CN114613423A - Biomarker for predicting curative effect of diffuse large B cell lymphoma chemotherapy - Google Patents

Abstract

The invention relates to a biomarker for predicting the curative effect of diffuse large B cell lymphoma chemotherapy, which comprises the combination of ARHGEF12, THAP3, SMYD3, OR2G2, ALKBH3, RNASEH2C, ZNF280D, SLC5A11, CTDP1, GPR15, GOLGB1, FBXL4 and LMBR 1. The biomarker provided by the invention can effectively predict the treatment effect of diffuse large B cell lymphoma chemotherapy, the sensitivity and specificity respectively reach 0.82 and 0.75, and the AUC value is 0.78. The biomarker is used for predicting the curative effect, has the advantages of safety, no wound, easy acquisition of samples, high accuracy, convenient operation and the like, and provides accurate judgment for the evaluation of the curative effect of the diffuse large B cell lymphoma chemotherapy.

Description

Biomarker for predicting curative effect of diffuse large B cell lymphoma chemotherapy

Technical Field

The invention relates to the technical field of biology, in particular to a biomarker for predicting the curative effect of diffuse large B cell lymphoma chemotherapy.

Background

Diffuse large B-cell lymphoma (DLBCL) is the major type of aggressive lymphoid tissue tumor, accounting for approximately 30% of non-hodgkin's lymphoma. Although most of the patients with DLBCL are old, the disease is found in all ages. The overall survival of DLBCL patients was significantly improved since rituximab (R) combined with cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP) chemotherapy regimen (R-CHOP) 10 years ago.

However, 30% to 50% of patients are not sensitive to this standard of care and the current methods do not accurately or efficiently predict the therapeutic effect of R-CHOP prior to treatment. Currently, Positron Emission Tomography (PET) -CT is the gold standard for evaluating the efficacy of different treatment regimens on DLBCL. However, it is usually used after treatment, and therefore the treatment effect cannot be predicted in advance. The International Prognostic Index (IPI) is the currently adopted method for assessing the major prognostic risk of DLBCL, especially in high risk patients, and is useful in R-CHOP chemotherapy. However, the accuracy of predicting the therapeutic effect of R-CHOP on DLBCL patients by IPI is low. Therefore, it is necessary to find an accurate method to predict in advance the therapeutic effect of the R-CHOP regimen of DLBCL.

Disclosure of Invention

The biomarker provided by the invention can effectively predict the curative effect of the diffuse large B cell lymphoma chemotherapy, the sensitivity and the specificity respectively reach 0.82 and 0.75, and the AUC value is 0.78. The biomarker is used for predicting the curative effect, has the advantages of safety, no wound, easy acquisition of samples, high accuracy, convenient operation and the like, and provides accurate judgment for the evaluation of the curative effect of the diffuse large B cell lymphoma chemotherapy.

Therefore, the invention adopts the following technical scheme.

A first aspect of the invention provides a biomarker comprising a combination of ARHGEF12, THAP3, SMYD3, OR2G2, ALKBH3, RNASEH2C, ZNF280D, SLC5a11, CTDP1, GPR15, GOLGB1, FBXL4 and LMBR 1.

In a second aspect, the invention provides the use of the biomarker in the manufacture of a product for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, said use comprising the use for constructing a model for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma.

Further, the chemotherapy regimen is: rituximab (R) is combined with a chemotherapeutic regimen of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP). This chemotherapeutic regimen is commonly referred to in the art as R-CHOP.

In a third aspect, the present invention provides a model for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, wherein the input variable of the model is the content of the biomarker according to the present invention.

Further, the method for measuring the content of the biomarker is 5hmC high-throughput detection; in a specific embodiment, the assay is 5 hmC-Seal.

