Characteristic Genes for Predicting the Carbapenem Susceptibility Phenotype of Pseudomonas aeruginosa

Abstract

This application belongs to the field of bioinformatics technology, and specifically relates to a screening method for the drug sensitivity phenotype resistance characteristics of Pseudomonas aeruginosa - carbapenem drugs, as well as the screened characteristic gene combinations, kits and their applications for predicting the drug sensitivity phenotype of Pseudomonas aeruginosa - carbapenem drugs, which can accurately predict the drug sensitivity phenotype.

Description

[0001] This application is a divisional application of Chinese Patent Application CN 202311184864.9 filed on September 14, 2023. Technical Field

[0002] This application belongs to the field of bioinformatics, and specifically relates to a characteristic gene, a kit and an application for predicting the carbapenem drug susceptibility phenotype of Pseudomonas aeruginosa. Technical Background

[0003] Due to the long-term extensive use and abuse of antibiotics, the increasing bacterial drug resistance has become one of the urgent problems to be solved in global public health security. Accurately and rapidly identifying the drug resistance of pathogenic bacteria to antibiotics has become an urgent need. At present, antimicrobial resistance (AMR) detection mainly includes bacterial drug resistance phenotype detection and bacterial drug resistance gene molecular detection. The drug resistance phenotype detection technology is mainly the traditional antimicrobial susceptibility testing (AST), which is the main means for clinical detection of whether pathogens are drug-resistant and is the "gold standard" for diagnosing drug-resistant pathogen infections. However, AST still has deficiencies in bacterial drug resistance detection, such as long time consumption and low coverage of antibiotic types. This detection method can no longer meet the current medical needs; while the bacterial drug resistance genotype detection technology can only identify the drug resistance genes existing in pathogenic bacteria, but bacterial drug resistance is usually caused by the synergistic action of multiple genes and multiple mechanisms. Therefore, it is impossible to accurately judge the drug resistance / sensitivity of bacteria to antibiotics only based on the presence or absence of drug resistance genes.

[0004] Artificial intelligence (AI), especially machine learning (ML), has gradually been introduced into the field of life sciences. Machine learning guides computers to use known data to establish appropriate prediction models through mathematical algorithms, and uses these models to make judgments on new variables. Common ones include random forest (RF), support vector machines (SVM), naive Bayes (NB), and artificial neural networks (ANN), etc. As the core technology of artificial intelligence, machine learning has important potential application value in fields such as predicting AMR based on bacterial genome sequencing data due to its unique advantages in processing high dimensions and big data. More and more studies have shown that without the prior knowledge of drug resistance mechanisms, when using a sufficiently large training data set for ML model training, constructing a drug resistance prediction model can provide accurate and rapid AST predictions.

[0005] Pseudomonas aeruginosa is a highly prevalent nosocomial infection bacterium, Gram-negative, and its 6.3 Mbp genome is much larger than that of other common pathogenic bacteria, such as Escherichia coli (4.6 Mbp), Mycobacterium tuberculosis (4.4 Mbp), Bacillus subtilis (4.2 Mbp), etc. Such a large gene pool also makes it have complex drug resistance mechanisms such as natural drug resistance, acquired drug resistance, and adaptive drug resistance ( Figure 1 ), and the production of multidrug resistance or even pan-drug resistance can significantly increase the mortality rate and hospitalization costs of patients. Carbapenem antibiotics are a class of atypical β-lactam antibiotics with a broad antibacterial spectrum, strong antibacterial activity, and high safety. This class of antibiotics has become an important drug for the clinical treatment of Pseudomonas aeruginosa infections. However, due to the irregular use of carbapenem antibiotics in clinical treatment, it has exacerbated the occurrence of drug resistance of Pseudomonas aeruginosa to them, and at the same time has accelerated the emergence and spread of carbapenem-resistant Pseudomonas aeruginosa, bringing great troubles to the clinical treatment of Pseudomonas aeruginosa infections and seriously affecting the clinical anti-Pseudomonas aeruginosa infection treatment work. Therefore, rapid and accurate detection of carbapenem-resistant Pseudomonas aeruginosa is of great significance for the rational use of carbapenem antibiotics and the prevention and control of the spread of drug-resistant strains.

[0006] Clinical microbiology laboratories usually determine whether Pseudomonas aeruginosa produces carbapenemase through drug susceptibility tests, and then judge the drug resistance of Pseudomonas aeruginosa to carbapenem antibiotics. Common carbapenemase phenotype detection methods mainly include minimum inhibitory concentration (MIC) determination method, disk diffusion method, Hodge method, Carba NP method, etc.; however, these methods have disadvantages such as long detection cycle, cumbersome operation, the results being easily affected by subjective factors, limited types of culture media, and limited types of carbapenemases detected, which have an adverse impact on the precise and rapid treatment of clinical Pseudomonas aeruginosa antibiotics. At present, a few studies have applied machine learning to the prediction study of Pseudomonas aeruginosa drug resistance, aiming to construct an accurate and rapid Pseudomonas aeruginosa drug resistance prediction model suitable for clinical effective treatment, such as: training and mining from multiple dimensional features such as gene mutations, gene expressions, and gene presence and absence based on machine learning models and achieving results. However, the number of features selected by the model is too large, with a risk of overfitting, resulting in a decline in the generalization performance of the constructed prediction model and low clinical application value. Therefore, based on machine learning, developing a Pseudomonas aeruginosa-carbapenem drug resistance prediction model with both big data support and clinical application prospects for the carbapenem antibiotics mainly used in the clinical treatment of Pseudomonas aeruginosa infections is of great significance for guiding the clinical treatment of Pseudomonas aeruginosa antibiotics.

