, the prognosis of breast cancer prognosis in the application is 3-10 years recurrence and/or death risk assessment is specifically: obtaining total RNA of the sample to be tested, obtaining cDNA by reverse transcription, using real-time PCR The method obtains the Ct value of the molecular marker and the reference gene, averages the Ct value of the reference gene, obtains an average Ct value (Ct') of the internal reference gene combination, and then Ct the molecular marker The values were subtracted from the internal reference gene combination Ct' value to normalize, and △Ct was obtained. The △Ct value and the subject's age, pT value and LN value were reconstructed by random forest algorithm. The annual recurrence or death risk prediction model was analyzed and the results were obtained. Among them, the pT value is the pathological stage, and the LN value is the number of lymph node metastasis. The value obtained by the analysis is compared with a threshold value, and the result is obtained, the threshold value being 5. The value obtained by the analysis ≥ 5 is a good prognosis, and the value obtained by the analysis < 5 is a poor prognosis.

In some embodiments of the present invention, the method for constructing a 3-10 year recurrence or death risk assessment test model for breast cancer in the application is: constructing a ΔCt value of the molecular marker of the sample to be tested and the age of the subject , pT value, LN value construct a mathematical matrix, randomly select 1/2 as the training set, 1/2 as the verification set, through the random forest The algorithm establishes a prediction model with 10,000 decision trees. The total random sampling is ≥1000 times, and ≥1000 prediction models are established. From the ≥1000 prediction model, ≥39 preferred models with the highest coincidence rate with follow-up information are selected as the final model. The model was used with a median of ≥39 submodels as the final prognostic risk predictor.

The random forest is composed of many decision trees. The decision tree is constructed by a method of double randomness of attributes and samples, so it is also called random decision tree. In a random forest, there is no correlation between decision trees. When the test data enters the random forest, it is classified by each decision tree. Finally, the class with the most classification results in all decision trees is the final result, that is, the result of the decision tree "voting", in other words, the random forest is a inclusion. A classifier for multiple decision trees, and the category of its output is determined by the mode of the category of the individual tree output. In the present invention, we have optimized on the basis of the traditional random forest algorithm. We randomly sample the samples 1000 times, establish 1000 models, and select 39 models with higher accuracy and specificity from the 1000 model. The submodel of the model's most final model is preferred, and the median of the 39 submodels is used as the final prediction.

Age, PT stage, number of LN metastases, and 14 molecular markers BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11 of untreated early- and mid-term ER or PR-positive breast cancer , CD68, BAG1, MAPT, MS4A1 and 5 housekeeping genes ACTB, GAPDH, GUSB, NUP214, VCAN Ct values, input into 39 prediction models for analysis, obtain predictive analysis results, get 3-10 years of recurrence or death risk Values, and based on a risk threshold (threshold of 5), predict a good prognosis or a poor prognosis.

In some embodiments of the invention, the test sample in the present invention is an FFPE sample of a newly diagnosed early or mid-term ER or PR positive breast cancer patient.

The present invention also provides a primer set for amplifying the gene composition, the sequence of which is shown in SEQ ID No. 1 to SEQ ID No. 28.

The present invention also provides a probe set for amplifying the gene composition, the sequence of which is shown in SEQ ID No. 29 to SEQ ID No. 42.

The present invention also provides a primer set for amplifying an internal reference gene of the gene composition, as shown in SEQ ID No. 43 to SEQ ID No. 47.

The present invention also provides a probe set for amplifying an internal reference gene of the gene composition, as shown in SEQ ID No. 48 to SEQ ID No. 52.

The invention also provides a test kit for predicting the risk of recurrence and/or mortality of breast cancer after 3-10 years, including the primer set and/or the probe set and reagents commonly used in the kit.

The invention also provides a method for constructing a risk assessment model for recurrence or death of breast cancer with a prognosis of 3-10 years, constructing a mathematical matrix of the ΔCt value of the molecular marker of the sample to be tested and the age, pT value and LN value of the subject, randomized Select 1/2 as the training set and 1/2 as the verification set. The prediction model with 10000 decision trees is established by the random forest algorithm. The total random sampling is ≥1000 times, and ≥1000 prediction models are established. The ≥1000 prediction models are selected. The ≥39 preferred models with the highest rate of coincidence with follow-up information were used as sub-models of the final model, and the median of ≥39 sub-models was used as the final prognostic risk predictor.

