Method and device for constructing prediction model of drug resistance phenotype of helicobacter pylori to metronidazole

Abstract

The invention discloses a prediction model construction method and device for the drug resistance phenotype of helicobacter pylori to metronidazole, and relates to the technical field of drug resistance prediction.The method comprises the steps that helicobacter pylori whole genome sequencing data is obtained and preprocessed, and an affiliated gene and an SNP matrix are obtained; respectively inputting the affiliated genes, the SNP matrix and the affiliated genes and the SNP matrix into a plurality of machine learning models for training, optimizing model parameters, selecting an optimal prediction model through model performance, globally explaining the optimal prediction model, extracting important features, establishing a network calculator according to the optimal prediction model, and performing prediction on the basis of the network calculator. And the network calculator predicts the drug resistance probability of the helicobacter pylori to the metronidazole through the extracted important features. The network calculator constructed by the application can predict the drug resistance probability of the helicobacter pylori to the metronidazole, and is used for guiding H.pylori clinical eradication decision and individualized medication scheme selection.

Description

Technical Field

[0001] This application relates to the field of drug resistance prediction technology, specifically to a method and apparatus for constructing a predictive model of Helicobacter pylori resistance to metronidazole. Background Technology

[0002] Helicobacter pylori (H. pylori) infection is one of the most common chronic bacterial infections worldwide, affecting more than half of the global population. In 1994, the World Health Organization (WHO) classified it as a Group 1 carcinogen. In 2017, the WHO classified clarithromycin-resistant H. pylori as a "high priority" bacterium, and in 2022, the U.S. Department of Health and Human Services classified H. pylori as a known carcinogen. Multiple randomized controlled trials and cohort studies have shown that eradicating H. pylori infection can reduce the risk of developing gastric cancer and is the most important controllable risk factor for gastric cancer prevention.

[0003] In clinical practice, only a few antibiotics, such as metronidazole (MTZ), clarithromycin (CLR), levofloxacin (LEV), amoxicillin (AMX), and tetracycline (TET), have been proven effective in eradicating H. pylori. Over the past 40 years, various treatment regimens using these commonly used drugs have been developed; however, global retrospective and meta-analyses of antibiotic resistance show that H. pylori eradication rates are declining year by year, while resistance rates are increasing. Therefore, understanding H. pylori drug susceptibility before using antibiotics and conducting individualized eradication therapy guided by drug susceptibility testing is essential.

[0004] Currently, the detection methods available for guiding H. pylori individualized treatment with drug susceptibility testing mainly include: culture-based phenotypic drug susceptibility testing methods, such as the agar dilution method (gold standard) and the E-test method, as well as molecular drug susceptibility testing methods based on resistance mutations, such as PCR. However, these methods have certain limitations, which restrict the clinical application of drug susceptibility-guided individualized eradication therapy.

[0005] Next-generation sequencing (NGS), as a comprehensive, cost-effective, and rapid tool for obtaining bacterial whole-genome information, can be used for monitoring drug-resistant infectious diseases and evolutionary analysis, and has the potential to discover rare drug resistance mechanisms in clinical isolates. In recent years, with the decrease in sequencing costs and the development of bioinformatics technology, a method combining bacterial NGS and machine learning to predict drug resistance outcomes and identify drug resistance-related features has been developed. This method does not rely on pre-existing databases of AMR genes or mutations; it uses a pan-genomic approach to associate NGS whole-genome sequences with drug resistance phenotypic information, ultimately obtaining a drug resistance prediction model. Its feasibility has been demonstrated in pathogenic bacteria such as *Escherichia coli*, *Mycobacterium tuberculosis*, *Salmonella*, and *Pseudomonas aeruginosa*. Furthermore, SHapley Additive exPlanation (SHAP) values ​​are used to visualize the "black box model" of machine learning, revealing each important feature affecting the machine learning model. However, whether the above-mentioned bacterial drug resistance prediction strategy can be applied to the prediction of metronidazole resistance in *H. pylori* remains a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] Therefore, this application provides a method and apparatus for constructing a predictive model of Helicobacter pylori resistance to metronidazole, in order to solve the problem that existing detection methods for H. pylori individualized treatment drug sensitivity guidance limit the clinical application of drug sensitivity-guided individualized eradication therapy.