[0007] In view of this, the present application is proposed. Summary of the Invention

[0008] To solve the technical problems in the prior art, the present application innovatively proposes a screening method for the resistant phenotype characteristics of Pseudomonas aeruginosa - carbapenem drugs, as well as characteristic genes for predicting the drug sensitivity phenotype of Pseudomonas aeruginosa - carbapenem drugs screened based on this method. The present application is based on the single - bacterium genome Contig sequence for the alignment detection and identification of Pseudomonas aeruginosa and its carried resistant genes, as well as the data analysis process for predicting the drug sensitivity phenotype of Pseudomonas aeruginosa - carbapenem drugs. First, obtain the genomic data of Pseudomonas aeruginosa strains, and at the same time collect the corresponding drug sensitivity test result data; then, based on the contig sequence of the Pseudomonas aeruginosa genome, conduct alignment with the CARD resistance database and annotate the resistant genes; at the same time, detect and annotate the gene mutation sites of the Oprd gene (which plays an important role in resistance to Pseudomonas aeruginosa), and grade the degree of functional variation, and integrate the HIGH - level characteristics with the highest degree of variation into a class of key characteristics; then, for carbapenem drugs (imipenem and meropenem), conduct correlation analysis of genotype and drug - resistant phenotype data, screen important characteristic genes related to the generation of drug resistance, and calculate the weight coefficients of these important characteristic genes; finally, evaluate through ROC analysis the performance of the Pseudomonas aeruginosa - carbapenem drug sensitivity prediction machine - learning model constructed based on the screened important characteristic genes.

[0009] Secondly, although in the prior art, some genes related to Pseudomonas aeruginosa resistance are disclosed. For example, for some common carbapenem - resistant enzymes such as KPC, NDM, and VIM, the coincidence rate of clinical drug - resistant phenotypes is relatively high and is clinically recognized. For genes such as PDC and oprD, although there are literature reports suggesting that they may be related to carbapenem resistance, for this type of special sequencing data of infected metagenomes, due to the usually high content of the sample host and the complex sample composition, the detection limit of metagenomic sequencing data is generally not ideal. In practice, even for genes that may be related to carbapenem resistance, due to problems such as the sequencing abundance of the target gene, they still cannot be used for drug sensitivity prediction in actual processes. Therefore, the present application combines metagenomic big data with drug - sensitive phenotype data, and uses the LASSO machine - learning method to screen characteristic genes / combinations that can be used for drug - sensitive phenotype prediction in practical metagenomic sequencing samples.

[0010] Specifically, the present application proposes the following technical solutions:

[0011] The present application first provides a screening method for the resistant phenotype characteristics of Pseudomonas aeruginosa - carbapenem drugs, including the following steps:

[0012] 1) Obtain the genomic data of Pseudomonas aeruginosa strains from public databases, and collect the corresponding drug sensitivity test result data;

[0013] 2) Align the contig sequences of the Pseudomonas aeruginosa genome with the CARD antibiotic resistance database and annotate the antibiotic resistance genes;

[0014] 3) Detect and annotate the variant sites based on the contig sequences of the Pseudomonas aeruginosa genome to obtain a variant site spectrum;

[0015] 4) For imipenem and / or meropenem, perform a correlation analysis between the detected and annotated antibiotic resistance genes and the variant site spectrum with the antibiotic resistance phenotype data, screen the important features related to the generation of antibiotic resistance, and calculate the weight coefficients of the important feature genes;

[0016] Preferably, it further includes:

[0017] 5) Use ROC analysis to evaluate the performance of the classification model constructed based on the selected feature genes.

[0018] Further, in step 1), the public database includes the NCBI database and / or the PATRIC database;

[0019] Further, in step 3), the variant site detection and annotation are specifically: break the Pseudomonas aeruginosa genome contig sequences into fragmented sequences, align the fragment sequences to the Pseudomonas aeruginosa reference genome, and then perform variant site detection to obtain a variant site spectrum;

[0020] Further, in step 2), the alignment and antibiotic resistance gene annotation are specifically: align the contig sequences with the CARD antibiotic resistance gene reference sequence library, filter out the hits with an identity less than 80% or a reference gene coverage less than 80%, select the best hit for the aligned region on each contig as the final alignment result for that contig region, and add the annotation information of the antibiotic resistance genes, count the detection of antibiotic resistance genes in each strain, and summarize them into a matrix table; preferably, summarize them into a 0-1 matrix table, where 0 indicates that the antibiotic resistance gene is not detected, and 1 indicates that the antibiotic resistance gene is detected;

[0021] Even further, step 3) also includes: detecting and annotating the gene variant sites of genes such as Oprd, grading the degree of functional variation as high, medium, and low, and integrating the Oprd(HIGH) with the highest level of variation degree as the key feature. The variations mainly include SNP variations that cause premature termination of CDS, InDels that cause CDS frame shift, and nonsense SNP mutations at the initiator; and for genes with PPV>0.9, additional templates will be used for further variant detection.