The invention also provides an evaluation method for the risk of recurrence or death of breast cancer with a prognosis of 3-10 years, obtaining total RNA of the sample to be tested, obtaining cDNA by reverse transcription, and obtaining the molecular marker and the reference gene by fluorescence quantitative PCR. Ct value, the Ct value of the internal reference gene is averaged to obtain the average Ct value (Ct') of the internal reference gene combination, and then the Ct value of the molecular marker is respectively subtracted from the internal reference gene combination Ct' value to be normalized. △Ct, the △Ct value and the age, pT value and LN value of the subject were analyzed by the random forest algorithm for 3-10 years postoperative recurrence or death risk prediction model of breast cancer, and the result was obtained, that is, 3-10 The annual recurrence or death risk value is predicted to be a good prognosis or a poor prognosis based on the risk threshold (risk threshold of 5).

The sample to be tested in the present invention is an FFPE sample of a newly diagnosed early or mid-stage ER or PR positive breast cancer patient.

The technical solution to solve the problem of the present invention includes: (1) selecting 192 breast cancer related candidate genes (not limited to breast cancer prognosis related, including internal reference genes), and customizing TLDA gene expression detection chip (Applied Biosystems) through literature and database research. (2) systematically collect complete demographic data, clinical data and follow-up data (recurrence and metastasis time, survival time), and select untreated early and mid-term ER or PR positive breast cancer FFPE samples for newly diagnosed patients, using customized TLDA chip was used to detect 192 genes, and the molecular markers related to prognosis and breast cancer prognosis were screened. (3) The candidate molecular markers and reference genes were screened in independent samples and constructed by random forest algorithm. A predictive model of the risk of recurrence or death 3-10 years after surgery, and a consensus rate between the predictive model and the follow-up results; (4) further validation using independent clinical samples: 19 cases of known clinical follow-up data for ER or PR positive The FFPE samples of newly diagnosed patients with metaphase breast cancer were evaluated for the coincidence rate between the test results and the follow-up results.

The present invention provides a prognostic evaluation gene detection system for recurrence or death 3 to 10 years after surgery in a newly diagnosed early or mid-stage ER or PR positive breast cancer patient. By extracting total RNA from formalin-Fixed and Parrffin-Embedded (FFPE) breast cancer tissue samples, reverse transcription by universal primers, PCR detection of breast cancer prognosis Ct values were expressed for 14 molecular markers and 5 internal reference genes. A predictive model of the risk of recurrence or death after 3-10 years of postoperative ER or PR-positive early-stage breast cancer patients with Ct values and subject age, pT value and LN number was determined by prognosis or poor prognosis. Compared with the follow-up information, the system achieved an accuracy of 70%. Except for patient age, pT stage, and LN number, there is no need to rely on other clinical pathological information.

The kit provided by the invention has a prediction accuracy of 81.1% for a newly diagnosed breast cancer patient with a low risk of recurrence or death of 3-10 years, and a pathological prediction accuracy of 71.9%, which is a risk of recurrence or death of 3-10 years. The accuracy of the prediction accuracy of the newly diagnosed patients with high breast cancer was 54.4%, which was close to the accuracy of the corresponding pathological prediction detection of 56.8%. Compared with clinical follow-up information, the kit has a concordance rate of 70%. Except for patient age, pT stage, and LN number, there is no need to rely on other clinical pathological information.

The detection system and the kit are superior to the clinical pathological prediction results in the prognosis evaluation performance of breast cancer, and can reduce the excessive treatment and improper treatment caused by pathological diagnosis errors to meet the individualized precision treatment of breast cancer patients. The need to further improve the technical methods of domestic breast cancer prognosis prediction.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the prior art description will be briefly described below.

