CN110656173A

CN110656173A - Breast cancer prognosis evaluation model and establishment method thereof

Info

Publication number: CN110656173A
Application number: CN201910843958.XA
Authority: CN
Inventors: 王昕�; 伍启熹; 孟祥志; 杨婧怡; 应建明; 张静波; 刘霞; 马劲枫; 王建伟; 王翔
Original assignee: Beijing Usci Medical Laboratory Co ltd; Cancer Hospital and Institute of CAMS and PUMC
Current assignee: Beijing Usci Medical Laboratory Co ltd; Cancer Hospital and Institute of CAMS and PUMC
Priority date: 2019-09-06
Filing date: 2019-09-06
Publication date: 2020-01-07

Abstract

The invention provides a breast cancer prognosis evaluation model and an establishment method thereof. According to the application, Chinese people are selected as research objects, differential expression genes are screened by adopting a transcriptome sequencing (RNA-seq) method, genes closely related to recurrence are screened, and then a prediction model suitable for the recurrence risk of the breast cancer of the Chinese people is obtained. The model included 190 genes shown in table 1 that were differentially expressed in the relapsing and non-relapsing groups. The model can provide more accurate individualized treatment effect prediction and 10-year recurrence risk prediction for early breast cancer patients of Chinese population, so that high-risk population can not quickly develop diseases due to insufficient postoperative adjuvant chemotherapy, and low-risk population can not suffer from unnecessary chemotherapy toxic and side effects due to over-treatment.

Description

Breast cancer prognosis evaluation model and establishment method thereof

Technical Field

The invention relates to the technical field of high-throughput sequencing, in particular to a breast cancer prognosis evaluation model and an establishment method thereof.

Background

For early stage breast cancer patients with hormone receptor positive, Her2 and lymph node negative, whether chemotherapy is needed after surgical removal of tumor tissue to reduce the chance of postoperative tumor recurrence and metastasis is always a big problem for clinicians. The risk of tumor recurrence and metastasis of a patient after operation is increased due to insufficient chemotherapy, unnecessary chemotherapy is suffered by the patient due to excessive chemotherapy, unnecessary medical resources and social wealth are wasted, and economic losses and burdens of the society and the patient are increased.

Breast cancer is a heterogeneous disease, with different immunohistochemistry, molecular characteristics, pathotyping and gene expression, often with different prognosis. Some patients with the same pathological type and clinical stage have different prognosis. Studies have shown that gene heterogeneity and diversity are the main causes of the same type of pathology and clinical staging in patients, but with different prognosis. Genotyping can reflect different clinical manifestations and prognoses of different breast cancer patients, and provides theoretical support for further treatment. Therefore, the proposal of breast cancer genotyping based on the gene expression difference of tumor tissues provides important basis for solving the heterogeneity of tumors, the rationality of staging, the accuracy of prognosis judgment, the necessity of postoperative chemotherapy of early patients and the like. At present, some relatively mature methods for early-stage patient chemotherapy benefit evaluation and prognosis evaluation based on breast cancer tissue gene expression difference exist internationally, and the most authoritative and most widely applied method comprises Oncotype DX breast cancer 21 gene detection and Mammaprint70 gene detection.

The detection of the oncogene type DX breast cancer 21 gene is an accurate medical detection product which is introduced to the American market by the American Genomic Health company in 2005. The product utilizes a patient tissue wax block specimen in the experimental research of a United states mammary gland and intestinal tract surgery adjuvant therapy research group (NSABP) B-14, utilizes an RT-PCR technology to extract RNA from a tumor fixed by 4 percent formaldehyde and embedded by paraffin and immediately carries out reverse transcription polymerase chain reaction, and 21 genes related to the long-term recurrence of a patient with the cancer of lymph node negative and ER positive breast after being treated by tamoxifen are selected. The 21 genes comprise 16 tumor-related functional genes and 5 reference genes. Wherein the tumor-related functional genes comprise: proliferation related genes (Ki-67, STK15, Survivin, CyclinB1, MYBL 2); invasion-associated genes (Stromelysin3, Cathepsin L2); her-2 related genes (GRB7, Her-2); hormone-related genes (ER, PR, Bcl-2, SCUBE 2); GSTM 1; BAG 1; CD 68. And the 5 reference genes are Beta-actin, GAPDH, RPLPO, GUS and TFRC. By using RT-qPCR technology, the expression conditions of 16 tumor-related genes (related genes such as increment, invasion, HER2 and hormone) and 5 reference genes are detected, and the detection result is quantified into a Recurrence Score (RS) (specifically shown below), so that the risk of distant recurrence and the chemotherapy benefit within 10 years are predicted. RS ranges from 0 to 100, with higher scores, higher probability of recurrence, and the benefit from chemotherapy.

