Model for predicting curative effect of abiraterone on metastatic castration-resistant prostate cancer

Abstract

The invention discloses a gene marker combination for predicting the curative effect of abiraterone in metastatic castration resistant prostate cancer (mCRPC). The gene marker combination comprises 16 genes, namely HDAC2, TP53, GEN1, FOXA1, ATR, MUTYH, SPOP, BRCA2, ATM, PALB2, RB1, CDK12, BARD1, PTEN, NCOR2 and MLH1. The invention further discloses an mCRPC abiraterone curative effect prediction model based on the marker combination and a construction method of the mCRPC abiraterone curative effect prediction model. According to the method, mCRPC gene markers highly related to abiraterone treatment response are screened based on detection data of related genes of prostate cancer in combination with clinical abiraterone treatment effect results, and an abiraterone treatment effect prediction model is constructed based on the markers, so that personalized prediction of the treatment effect of an mCRPC patient using abiraterone treatment is realized, and the treatment effect of the patient is improved. And an effective reference can be provided for accurate medication.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of prostate cancer, in particular to metastatic castration-resistant prostate cancer (mCRPC), and more specifically to the prediction of the efficacy of abiraterone in the treatment of mCRPC. BACKGROUND

[0002] Prostate cancer (PCa) is one of the common urogenital diseases in men over 50 years old, and has become the second most common malignant tumor in men worldwide, with an incidence rate second only to lung cancer. Prostate cancer is also the fifth leading cause of cancer-related deaths worldwide. Prostate cancer usually has no significant clinical symptoms in the early stage.

[0003] Metastatic hormone-sensitive prostate cancer (mHSPC) refers to PCa patients who have certain response to endocrine therapy and are accompanied by bone or other organ metastasis. Metastatic prostate cancer initially has a significant response to androgen deprivation therapy, but almost all patients will develop incurable metastatic castration-resistant prostate cancer (mCRPC). Despite the existence of numerous drugs (such as new endocrine therapy drugs, chemotherapy, PARP inhibitors, etc.), mCRPC is still incurable. Previous research results found that among mCRPC patients treated with abiraterone, 46.55% of patients showed effective treatment response at 12 weeks, i.e., prostate-specific antigen (PSA) decreased by more than 50% after treatment. Studies have also found that patients carrying different mutations show significant differences in PFS (Progression Free Survival, Progression Free Survival): CDK12 (- / +, 1.6 / 10.4 months, P=0.001), TP53 / RB1 (- / +, 2.0 / 11 months, P=0.001), AR (- / +, 10.4 / 3.0 months, P=0.007), however, whether the AR gene has a mutation does not affect the PFS of abiraterone treatment. It shows that mCRPC has obvious interpatient and intratumor genomic heterogeneity, and the mCRPC mutation map changes dynamically during treatment, which explains its different molecular characteristics at different stages of treatment, and also shows that the same treatment strategy has different efficacy for different patients. At present, the European Association of Urology (EAU), the American Urological Association (AUA) and other major guidelines recommend that abiraterone is an effective treatment for mCRPC patients. Therefore, it is particularly important to select appropriate molecular markers to screen potential abiraterone treatment advantage population. In addition, due to the differences in genetic background between different patients, the information data provided by a single gene mutation is limited, and it is difficult to provide accurate and effective efficacy prediction for individuals. SUMMARY

[0004] One of the technical problems to be solved by the present application is to provide a set of efficacy prediction gene markers for metastatic castration-resistant prostate cancer (mCRPC) treated with abiraterone, which is highly related to the response of abiraterone treatment and can be used to effectively predict the efficacy of mCRPC treated with abiraterone.

[0005] To solve the above technical problems, the metastatic castration-resistant prostate cancer abiraterone efficacy prediction gene marker combination of the present application comprises the following 16 genes: HDAC2, TP53, GEN1, FOXA1, ATR, MUTYH, SPOP, BRCA2, ATM, PALB2, RB1, CDK12, BARD1, PTEN, NCOR2, and MLH1.

[0006] The second technical problem to be solved by the present application is to provide a kit comprising the detection reagent of the gene marker combination.

[0007] The third technical problem to be solved by the present application is to provide a metastatic castration-resistant prostate cancer abiraterone efficacy prediction model based on the gene marker combination, which comprises the following calculation formula:

[0008]

[0009] wherein, F Score is the efficacy score; A is the weight coefficient of the gene; B is the mutation condition of the gene, B=1 if the gene has a mutation, and B=0 if the gene has no mutation; and n is the number of the gene. The mutation includes non-synonymous mutation or copy number variation in the exon or splicing site region of the gene.