Figure 1 shows the results of correlation analysis between the internal reference gene and the test gene;

Figure 2 shows the establishment of a risk assessment model for 3-10 years of recurrence or death in breast cancer.

detailed description

The invention discloses a molecular marker, an internal reference gene and an application thereof, a detection kit and a construction method of the detection model, and those skilled in the art can learn from the contents of the paper and appropriately improve the process parameters. It is to be understood that all such alternatives and modifications are obvious to those skilled in the art and are considered to be included in the present invention. The method and the application of the present invention have been described by the preferred embodiments, and it is obvious that the method and application described herein may be modified or appropriately modified and combined without departing from the scope of the present invention. The technique of the present invention is applied.

The technical solution to solve the problem of the present invention includes: (1) selecting 192 breast cancer related candidate genes (not limited to breast cancer prognosis related, including internal reference genes), and customizing TLDA gene expression detection chip (Applied Biosystems) through literature and database research. (2) systematically collect complete demographic data, clinical data and follow-up data (recurrence and metastasis time, survival time), and select untreated early and mid-term ER or PR positive breast cancer FFPE samples for newly diagnosed patients, using customized TLDA chip was used to detect 192 genes, and the molecular markers related to prognosis and breast cancer prognosis were screened. (3) The candidate molecular markers and reference genes were screened in independent samples and constructed by random forest algorithm. A predictive model of the risk of recurrence or death 3-10 years after surgery, and a consensus rate between the predictive model and the follow-up results; (4) further validation using independent clinical samples: 19 cases of known clinical follow-up data for ER or PR positive The FFPE samples of newly diagnosed patients with metaphase breast cancer were evaluated for the coincidence rate between the test results and the follow-up results.

1. Study sample selection

(1) Untreated early and mid-stage breast cancer patients;

(2) Untreated early and mid-term ER or (and) PR-positive breast cancer patients;

(3) LN has or (and) no transfer, and the number of LN transfers;

(4) Have accurate and detailed follow-up information;

A total of 339 samples meeting the criteria were used in this study.

2. FFPE sample total RNA extraction

Total RNA was extracted from FFPE samples using High Pure FFPET RNA Isolation Kit (Roche) at concentrations ranging from 25 ng to 400 ng/μL, OD260/280 in the range of 1.8-2.0, and OD260/230 between 1.5-2.0.

3. TLDA (Applied Biosystems) chip detection.

The following experiments were performed on 26 samples with good prognosis and poor prognosis using known clinical follow-up information.

(1) Total RNA is subjected to reverse transcription reaction to obtain a cDNA sample;

(2) The cDNA product is subjected to TLDA chip detection;

(3) Data analysis and processing to obtain candidate molecular markers and internal reference genes.

4. Real-time quantitative RT-PCR (qRT-PCR) method

Candidate molecular markers were validated using 289 samples of known clinical follow-up information.

(1) Total RNA is subjected to reverse transcription reaction to obtain a cDNA sample;

(2) The cDNA product is subjected to RT-PCR detection;

(3) Data analysis and processing.

5. Diagnostic kit preparation method

Through the custom TLDA chip detection method, the difference of internal reference gene and gene expression between 26 cases of breast cancer prognosis and 26 cases of breast cancer prognosis samples were determined, and the internal reference genes and differentially expressed genes were selected. Candidate molecular markers were verified by large-sample quantities by reverse transcription fluorescent quantitative PCR. The final screening of 14 genes and 5 internal reference genes for diagnosis of breast cancer prognosis (BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT, MS4A1 ;ACTB, GAPDH, GUSB, NUP214, VCAN). Diagnostic kits include primers for these genes, probes, and other conventional reagents for qRT-PCR. The kit further comprises a predictive model, wherein the expression levels of BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT and MS4A1 are ACTB, GAPDH, GUSB. The mean values of NUP214 and VCAN were detected as reference genes. At the same time, the clinical information of age, PT and LN of breast cancer patients were comprehensively evaluated for the prognosis of recurrence or death after 3-10 years, and the prognosis of poor prognosis was poor.

6. Establishment of risk assessment test model

(1) Establishment of a predictive model for risk of recurrence or death after 3-10 years of breast cancer surgery.