1. Internal reference group: ACTB, GAPDH, GUS, RPLPO and TFRC

HER2 group 0.9 xgrb 7+0.1 xher 2

Remarking: if the value of this group is < 8, then the value is 8

ER group ═ 0.8 XER +1.2 XPGR + BCL2+ SCUBE2)/4

4. Proliferation group (Survivin + Ki67+ STK15+ CCNB1+ MYBL2)/5

Remarking: if this value is < 6.5, then this value is 6.5

5. Attack group (MMP11+ CTSL2)/2

RSU-0.47 × HER2 group-0.34 × ER group +1.04 × proliferation group +0.1 × invasion group

+0.05×CD68-0.08×GSTM1-0.07×BAG1

7. If RSU < 0, then RS ═ 0

If RSU is 0 ≦ RSU ≦ 100, then RS × (RSU-6.7)

The Mammaprint70 gene assay is a product of Agendia. In 2002, researchers used gene chip technology to study the gene expression profiles of 98 patients with negative metastasis to breast cancer lymph nodes. About 25000 genes are involved on the gene chip, and 70 genes are closely related to the prognosis of the patient (see table 1 in particular).

Table 1:

BBC3	DCK	KNTC2	MMP9	SLC2A3	C2orf46
						TGFB3	MELK	MCM6	GPR126	RAB6B	LOC730018
ESM1	EXT1	NUSAP1	*RTN4PL1	IGFBP5	LOC1001311053
						IGFBP5	GNAZ	ORC6L	DIAPH3	COL4A2	AA555029_RC
FGF18	EBF4	TSPYL5	CDC42BPA	PECI	LGP2
						*SCUBE2	MTDH	*RUNDC1	PALM2	EGLN1	NMU
DIAPH3	PITRM1	PRC1	ALDH4A1	DIAPH3	UCHL5
						WISP1	QSCN6L1	RFC4	AYTL2	LOC100288906	JHDM1D
FLT1	CCNE2	RECQL5	QXCT1	C9orf30	AP2B1
						HRASLS	ECT2	*CDCA7	PECI	ZNF533	MS4A7
STK32B	CENPA	DTL	GMPS	C16orf61
						RASSF7	LIN9	GPR180	GSTM3	SERF1A

subsequently, the researchers used 70 genes as identification criteria to determine the prognosis of 298 breast cancer patients (without considering lymph node status when they were enrolled). Of these 118 patients suggested a good prognosis, and the results of the other 180 patients suggested a poor prognosis and matched the actual results. The 70 genes form a small gene expression profile, and can be used as a rapid evaluation method for clinical judgment of patient prognosis.

Breast cancer is a highly heterogeneous tumor with regional, population, and ethnic differences. The data statistics show that compared with the statistics data of the United states and Europe in 2000, Chinese breast cancer patients have larger tumor load (P < 0.001), poorer differentiation degree (P < 0.001), relatively lower ER positive rate (P < 0.001), but relatively higher HER2 positive rate (P < 0.001), and the breast cancer of Chinese people is probably more invasive than that of western people. Meanwhile, the onset age of breast cancer patients in China is 10-20 years earlier than that of western countries, more than half of the patients before menopause are different from the characteristics of breast cancer onset of European and American women. In addition, the incidence rate of triple negative breast cancer (TNBC, HER2-/ER-/PR-) in Chinese population and the positive expression rate of ER in 50+ age group are lower than those reported abroad, the positive expression rate of PR is higher, the proportion of patients with age more than or equal to 50 in HER2 over-expressed breast cancer is larger, and the TNBC lymph node metastasis rate is lower than that of other three subtype breast cancers. The detection products of 'breast cancer 21 gene' and 'breast cancer 70 gene' widely applied to European and American countries are calculated according to gene expression results of European and American groups, while the expression results of the genes in Asian regions are possibly different from those in European and American regions due to ethnic differences, so that the detection products are not very suitable for Chinese breast cancer patients according to the statistical results of European and American countries.

In terms of a detection method, the Oncotype DX detects the expression quantity of 21 genes by using a fluorescence quantitative PCR method, so that the sample quality has a great influence on the accuracy of a detection result, particularly in a clinically common FFPE sample, if the sample is degraded too severely, the extracted total RNA fragment is too short, the qPCR quantitative result is inaccurate, and the detection result is influenced. The MammaPrint utilizes a gene chip to detect the expression level of 70 related genes of tumor tissues, and the biggest limitation of the gene chip technology is that only known expressed genes can be detected, which results in the relative limitation of research content. In addition, the dynamic range of chip detection is narrow, selection must be made between high-abundance transcripts and low-abundance transcripts, detection of rare transcripts is not easy, and the common chip cannot identify shear mutations or allele-specific mutations, so that the gene combination and prediction model obtained by the MammaPrint may not be the optimal results.

