Establishment method of antituberculous drug-induced liver injury prediction model and prediction system thereof

Abstract

The invention provides an establishment method of an antituberculous drug-induced liver injury prediction model and a prediction system thereof, and belongs to the technical field of biomedicine. Aiming at the influence of multiple factors on antituberculosis drug-induced liver injury (ADLI), the prediction effects of different types of variables are integrated by utilizing a machine learning method, meaningful potential prediction variables are obtained after screening, the method is not limited to existing possible indexes, meanwhile, new potential variables can be found, and the method has a good application prospect. The occurrence of the anti-tuberculosis drug-induced liver injury can be accurately predicted. Therefore, the ADLI occurrence risk of a pulmonary tuberculosis patient is evaluated through a machine learning method before the antituberculosis drug is applied, that is, the ADLI occurrence risk of different patients can be preliminarily known before treatment by utilizing the antituberculosis drug-induced liver injury prediction model established by the invention, and information is provided for personalized treatment of the patients.

Description

Technical Field

[0001] The invention belongs to the field of biomedical technology, and specifically relates to a method for establishing a prediction model for anti-tuberculosis drug-induced liver injury and a prediction system thereof. Background Art

[0002] Tuberculosis is a chronic infectious disease caused by Mycobacterium tuberculosis. Its pathogens can remain dormant in the human body for a long time, bringing a serious disease burden to the world. During the anti-tuberculosis treatment, a variety of adverse reactions may be induced, among which anti-tuberculosis drug-induced liver injury (ADLI) is particularly harmful and is one of the most serious adverse reactions. Severe cases may lead to liver failure or even death.

[0003] Liver damage caused by anti-tuberculosis drugs seriously limits the treatment of tuberculosis, leading to the emergence of more drug-resistant tuberculosis, which in turn limits the effective control of tuberculosis. Early identification of anti-tuberculosis drug-induced liver damage is of great significance for controlling the spread of tuberculosis, shortening the treatment time, reducing drug resistance and reducing the burden on the health sector. At present, ADLI lacks early biomarkers for predicting treatment effects and the occurrence of damage. Individual differences in drug metabolism can affect the treatment effect and cause adverse reactions. Therefore, it has a significant impact on the selection and adjustment of tuberculosis treatment plans.

[0004] Studies have shown that the plasma concentration of anti-tuberculosis drugs is related to hepatotoxicity, and the drug metabolism process involves multiple reaction types and changes in drug metabolizing enzymes. The rate-limiting step for drug clearance from the body is the phase I metabolic reaction. During the phase II metabolic reaction, the drug and its metabolites bind to endogenous substances and are released from the body. The liver is the main site for drug metabolism, and the cytochrome oxidase P450 family is most widely involved in metabolism. Among them, members of the cytochrome oxidase family, CYP1, 2, and 3 are closely related to the occurrence and development of ADLI. CYP3A5*3 and CYP3A4*18B gene polymorphisms are associated with the occurrence of anti-tuberculosis drug-induced liver injury; the Dra I polymorphism of the cytochrome oxidase 2E1 (Cytochrome P4502E1, CYP2E1) gene is closely related to the risk of ADLI. In the phase II metabolism process, the slow acetylation phenotype of N-deacetyltransferase 2 (NAT2) is closely related to the process of drug metabolism in the liver. There are two main types of NAT family, and NAT2 gene polymorphism presents two characteristics, namely fast acetylation and slow acetylation phenotypes. The genotyping of NAT2 and CYP2E1 genes is closely related to the occurrence of ADLI and is a candidate indicator for predicting ADLI.

[0005] It can be seen that anti-tuberculosis drug-induced liver injury is affected by multiple factors, and clinically it is impossible to simultaneously consider the impact of multiple variables on the occurrence of ADLI. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a method for establishing a prediction model for anti-tuberculosis drug-induced liver injury. By using machine learning methods, the risk of ADLI in patients with pulmonary tuberculosis is evaluated before the use of anti-tuberculosis drugs. The risk of ADLI in different patients can be preliminarily understood before treatment, providing information for personalized treatment of patients.

[0007] The present invention provides a method for establishing a prediction model of anti-tuberculosis drug-induced liver injury, comprising the following steps: extracting genomic DNA from tuberculosis patients;

[0008] Using the extracted genomic DNA as a template, GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were amplified to obtain DNA amplification products;

[0009] The DNA amplification products were analyzed for polymorphism using PCR-FLP technique, and the genotypes of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were obtained;

[0010] The genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B were used as independent variables, and the occurrence of anti-tuberculosis drug-induced liver injury after anti-tuberculosis treatment was used as the dependent variable, and a prediction model was constructed using multivariate logistic regression analysis.

[0011] Preferably, the primers for amplifying GSTM1 include a forward primer having a nucleotide sequence as shown in SEQ ID NO: 1 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO: 2;

[0012] The primers for amplifying GSTT1 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:3 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:4;

[0013] The primers used to amplify rs1799929 and rs1799931 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:7 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:8;

[0014] The primers used to amplify rs2031920 include a forward primer having a nucleotide sequence as shown in SEQ ID NO: 9 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO: 10;

[0015] The primers for amplifying CYP3A4*18B include a forward primer having a nucleotide sequence as shown in SEQ ID NO:11 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:12.

[0016] Preferably, the reaction procedure for amplifying GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B is 94°C for 3 min; 94°C for 60 s, 56°C for 45 s, 72°C for 60 s, 35 cycles; 72°C for 7 min.