[0010] The calculation formula is preferably:

[0011] F Score=(1.45550198×B BARD1 )+(1.44867163×B CDK12 )+(1.43994884×B TP53 )+(0.89126422

[0012] ×B MLH1 )+(0.79292448×B HDAC2 )+(0.22807960×B SPOP )+(0.21785647×B RB1 )+(-0.06696785×B BRCA2 )+(-0.31043467×B ATR )+(-0.31350703×B PALB2 )+(-0.33076422×B NCOR2 )+(-0.35784534×B ATM )+(-0.41346502×B PTEN )+(-0.50077044×B FOXA1 )+(-0.57777498×B MUTYH )+(-0.78258597×BGEN1 )。

[0013] The fourth technical problem to be solved by the present application is to provide a construction method of the abovementioned abiraterone efficacy prediction model for metastatic castration-resistant prostate cancer, which comprises the following steps:

[0014] 1) Obtain PSA data and gene detection data of mCRPC patients who have received abiraterone treatment after surgery;

[0015] 2) Screen genes related to prostate cancer as candidate genes;

[0016] 3) For the training set patient samples, input the PSA data and mutation information of all candidate genes of each sample, and perform model training;

[0017] 4) Obtain genes related to abiraterone efficacy and their corresponding weight coefficients, and establish an abiraterone efficacy prediction model for metastatic castration-resistant prostate cancer.

[0018] The candidate genes in step 2) above include the following 50 genes: AR, MUTYH, ATM FANCA, PMS2, BARD1, NBN, ATR, FANCD2, POLD1, BRIP1, NCOR1, BRCA1, HDAC2, POLE, CDH1, NCOR2, BRCA2, MLH1, RAD51, EPCAM, PTEN, CDK12, MLH3, RAD51B, ESR1, SPOP, CHEK2, MRE11A, RAD51C, FAM175A, STK11, ERCC2, MSH2, XRCC4, FOXA1, ZBTB16, ERCC3, MSH6, RAD50, GEN1, RB1, ERCC4, PALB2, RAD51D, IDH1, TP53, ERCC5, PMS1, HOXB13.

[0019] In step 3) above, the model training can adopt LASSO regression method, preferably LASSO 10-fold cross-validation regression method.

[0020] The present application is based on Chinese multi-center prostate cancer related gene detection data, combined with the clinical abiraterone treatment effect result, using machine learning method, screening out 16 metastatic hormone-resistant prostate cancer (mCRPC) gene markers highly related to endocrine drug-abiraterone treatment response (PAS decrease more than 50%), and constructing an mCRPC abiraterone efficacy prediction model based on these gene markers, which can not only individualize the efficacy of mCRPC patients treated with abiraterone, but also has high efficiency and accuracy of efficacy prediction and low cost, thereby helping to stratify patients and effectively screen potential treatment advantage population, providing a reference for clinical precision medicine treatment. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is the LASSO coefficient column chart of 16 genes related to abiraterone efficacy;

[0022] Figure 2 is the correlation graph between the efficacy score of the training set and the drug treatment response evaluation index PSA (Spearman correlation analysis, R=0.73, Wilcoxon rank sum test p=1e-07);

[0023] Figure 3 is the ROC curve of the training set, AUC=0.775;

[0024] Figure 4 is the abiraterone treatment response difference comparison chart of the low score group and the high score group in the training set (Wilcoxon rank sum test p=2.8e-4);

[0025] Figure 5 is the efficacy score comparison chart of the abiraterone treatment effective group (PSA<0.5) and the abiraterone treatment ineffective group (PSA>0.5) in the training set; Wilcoxon rank sum test p=7e-4;

[0026] Figure 6 is the Spearman correlation analysis of the verification set, R=0.52, Wilcoxon rank sum test p=2.4e-2;

[0027] Figure 7 is the ROC curve of the verification set, AUC=0.8624. DETAILED DESCRIPTION

[0028] In order to have a more specific understanding of the technical content, characteristics and effects of the present application, the technical solutions of the present application will be further described in detail in combination with the drawings and specific embodiments.

[0029] Example 1: Abiraterone efficacy prediction model for metastatic hormone-resistant prostate cancer

[0030] 1. Construction of the model

[0031] The mCRPC patient sample data used in this embodiment comes from the clinical research and clinical treatment sample data of prostate cancer in eight hospitals, namely, Renji Hospital Affiliated to School of Medicine, Shanghai Jiao Tong University, the Tenth People's Hospital of Shanghai, the First Affiliated Hospital of Wenzhou Medical University, the Department of Urology, Center for Cancer Prevention and Treatment, Sun Yat-sen University, the First Affiliated Hospital of Xi'an Jiaotong University, Beijing Friendship Hospital Affiliated to Capital Medical University, the Ninth Three Hundred Hospital of the Joint Logistics Support Force of the Chinese People's Liberation Army, and Fuzhou General Hospital of Nanjing Military Region of the Chinese People's Liberation Army (the research results are published in JNCCN Journal, DOI: 10.6004 / jnccn.2020.7663). This embodiment selects 58 mCRPC patients who have received abiraterone treatment after surgery, obtains the gene detection data of the blood samples of the 58 patients from the prostate cancer sample data of the above eight hospitals, and obtains the PSA change data reflecting the efficacy of abiraterone, and then screens the signal pathway genes closely related to the occurrence and development of prostate cancer (including endocrine therapy and novel endocrine therapy related pathways, neural endocrine differentiation pathways, DNA damage repair pathways, PI3K / AKT signaling pathways, etc.) and other important genes according to the literature reports, a total of 50 genes are screened (see Table 1). The 58 mCRPC patients are grouped according to the training set: validation set = 2: 1, and 39 patients are randomly selected as the training set, and the rest are as the validation set.