In a fourth aspect, the present invention provides a method for constructing a model for predicting the therapeutic effect of diffuse large B-cell lymphoma chemotherapy, comprising the following steps:

(1) respectively detecting samples from a plurality of patients with the diffuse large B cell lymphoma subjected to chemotherapy to obtain 5hmC sequencing data of DNA;

(2) sequentially carrying out first filtration, screening and second filtration on the sequencing data obtained in the step (1) to obtain a biomarker for predicting the curative effect of the diffuse large B cell lymphoma chemotherapy;

the first filtering includes: removing spike information only occurring in 10 samples or less;

the screening comprises the following steps: comparing sequencing data from different samples by using DESeq2 software, reserving a 5hmC peak region with the reads number larger than 50, and finding out a difference biomarker of 5hmC up-down regulation according to FoldChage > being 0.5 and pvalue being less than 0.01;

the second filtering includes: for the differential biomarkers obtained by said screening, filtration was performed using the recursive feature elimination algorithm (RFECV) in Scikit-leran to obtain biomarkers for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma: ARHGEF12, THAP3, SMYD3, OR2G2, ALKBH3, RNASEH2C, ZNF280D, SLC5a11, CTDP1, GPR15, GOLGB1, FBXL4, and LMBR 1;

(3) inputting the data of the biomarkers obtained in the step (2) into a machine learning model, training the model, storing the trained model, and obtaining a model for predicting the curative effect of the diffuse large B cell lymphoma chemotherapy.

Further, the sample is plasma; in a specific embodiment, the sample is peripheral blood.

Further, the patients with chemotherapy-induced diffuse large B-cell lymphoma include patients evaluated for Complete Remission (CR), Partial Remission (PR), Stable Disease (SD), and Progressive Disease (PD) by treatment effect.

Further, in the second filtering, the parameters used are: logistic regressoncv (class _ weight: ' balanced ', cv ═ 2, max _ iter ═ 1000), rating ═ accuracy ')

Further, in step (3), the machine learning model includes: CV model (LR) was regressed by training logistic.

Further, the chemotherapy regimen is: rituximab (R) is combined with a chemotherapeutic regimen of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP).

The invention discloses a method for predicting the chemotherapy curative effect of diffuse large B cell lymphoma, which is characterized in that the method utilizes the logistic regression analysis to carry out data mining, discusses factors related to the chemotherapy curative effect of the diffuse large B cell lymphoma, and predicts the chemotherapy curative effect of the diffuse large B cell lymphoma according to the factors. Through logistic regression analysis, the weight of independent variable can be obtained, so that the factors which are closely related to the curative effect of diffuse large B cell lymphoma chemotherapy can be known. Meanwhile, the chemotherapy curative effect of diffuse large B cell lymphoma of one person can be predicted according to the weight value and the factor. For the model provided by the invention, the performance of the model is evaluated by adopting Receiver Operating Characteristic (ROC) analysis, the area under the curve (AUC) is calculated by using skleran, the sensitivity and the specificity of the model respectively reach 0.82 and 0.75, and the ROC value is 0.78.

In a fifth aspect, the present invention provides a product for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, said product predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma based on said model for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma.

DNA 5-methylcytosine (5mCs) is an important epigenetic feature that plays an important role in gene expression and tumorigenesis development. According to the invention, through a high-throughput sequencing technology, the 5-hydroxymethylcytosine biomarker for predicting the chemotherapy effect of diffuse large B-cell lymphoma is found, and the R-CHOP treatment effect of a patient with diffuse large B-cell lymphoma is effectively predicted in advance.

Compared with the prior art, the invention has the following beneficial effects:

(1) the invention provides a biomarker for predicting the chemotherapy curative effect of diffuse large B cell lymphoma, which can effectively predict the chemotherapy curative effect of a patient with diffuse large B cell lymphoma in advance.

(2) The invention also provides a model for predicting the curative effect of the diffuse large B cell lymphoma chemotherapy, the sensitivity and the specificity of the model respectively reach 0.82 and 0.75, the AUC value is 0.78, and the model has the advantages of strong specificity and high sensitivity. By applying the biomarker and/or the model, the safe, noninvasive and high-accuracy prediction of the curative effect of the diffuse large B cell lymphoma of a patient can be realized.