[0017] Preferably, the PCR-FLP technique is used to perform polymorphism analysis on the DNA amplification products, including restriction endonuclease digestion of the amplified products of rs1799929, rs1799931, rs2031920 and CYP3A4*18B:

[0018] The amplified product of the rs1799929 polymorphic site was digested with Kpn I-HF endonuclease;

[0019] The amplified product of the rs1799931 polymorphic site was digested with BamHI-HF endonuclease;

[0020] The amplified products of rs2031920 and CYP3A4*18B polymorphisms were digested with Rsa I endonuclease.

[0021] Preferably, the genotype of GSTM1 includes a deletion type and a non-deletion type; the deletion type has an amplification product fragment length of 268 bp; the non-deletion type has an amplification product fragment length of 268 bp and 219 bp;

[0022] The genotype of GSTT1 includes a deletion type and a non-deletion type; the deletion type has an amplification product fragment length of 268 bp; the non-deletion type has an amplification product fragment length of 268 bp and 459 bp;

[0023] The genotype of rs1799929 includes CC genotype, CT genotype and TT genotype; the enzyme digestion product of the CC genotype includes 60bp and 437bp; the enzyme digestion product of the CT genotype includes 60bp, 437bp and 497bp; the enzyme digestion product of the TT genotype includes 497bp;

[0024] The genotype of rs1799931 includes GG genotype, GA genotype and AA genotype; the enzyme digestion product of the GG genotype includes 61bp and 436bp; the enzyme digestion product of the GA genotype includes 61bp, 436bp and 497bp; the enzyme digestion product of the AA genotype includes 497bp;

[0025] The enzyme digestion products of rs2031920 include GG genotype, GA genotype and AA genotype; the enzyme digestion products of the GG genotype include 155bp and 219bp; the enzyme digestion products of the GA genotype include 155bp, 219bp and 374bp; the enzyme digestion products of the AA genotype include 374bp;

[0026] The enzyme cleavage products of CYP3A4*18B include GG genotype, GA genotype and AA genotype; the enzyme cleavage products of the GG genotype include 70bp and 205bp; the enzyme cleavage products of the GA genotype include 70bp, 205bp and 275bp; the enzyme cleavage products of the AA genotype include 275bp.

[0027] Preferably, before constructing the prediction model, the method further includes performing a balanced test on the distribution of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B polymorphic sites in the case group and the control group and analyzing the polymorphic sites using a univariate logistic regression model to exclude risk factors and determine that the polymorphic sites are significantly correlated with liver damage caused by drug-resistant pulmonary tuberculosis;

[0028] The samples of the case group were from tuberculosis patients who developed anti-tuberculosis drug-induced liver injury;

[0029] The samples of the control group were from tuberculosis patients who did not suffer from anti-tuberculosis drug-induced liver injury.

[0030] Preferably, after obtaining the prediction model, the method further comprises performing model verification on the prediction model;

[0031] The method for model verification is to use the verification set data to calculate the AUC of the prediction model: when the AUC is greater than 0.7, the prediction model is considered to have moderate diagnostic value; when the AUC exceeds 0.8, it indicates that the prediction model has high diagnostic value.

[0032] The present invention provides a prediction model for anti-tuberculosis drug-induced liver injury obtained by the establishment method.

[0033] The present invention provides a prediction system for anti-tuberculosis drug-induced liver injury, comprising a data preprocessing module, a model processing module and a prediction result output module;

[0034] The data preprocessing module is used to obtain the medical record data to be tested of the target object, wherein the medical record data to be tested includes the genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B of the target object;

[0035] The model processing module is used to calculate and predict the result of anti-tuberculosis drug-induced liver injury based on the medical record data to be tested and the anti-tuberculosis drug-induced liver injury prediction model;

[0036] The prediction result output module is used to output the prediction result of anti-tuberculosis drug-induced liver injury.

[0037] The present invention provides a device equipped with the anti-tuberculosis drug-induced liver injury prediction system.

[0038] The present invention provides a method for establishing a prediction model of anti-tuberculosis drug-induced liver injury. The present invention combines GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B multiple variables, and can identify the subtle relationship between risk factors and results under the application of machine learning methods, and find new patterns between variables, and generate prediction relationships. In a complex disease such as liver disease, anti-tuberculosis drug-induced liver injury is affected by multiple factors. The prediction effects of different types of variables are integrated by machine learning methods, and after screening, meaningful potential predictive variables are obtained, which are not limited to existing possible indicators, and new potential variables can be found at the same time, and the occurrence of anti-tuberculosis drug-induced liver injury is more accurately predicted. It can be seen that the risk of ADLI in patients with pulmonary tuberculosis is evaluated before the application of anti-tuberculosis drugs by machine learning methods. The prediction model of anti-tuberculosis drug-induced liver injury obtained by the establishment method provided by the present invention can preliminarily understand the risk of ADLI in different patients before treatment, and provide information for personalized treatment of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is the ROC curve of the prediction model of anti-tuberculosis drug-induced liver injury in the validation group. DETAILED DESCRIPTION

[0040] The present invention provides a method for establishing a prediction model of anti-tuberculosis drug-induced liver injury, comprising the following steps: extracting genomic DNA from tuberculosis patients;

[0041] Using the extracted genomic DNA as a template, GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were amplified to obtain DNA amplification products;

[0042] The DNA amplification products were analyzed for polymorphism using PCR-FLP technique, and the genotypes of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were obtained;

[0043] A prediction model was constructed using multivariate logistic regression analysis, with the polymorphic sites of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B as independent variables and the occurrence of anti-tuberculosis drug-induced liver injury after anti-tuberculosis treatment as the dependent variable.

[0044] The invention extracts genomic DNA from tuberculosis patients.

[0045] In the present invention, the number of samples from tuberculosis patients is not less than 200, and more preferably 250 to 400. In an embodiment of the present invention, the number of samples from tuberculosis patients is preferably 275.