The random forest algorithm in the machine learning method was used to evaluate the risk of recurrence or death after 3-10 years of postoperative detection, and a gene detection model for breast cancer prognosis evaluation was established. The random forest is composed of many decision trees. The decision tree is constructed by a method of double randomness of attributes and samples, so it is also called random decision tree. In a random forest, there is no correlation between decision trees. When the test data enters the random forest, it is classified by each decision tree. Finally, the class with the most classification results in all decision trees is the final result, that is, the result of the decision tree "voting", in other words, the random forest is a inclusion. a classifier for multiple decision trees, and the category of its output is the category output by the individual tree The majority depends on the number. In the present invention, we have optimized on the basis of the traditional random forest algorithm. We randomly sample the samples 1000 times, establish 1000 models, and select the 39 best models with the highest accuracy from the 1000 model. The submodel is used and the median of the 39 submodels is used as the final prediction.

The following is a further description of the invention:

In the first phase of the study, 192 candidate genes for breast cancer were detected by TLDA detection technology. The gene expression differences of 26 breast cancer prognosis samples and 26 breast cancer prognosis samples were detected, and differentially expressed genes were screened. The different expression levels of the gene are expressed as 2 - ^ΔCt , wherein ^ΔCt = CT sample-CT reference, and the selected reference gene is normalized as a reference to calculate the relative expression amount. The screening process of the internal reference gene: using the genetic algorithm based on genorm, bestkeeper, normfinder, delta Ct and considering the biological function of the less fluctuating gene and its relationship with the tumor, screening candidate internal reference genes; calculating all candidate internal reference gene combinations Ct The correlation between the mean and the mean Ct of 192 genes, the most relevant combination is the internal reference genes including: ACTB, GAPDH, GUSB, NUP214, VCAN. Candidate gene screening criteria: (1) overall analysis - good prognosis and poor prognosis, the difference between the two groups is 2 times or less, and the proportion of cases with Ct < 35 is 50%; (2) stratified analysis - no lymph nodes The prognosis of the metastatic group was better than that of the poor prognosis. The difference between the two groups was more than 2 times, and the statistical difference was <0.05. (3) The difference between the two groups was not significant, but it was reported in the prognosis of breast cancer, and Ct< The proportion of cases in 35 reached 90%. Genes satisfying the above screening criteria include: BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT, MS4A1

Firstly, the present invention provides a prognostic evaluation gene for breast cancer in China: at present, foreign similar products are developed based on European and American populations, and different ethnic groups have different gene expressions. In the present invention, 19 genes are screened, wherein MAPT and MS4A1 are based on The genes related to the recurrence or death assessment of female breast cancer patients in China after 3-10 years postoperatively have been reported. Although this gene has been reported to be associated with breast cancer, no direct report related to the prognosis of breast cancer has been found. Secondly, the present invention establishes a new internal reference gene combination different from other inventions and products, and the gene combination is less affected by the RNA quality in the FFPE sample, so that the detection result of the molecular marker is more reliable. Third, the predictive model of the random forest algorithm was used for comprehensive analysis. The model predicts the risk of recurrence or death after 3-10 years of postoperative diagnosis of ER+ or PR+ breast cancer in the early and middle stages.

In summary, the present invention provides a prognostic evaluation gene detection system for relapse or death of patients with stage I and stage II ER or PR positive for untreated breast cancer who are 3-10 years postoperatively. Compared with the follow-up information, the system achieved an accuracy of 70%. Except for patient age, pT stage, and LN number, there is no need to rely on other clinical pathological information.

The materials and reagents used in the molecular markers, internal reference genes and their applications, detection kits, and detection methods for the detection models provided by the present invention are all commercially available.

The present invention is further illustrated below in conjunction with the embodiments:

Example 1 Collection of samples and preparation of sample data

Using the FFPE samples of untreated newly diagnosed breast cancer patients, the system collected complete clinical follow-up data. Through the collation of the sample data, the inventors selected 341 samples that meet the following criteria as TLDA (Taqman Low Density Array, TLDA) chips. Test and subsequent series of experimental samples verified by qRT-PCR:

(1) Untreated early and mid-stage breast cancer patients;

(2) Untreated early and mid-term ER or (and) PR-positive breast cancer patients;

(3) LN has or (and) no transfer, and the number of LN transfers;

(4) Have accurate and detailed follow-up information.