[0007] To achieve the above objectives, this application provides the following technical solution:

[0008] Firstly, a method for constructing a predictive model for Helicobacter pylori resistance to metronidazole includes:

[0009] Step 1: Obtain Helicobacter pylori whole genome sequencing data;

[0010] Step 2: Preprocess the Helicobacter pylori whole genome sequencing data to obtain accessory genes and SNP matrix;

[0011] Step 3: Input the accessory gene, the SNP matrix, and the accessory gene and SNP matrix into multiple machine learning models for training, optimize the model parameters, and select the optimal prediction model based on model performance;

[0012] Step 4: Perform a global interpretation of the optimal prediction model and extract important features;

[0013] Step 5: Establish a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

[0014] Preferably, step 2 includes:

[0015] Step 201: Perform genome annotation on the whole genome sequencing data of Helicobacter pylori to obtain core genes and accessory genes;

[0016] Step 202: Identify single nucleotide polymorphisms in the core gene to obtain the SNP matrix.

[0017] Preferably, in step 201, Prokka is used to perform genome annotation on the whole genome sequencing data of Helicobacter pylori.

[0018] Preferably, in step 3, various machine learning models include KNN, LR, SVM, RF, GBDT, and XGBoost.

[0019] Preferably, in step 3, the model parameters are optimized using grid search and cross-validation.

[0020] Preferably, in step 4, SHAP is used to perform a global interpretation of the optimal prediction model.

[0021] Preferably, the key features extracted in step 4 include HP_0040Val241Ile, HP_0642Ala70fs, HP_0887Thr439Ser, group_913, HP_0060Glu475Thr, tdhF Ala217Thr, group_1110, group_641, group_435, and nolkSer97Leu.

[0022] Secondly, a device for constructing a predictive model for Helicobacter pylori resistance to metronidazole includes:

[0023] The data acquisition module is used to acquire Helicobacter pylori whole genome sequencing data;

[0024] The data preprocessing module is used to preprocess the Helicobacter pylori whole genome sequencing data to obtain accessory genes and SNP matrix;

[0025] The training module is used to input the accessory genes, the SNP matrix, and the accessory genes and SNP matrix into various machine learning models for training, and to optimize the model parameters and select the optimal prediction model based on the model performance.

[0026] The global interpretation module is used to perform a global interpretation of the optimal prediction model and extract important features;

[0027] The prediction module is used to build a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

[0028] Thirdly, a computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a method for constructing a predictive model of Helicobacter pylori resistance to metronidazole.

[0029] Fourthly, a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a method for constructing a predictive model of Helicobacter pylori resistance to metronidazole.

[0030] Compared with the prior art, this application has at least the following beneficial effects:

[0031] This application provides a method and apparatus for constructing a predictive model for Helicobacter pylori resistance to metronidazole. The method involves acquiring whole-genome sequencing data of Helicobacter pylori; preprocessing the whole-genome sequencing data to obtain accessory genes and SNP matrices; inputting the accessory genes and SNP matrices, and vice versa, into various machine learning models for training, optimizing model parameters, and selecting the optimal predictive model based on model performance; globally interpreting the optimal predictive model and extracting important features; establishing a network calculator based on the optimal predictive model; and using the extracted important features to predict the probability of Helicobacter pylori resistance to metronidazole. The network calculator constructed using this application can predict the probability of Helicobacter pylori resistance to metronidazole and can be used to guide H. pylori clinical eradication decisions and individualized medication regimen selection. Attached Figure Description

[0032] To more intuitively illustrate the prior art and this application, several exemplary figures are provided below. It should be understood that the specific shapes and structures shown in the figures should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary figures, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0033] Figure 1 A flowchart illustrating a method for constructing a predictive model for Helicobacter pylori resistance to metronidazole, as provided in Embodiment 1 of this application;

[0034] Figure 2 The SHAP importance matrix diagram based on the optimal prediction model for metronidazole provided in Embodiment 1 of this application;

[0035] Figure 3 This is a SHAP summary diagram of the optimal prediction model for metronidazole provided in Embodiment 1 of this application;

[0036] Figure 4 This is a schematic diagram of an online calculator for predicting H. pylori metronidazole phenotype resistance, provided in Embodiment 1 of this application. Detailed Implementation

[0037] The present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] In the description of this application: unless otherwise stated, "a plurality of" means two or more. The terms "first," "second," "third," etc., in this application are intended to distinguish the objects referred to and do not have any special meaning in terms of technical connotation (e.g., they should not be construed as an emphasis on importance or order). Expressions such as "comprising," "including," and "having" also mean "not limited to" (certain units, components, materials, steps, etc.).

[0039] The terms used in this application, such as "upper," "lower," "left," "right," and "middle," are generally used to facilitate intuitive understanding by referring to the accompanying drawings, and are not absolute limitations on the positional relationships in the actual product. Changes in these relative positional relationships, without departing from the technical concept disclosed in this application, should also be considered within the scope of this application.