[0032] Table 1 List of 50 candidate genes

[0033] AR MUTYH ATM FANCA PMS2 BARD1 NBN ATR FANCD2 POLD1 BRIP1 NCOR1 BRCA1 HDAC2 POLE CDH1 NCOR2 BRCA2 MLH1 RAD51 EPCAM PTEN CDK12 MLH3 RAD51B ESR1 SPOP CHEK2 MRE11A RAD51C FAM175A STK11 ERCC2 MSH2 XRCC4 FOXA1 ZBTB16 ERCC3 MSH6 RAD50 GEN1 RB1 ERCC4 PALB2 RAD51D IDH1 TP53 ERCC5 PMS1 HOXB13

[0034] Using the glment software package of R language, the abiraterone treatment response PSA information of each sample in the training set and the mutation information of 50 genes in each sample are input, nlambda = 100, alpha = 1, and the rest of the parameters are set to the default value, LASSO 10-fold cross-validation regression analysis is performed, and genes related to the efficacy of abiraterone treatment are screened from the above 50 genes.

[0035] The feature of LASSO (Least absolute shrinkage and selection operator) regression is that it can be used to establish a generalized linear model regardless of whether the dependent variable is continuous or discrete. It has extremely low requirements for data, so it is widely used. In addition, LASSO can also screen variables and reduce the complexity of the model. When dealing with high-dimensional data sets, when the number of features is much larger than the number of samples, by introducing a regularization term, the coefficients of irrelevant or redundant features can be set to zero, and only important features related to the target variable are retained, thereby reducing the complexity of the model and the influence of noise and overfitting of the results.

[0036] The basic idea of LASSO is to minimize the sum of squared residuals under the constraint that the sum of the absolute values of the regression coefficients is less than a constant, so that some regression coefficients are strictly equal to 0, and an interpretable model is obtained.

[0037] In this embodiment, 16 gene markers related to the efficacy of abiraterone and their corresponding LASSO coefficients are obtained by LASSO regression analysis, as shown in Table 2, Figure 1

[0038] Table 2 16 gene markers related to the efficacy of abiraterone and their LASSO weight coefficients

[0039]

[0040]

[0041] The final calculation formula of the abiraterone efficacy prediction model for metastatic castration-resistant prostate cancer is as follows:

[0042]

[0043] Wherein, F Score is the efficacy score; A is the weight coefficient of the gene (i.e. the LASSO coefficient); B is the mutation status of the gene (if the gene has a mutation, B = 1; if the gene has no mutation, B = 0); n is the number of the gene. In this embodiment, the gene mutation includes exon or splice site region nonsynonymous mutation or copy number variation.

[0044] 2. Performance verification of the model

[0045] The correlation between the efficacy score F Score and the treatment response PSA of the patients in the training set was evaluated by Spearman correlation analysis, and the results are shown in Table 3. Figure 2 ​As shown, the efficacy score was significantly positively correlated with PSA, the lower the efficacy score, the greater the PSA reduction after abiraterone treatment. Spearman correlation analysis R=0.73, Wilcoxon rank-sum test p=1e-07.

[0046] The performance of the abovementioned constructed abiraterone efficacy prediction model for metastatic castration-resistant prostate cancer was evaluated by ROC curve, as shown in Figure 3 As shown, the area under the ROC curve AUC=0.775. The best threshold point on the ROC curve is the basis for calculating the sensitivity (Sensitivity), specificity (Specificity) and accuracy and other indicators. In this embodiment, the best threshold is selected by the Youden index. The Youden index, also known as the correct index, is the sum of sensitivity and specificity minus 1. Through calculation, the best threshold of the model efficacy score F Score is-0.151. According to the best threshold, the training set patients are divided into a low score group (F Score<-0.151) and a high score group (F Score>-0.151), and the PSA difference of the low score group and the high score group is statistically analyzed and compared, and the difference in treatment response between the two groups is verified by box plot comparison, as shown in Figure 4 , Table 3, the PSA level of the low score group after abiraterone treatment is significantly lower than that of the high score group, confirming that the low score group has effective response to abiraterone treatment, and the high score group has no effective response to abiraterone treatment. Wilcoxon rank-sum test p=2.8e-4.