Example 2 TLDA chip screening for molecular markers and internal reference genes

Twenty-six patients with good prognosis of breast cancer and 26 samples with poor prognosis of breast cancer were tested by TLDA chip and the relevant results were obtained. The specific steps are:

(1) RNA extraction from FFPE samples: 4 samples of 20 μm slices per sample or 8 slices of 10 μm slices were taken, and RNA was extracted according to the instructions of High Pure FFPET RNA Isolation Kit (Roche). The extracted RNA was quantified by NanoDrop-2000. Downstream reverse transcription experiments were performed after control.

(2) Total RNA is subjected to reverse transcription reaction to obtain cDNA sample: 1 μg of total RNA is taken according to

VILO ^TM Master Mix kit (Invitrogen) instructions for reverse transcription.

(3) TLDA detection of cDNA samples: the above cDNA products and

After the Universal PCR Master Mix was thoroughly mixed, the assay was performed on an ABI 7900 fluorescence quantitative PCR machine according to the TLDA standard procedure. (4) Data analysis and processing:

In the first phase of the study, 192 candidate genes derived from breast cancer were detected by TLDA detection technology. The gene expression differences of 26 breast cancer prognosis samples and 26 breast cancer prognosis samples were detected, and differentially expressed genes were screened. The different expression levels of the gene are expressed as 2 - ^ΔCt , wherein ^ΔCt = CT sample-CT reference, and the selected reference gene is normalized as a reference to calculate the relative expression amount. The screening process of the internal reference gene: using the genetic algorithm based on genorm, bestkeeper, normfinder, delta Ct and considering the biological function of the less fluctuating gene and its relationship with the tumor, screening candidate internal reference genes; calculating all candidate internal reference gene combinations Ct The correlation between the mean and the mean Ct of 192 genes, the most relevant combination is the internal reference genes including: ACTB, GAPDH, GUSB, NUP214, VCAN. Candidate gene screening criteria: (1) overall analysis - good prognosis and poor prognosis, the difference between the two groups is 2 times or less, and the proportion of cases with Ct < 35 is 50%; (2) stratified analysis - no lymph nodes The prognosis of the metastatic group was better than that of the poor prognosis. The difference between the two groups was more than 2 times, and the statistical difference was <0.05. (3) The difference between the two groups was not significant, but it was reported in the prognosis of breast cancer, and Ct< The proportion of cases in 35 reached 90%. The genes satisfying the above criteria include: BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT, MS4A1 and above.

Table 1 gene function analysis

Serial number	Gene name	Functional relevance
1.	BCL2	Estrogen related
2.	PGR	Estrogen related
3.	SCUBE2	Estrogen related
4.	ESR1	Estrogen related
5.	MKi67	Proliferation related
6.	CCNB1	Proliferation related
7.	MYBL2	Proliferation related
8.	GRB7	Her-2 related
9.	ERBB2	Her-2 related
10.	MMP11	Invasion related
11.	CD68	Differentiation group 68
12.	BAG1	BCL2 binds to anti-apoptotic gene 1
13.	MAPT	Microtubule-associated protein tau
14.	MS4A1	Transmembrane 4 domain subfamily A member 1

Example 3 Large sample size qRT-PCR verification of molecular markers

14 molecular markers and 5 internal reference genes screened by TLDA: BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT, MS4A1; ACTB, GAPDH, GUSB, NUP214 , VCAN. A single tube qRT-PCR was performed using 289 FFPE samples that met the above sample collection requirements and clinical follow-up information.

(1) RNA extraction of 289 FFPE samples: 4 samples of 20 μm slices per sample or 8 slices of 10 μm slices, RNA was extracted according to the instructions of High Pure FFPET RNA Isolation Kit (Roche), and the extracted RNA was quantified by NanoDrop-2000. Downstream reverse transcription experiments were performed after quality control.

(2) 289 FFPE samples were reverse transcribed into cDNA: 1 μg of total RNA was used.

VILO ^TM Master Mix kit (Invitrogen) instructions for reverse transcription.

(3) 289 FFPE sample cDNA products were subjected to qPCR detection: cDNA products, probes and primers of each sample,

After mixing the Universal Master Mix II, the assay was performed on an ABI 7900 real-time PCR machine. The qPCR primers and probe sequences are shown in Tables 2 to 5.