[0040] Example 1

[0041] Please see Figure 1 This embodiment provides a method for constructing a predictive model for Helicobacter pylori resistance to metronidazole, including:

[0042] S1: Obtain whole genome sequencing data of Helicobacter pylori;

[0043] Specifically, the FASTA files of the complete genomes of 301 Helicobacter pylori strains were downloaded from the Bacterial and Viral Bioinformatics Resource Center (BV-BRC) and the NCBIGenBank database. These strains met the following criteria:

[0044] (1) The genome assembly level must be complete;

[0045] (2) The bacteria must be isolated from a human host;

[0046] (3) Phenotypic resistance information of metronidazole can be obtained.

[0047] S2: Preprocess the whole genome sequencing data of Helicobacter pylori to obtain accessory genes and SNP matrix;

[0048] Specifically, it includes:

[0049] S201: Genetic annotation of Helicobacter pylori whole genome sequencing data to obtain core genes and accessory genes;

[0050] This step used Prokka to annotate the genome sequences of 301 strains. Based on whether the gene clusters came from all strains, the genomes were divided into "core" genomes and "accessory" genomes.

[0051] S202: Identify single nucleotide polymorphisms in the core gene to obtain the SNP matrix.

[0052] This step uses Snippyv4.6.0 to identify gene variations in 301 strains, including single nucleotide polymorphisms (SNPs), insertions, deletions, multiple nucleotide polymorphisms, and complex variations.

[0053] When using Snippyv4.6.0 to identify genetic variations in 301 strains, Helicobacter pylori 26695 (genome assembly ID: GCF_000307795.1) was used as the reference genome.

[0054] S3: Input the accessory genes, SNP matrix, and accessory genes and SNP matrix into multiple machine learning models for training, optimize the model parameters, and select the optimal prediction model based on model performance;

[0055] In this step, all data are randomly grouped, with 80% used for the training set and 20% for the validation set. The associated genes, SNP matrix, and associated gene + SNP matrix are then input into six machine learning models for training: K-Nearest Neighbor (KNN), Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), and eXtreme Gradient Boosting (XGBoost). During training, grid search and cross-validation are used to optimize the parameters of each machine learning model. Accuracy, Precision, Recall, F1 score, and AUC are then used to evaluate the performance of each learning model on the test set, and the optimal prediction model is selected. The performance parameters of each learning model are shown in Table 1.

[0056] Table 1: Comparison of the predictive performance of six machine learning algorithms for metronidazole phenotype resistance in Helicobacter pylori

[0057]

[0058] As shown in Table 1:

[0059] (1) When drug resistance prediction is based solely on accessory genes, XGBoost shows strong predictive ability with an accuracy of 0.8033, an F1 score of 0.7989, and an AUC of 0.8174, which is better than other models.

[0060] (2) When drug resistance is predicted based solely on SNPs, SVM shows strong predictive ability, with an AUC of 0.8459, an Accuracy of 0.7759, and an F1 score of 0.7705.

[0061] (3) When drug resistance prediction was based on accessory genes + SNPs, the performance of all models was significantly improved compared to the accessory gene or SNP group alone. This was reflected in the fact that the AUC values ​​of all models reached above 0.9, with XGBoost performing the best. It performed best in all five performance evaluation indicators, with Accuracy of 0.9180, Precision of 0.9169, Recall of 0.9180, F1 of 0.9172, and AUC of 0.9535.

[0062] Based on the above results, the accessory gene + SNP group model performed best, and the XGBoost model outperformed other prediction models. Therefore, the above-mentioned optimal models were selected for further analysis.

[0063] S4: Perform a global interpretation of the optimal prediction model and extract important features;

[0064] To identify the features that have the greatest impact on the prediction model and to make the model's output more intuitive and visual, this embodiment uses SHapley Additive exPlanation (SHAP) to interpret the above-mentioned optimal antibiotic resistance prediction "black box model" from both global and local perspectives. Figure 2 and Figure 3 As shown, the matrix diagram and summary diagram represent the top 10 features in metronidazole resistance prediction and their contribution directions, respectively.

[0065] The top 10 characteristics predicting metronidazole resistance all indicate that the presence of a specific SNP mutation or gene increases the likelihood of *H. pylori* resistance to MTZ. The Val241Ile mutation in the HP0040 gene ranks first, followed by a frameshift mutation at amino acid 70 in the HP0642 (frxA) gene encoding NADPH flavin oxidoreductase, HP_0887Thr439Ser,group_913, HP_0060Glu475Thr,tdhF Ala217Thr,group_1110,group_641,group_435, and nolkSer97Leu. Except for the nolkSer97Leu mutation, which reduces the likelihood of *H. pylori* resistance to metronidazole, the presence of the other characteristics increases the likelihood of *H. pylori* resistance to metronidazole.