[0046] The present invention has no particular restrictions on the method for extracting genomic DNA from tuberculosis patients, and any method known in the art for extracting genomic DNA may be used, such as extraction using a commercial kit. In the present embodiment, the commercial kit is a small amount of whole blood genomic DNA rapid extraction kit, purchased from Beijing Jinbaite Biotechnology Co., Ltd., with a product number of Cat. NO: D011-100.

[0047] After extraction, the present invention uses the extracted genomic DNA as a template to amplify GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B respectively to obtain DNA amplification products.

[0048] In the present invention, the primers for amplifying GSTM1 preferably include a forward primer having a nucleotide sequence as shown in SEQ ID NO:1 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:2. The primers for amplifying GSTT1 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:3 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:4. The primers for amplifying rs1799929 and rs1799931 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:7 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:8. The primers for amplifying rs2031920 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:9 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:10. The primers for amplifying CYP3A4*18B include a forward primer having a nucleotide sequence as shown in SEQ ID NO:11 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:12. In an embodiment of the present invention, the above-mentioned amplification primers are commissioned to be synthesized by Sangon Biotech (Shanghai) Co., Ltd. Since GSTM1 and GSTT1 polymorphisms are expressed as gene deletions, primers for internal reference gene amplification are added to the amplification system for correction. The internal reference gene is the β-globulin gene. The primers for β-globulin gene amplification preferably include a forward primer having a nucleotide sequence as shown in SEQ ID NO:5 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:6.

[0049] In the present invention, the reaction procedure for amplifying GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B is preferably 94°C for 3 min; 94°C for 60 s, 56°C for 45 s, 72°C for 60 s, 35 cycles; 72°C for 7 min. The reaction system for amplifying GSTM1 or GSTT1 is preferably 20 μL, including 10 μL PCR MasterMix, 0.5 μL F1, 0.5 μL R1, 0.5 μL Fβ, 0.5 μL Rβ, 1 μL DNA template, and 20 μL water. The reaction system for amplifying rs1799929, rs1799931, rs2031920 and CYP3A4*18B is preferably 10 μL PCR MasterMix, 0.5 μL forward primer, 0.5 μL reverse primer, 1 μL DNA template, ddH 2 Make up to 20 μL with HO.

[0050] After obtaining the DNA amplification product, the present invention uses PCR-FLP technology to perform polymorphism analysis on the DNA amplification product to obtain the genotypes of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B.

[0051] In the present invention, the PCR-FLP technique is used to perform polymorphism analysis on the DNA amplification products, preferably including the amplification products of rs1799929, rs1799931, rs2031920 and CYP3A4*18B, and restriction endonuclease digestion: the amplification product of the rs1799929 polymorphic site is digested by Kpn I-HF endonuclease; the amplification product of the rs1799931 polymorphic site is digested by BamHI-HF endonuclease; the amplification product of the rs2031920 and CYP3A4*18B polymorphic site is digested by Rsa I endonuclease. The reaction system of endonuclease digestion is preferably endonuclease 1μl, 10×NEBuffer 5μl, DNA amplification product 1μl, ddH 2 O supplemented with 50 μl.

[0052] In the present invention, the genotype of GSTM1 preferably includes a deletion type and a non-deletion type; the deletion type preferably has an amplified product fragment length of 268bp; the non-deletion type preferably has an amplified product fragment length of 268bp and 219bp. The genotype of GSTT1 preferably includes a deletion type and a non-deletion type; the deletion type preferably has an amplified product fragment length of 268bp; the non-deletion type preferably has an amplified product fragment length of 268bp and 459bp. The genotype of rs1799929 preferably includes CC genotype, CT genotype and TT genotype; the enzyme digestion products of the CC genotype include 60bp and 437bp; the enzyme digestion products of the CT genotype include 60bp, 437bp and 497bp; the enzyme digestion products of the TT genotype include 497bp. The genotype of rs1799931 includes GG genotype, GA genotype and AA genotype; the enzyme digestion product of the GG genotype includes 61bp and 436bp; the enzyme digestion product of the GA genotype includes 61bp, 436bp and 497bp; the enzyme digestion product of the AA genotype includes 497bp. The enzyme digestion products of rs2031920 include GG genotype, GA genotype and AA genotype; the enzyme digestion product of the GG genotype includes 155bp and 219bp; the enzyme digestion product of the GA genotype includes 155bp, 219bp and 374bp; the enzyme digestion product of the AA genotype includes 374bp. The enzyme cleavage products of CYP3A4*18B include GG genotype, GA genotype and AA genotype; the enzyme cleavage products of the GG genotype include 70bp and 205bp; the enzyme cleavage products of the GA genotype include 70bp, 205bp and 275bp; the enzyme cleavage products of the AA genotype include 275bp.

[0053] In the present invention, after obtaining the genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B, it is also preferably included to perform data review, classification and database construction, so as to improve the data, modify the data type to meet the processing requirements of SPSS software, eliminate incomplete data, and ensure the integrity of the included data. SPSS is used to process univariate analysis to obtain risk factors. The model is established with a 1:1 case control, and interference from other factors except risk factors is excluded to ensure that there is no statistical difference between the case group and the control group except risk factors.