Disclosure of Invention

The invention mainly aims to provide a breast cancer prognosis evaluation model and an establishment method thereof, so that the breast cancer prognosis evaluation model more suitable for Chinese people can be provided.

To achieve the above object, according to one aspect of the present invention, there is provided a use of a gene set for establishing a breast cancer prognosis evaluation model, the gene set being shown in table 2:

TABLE 2

NFYA

MDN1

CYP19A1

ERAP1

CHFR

PSMF1

ARHGAP26

CHST1S

NFKBIA

B4GALNT1

STAB1

PERP

TACC2

CDK5

PH4A2

DTD1

KCNMB1

FAM38

PHEX

SEMA4F

TSPAN9

STXBP3

PLCE1

RASEF

GPATCH1

ID1

LRRC27

SEPTINS

RABAC1

CAPN9

CALCOCO1

SWT1

DNAJC13

KLHDC2

CTTNBP2

GAL3ST1

SIDT2

MUC6

MPDZ

ABCB10

MIPEP

MYCL

CHRNA1

AP1G1

APBB1IP

MPST

PDCD4

BRF1

ATRNL1

TBRG4

DNAHS

RP56KA1

FAM117B

CATSPER2

PGR

PLD2

TRAPPC8

TARSL2

HSD17B1

SAP130

JADE2

SGIP1

PRDM5

KATNAL2

P2RXS

KIF1C

USP43

HLA-DRB1

CCDC86

MYC

PHPT1

PHF3

CASP6

TLCD3A

DDX18

GNL3L

LRGUK

CFAP43

CAPRIN2

SPAG8

KIF1B

SPP1

KLRG1

IRF2

SLC4A11

DOCK5

SAMD8

CD2AP

5100A1

WDR47

TRAF1

PTK2B

COL2A1

LUZP1

C20orf194

ECSrr

SCUBE1

SLC34A3

ITGA5

GTF2H4

NCKAP1

KIF18A

TICRR

MGMT

MYL6

SLC6A6

TMPRSS3

PLXNB3

CCDC78

SMTNL1

RFXANK

LY9

ELMO2

CLSTN1

SH2D3C

WDR44

RRP1B

TSBP1

SNED1

HLA-DPB1

ATG2B

RPL5

PRXACB

WASHC2C

SCL25A1

IQCA1

LSS

EHMT2

ACP6

SLC26A6

PYGM

WWP1

HMCN1

TNFRSP10D

RANGAP1

TESMIN

DIP2A

ZFP57

FZD5

C14orf132

SDK2

DNPEP

CAMKMT

C11orf45

CTSG

LP1N1

S100B

SLFN12L

CD51

RNF214

PTPN3

OARD1

TMM44

FUT1

SPTLC2

DOCK2

ST6GALNAC6

SNX2

FYCO1

GDPD5

ALDH2

PGBD1

NMNAT3

FMN1

CoG3

FAM205CP

SEC24D

EPR2

SORT1

NPIPA8

RP11-189E144

PHFSA

LIMD2

DOPEY2

ENDOV

NOP2

DBX2

5LC25A6

RP11-95H11

CATSPERB

RAPGEF5

VPSS1

ENKD1

C10orf95

XPOS

POLR3A

SLC15A4

RBBP4

FAM177A1

SPRYD3

。

In order to achieve the above object, according to a second aspect of the present invention, there is provided a breast cancer prognosis evaluation model including 190 genes differentially expressed in a recurrent group and a non-recurrent group as shown in table 2.

According to a third aspect of the present invention, there is provided a method for establishing a breast cancer prognosis evaluation model, the method comprising: obtaining genes differentially expressed in a recurrent group and a non-recurrent group of Chinese population breast cancer patients after treatment; establishing a breast cancer prognosis evaluation model by using differentially expressed genes by adopting a machine learning method; among them, the differentially expressed genes are 190 genes shown in table 2.

Further, a breast cancer prognosis evaluation model is established by adopting a support vector machine method.

Further, the method for establishing the breast cancer prognosis evaluation model by using the support vector machine comprises the following steps: adopting logTPM of 190 genes of a training set sample as input, adopting a Gaussian kernel function, and learning the training set by using an SVC function in sklern in python, thereby obtaining a breast cancer prognosis evaluation model; the training set samples included samples from the relapsing group and samples from the non-relapsing group.