Table 2 qRT-PCR primer sequences

Table 3 qRT-PCR probe sequence

Table 4 housekeeping gene qRT-PCR primer sequence

Table 5 housekeeping gene qRT-PCR probe sequence

Example 4 Breast cancer prognosis 3-10 years recurrence or death risk prediction model establishment

The random forest algorithm in the machine learning method was used to evaluate the risk of recurrence or death after 3-10 years of postoperative detection, and a gene detection model for breast cancer prognosis evaluation was established. The random forest is composed of many decision trees. The decision tree is constructed by a method of double randomness of attributes and samples, so it is also called random decision tree. In a random forest, there is no correlation between decision trees. When the test data enters the random forest, it is classified by each decision tree. Finally, the class with the most classification results in all decision trees is the final result, that is, the result of the decision tree "voting", in other words, A random forest is a classifier that contains multiple decision trees, and the category of its output is determined by the mode of the category of the individual tree output. In the present invention, we have optimized on the basis of the traditional random forest algorithm. We randomly sample the samples 1000 times, establish 1000 models, and select the 39 best models with the highest accuracy from the 1000 model. The submodel is used and the median of the 39 submodels is used as the final prediction.

Example 5 Further verification of independent clinical samples

19 FFPE samples of ER or PR-positive early-stage breast cancer patients with known clinical follow-up data: PT staging was stage 1 and 2, of which patients were operated between 2004 and 2008, followed up from 2011 to 2015. In the year, the follow-up period was 3-10 years.

The high-purity FFPET RNA Isolation Kit (Roche) was used to extract the total RNA from the above 19 FFPE samples. After the quality control, the RNA was subjected to reverse transcription reaction to obtain cDNA samples. The cDNA products were subjected to qRT-PCR reaction to detect the internal reference genes ACTB, GAPDH, GUSB, NUP214. , VCAN, and BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT, MS4A1 genes. The Ct value of the above genes was introduced into the randomized forest method to establish a 3-10 year recurrence or death risk assessment model for breast cancer recurrence or death risk assessment model, and the risk threshold was predicted to be good prognosis or poor prognosis. The agreement analysis result and the known follow-up information were 73.6%. The specific results are shown in Table 6:

Table 6 Results of prognosis evaluation of 19 patients with breast cancer

Table 7

Example 6

At present, the decision of clinical treatment and treatment plan for breast cancer ultimately depends on the results of pathological examination, and the results of pathological examination are also the most important objective basis for judging prognosis. The detection system used FFPE samples of 289 newly diagnosed breast cancer patients with known clinical follow-up data collected by Tianjin Medical University Cancer Hospital and Henan Cancer Hospital. Five internal reference genes and 14 molecular markers were detected.

The test results are shown in Table 8.

Table 8

The kit provided by the invention has an accuracy rate of 81.1% for a newly diagnosed breast cancer patient with a low risk of recurrence or death of 3-10 years, and a pathological detection accuracy of 71.8%, which has a high risk of recurrence or death of 3-10 years. The accuracy rate of the newly diagnosed patients with breast cancer was 54.4%, which was close to the corresponding pathological detection accuracy of 56.8%. Compared with clinical follow-up information, the kit has a concordance rate of 70%. Except for patient age, pT stage, and LN number, there is no need to rely on other clinical pathological information.

The detection system and the kit are superior to the clinical pathological diagnosis result in the prognosis evaluation performance of breast cancer, and can reduce the excessive treatment and improper treatment caused by the pathological diagnosis error to meet the individualized precision of the breast cancer patient to a certain extent. The need for treatment has further improved the technical methods for predicting the prognosis of breast cancer in China.

The molecular markers provided by the present invention and their applications, detection kits, and methods for constructing detection models are described in detail above. The principles and embodiments of the present invention have been described with reference to specific examples, and the description of the above embodiments is only to assist in understanding the method of the present invention and its core idea. It should be noted that those skilled in the art can make various modifications and changes to the present invention without departing from the spirit and scope of the invention.