[0054] In the present invention, before constructing the prediction model, it is preferred to also include a balanced test of the distribution of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B polymorphic sites in the case group and the control group and analyze the polymorphic sites using a univariate logistic regression model, exclude risk factors, and determine that the polymorphic sites are significantly associated with liver damage caused by drug-resistant pulmonary tuberculosis. The samples of the case group are from tuberculosis patients with anti-tuberculosis drug-induced liver damage; the samples of the control group are from tuberculosis patients without anti-tuberculosis drug-induced liver damage. The results of the embodiment of the present invention show that there are significant differences in the distribution of mutations in the polymorphic sites of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B in the control group and the case group, and the univariate logistic analysis also shows that there are significant differences in the distribution of mutations in each gene site. The balance test includes a paired t-test for normally distributed quantitative data; and a univariate logistic regression analysis for data containing categorical variables. Univariate logistic regression model analysis is beneficial to exclude the influence of baseline data on it, ensuring that there is no statistical difference in baseline data between the case group and the control group. Univariate logistic regression model analysis calculates OR value and 95% CI. OR value (odds ratio), also known as odds ratio, advantage ratio, mainly refers to the ratio of the number of exposed people to the number of non-exposed people in the case group divided by the ratio of the number of exposed people to the number of non-exposed people in the control group, and is a commonly used indicator in case-control studies in epidemiological studies. The calculation method of OR value preferably refers to the ratio of the number of exposed people to the number of non-exposed people in the case group divided by the ratio of the number of exposed people to the number of non-exposed people in the control group. An OR value equal to 1 indicates that the factor does not play a role in the occurrence of the disease; an OR value greater than 1 indicates that the factor is a risk factor; an OR value less than 1 indicates that the factor is a protective factor. 95% CI means that if you use the same steps to select samples and calculate confidence intervals, then after 100 such independent processes, there is a 95% probability that the calculated interval will contain the true parameter value, that is, approximately 95 confidence intervals will contain the true value.

[0055] A prediction model was constructed using multivariate logistic regression analysis, with the polymorphic sites of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B as independent variables and the occurrence of anti-tuberculosis drug-induced liver injury after anti-tuberculosis treatment as the dependent variable.

[0056] In the present invention, the method of anti-tuberculosis treatment preferably includes the combined use of first-line drugs isoniazid, rifampicin, ethambutol and pyrazinamide for 2 months of intensive treatment, and then the use of isoniazid and rifampicin for 4 months of consolidation treatment. In the embodiment of the present invention, the included patients are all patients in the intensive treatment stage of using first-line drugs isoniazid, rifampicin, ethambutol and pyrazinamide for 2 months.

[0057] In the present invention, the diagnostic criteria for anti-tuberculosis drug-induced liver injury are preferably: ADLI is diagnosed when at least one of the following criteria exists: when there are clinical symptoms of liver injury, the increase in serum ALT or AST is greater than 3 times the upper limit of normal (ULN) or when symptoms of liver injury occur, the increase in TBIL exceeds 2 times the ULN; regardless of clinical symptoms, the ULN of serum ALT, AST or TBIL increases by 5 times. Abnormal liver function judgment criteria: after medication, serum AST or ALT is higher than the normal value, lower than 3 times the normal value, and does not meet the injury criteria, it is judged as abnormal liver function.

[0058] In the present invention, after obtaining the prediction model, it is preferred to further include model validation of the prediction model. The method of model validation is to use the validation set data to perform AUC measurement on the prediction model: when AUC is greater than 0.7, the prediction model is considered to have moderate diagnostic value; when AUC exceeds 0.8, it indicates that the prediction model has high diagnostic value. The number of samples in the validation set is preferably 1 / 3 to 1 / 4 of the number of samples when modeling.

[0059] The present invention provides a prediction model for anti-tuberculosis drug-induced liver injury obtained by the establishment method.

[0060] In an embodiment of the present invention, the prediction model constructed by the present invention is verified, and the AUC value is 0.8076, indicating that the prediction ability of the prediction model constructed by the present invention is highly accurate. The prediction probability cutoff value of the prediction model is 0.9809. The judgment method for prediction based on the prediction model is preferably that the cases with a prediction probability cutoff value ≥ 0.9809 are predicted to have anti-tuberculosis drug-induced liver injury, and the cases with a prediction probability cutoff value < 0.9809 are predicted to have not had anti-tuberculosis drug-induced liver injury. The sensitivity of the prediction model is 0.8607, and the specificity of the prediction model is 0.6064. The positive predictive value of the prediction model is 70.45%, and the negative predictive value is 80%. The positive predictive value refers to the proportion of samples predicted by the classification model as positive examples that are truly positive examples. The negative predictive value refers to the proportion of samples predicted by the classification model as negative examples that are truly negative examples.

[0061] The present invention provides a prediction system for anti-tuberculosis drug-induced liver injury, comprising a data preprocessing module, a model processing module and a prediction result output module;

[0062] The data preprocessing module is used to obtain the medical record data to be tested of the target object, wherein the medical record data to be tested includes the genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B of the target object;

[0063] The model processing module is used to output the predicted anti-tuberculosis drug-induced liver injury result based on the medical record data to be tested and the anti-tuberculosis drug-induced liver injury prediction model according to claim 8;

[0064] The prediction result output module is used to output the prediction result of anti-tuberculosis drug-induced liver injury.

[0065] The present invention provides a device equipped with the anti-tuberculosis drug-induced liver injury prediction system.

[0066] The present invention also provides the application of the anti-tuberculosis drug-induced liver injury prediction model in predicting the occurrence of anti-tuberculosis drug-induced liver injury in tuberculosis patients.