Further, the genes which are obtained by the differential expression of the recurrent group and the non-recurrent group after the treatment of the breast cancer patients of the Chinese population comprise: acquiring complete transcriptome sequencing data of tumor tissues of early breast cancer patients in a recurrent group and a non-recurrent group; acquiring TPM values of respective sequencing data of a recurrent group and a non-recurrent group; and (4) statistically analyzing the differentially expressed genes in the recurrent group and the non-recurrent group according to the TPM value.

Further, obtaining TPM values for respective sequencing data of the relapsing group and the non-relapsing group comprises: comparing the sequencing data with a human reference genome sequence to obtain the Count of each gene; correcting the Count of each gene according to R1/(L1/1000) to obtain corrected Count; calculating the sum of the count numbers after gene correction to obtain R_{General assembly}(ii) a According to R1 × 1000 × 1000000/(L1 × R_{General assembly}) The TPM is computed.

Further, differentially expressed genes were obtained using the R toolkit edgeR analysis.

According to a third aspect of the present invention, there is provided an apparatus for building a breast cancer prognosis evaluation model, the apparatus comprising: the acquisition module is used for acquiring genes which are differentially expressed in a recurrent group and a non-recurrent group after treatment of breast cancer patients of Chinese population; the model establishing module is used for adopting a machine learning device to establish a breast cancer prognosis evaluation model by using the differentially expressed genes; among them, the differentially expressed genes are 190 genes shown in table 2.

Further, the model building module comprises a support vector machine module, wherein the support vector machine module is used for learning the training set by using the logTPM of 190 genes of the training set sample as input and using a Gaussian kernel function in python and using the SVC function in sklern, so as to obtain a breast cancer prognosis evaluation model; the training set samples included samples from the relapsing group and samples from the non-relapsing group.

Further, the acquisition module includes: the system comprises a first acquisition unit, a second acquisition unit and a difference statistical unit, wherein the first acquisition unit is used for acquiring the whole transcriptome sequencing data of the tumor tissues of early-stage breast cancer patients in a recurrent group and a non-recurrent group; the second acquisition unit is used for acquiring TPM values of sequencing data of the recurrent group and the non-recurrent group respectively; and the difference statistical unit is used for statistically analyzing and obtaining the genes with different expressions in the recurrent group and the non-recurrent group according to the TPM value.

Further, the second acquisition unit includes: the system comprises an alignment submodule, a correction submodule, a first calculation submodule and a second calculation submodule, wherein the alignment submodule is used for aligning sequencing data with a human reference genome sequence to obtain the Count number of each gene; the correction submodule is used for correcting the Count of each gene according to R1/(L1/1000) to obtain a corrected Count; a first calculation submodule for calculating the sum of the count numbers after gene correction to obtain R_{General assembly}(ii) a A second calculation submodule for calculating R1 1000 1000000/(L1R)_{General assembly}) The TPM is computed.

Further, the variance statistic unit is edgeR.

By applying the technical scheme of the invention, Chinese population is selected as a research object, and a transcriptome sequencing (RNA-seq) method is adopted to screen the differentially expressed genes, so that the genes closely related to the recurrence are screened, and further, a prediction model suitable for the recurrence risk of the breast cancer of the Chinese population is obtained. The model included 190 genes shown in table 2 that were differentially expressed in the recurrent and non-recurrent groups. The model can provide more accurate individualized treatment effect prediction and 10-year recurrence risk prediction for early breast cancer patients of Chinese population, so that high-risk population can not quickly develop diseases due to insufficient postoperative adjuvant chemotherapy, and low-risk population can not suffer from unnecessary chemotherapy toxic and side effects due to over-treatment.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:

FIG. 1 is a schematic flow chart illustrating a method for establishing a breast cancer prognosis evaluation model according to an alternative embodiment of the present application;

fig. 2 is a schematic structural diagram of an apparatus for establishing a breast cancer prognosis evaluation model according to an alternative embodiment of the present application.

Detailed Description

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present invention will be described in detail with reference to examples.

As mentioned in the background art, the existing breast cancer prognosis evaluation model is not suitable for the diseased characteristics of Chinese population, so that the existing breast cancer prognosis evaluation model has the defect of poor accuracy when being used for guiding the prognosis treatment scheme of Chinese patients. In order to provide a prognosis evaluation model more suitable for Chinese people, the application selects Chinese people as research objects, screens differentially expressed genes by adopting a transcriptome sequencing (RNA-seq) method, screens out genes closely related to recurrence, and further obtains a prediction model suitable for the recurrence risk of breast cancer of the Chinese people.