(3) According to the biomarker and the model provided by the invention, the sample for predicting the chemotherapeutic effect is peripheral blood, and is easy to obtain; the prediction method is based on high-throughput sequencing, and the detection efficiency is high; the prediction model is reliable, and the accuracy of the prediction result is high.

Drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. In the drawings:

FIG. 1: the prediction accuracy of a single marker in the 13 feature 5hmC markers in the training set;

FIG. 2: and verifying the prediction accuracy of a single marker in the 13 characteristic 5hmC markers in the group. Of these, ARHGEF12(AUC 0.76) and ZNF280D (AUC 0.76) showed the highest prediction accuracy.

FIG. 3: feature selection for 5hmC differential markers. The recursive feature selection algorithm selects 13 features as the minimum feature number to obtain the best cross-validation score. The x-axis is the number of features selected and the y-axis is the cross-validation score for model performance.

FIG. 4: receiver Operating Characteristics (ROC) curves for the prediction models constructed using 135 hmC feature Markers in the training and validation cohort. The true positive rate (sensitivity) and false positive rate (1-specificity) were plotted as a function. In the training set, the prediction model constructed by the 135 hmC feature Markers can effectively predict responders and non-responders (AUC 1.00) to the R-CHOP treatment, and in the verification set, the prediction model based on the 135 hmC feature Markers can also effectively predict responders and non-responders (AUC 0.78) to the R-CHOP treatment.

FIG. 5: confusion matrices for predictive model performance in the validation set (22 patients with treatment effect and 8 patients with treatment non-effect). The results show that: in 22 cases of treatment effective patients, the treatment effectiveness of 18 patients is predicted based on 13 prediction models of 5hmC characteristic Markers, and the sensitivity reaches 0.82; of the 8 treatment-ineffectual patients, 6 were predicted to be ineffectual based on a predictive model of 135 hmC feature Markers, with a specificity of 0.75.

Detailed Description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

The invention provides a model for predicting the curative effect of diffuse large B cell lymphoma chemotherapy, and relevant parameters of 135 hmC characteristic biomarkers in the model are shown in a table 1:

TABLE 1

Markers	GeneID	Coefficients	SE	z.value	p.value
						Intercept		-5.5704	0.867	2.652	<0.01
chr1_6721489_6721898	THAP3	0.7712	0.145	1.865	<0.05
						chr1_246290825_246291238	SMYD3	0.39	0.149	1.955	<0.05
chr1_247755954_247756505	OR2G2	3.1779	0.108	1.344	<0.05
						chr11_43905400_43905804	ALKBH3	3.3423	0.128	1.306	<0.05
chr11_65511519_65512429	RNASEH2C	1.5211	0.072	2.061	<0.05
						chr11_120211662_120212234	ARHGEF12	-3.8797	0.115	-3.225	<0.001
chr15_56982146_56982638	ZNF280D	-1.2266	0.177	-3.250	<0.001
						chr16_24916341_24916920	SLC5A11	0.6683	0.076	0.149	<0.05
chr18_77500908_77501376	CTDP1	-2.573	0.103	-3.182	<0.001
						chr3_98270705_98271079	GPR15	0.1052	0.167	2.348	<0.05
chr3_121430838_121431239	GOLGB1	0.8526	0.178	2.982	<0.01
						chr6_99461404_99461922	FBXL4	1.7188	0.101	2.165	<0.05
chr7_156700537_156701031	LMBR1	1.0942	0.078	0.579	<0.05

In the examples of the present invention, samples were sequenced using 5hmC-Seal high throughput sequencing, and the 5hmC-Seal high throughput sequencing used in the following examples was explained as follows:

5hmC-Seal is a high throughput sequencing method based on 5 hmC. The method is based on the improved chemical glycosylation mark and combines the second generation high-throughput sequencing technology to obtain the distribution information of 5hmC on the genome DNA.