Claims (15)

1. A genetic composition characterized by comprising the molecular markers MAPT and/or MS4A1.
2. A gene composition consisting of the molecular markers BCL2, PGR, SCUBE2, ESR1, MKi67, CCNB1, MYBL2, GRB7, ERBB2, MMP11, CD68, BAG1, MAPT and MS4A1.
3. The genetic composition according to claim 2, further comprising internal reference genes ACTB, GAPDH, GUSB, NUP214, VCAN.
4. Use of the genetic composition according to claim 1 or 2 or 3 for the preparation of a device for predicting the risk of recurrence and/or mortality after 3-10 years of breast cancer surgery.
5. The use according to claim 4, wherein the risk of recurrence and/or death of the breast cancer after 3-10 years is specifically obtained by obtaining total RNA of the sample to be tested, obtaining cDNA by reverse transcription, and using fluorescence quantitative The Ct value of the molecular marker and the internal reference gene is obtained by a PCR method, and the Ct value of the internal reference gene is averaged to obtain an average Ct value (Ct') of the internal reference gene combination, and then the molecular marker is The Ct value was subtracted from the internal reference gene combination Ct' value to normalize, and △Ct was obtained. The △Ct value and the subject's age, pT value and LN value were reconstructed by random forest algorithm. A 10-year recurrence or death risk prediction model was analyzed and the results were obtained.
6. The use according to claim 5, wherein the predictive model for the risk of recurrence or death after 3-10 years of breast cancer is constructed by: ΔCt value of the molecular marker of the sample to be tested and the age of the subject , pT value, LN value construct mathematical matrix, randomly select 1/2 as training set, 1/2 as verification set, establish a prediction model containing 10000 decision trees through random forest algorithm, total random sampling ≥1000 times, establish ≥ In 1000 prediction models, ≥39 optimal models with the highest rate of coincidence with follow-up information were selected as the submodel of the final model from ≥1000 predictive models, and the median of ≥39 submodels was used as the final prognostic risk predictor.
7. The use according to claim 5 or 6, wherein the sample to be tested is an FFPE sample of a newly diagnosed early or mid-stage ER or PR positive breast cancer patient.
8. A primer set for amplifying the gene composition according to claim 1 or 2, wherein the sequence is shown in SEQ ID No. 1 to SEQ ID No. 28.
9. A probe set for amplifying the gene composition according to claim 1 or 2, wherein the sequence is represented by SEQ ID No. 29 to SEQ ID No. 42.
10. A primer set for amplifying an internal reference gene of the gene composition according to claim 3, which is represented by SEQ ID No. 43 to SEQ ID No. 47.
11. A probe set for amplifying an internal reference gene of the gene composition according to claim 3, which is represented by SEQ ID No. 48 to SEQ ID No. 52.
12. A kit for predicting the risk of recurrence and/or mortality after 3-10 years of breast cancer, comprising the primer set according to claim 8 and/or 10 and/or according to claims 9 and/or 11 The probe set described and the reagents commonly used in the kit.
13. A method for constructing a predictive model for recurrence or death risk of breast cancer 3-10 years after surgery, which is characterized in that a mathematical matrix is constructed by constructing a mathematical matrix of the ΔCt value of the molecular marker of the sample to be tested and the age, pT value and LN value of the subject, and randomly selecting 1/2 is used as the training set, 1/2 is used as the verification set, and the prediction model containing 10000 decision trees is established by the algorithm of random forest. The total random sampling is ≥1000 times, ≥1000 prediction models are established, and ≥1000 prediction models are selected. The ≥39 preferred models with the highest rate of follow-up information were the sub-models of the final model, and the median of ≥39 sub-models was used as the final prognostic risk predictor.
14. A method for detecting the risk of recurrence or death after 3-10 years of breast cancer, characterized in that the total RNA of the sample to be tested is obtained, the cDNA is obtained by reverse transcription, and the molecular marker and the reference gene are obtained by real-time PCR. The Ct value of the internal reference gene is averaged to obtain an average Ct value (Ct') of the internal reference gene combination, and then the Ct value of the molecular marker is subtracted from the internal reference gene combination Ct' value. Normalization, △Ct was obtained, and the △Ct value and the age, pT value and LN value of the subject were analyzed by the random forest algorithm for 3-10 years postoperative recurrence or death risk prediction model of breast cancer.
15. The method according to claim 13 or the method according to claim 14, wherein the sample to be tested is an FFPE sample of a newly diagnosed early or mid-stage ER or PR-positive breast cancer patient.

Images