Example 1

Based on the above research results, the present application proposes the use of a gene set in establishing a breast cancer prognosis evaluation model, wherein the gene set is shown in table 2.

190 genes in the table 2 are closely related to relapse, and the 190 genes are utilized to establish a prediction model suitable for the relapse risk of the breast cancer of Chinese people, so that the prediction accuracy of the established model is improved, more accurate individualized treatment effect prediction and 10-year relapse risk prediction are provided for early breast cancer patients of Chinese people, the high-risk people can not quickly develop diseases due to insufficient postoperative adjuvant chemotherapy, the low-risk people can not bear unnecessary toxic and side effects of chemotherapy due to over-treatment, even the disease-free survival period is shortened due to the hepatorenal toxicity and the immunity decline of chemotherapy, meanwhile, the waste of unnecessary medical resources and social wealth is avoided, and the economic loss and burden of society and patients are reduced.

Example 2

In another exemplary embodiment, a breast cancer prognosis evaluation model is provided, which comprises 190 genes shown in table 2 that are differentially expressed in the recurrent group and the non-recurrent group. The model can provide more accurate individualized treatment effect prediction and 10-year recurrence risk prediction for early breast cancer patients of Chinese population, so that high-risk population can not quickly develop diseases due to insufficient postoperative adjuvant chemotherapy, and low-risk population can not suffer from unnecessary chemotherapy toxic and side effects due to over-treatment.

In the above model, the P value of the differential expression of the differentially expressed genes can be selected to be less than 0.05.

The model is established based on sequencing data of genes with significant differential expression in the recurrent group and the non-recurrent group, and therefore, any method capable of establishing a prognosis model using the relevant sequencing data of two groups of differentially expressed genes is suitable for the present application.

Example 3

In a preferred embodiment of the present application, a method for establishing a breast cancer prognosis evaluation model is provided, as shown in fig. 1, the method comprising:

step S101, obtaining genes which are differentially expressed in a recurrent group and a non-recurrent group of Chinese population breast cancer patients after treatment;

step S102, establishing a breast cancer prognosis evaluation model by using differentially expressed genes by adopting a machine learning method; among them, the differentially expressed genes are 190 genes shown in table 2.

The model is trained and learned by using a machine learning method and taking genes with differential expression in two groups of samples with known relapse and without relapse as training sets, so that the model conforming to a certain rule is established. By utilizing the model, more accurate individual treatment effect prediction and recurrence risk assessment can be accurately provided for patients in Chinese population.

In the above model establishing method, any machine learning method is suitable for the present application as long as the indexes of the two training sets can be distinguished. In a preferred embodiment of the present application, a breast cancer prognosis evaluation model is established by using a support vector machine method.

The support vector machine belongs to a two-classification model in a machine learning method. The basic idea is to map a set of multidimensional data into a multidimensional feature space. A unique hyperplane is then determined, separating the two sets of data in the training set, and maximizing the geometric separation of each data into hyperplanes. An example of a sample point of the training data set and a sample point closest in separation hyperplane distance is a support vector. When the plane exists, the plane is a linear branch support vector machine, and if only a hypersurface can be used for separating positive and negative examples, the plane is a nonlinear support vector machine. At this time, a transformation function can be used to map the points in the original space to a new space, so that the hyperplane separating the positive and negative cases is found in the new space. And the inner product of the mapping function is called a kernel function.

The fact that the support vector machine can find the hyperplane with the largest geometric interval for the training data set means that the training data are classified with sufficient certainty factor, particularly points which are the most difficult to be classified are separated with sufficient certainty factor, and therefore the prediction effect of the constructed model on Chinese people is more accurate.

In a preferred embodiment, the method of using a support vector machine to establish a breast cancer prognosis evaluation model comprises: adopting logTPM (globally called Transcripts per mileon reads) of 190 genes of a training set sample as input, adopting a Gaussian kernel function, and learning the training set by using an SVC (static var compensator) function in sklern in python, thereby obtaining a breast cancer prognosis evaluation model; the training set samples included samples from the relapsing group and samples from the non-relapsing group. The model established in the way can predict the relapse risk of the Chinese population more accurately.

The method for obtaining the genes which are differentially expressed in the two groups of samples can be an existing method or can be obtained by improving the existing method. In a preferred embodiment of the present application, the genes that are differentially expressed in the recurrent group and the non-recurrent group of Chinese people breast cancer patients after treatment comprise: obtaining whole transcriptome sequencing data of tumor tissues of patients with early (stage I and/or stage II) breast cancer in a relapsing group and a non-relapsing group; acquiring TPM values of respective sequencing data of a recurrent group and a non-recurrent group; and (4) statistically analyzing the differentially expressed genes in the recurrent group and the non-recurrent group according to the TPM value.