Due to the high sensitivity of the chemical labeling method, the input DNA can be as low as 1-10ng, the DNA can be fragmented genome DNA or small fragments such as cfDNA, filling up is carried out AT two ends of the DNA fragment according to the requirement of second-generation sequencing, then an A tail is connected AT the 3' end, and a sequencing Y-shaped joint is connected AT two ends of each DNA fragment by AT specificity, wherein the sequencing Y-shaped joint comprises index information for distinguishing samples, amplification primers and other sequences. Then, entering a labeling step of 5hmC, firstly adding UDP-6-N3-Glc, reacting 5hmC on DNA to obtain N3-5ghmC under a specific condition, then adding DBCO-PEG4-Biotin, connecting N3-5ghmC with Biotin, finally screening to obtain all DNA fragments containing 5hmC sites through efficient specific binding of Biotin-magnetic beads, and completing construction of a DNA library based on 5hmC after PCR amplification and purification; DNA band size and distribution of each sample were analyzed by Fragment Analyzer quality control, and after library was accurately quantified by qPCR, high throughput sequencing was performed using Nextseq500 sequencer from Illumina to obtain base sequences of all DNA fragments in the library.

Example 1

86 patients with R-CHOP chemotherapy diffuse large B cell lymphoma, including patients with Complete Remission (CR), Partial Remission (PR), Stable Disease (SD) and Progressive Disease (PD) assessed by treatment efficacy. Peripheral blood samples (3-4mL) of the above 86 patients were taken, cfDNA was extracted from the plasma and subjected to 5hmC-Seal high-throughput sequencing, the sequencing throughput of each sample was 1.5Gb, and the sequencing band size was 38 bp. The method comprises the following specific steps:

(1) collecting samples:

at the time of confirmed diagnosis of patient admission (before R-CHOP standard chemotherapy), a total of 8mL of blood was collected from 86 patients with diffuse large B cell lymphoma, and plasma samples were collected within 24 hours after plasma separation. First centrifugation was performed at 1350g and 4 ℃ for 12min, and the pale yellow supernatant was carefully removed and transferred to a 2ml DNA free sterile EP tube to avoid contaminating the buffy coat. Secondly, carrying out secondary centrifugation at 1350g and 4 ℃ for 5 min; the supernatant was carefully removed, transferred to a 2mL DNase free sterile EP tube, and the remaining erythrocytes were removed again. Thirdly, centrifuging at 13500g and 4 ℃ for 5 min; the supernatant was carefully removed and transferred to a 2mL DNase free sterile centrifuge tube to obtain about 4mL of plasma, which was labeled and then frozen in a-80 ℃ freezer for use.

(2) Extracting plasma DNA:

8-10ng of Plasma cfDNA was extracted from each of the Plasma samples from 86 patients with diffuse large B-cell lymphoma using the Quick-cfDNA Serum & Plasma Kit (ZYMO) Kit.

(3) End-filling and ligation of cfDNA to sequencing adapters:

a reaction mixture (total volume 24uL) containing 20uL cfDNA, 2.8uL End Repair & A-labeling Buffer, and 1.2uL End Repair & A-labeling Enzyme mix was prepared according to the KAPA Hyperplus Library Preparation Kit (96) instructions, and was incubated at 20 for 30 minutes and then at 65 for 30 minutes. Next, a Ligation reaction mixture of 2uL of nucleic acid free water,12uL of Ligation Buffer and 4uL of DNA Ligase was placed in a 1.5mL low adsorption EP tube. To 18uL ligation reaction mixture was added 2uL of sequencing adapter, mixed, added to 24uL reaction samples, and heated at 20 ℃ for 4 hours. The reaction product was purified using a DNA Clean & Concentrator 5(ZYMO) purification kit, and eluted with 20uL of an elution buffer to obtain a final DNA ligation sample.

(4)5hmC marker:

a total volume of 4uL of labeling reaction mix, 1uL of 50uM UDP-N3-Glu, 1uL of β GT enzyme (NEB), and 2uL of HEPES buffer (pH 8.0, 50mM final concentration), was prepared and 4uL of the labeling reaction mix was added to 20uL of the DNA ligation sample. The mixture was incubated at 37 ℃ for 2 hours in a water bath. The mixture was taken out, and the reaction product was purified with a DNA Clean & Concentrator 5(ZYMO) purification kit to obtain purified 30uL DNA. Then, 1uL of DBCO-PEG4-biotin (click Chemistry tools) was added to the above purified 30uL of DNA, and the reaction product was purified by DNA Clean & Concentrator 5(ZYMO) purification kit after 1 hour of water bath at 37 ℃ to obtain 30uL of purified labeled product.