[0067] The following is a detailed description of a method for establishing a prediction model for anti-tuberculosis drug-induced liver injury and a prediction system thereof provided by the present invention in conjunction with the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0068] Example 1

[0069] A prediction method for anti-tuberculosis drug-induced liver injury

[0070] 1. Sample Collection

[0071] The questionnaire was designed to obtain the general situation of the subjects by reviewing the medical records, and 3 ml of anticoagulated blood samples were collected. The baseline information of the participants was systematically collected by reviewing the hospitalization medical records of the subjects included in the study. The baseline information collected includes but is not limited to individual demographic characteristics, such as age, gender, marital status, place of residence, and occupational classification. The lifestyle and behavioral habits of the participants were also recorded, including the history of smoking and drinking, potential environmental exposure factors, and occupational health risks. During the study, all patients were followed up regularly to track and record the liver damage that occurred during their anti-tuberculosis treatment. Therefore, the information on clinical liver function and blood routine indicators was used to observe and record the changes in relevant indicators to provide information for subsequent screening of potential risk factors.

[0072] A total of 344 patients with pulmonary tuberculosis who were newly admitted to Shijiazhuang Fifth Hospital from January to December 2022 were included, and their demographic data, biochemical tests, and blood specimens were collected.

[0073] 2. Total RNA Extraction and Gene Polymorphism Identification

[0074] 2.1 Kit method for extracting DNA from plasma of tuberculosis patients

[0075] (1) Destruction of red blood cells:

[0076] ① Thaw the whole blood, take 300μl and mix it with 1000μl red blood cell lysis buffer;

[0077] ②Stand for 10 min, centrifuge at 12000 rpm for 20 s, and discard the supernatant;

[0078] ③Repeat step ②;

[0079] ④ Add 200 μl buffer BB and vortex thoroughly.

[0080] (2) Digestion of proteins and RNA:

[0081] ① Add proteinase K to 200 μl of the mixture and mix it thoroughly;

[0082] ② Add 20 μl of 25 mg / ml RNase A solution to the above system, vortex thoroughly, and let stand for 5 minutes.

[0083] (3) DNA extraction:

[0084] ① Add 200 μl of binding solution CB, mix thoroughly immediately, place in a 70℃ water bath for 10 min, and cool at -20℃ for 2 min;

[0085] ② Add 100 μl of isopropanol, vortex and mix immediately, add to adsorption column AC (100 μl of equilibrium solution, centrifuge at 13000 rpm for 1 min, discard the waste liquid), centrifuge at 13000 rpm for 60 s, and discard the waste liquid;

[0086] ③ Add 500 μl inhibitor removal solution IR, centrifuge at 12000 rpm for 30 seconds, and discard the waste liquid;

[0087] ④Add 600μl of washing solution WB, centrifuge as in ③, and discard the waste liquid. Repeat this step 2 to 3 times;

[0088] ⑤Put the adsorption column AC into an empty tube, centrifuge at 13000rpm for 2min, and remove the rinse solution;

[0089] ⑥Take out the adsorption column AC, change the centrifuge tube, add 100 μl elution buffer EB (the elution buffer was pre-bathed at 70°C) in the middle of the adsorption membrane, let it stand for 2 minutes, and centrifuge at 12000 rpm for 1 minute;

[0090] ⑦DNA was stored in -20℃ refrigerator.

[0091] 2.2 Identification of gene polymorphism

[0092] (1) The polymerase chain reaction-restriction fragment length polymorphism (PCR-FLP) technique was used to analyze the polymorphism of the DNA amplification products and determine the genotypes of each locus of GSTM1, GSTT1, NAT2 genes (rs1799929 and rs1799931), rs2031920, and CYP3A4*18B.

[0093] In this example, the software Premier 5.0 was used to design primers for GSTM1, GSTT1, NAT2, rs2031920 and CYP3A4*18B gene sequences. The designed primer sequences were commissioned to Sangon Biotech (Shanghai) Co., Ltd. for synthesis. The primer sequences of each gene are shown in Table 1.

[0094] Table 1 Primer sequences and internal reference sequences for PCR amplification of each gene

[0095]

[0096] (2) PCR reaction system

[0097] The amplification of each gene was carried out in a 20 μL amplification system. The amplification system configurations of GSTM1 and GSTT1 genes are shown in Table 2 , the amplification system configurations of NAT2, rs2031920 and CYP3A4*18B genes are shown in Table 3 , and the PCR reaction parameters of the five genes are shown in Table 4 .

[0098] Table 2 GSTM1, GSTT1 gene amplification system

[0099]

[0100]

[0101] Table 3 NAT2, rs2031920 and CYP3A4*18B gene amplification system

[0102] Reagents Dosage (μl) Pure water 8 PCRMasterMix 10 F 0.5 R 0.5 DNA template 1

[0103] Table 4 PCR amplification parameters of GSTM1, GSTT1, NAT2, rs2031920 and CYP3A4*18B genes

[0104] Expansion phase Temperature(℃) Time(s) Number of cycles Pre-denaturation 94 180 1 Deformation 94 60 35 annealing 56 45 35 extend 72 60 35 Terminal extension 72 420 1

[0105] 2.3 Electrophoresis analysis

[0106] Prepare a 2% agarose gel, add 2μL DNAMaker to the first loading well, mix the sample DNA PCR amplification product with 0.2μL 10× Loading Buffer, place it in the corresponding spotting well on the agarose gel, and run it at 100V for 40min. Then image it under the gel imaging system. If further restriction enzyme analysis is required, it is necessary to confirm that a clear band appears at the corresponding position, which serves as the basis for successful amplification and continued restriction fragment length polymorphism analysis.

[0107] 2.4 RFLP analysis

[0108] The polymorphism of GSTM1 and GSTT1 genes is manifested as gene deletion, so internal reference was added to identify the genotypes of GSTM1 and GSTT1.

[0109] For each site of rs1799929, rs1799931, rs2031920, and CYP3A4*18B on the NAT2 gene, the amplified products were digested with the corresponding restriction endonucleases in a water bath for 0.5 h. The operation was carried out according to the product instructions. The enzyme digestion reaction system and reaction conditions are shown in Table 5.