In a preferred embodiment of the present application, obtaining TPM values of sequencing data of each of the recurrent group and the non-recurrent group comprises: comparing the sequencing data with a human reference genome sequence to obtain the Count of each gene; correcting the Count of each gene according to R1/(L1/1000) to obtain corrected Count; calculating the sum of the count numbers after gene correction to obtain R_{General assembly}(ii) a According to R1 × 1000 × 1000000/(L1 × R_{General assembly}) The TPM is computed.

In the establishing method of the model, the differentially expressed genes are obtained by adopting an R tool package edgeR analysis.

Example 4

RNA extraction

Extracting total RNA by using a commercial nucleic acid extraction kit, detecting the RNA degradation degree by electrophoresis after extraction, and detecting the RNA pollution condition by using Nanodrop.

Ribo-zero-kit removal of ribosomal RNA (rRNA)

Since the ratio of rRNA in the total RNA extracted exceeds 95%, it is necessary to remove the influence of rRNA on transcriptome-specific data by experimental means, and to specifically remove rRNA using Ribo-zero-kit.

3. Library construction

After rRNA is removed from the total RNA, one-strand and two-strand synthesis is carried out to obtain double-stranded cDNA, then the steps of end repair of cDNA fragments, A addition, joint addition and the like are carried out, and finally index is introduced through PCR and amplified to obtain a final transcriptome library.

4. Library quality control and on-machine sequencing

The constructed library is subjected to 2100 detection and QPCR quantification before on-machine sequencing, the size of an insert fragment is detected, the concentration of an effective fragment is calculated, and each library to be detected is Pooling according to the concentration of the effective fragment and the required on-machine data quantity.

5. Data analysis

5.1 using STAR software to compare the sequence obtained by sequencing with the human genome hg19, and using STAR software to quantify the bam file after comparison to obtain the number of genes of each gene.

5.2 calculating the TPM value of the gene according to the Counts value of the gene, wherein the TPM is named as Transcripts per milliontreads, and the calculation method is as follows:

(1) correcting the Count value according to the gene length, wherein if the Count value of a certain gene is R1 and the gene length is L1, the corrected Count value is R1/(L1/1000);

(2) calculating the corrected total value of the total, i.e. the sum R of corrected total values of all genes_{General assembly}；

(3) Calculating TPM, the result of TPM is R1 1000 1000000/(L1R_{General assembly})

6. Finding prognosis related gene and establishing classification model

6.1 genes whose expression differed significantly between the relapsed and non-relapsed groups were determined by the R kit edgeR. 190 genes were screened by gene pathway analysis in combination with gene function (as shown in table 2).

6.2 establishing a breast cancer prognosis evaluation model by using a support vector machine method.

And (3) learning the training set by using the logTPM of 190 genes of the training set sample as input and using the Gaussian kernel function and the SVC function in the skearn in python, and classifying to obtain the breast cancer prognosis evaluation model.

Example 5

281 samples were randomly divided into training and test sets at a ratio of 7: 3. 190 prognostic related genes are obtained from 195 samples, a support vector machine is used for learning training set data, the obtained model is used for predicting 85 test set samples, and the classification specificity and sensitivity of the model are calculated according to the observed value and the true value of the test set. Sensitivity was 1 and accuracy was 0.823.

Through the above description of the embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (e.g., a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.

Example 6

In another alternative embodiment, the present application further provides an apparatus for establishing a breast cancer prognosis evaluation model, as shown in fig. 2, the apparatus comprising: the system comprises an acquisition module 20 and a model establishing module 40, wherein the acquisition module 20 is used for acquiring genes which are differentially expressed in a recurrent group and a non-recurrent group of treated breast cancer patients of Chinese population; the model establishing module 40 is used for adopting a machine learning device to establish a breast cancer prognosis evaluation model by using the differentially expressed genes; the differentially expressed genes were 190 genes as shown in table 1.

The device utilizes the 190 genes to establish a prediction model suitable for the recurrence risk of the breast cancer of Chinese population, which is beneficial to improving the prediction accuracy of the established model, and further provides more accurate individualized treatment effect prediction and 10-year recurrence risk prediction for early-stage breast cancer patients of Chinese population, so that high-risk population can not quickly develop disease due to insufficient postoperative auxiliary chemotherapy, low-risk population can not bear unnecessary toxic and side effects of chemotherapy due to over-treatment, even the disease-free life cycle is shortened due to the hepatotoxicity and the renal toxicity of chemotherapy and the decline of immunity, meanwhile, the waste of unnecessary medical resources and social wealth is avoided, and the economic loss and burden of society and patients are reduced.