(5) Enrichment of 5 hmC:

first, bound beads were prepared as follows: 2.5uL of Dynabeads (Invitrogen) were removed and 100uL of wash buffer (5mM Tris (pH 7.5), lM NaCl and 0.02% Tween20) was added, vortexed, placed on a 1.5mL magnetic rack, washed repeatedly with 100uL of wash buffer for 3 times, and finally 32uL of binding buffer (10mM Tris (pH 7.5), 2M NaCl and 0.04% Tween20) was added and vortexed. Then, the purified labeled product obtained in the above step was added to the mixture of magnetic beads and mixed in a rotary mixer for 30min to allow sufficient binding. Finally, the beads were washed 5 times with 100uL of washing solution (5mM Tris (pH 7.5), lM NaCl and 0.02% Tween20), the supernatant was centrifuged off, and 23.8uL RNase-Free water was added.

(6) And (3) PCR amplification:

to the final system of the above steps, 25uL of 2X PCR master mix and 1.25uL of PCR primers (total volume 50u L) were added and amplification was performed according to the following PCR reaction cycle temperatures and conditions:

and (3) purifying the amplification product by using AmpureXP beads to obtain a final 5hmC library, and determining the library concentration by using the Qubit 3.0.

(6) High-throughput sequencing after quality control of the 5hmC library:

performing quality control on the obtained 5hmC library by a Fragment Analyzer (TM) full-automatic capillary electrophoresis system, and determining the size of a DNA Fragment in the library and whether the library contains impurities; concentration determination was performed by qPCR. The sequencing libraries that passed the quality control were mixed at the same concentration and sequenced with 11 lumine NextSeq500 using a sequencing Kit of High Output Kit v2(75cycles), with a sequencing throughput of 1.5Gb for each sample and a sequencing band size of 75 bp.

(7) Raw sequencing data alignment:

each original sequencing FASTQ data is firstly trimmed by low-quality data through Trimmomatic software, then is aligned to human genome hg19 through Bowtie2 software, and is identified by reads number peaks containing 5hmC through MACS software, wherein the parameters are as follows: effective genome size 2.72e + 09; tag size 38; band width is 100; model fold is 10; p value cutoff is 1.00 e-05.

The 86 samples were randomly split into a training group (training cohort) and a validation group (validation cohort) at a ratio of 2: 1. Comparing the peak information of plasma cfDNA of different patients in the training set (56 cases) and performing filtering, provided that each peak appears in at least 10 samples; comparing DNA sequencing data from different samples with DEseq2 software, finding 5hmC peak area with reads number greater than 50, finding difference biomarker of 5hmC up-down regulation according to | log2FoldChange | > 0.5 and pvalue < 0.01. The differential biomarkers were then further filtered using the recursive feature elimination algorithm (RFECV) in Scikit-Learn, and the feature biomarkers were screened (using the parameters: estimator ═ logistic regression cv (class _ weight ═ balanced ', cv ═ 2, max _ iter ═ 1000), and sequencing ═ accuracycacy'). Further, model construction is carried out by using the screened feature biomarkers, a logistic regression CV model (LR) is trained, the performance of the model is analyzed and evaluated by adopting the operating characteristics (ROC) of a subject, the area under the curve (AUC), the sensitivity and the specificity are calculated by using skleran, wherein the prediction accuracy of a single marker in the 5hmC markers with 13 features in a training group can be shown in figure 1, and the prediction accuracy of a single marker in the 5hmC markers with 13 features in a verification group can be shown in figure 2. Finally, after 100 times of training, it is found that the prediction model constructed by the 5hmC markers with 13 characteristics can reach the optimal cross validation score, and the result is shown in FIG. 3. The constructed LR prediction model was used to predict the treatment outcome for patients in the validation group, see fig. 4. The constructed LR prediction model was used in example 2 to predict the treatment outcome of patients in the validation group.