[0110] Table 5 Reaction system and reaction conditions of restriction enzyme cutting sites

[0111]

[0112]

[0113] 2.5 Gene polymorphism analysis

[0114] Prepare a 2% agarose gel, add 2μL DNAMaker to the first loading well, mix the sample DNA PCR amplification product with 0.2μL 10× Loading Buffer, place it in the corresponding spot well on the agarose gel, and run it at 100V for 40min. Then image it under the gel imaging system. Image and record the enzyme digestion products under the gel imaging system, and judge the genotype of each locus according to the size and number of bands in the electrophoresis diagram after enzyme digestion of each gene locus, see Table 6 and Table 7.

[0115] Table 6 GSTM1 and GSTT1 gene locus polymorphism analysis and judgment method

[0116]

[0117] Table 7 Analysis and judgment methods of polymorphism of rs1799929, rs1799931, rs2031920 and CYP3A4*18B gene loci

[0118]

[0119]

[0120] 3. Data collation and analysis

[0121] EXCEL 2019 and SPSS23.0 were used to review, classify and build a database for the field survey and experimental test data, and then statistical methods were used for data input and processing. The statistical methods used were based on two-sided tests, and the test level was set at α = 0.05.

[0122] (1) Balanced analysis of the general conditions of cases and controls

[0123] The baseline data of the case group and the control group were tested for balance. For the quantitative data with normal distribution, the paired t test was used; for the data containing categorical variables, the univariate logistic regression analysis was used.

[0124] (2) Univariate analysis

[0125] In this study, the chi-square test was used to compare the distribution of drug metabolizing enzyme gene polymorphisms in the case group and the control group; then a univariate logistic regression model was used to explore the baseline characteristics and risk factors of the subjects, as well as the relationship between the polymorphisms of each gene locus and liver damage caused by drug-resistant pulmonary tuberculosis, and the OR value and 95% CI were calculated.

[0126] (3) Multifactor analysis

[0127] Whether anti-tuberculosis drug-induced liver injury occurred after anti-tuberculosis treatment was taken as the dependent variable, and the relevant risk factors with significant influence (P<0.05) in the study were taken as independent variables. Multivariate logistic regression analysis was used to calculate the OR value and its 95% CI to construct a model, and then the area under the ROC curve (AUC) of the prediction model was further plotted to evaluate the discriminative ability of the model.

[0128] (4) Prediction model

[0129] Model construction: 344 samples were randomly divided into a modeling group and a validation group at a ratio of 8:2, with 275 samples in the modeling group and 69 samples in the validation group. R software was used in the model construction process, and SPSS23.0 was used for data analysis. In the model construction stage, univariate logistic analysis was performed on the variables, and variables with statistically significant differences were selected, and then these variables were included in the multivariate logistic regression analysis.

[0130] Model validation: The AUC is calculated using the model validation set data. This indicator can fully reflect the sensitivity and specificity of the model. The higher the AUC, the better the model prediction performance. If the AUC is greater than 0.7, the model can be considered to have moderate diagnostic value; if it exceeds 0.8, it is considered to have high diagnostic value. No improvements were made to the bioinformatics analysis method.

[0131] 4. Quality Control

[0132] (1) This example strictly implements the established inclusion and exclusion criteria to screen the patient population participating in the study and minimize selection bias.

[0133] (2) When collecting basic information about patients, ensure that the survey items have clear definitions and conduct preliminary communication with the attending physician to reduce bias in the information collection process.

[0134] (3) While ensuring the standardization of the patient treatment process, effective publicity and education as well as emotional comfort can be used to improve the patient's treatment compliance and cooperation, thereby reducing possible confounding biases that may occur during the treatment process.

[0135] (4) Double data entry and cross-validation methods are used to reduce the possibility of data entry errors; the entered data are regularly compared with the original data to ensure the integrity of the collected data.

[0136] 5. Results

[0137] For the case group and the control group, the distribution of age, gender, smoking status and drinking status was sorted out, and the results are shown in Table 8.

[0138] This study used the Logistic regression analysis method to comprehensively evaluate multiple variables including age, gender, blood pressure, smoking status, drinking habits, and ALT levels. The results of the correlation analysis are summarized in Table 9. The study found that there was no statistically significant difference in the distribution of age, gender, blood pressure, and smoking and drinking behaviors between the case group and the control group (P>0.05), indicating that the two groups were balanced in the distribution of these factors. However, the ALT level showed a significant difference in the comparison between the two groups (P<0.05), revealing the difference in ALT levels between the case group and the control group.

[0139] Table 8 Overall distribution of general influencing factors in the case group and the control group

[0140]

[0141] Table 9 Univariate conditional Logistic regression study on factors related to liver injury caused by anti-tuberculosis drugs

[0142]

[0143]

[0144] According to the data shown in Table 10, the distribution of mutations at each gene locus between the two groups showed significant differences. After univariate logistic analysis, it was further determined that there were significant differences in the distribution of mutations at each gene locus, and this difference was statistically significant (P<0.05), and the results are shown in Table 11.

[0145] Table 10 Comparison of genotype distribution between case group and control group

[0146]

[0147] Table 11 Univariate Logistic regression analysis of each gene locus and liver injury caused by anti-tuberculosis drugs

[0148]

[0149] Multivariate Logistic regression analysis was used to reduce the interference of confounding factors. The specific results are shown in Table 12. The mutation status of the six gene loci was the independent variable, and whether anti-tuberculosis drug-induced liver injury occurred was the dependent variable. Among them, the five independent variables of GSTM1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were still significantly correlated, with ORs of 2.182, 2.699, 2.905, 2.166, and 1.878, respectively.