Optionally, the model building module comprises: the support vector machine module is used for learning the training set by using an SVC function in sklern in python by using a Gaussian kernel function and taking logTPM of 190 genes of a training set sample as input so as to obtain a breast cancer prognosis evaluation model; the training set samples included samples from the relapsing group and samples from the non-relapsing group.

Optionally, the obtaining module includes: a first obtaining unit, a second obtaining unit and a difference statistic unit, wherein the first obtaining unit is used for obtaining the whole transcriptome sequencing data of the tumor tissues of the patients with early-stage (I stage and/or II stage) breast cancer of a relapse group and a non-relapse group; the second acquisition unit is used for acquiring TPM values of sequencing data of the recurrent group and the non-recurrent group respectively; and the difference statistical unit is used for statistically analyzing and obtaining the genes with different expressions in the recurrent group and the non-recurrent group according to the TPM value.

Optionally, the second obtaining unit includes: the system comprises an alignment submodule, a correction submodule, a first calculation submodule and a second calculation submodule, wherein the alignment submodule is used for aligning sequencing data with a human reference genome sequence to obtain the Count number of each gene; the correction submodule is used for correcting the Count of each gene according to R1/(L1/1000) to obtain a corrected Count; a first calculation submodule for calculating the basisR is obtained from the sum of corrected count numbers_{General assembly}(ii) a A second calculation submodule for calculating R1 1000 1000000/(L1R)_{General assembly}) The TPM is computed.

Optionally, the difference statistic unit is edgeR.

From the above description, it can be seen that the above-described embodiments of the present invention achieve the following technical effects: the research object aimed at by the method is Chinese population, so that the obtained data and model are more suitable for Chinese population patients, and when the model is constructed, the hyperplane with the largest geometric interval can be found for the training data set by adopting the support vector machine, so that the training data are classified with sufficient certainty factor, particularly, points which are the hardest to be classified have large enough certainty factor to separate the points, and the constructed model has more accurate prediction effect on the Chinese population.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The application of a gene combination in establishing a breast cancer prognosis evaluation model, wherein the gene combination is shown in the following table:

NFYA MDN1 CYP19A1 ERAP1 CHFR PSMF1 ARHGAP26 CHST1S NFKBIA B4GALNT1 STAB1 PERP TACC2 CDK5 PH4A2 DTD1 KCNMB1 FAM38 PHEX SEMA4F TSPAN9 STXBP3 PLCE1 RASEF GPATCH1 ID1 LRRC27 SEPTINS RABAC1 CAPN9 CALCOCO1 SWT1 DNAJC13 KLHDC2 CTTNBP2 GAL3ST1 SIDT2 MUC6 MPDZ ABCB10 MIPEP MYCL CHRNA1 AP1G1 APBB1IP MPST PDCD4 BRF1 ATRNL1 TBRG4 DNAHS RP56KA1 FAM117B CATSPER2 PGR PLD2 TRAPPC8 TARSL2 HSD17B1 SAP130 JADE2 SGIP1 PRDM5 KATNAL2 P2RXS KIF1C USP43 HLA-DRB1 CCDC86 MYC PHPT1 PHF3 CASP6 TLCD3A DDX18 GNL3L LRGUK CFAP43 CAPRIN2 SPAG8 KIF1B SPP1 KLRG1 IRF2 SLC4A11 DOCK5 SAMD8 CD2AP 5100A1 WDR47 TRAF1 PTK2B COL2A1 LUZP1 C20orf194 ECSIT SCUBE1 SLC34A3 ITGA5 GTF2H4 NCKAP1 KIF18A TTCRR MGMT MYL6 SLC6A6 TMPRSS3 PLXNB3 CCDC78 SMTNL1 RFXANK LY9 ELMO2 CLSTN1 SH2D3C WDR44 RRP1B TSBP1 SNED1 HLA-DPB1 ATG2B RPL5 PRKACB WASHC2C SCL25A1 IQCA1 LSS EHMT2 ACP6 SLC26A6 PYGM WWP1 HMCN1 TNFRSP10D RANGAP1 TESMIN DIP2A ZFP57 FZD5 C14orf132 SDK2 DNPEP CAMKMT C11orf45 CTS6 LP1N1 S100B SLFN12L CD51 RNF224 PTPN3 OARD1 TMEM44 FUT1 SPTLC2 DOCK2 ST6GALNAC6 SNX2 FYCo1 GDPD5 ALDH2 PGBD1 NMNAT3 FMN1 COG3 FAM205CP SEC24D EPR2 SORT1 NPIPA8 RP11-189E14.4 PHFSA LMD2 DOPEY2 ENDOV NOP2 DBX2 5LC25A6 RP11-95H11 CATSPERB RAPGEF5 VPSS1 ENKD1 C10orf95 XPOS POLR3A SLC15A4 RBBP4 FAM177A1 SPRYD3