Example 2

Peripheral blood of patients in the validation group (30 cases, wherein CR 14 case, PR 8 case, SD 3 case, PD 5 case) in example 1 was taken for sample collection/cfDNA extraction/library construction/library sequencing according to the experimental method in example 1, the sequencing data was processed and analyzed, and 5hmC enrichment sites were screened. Then, efficacy prediction is performed by the prediction model obtained in example 1 according to the selected 5 hmC-enriched site. The prediction results are shown in fig. 5, and the results show that 18 of 22 PRCR patients are predicted to be PRCR, with a sensitivity of 0.82; of the 8 PDSD patients, 6 predicted PDSD with a specificity of 0.75.

Therefore, the prediction model constructed based on the 13 characteristic biomarkers of 5hmC provided by the invention can effectively predict the treatment effect of R-CHOP chemotherapy of patients with diffuse large B-cell lymphoma in advance.

The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (10)

1. A biomarker for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, wherein the biomarker comprises a combination of ARHGEF12, THAP3, SMYD3, OR2G2, ALKBH3, RNASEH2C, ZNF280D, SLC5a11, CTDP1, GPR15, GOLGB1, FBXL4 and LMBR 1.

2. Use of a biomarker according to claim 1 in the manufacture of a product for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, said use comprising constructing a model for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma.

3. Use of a biomarker according to claim 2, wherein the chemotherapy regimen is: rituximab in combination with a chemotherapeutic regimen of cyclophosphamide, doxorubicin, vincristine and prednisone.

4. A product based on a model for predicting the efficacy of chemotherapy for diffuse large B-cell lymphoma, wherein the input variables for said model are the levels of the biomarkers of claim 1.

5. The product of claim 4, wherein the biomarker level is determined by a 5hmC high throughput assay; preferably 5 hmC-Seal.

6. The product of claim 4, wherein the model is constructed by a method comprising the steps of:

(1) respectively detecting samples from a plurality of patients with the diffuse large B cell lymphoma subjected to chemotherapy to obtain 5hmC sequencing data of DNA;

(2) sequentially carrying out first filtration, screening and second filtration on the sequencing data obtained in the step (1) to obtain a biomarker for predicting the curative effect of the diffuse large B cell lymphoma chemotherapy;

the first filtering includes: removing peak information that appears only in 10 samples or less;

the screening comprises the following steps: comparing sequencing data from different samples by using DESeq2 software, reserving a 5hmC peak region with the reads number larger than 50, and obtaining a differential biomarker of 5hmC up-and-down regulation according to the conditions that log2FoldChange | > (0.5) and pvalue (0.01);

the second filtering includes: for the differential biomarkers obtained by said screening, filtering was performed using the recursive feature elimination algorithm in Scikit-leran to obtain biomarkers for predicting the efficacy of chemotherapy on diffuse large B-cell lymphoma: ARHGEF12, THAP3, SMYD3, OR2G2, ALKBH3, RNASEH2C, ZNF280D, SLC5a11, CTDP1, GPR15, GOLGB1, FBXL4, and LMBR 1;

(3) inputting the data of the biomarkers obtained in the step (2) into a machine learning model, training the model, storing the trained model, and obtaining a model for predicting the curative effect of the diffuse large B cell lymphoma chemotherapy.

7. The product of claim 6, wherein the sample is plasma; peripheral blood is preferred.

8. The product of claim 6, wherein said chemotherapy-treated diffuse large B-cell lymphoma patients comprise patients evaluated for complete remission, partial remission, stable disease and disease progression.

9. The product according to claim 6, characterized in that in the second filtration the parameters used are: the identifier is logisticregressoncv (class _ weight is 'balanced', cv is 2, max _ iter is 1000), and the calibrating is 'accuracy').

10. The product of claim 6, wherein in step (3), the machine learning model comprises: the CV model was regressed by training logistic.

Images