[0150] Table 12 Results of multivariate logistic regression analysis of influencing factors of anti-tuberculosis drug-induced liver injury

[0151] factor β SE <![CDATA[Waldχ 2 ]]> P OR 95% CI GSTM1 0.780 0.245 10.179 0.001 2.182 (1.351，3.524) GSTT1 0.520 0.293 3.151 0.076 1.683 (0.947，2.989) rs1799929 0.993 0.275 13.047 <0.001 2.699 (1.575，4.624) rs1799931 1.066 0.274 15.165 <0.001 2.905 (1.698，4.969) rs2031920 0.773 0.281 7.584 0.006 2.166 (1.250，3.753) CYP3A4*18B 0.630 0.241 6.846 0.009 1.878 (1.171，3.012)

[0152] The 344 samples were randomly divided into a modeling group and a validation group according to a quantitative ratio of 8:2. The specific assignment methods of each variable are detailed in Table 13, and the results are detailed in Table 14. The results showed that there was no statistically significant difference in each variable between the two groups (P>0.05), proving that the modeling group and the validation group were basically randomly grouped.

[0153] Table 13 General influencing factors and assignment methods

[0154] variable grade Assignment method gender 0～1 0=male 1=female address 0～1 0 = urban 1 = rural Smoking status 0～1 0 = No smoking 1 = Smoking Drinking habits 0～1 0 = No drinking 1 = Drinking

[0155] Table 14 Comparison of variables between the modeling group and the validation group

[0156]

[0157] Univariate analysis of each gene locus in the modeling group and anti-tuberculosis drug-induced liver injury showed that the mutation of each gene locus in the modeling group was statistically significant (P<0.05). The results are shown in Table 15.

[0158] Table 15 Results of univariate logistic regression analysis of influencing factors of anti-tuberculosis drug-induced liver injury in the modeling group

[0159]

[0160] Multivariate logistic regression was performed with the indicators with statistically significant differences as independent variables and whether or not the patient suffered from anti-tuberculosis drug-induced liver injury as the dependent variable. The results are shown in Table 16. The six independent variables, including the mutation status of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B, were independent variables, and whether or not the patient suffered from anti-tuberculosis drug-induced liver injury was the dependent variable, were still significantly correlated.

[0161] The prediction model established is: Y = 0.545X 1 +0.666X 2 +0.986X 3 +0.919X 4 +0.630X 5 +0.531X 6 -1.365

[0162] Where X 1 is the genotype data of GSTM1; X 2 is the GSTT1 genotype data; X 3 is the genotype data of rs1799929; X 4 is the genotype data of rs1799931; X 5 is the genotype data of rs2031920; X 6 CYP3A4*18B genotype data. Specifically, for GSTM1 gene and GSTT1 gene, when a deletion mutation occurs, a value of 1 is assigned, and when no deletion mutation occurs, a value of 2 is assigned; for rs1799929, rs1799931, rs2031920, CYP3A4*18B polymorphic sites, when a mutation (including homozygous mutation and heterozygous mutation) occurs, a value of 1 is assigned; when no mutation occurs, a value of 2 is assigned.

[0163] The risk of anti-tuberculosis drug-induced liver injury in the sample to be tested is judged based on whether the Y value is closer to 1 or 2: when the Y value is closer to 1, the risk of anti-tuberculosis drug-induced liver injury in the sample to be tested is higher, and medication guidance such as changing the type of anti-tuberculosis drug or taking liver protection drugs should be adopted; when the Y value is closer to 2, the risk of anti-tuberculosis drug-induced liver injury in the sample to be tested is lower, and normal anti-tuberculosis drug clinical treatment can be given.

[0164] Table 16 Results of multivariate logistic regression analysis of influencing factors of anti-tuberculosis drug-induced liver injury in the modeling group

[0165]

[0166]

[0167] Use the validation group data to validate the model and draw the ROC curve, see Figure 1 As shown in the figure, its AUC is 0.8076, P < 0.05. The ROC curve takes into account the sensitivity and specificity of the model and can intuitively show the predictive ability of the model. The AUC value > 0.8 indicates that the predictive ability of the model is high.

[0168] The predicted probability values ​​of the prediction model of the present invention are shown in Table 17 below. The maximum value of the Youden index obtained by using R software is 0.4672. Taking the maximum value of the Youden index as the cutoff point, the corresponding probability cutoff value is 0.9809. That is, cases with a predicted probability cutoff value ≥ 0.9809 are predicted to have anti-tuberculosis drug-induced liver injury, and cases with a predicted probability cutoff value < 0.9809 are predicted to have no anti-tuberculosis drug-induced liver injury.

[0169] Table 17 Prediction model threshold table

[0170] Prediction probability cutoff Sensitivity 1-Specificity Youden Index 0 1 1 0 0.9587 1 0.8788 0.1212 0.9692 0.9167 0.6970 0.2197 0.9706 0.9166 0.6363 0.2803 0.9723 0.9162 0.6056 0.3106 0.9758 0.9161 0.5450 0.3712 0.9794 0.8611 0.4243 0.4368 0.9809 0.8607 0.3936 0.4671 0.9825 0.7222 0.3031 0.4191 0.9833 0.6666 0.3030 0.3636

[0171] The data of the validation group were used to compare whether the outcome of anti-tuberculosis drug-induced liver injury actually occurred. The maximum value of the Youden index was found to be 0.4671 in the threshold table, and the corresponding model sensitivity was 0.8607. The specificity of the model was calculated to be 0.6064. The data was back-substituted in the validation group, and the positive predictive value of the model was calculated to be 70.45%, and the negative predictive value was 80%. The specific results are shown in Table 18.