[0047] Table 3 Abiraterone treatment response of training set patients and efficacy score predicted by the model

[0048]

[0049]

[0050] Prostate-specific antigen (PSA) reduction is an important indicator for evaluating the efficacy of treatment of prostate cancer at all stages, and a decrease in PSA by more than 50% compared with the baseline is a commonly used standard for effective treatment. Take the 12-week PSA reduction of 50% after abiraterone treatment as the threshold, and divide the training set patients into two groups (PSA<0.5, PSA>0.5), as shown in Figure 5 As shown, the efficacy score F Score of the effective abiraterone treatment group (PSA<0.5) is significantly lower than that of the ineffective abiraterone treatment group (PSA>0.5), indicating that the model constructed in this embodiment can effectively predict the efficacy of abiraterone treatment for mCRPC patients.

[0051] The above model was tested for its ability to predict the efficacy of abiraterone in the validation set: the F Score of each sample in the validation set was calculated using the above calculation formula of the model, and the results showed that the F Score of the validation set was significantly correlated with the change in PSA, see Figure 6 Spearman correlation analysis R = 0.52, Wilcoxon rank sum test p = 2.4e-2; the prediction performance of the model was evaluated by ROC curve, and the area under the ROC curve AUC = 0.8624 (see Figure 7 ), which shows that the model constructed in this embodiment can effectively predict the efficacy of abiraterone in the validation set.

[0052] The above embodiments are only feasible or preferred embodiments of the present application, which are used to illustrate the present application, and are not used to limit the scope of the patent application. Therefore, any equivalent changes and modifications made in the scope of the patent application of the present application should belong to the scope covered by the present patent.

Claims

1. A combination of gene markers for predicting the efficacy of abiraterone in metastatic castration-resistant prostate cancer, characterized in that, The 16 genes include: HDAC2, TP53, GEN1, FOXA1, ATR, MUTYH, SPOP, BRCA2, ATM, PALB2, RB1, CDK12, BARD1, PTEN, NCOR2, MLH1.

2. A kit characterized in that, The reagent for detecting the combination of the gene markers in claim 1 is included.

3. A prognostic model for the prediction of the abiraterone treatment efficacy in metastatic castration-resistant prostate cancer based on the combination of genetic markers according to claim 1, characterized in that, The model comprises the following calculation formula: Wherein, F Score is the efficacy score; A is the weight coefficient of the gene; B is the mutation of the gene, B=1 if the gene has mutation, and B=0 if the gene has no mutation; n is the number of the gene.

4. The model of claim 3, wherein, The mutation includes non-synonymous mutation or copy number variation in the exon or splicing site region of the gene.

5. The model of claim 3, wherein, The calculation formula is: F Score = (1.45550198 x B BARD1 ) + (1.44867163 x B CDK12 ) + (1.43994884 x B TP53 ) + (0.89126422 ×B MLH1 )+(0.79292448×B HDAC2 )+(0.22807960×B SPOP )+(0.21785647×B RB1 )+(-0.06696785× B BRCA2 )+(-0.31043467×B ATR )+(-0.31350703×B PALB2 )+(-0.33076422×B NCOR2 )+(-0.35784534× B ATM )+(-0.41346502×B PTEN )+(-0.50077044×B FOXA1 )+(-0.57777498×B MUTYH )+(-0.78258597× B GEN1 )。 6. A method of constructing a model according to any one of claims 3 to 5, characterised in that, The method comprises the following steps: 1) obtaining PSA data and gene detection data of mCRPC patients receiving abiraterone treatment after surgery; 2) screening genes related to prostate cancer as candidate genes; 3) inputting the PSA data and mutation information of all candidate genes of each sample into the training set patient sample, and performing model training; 4) obtaining genes related to abiraterone efficacy and their corresponding weight coefficients, and establishing a metastatic castration-resistant prostate cancer abiraterone efficacy prediction model.

7. The method of claim 6, wherein, The candidate genes include the following 50 genes: AR, MUTYH, ATM FANCA, PMS2, BARD1, NBN, ATR, FANCD2, POLD1, BRIP1, NCOR1, BRCA1, HDAC2, POLE, CDH1, NCOR2, BRCA2, MLH1, RAD51, EPCAM, PTEN, CDK12, MLH3, RAD51B, ESR1, SPOP, CHEK2, MRE11A, RAD51C, FAM175A, STK11, ERCC2, MSH2, XRCC4, FOXA1, ZBTB16, ERCC3, MSH6, RAD50, GEN1, RB1, ERCC4, PALB2, RAD51D, IDH1, TP53, ERCC5, PMS1, HOXB13.

8. The method of claim 6, wherein, In step 3), the model training adopts LASSO regression method.

Images