。

2. a breast cancer prognosis evaluation model, comprising 190 genes shown in table 2 which are differentially expressed in a recurrent group and a non-recurrent group.

3. A method for establishing a breast cancer prognosis evaluation model is characterized by comprising the following steps:

obtaining genes differentially expressed in a recurrent group and a non-recurrent group of Chinese population breast cancer patients after treatment;

establishing a breast cancer prognosis evaluation model by using the differentially expressed genes by adopting a machine learning method;

wherein the differentially expressed genes are 190 genes shown in Table 2.

4. The method for establishing the breast cancer prognosis evaluation model according to the claim 3, wherein a method of a support vector machine is adopted to establish the breast cancer prognosis evaluation model.

5. The method for establishing the breast cancer prognosis evaluation model according to the claim 4, wherein the establishing the breast cancer prognosis evaluation model by using a support vector machine comprises the following steps:

adopting logTPM of 190 genes of a training set sample as input, adopting a Gaussian kernel function, and learning the training set by using an SVC function in sklern in python, thereby obtaining the breast cancer prognosis evaluation model;

the training set samples include samples of the relapsing group and samples of a non-relapsing group.

6. The method of claim 4, wherein the obtaining of the genes differentially expressed in the recurrent group and the non-recurrent group of Chinese people after treatment for breast cancer patients comprises:

acquiring complete transcriptome sequencing data of tumor tissues of early breast cancer patients in a recurrent group and a non-recurrent group;

obtaining TPM values of respective sequencing data of the recurrent group and the non-recurrent group;

and statistically analyzing the differentially expressed genes in the recurrent group and the non-recurrent group according to the TPM value.

7. The method of establishing according to claim 5, wherein obtaining TPM values for respective sequencing data of the recurrent group and the non-recurrent group comprises:

comparing the sequencing data with a human reference genome sequence to obtain the Count of each gene;

correcting the Count number of each gene according to R1/(L1/1000) to obtain a corrected Count number;

calculating the sum of the corrected count numbers of the genes to obtain R_{General assembly}；

According to R1 × 1000 × 1000000/(L1 × R_{General assembly}) And calculating the TPM.

8. The method of claim 5, wherein the differentially expressed genes are obtained using an R toolkit edgeR analysis.

9. An apparatus for building a breast cancer prognosis evaluation model, the apparatus comprising:

the acquisition module is used for acquiring genes which are differentially expressed in a recurrent group and a non-recurrent group of Chinese population breast cancer patients after treatment;

the model establishing module is used for adopting a machine learning device to establish a breast cancer prognosis evaluation model by using the differentially expressed genes;

wherein the differentially expressed genes are 190 genes shown in Table 2.

10. The apparatus according to claim 9, wherein the model building module comprises:

the support vector machine module is used for learning the training set by using an SVC function in skearn in python by using a Gaussian kernel function and taking logTPM of 190 genes of a training set sample as input so as to obtain the breast cancer prognosis evaluation model;

11. The apparatus according to claim 9, wherein the obtaining module comprises:

a first obtaining unit, configured to obtain transcriptome-wide sequencing data of tumor tissues of the patients with early breast cancer in the relapsing group and the non-relapsing group;

a second obtaining unit, configured to obtain TPM values of respective sequencing data of the recurrent group and the non-recurrent group;

and the difference statistical unit is used for statistically analyzing and obtaining the genes which are differentially expressed in the recurrent group and the non-recurrent group according to the TPM value.

12. The apparatus according to claim 11, wherein the second obtaining unit comprises:

the alignment submodule is used for aligning the sequencing data with a human reference genome sequence to obtain the Count of each gene;

the correction submodule is used for correcting the Count number of each gene according to R1/(L1/1000) to obtain a corrected Count number;

a first calculation submodule for calculating the sum of corrected count numbers of genes to obtain R_{General assembly}；

A second calculation submodule for calculating R1 1000 1000000/(L1R)_{General assembly}) And calculating the TPM.

13. The apparatus according to claim 11, wherein the difference statistic unit is edgeR.