[0172] Table 18 Verification model discrimination and actual disease results of patients with anti-tuberculosis drug-induced liver injury

[0173]

[0174] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for establishing a prediction model for anti-tuberculosis drug-induced liver injury, characterized in that: The following steps are involved: Extract genomic DNA from tuberculosis patients; Using the extracted genomic DNA as a template, GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were amplified to obtain DNA amplification products; The DNA amplification products were analyzed for polymorphism using PCR-FLP technique, and the genotypes of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920, and CYP3A4*18B were obtained; The genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B were used as independent variables, and the occurrence of anti-tuberculosis drug-induced liver injury after anti-tuberculosis treatment was used as the dependent variable, and a prediction model was constructed using multivariate logistic regression analysis.

2. The method according to claim 1, characterized in that: The primers for amplifying GSTM1 include a forward primer having a nucleotide sequence as shown in SEQ ID NO: 1 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO: 2; The primers for amplifying GSTT1 include a forward primer having a nucleotide sequence as shown in SEQ ID NO: 3 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO: 4; The primers used to amplify rs1799929 and rs1799931 include a forward primer having a nucleotide sequence as shown in SEQ ID NO:7 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:8; The primers used to amplify rs2031920 include a forward primer having a nucleotide sequence as shown in SEQ ID NO: 9 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO: 10; The primers for amplifying CYP3A4*18B include a forward primer having a nucleotide sequence as shown in SEQ ID NO:11 and a reverse primer having a nucleotide sequence as shown in SEQ ID NO:

12.

3. The method according to claim 1, characterized in that: The reaction program for amplification of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B was 94°C for 3 min; 94°C for 60 s, 56°C for 45 s, and 72°C for 60 s, 35 cycles; 72°C for 7 min.

4. The method according to claim 1, characterized in that: The PCR-FLP technique was used to analyze the polymorphism of the DNA amplification products, including rs1799929, rs1799931, rs2031920 and CYP3A4*18B amplification products, and restriction endonuclease digestion was performed: The amplified product of the rs1799929 polymorphic site was digested with Kpn I-HF endonuclease; The amplified product of the rs1799931 polymorphic site was digested with BamHI-HF endonuclease; The amplified products of rs2031920 and CYP3A4*18B polymorphisms were digested with Rsa I endonuclease.

5. The method according to claim 1, characterized in that: The genotype of GSTM1 includes a deletion type and a non-deletion type; the deletion type has an amplification product fragment length of 268 bp; the non-deletion type has an amplification product fragment length of 268 bp and 219 bp; The genotype of GSTT1 includes a deletion type and a non-deletion type; the deletion type has an amplification product fragment length of 268 bp; the non-deletion type has an amplification product fragment length of 268 bp and 459 bp; The genotype of rs1799929 includes CC genotype, CT genotype and TT genotype; the enzyme digestion product of the CC genotype includes 60bp and 437bp; the enzyme digestion product of the CT genotype includes 60bp, 437bp and 497bp; the enzyme digestion product of the TT genotype includes 497bp; The genotype of rs1799931 includes GG genotype, GA genotype and AA genotype; the enzyme digestion product of the GG genotype includes 61bp and 436bp; the enzyme digestion product of the GA genotype includes 61bp, 436bp and 497bp; the enzyme digestion product of the AA genotype includes 497bp; The enzyme digestion products of rs2031920 include GG genotype, GA genotype and AA genotype; the enzyme digestion products of the GG genotype include 155bp and 219bp; the enzyme digestion products of the GA genotype include 155bp, 219bp and 374bp; the enzyme digestion products of the AA genotype include 374bp; The enzyme cleavage products of CYP3A4*18B include GG genotype, GA genotype and AA genotype; the enzyme cleavage products of the GG genotype include 70bp and 205bp; the enzyme cleavage products of the GA genotype include 70bp, 205bp and 275bp; the enzyme cleavage products of the AA genotype include 275bp.

6. The method according to claim 1, characterized in that: Before constructing the prediction model, the distribution of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B polymorphisms in the case group and the control group was tested for balance and analyzed using a univariate logistic regression model to exclude risk factors and determine that the polymorphisms were significantly associated with liver damage caused by drug-resistant pulmonary tuberculosis. The samples of the case group were from tuberculosis patients who developed anti-tuberculosis drug-induced liver injury; The samples of the control group were from tuberculosis patients who did not suffer from anti-tuberculosis drug-induced liver injury.

7. The method according to any one of claims 1 to 6, characterized in that: After the prediction model is obtained, the method further includes performing model verification on the prediction model; The method for model verification is to use the verification set data to calculate the AUC of the prediction model: when the AUC is greater than 0.7, the prediction model is considered to have moderate diagnostic value; when the AUC exceeds 0.8, it indicates that the prediction model has high diagnostic value.

8. A prediction model for anti-tuberculosis drug-induced liver injury obtained by the method according to any one of claims 1 to 7.

9. A system for predicting anti-tuberculosis drug-induced liver injury, characterized in that: It includes data preprocessing module, model processing module and prediction result output module; The data preprocessing module is used to obtain the medical record data to be tested of the target object, wherein the medical record data to be tested includes the genotype data of GSTM1, GSTT1, rs1799929, rs1799931, rs2031920 and CYP3A4*18B of the target object; The model processing module is used to calculate and predict the result of anti-tuberculosis drug-induced liver injury based on the medical record data to be tested and the anti-tuberculosis drug-induced liver injury prediction model according to claim 8; The prediction result output module is used to output the prediction result of anti-tuberculosis drug-induced liver injury.

10. A device equipped with the anti-tuberculosis drug-induced liver injury prediction system according to claim 9.

Images