[0066] S5: Build a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

[0067] This step involves building a web-based calculator using the optimal prediction model. First, the optimal prediction model is saved as a PKL document using Python software. Second, both the optimal model and its corresponding Python code are uploaded to GitHub. Finally, after linking the Streamlit application to GitHub, a web calculator is deployed. When 10 features of a Helicobacter pylori genome are input and the user clicks "Predict," the calculator provides a predicted probability of the strain's resistance to metronidazole. Figure 4 As shown.

[0068] The network calculator constructed using the predictive model construction method for Helicobacter pylori resistance to metronidazole provided in this embodiment can predict the probability of Helicobacter pylori resistance to metronidazole and be used to guide H. pylori clinical eradication decisions and individualized medication regimen selection.

[0069] Example 2

[0070] This embodiment provides a device for constructing a predictive model of Helicobacter pylori resistance to metronidazole, including:

[0071] The data acquisition module is used to acquire Helicobacter pylori whole genome sequencing data;

[0072] The data preprocessing module is used to preprocess the Helicobacter pylori whole genome sequencing data to obtain accessory genes and SNP matrix;

[0073] The training module is used to input the accessory genes, the SNP matrix, and the accessory genes and SNP matrix into various machine learning models for training, and to optimize the model parameters and select the optimal prediction model based on the model performance.

[0074] The global interpretation module is used to perform a global interpretation of the optimal prediction model and extract important features;

[0075] The prediction module is used to build a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

[0076] For specific limitations on the device for constructing a predictive model of Helicobacter pylori resistance to metronidazole, please refer to the limitations on the method for constructing a predictive model of Helicobacter pylori resistance to metronidazole mentioned above, which will not be repeated here.

[0077] Example 3

[0078] This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of a method for constructing a predictive model of Helicobacter pylori resistance to metronidazole.

[0079] Example 4

[0080] This embodiment discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of a method for constructing a predictive model of Helicobacter pylori resistance to metronidazole.

[0081] The technical features of the above embodiments can be combined in any way (as long as there is no contradiction in the combination of these technical features). For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written should also be considered to be within the scope of this specification.

[0082] The present application has been described in a relatively specific and detailed manner above through general descriptions and specific embodiments. It should be understood that, based on the technical concept of the present application, several conventional adjustments or further innovations can be made to these specific embodiments; however, as long as they do not depart from the technical concept of the present application, the technical solutions obtained by these conventional adjustments or further innovations also fall within the protection scope of the claims of the present application.

Claims

1. A method for constructing a predictive model for Helicobacter pylori resistance to metronidazole, characterized in that, include: Step 1: Obtain Helicobacter pylori whole genome sequencing data; Step 2: Preprocess the Helicobacter pylori whole genome sequencing data to obtain accessory genes and SNP matrix; Step 3: Input the accessory gene, the SNP matrix, and the accessory gene and SNP matrix into multiple machine learning models for training, optimize the model parameters, and select the optimal prediction model based on model performance; Step 4: Perform a global interpretation of the optimal prediction model and extract important features; Step 5: Establish a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

2. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 1, characterized in that, Step 2 includes: Step 201: Perform genome annotation on the whole genome sequencing data of Helicobacter pylori to obtain core genes and accessory genes; Step 202: Identify single nucleotide polymorphisms in the core gene to obtain the SNP matrix.

3. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 2, characterized in that, In step 201, Prokka is used to perform genome annotation on the whole genome sequencing data of Helicobacter pylori.

4. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 1, characterized in that, In step 3, various machine learning models are used, including KNN, LR, SVM, RF, GBDT, and XGBoost.

5. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 1, characterized in that, In step 3, grid search and cross-validation are used to optimize the model parameters.

6. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 1, characterized in that, In step 4, SHAP is used to perform a global interpretation of the optimal prediction model.

7. The method for constructing a predictive model for Helicobacter pylori resistance to metronidazole according to claim 1, characterized in that, In step 4, the key features extracted include HP_0040Val241Ile, HP_0642Ala70fs, HP_0887Thr439Ser, group_913, HP_0060Glu475Thr, tdhF Ala217Thr, group_1110, group_641, group_435, and nolk Ser97Leu.

8. A device for constructing a predictive model for Helicobacter pylori resistance to metronidazole, characterized in that, include: The data acquisition module is used to acquire Helicobacter pylori whole genome sequencing data; The data preprocessing module is used to preprocess the Helicobacter pylori whole genome sequencing data to obtain accessory genes and SNP matrix; The training module is used to input the accessory genes, the SNP matrix, and the accessory genes and SNP matrix into various machine learning models for training, and to optimize the model parameters and select the optimal prediction model based on the model performance. The global interpretation module is used to perform a global interpretation of the optimal prediction model and extract important features; The prediction module is used to build a network calculator based on the optimal prediction model; the network calculator predicts the probability of Helicobacter pylori resistance to metronidazole by extracting important features.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images