[0022] Further, in step 4), the correlation analysis is as follows: According to the antibiotic drug classification information corresponding to each drug-resistant gene recorded in the CARD library, based on the gene detection matrix table obtained in step 2), the mutation site spectrum obtained in step 3), and the three oprD gene mutation levels obtained in step 3), a sub-matrix table of imipenem- or meropenem-related drug-resistant genes is selected, and genes with relatively low detection frequencies (preferably less than 3) and relatively low PPV (preferably less than 0.9) are filtered out. The filtered table data is used to perform correlation analysis with the imipenem or meropenem drug sensitivity results using the lasso regression model to screen for important features related to imipenem or meropenem resistance, and the weight coefficients of the important feature genes are calculated.

[0023] Further, in step 5), the evaluation is specifically as follows: Define the positive and negative interpretation index Score index: Where arg_W represents the weight coefficient value of the detected corresponding gene; based on the important gene weight coefficient matrix screened by the model, combined with the actual detection situation of the sample drug-resistant genes, the Score value of each sample is calculated, and ROC curve analysis is performed to obtain the AUC value of the training set model; further, ROC analysis is performed using the validation set to obtain the AUC value of the validation set model; the higher the AUC of the training set and validation set models, the better the performance of the method.

[0024] Further, the important features screened for imipenem or meropenem resistance are as follows:

[0025] For imipenem, the important features are: KPC beta-lactamase, VIM beta-lactamase, oprD(p.Ser278Pro), ampR(p.Ser179Thr), GES beta-lactamase, mexT(p.Asp59Ala), parS(p.Leu137Pro), oprD(p.Val359Leu), PA2020(p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD(p.Leu229Phe), ampD(p.Val10Gly), IMP beta-lactamase, ftsI(p.Phe533Leu) and / or oprD(HIGH);

[0026] For meropenem, the important characteristics are: KPC beta-lactamase, parS (p.Ala13Thr), oprD (p.Ser278Pro), IMP beta-lactamase, oprD (p.Val359Leu), ftsI (p.Ala244Thr), nalD (p.Val151Leu), oprD (p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047 (p.Phe171Leu), ampD (p.Gly121Glu), ftsI (p.Arg504Cys), mpl (p.Val124Gly), PA2020 (p.Arg87Pro), parS (p.Ala13Val), PA2020 (p.Lys55Glu), mexR (p.Ala108fs), mexT (p.Gly191Arg), PA2020 (p.Gly137Asp), parS (p.Val152Ala), ftsI (p.Pro527Ser), mexR (p.Gly101Arg), parS (p.Ala149Thr), ampD (p.Ala96Thr), ftsI (p.Phe533Leu), mexT (p.Ala143Thr), parS (p.Arg383Ser), oprD (HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase and / or OXA beta-lactamase Cluster-40.

[0027] Based on the above-mentioned important characteristic genes related to the drug-resistant phenotype of Pseudomonas aeruginosa against carbapenem drugs that have been screened and determined, this application constructs a data analysis method for directly comparing and detecting target Pseudomonas aeruginosa pathogens and their carried drug-resistant genes based on the mNGS sequencing reads sequence, and predicting the drug-resistant phenotype against carbapenem drugs. Specifically: For all Pseudomonas aeruginosa genomes in the training set included in the previous BGWAS, the target drug-resistant genes are detected by simulating NGS sequencing reads sequence alignment (reads-based) and genome Contig sequence alignment (assembly-based). Taking the detection result of the assembly-based method as a reference, the reads-based detection process is verified and optimized to achieve the purpose of accurate reads-based gene typing detection; then, a custom formula is used to calculate the Score, and this Score is used as the judgment index for predicting the drug sensitivity properties of carbapenem drugs. Combined with the reads sequence simulation test, ROC analysis is performed to determine the optimal cutoff threshold, and at the same time, the accuracy and performance of the prediction model are evaluated.

[0028] This application also provides an electronic device, including: a processor and a memory; the processor is connected to the memory, wherein the memory is used to store a computer program, and the processor is used to call the computer program to execute the method described in any one of the above.

[0029] This application also provides a computer storage medium, which stores a computer program. The computer program includes program instructions, and when the program instructions are executed by a processor, the method described in any one of the above is executed.

[0030] This application also provides the use of a reagent or component for detecting genes KPC beta-lactamase, VIM beta-lactamase, oprD (p.Ser278Pro), ampR (p.Ser179Thr), GES beta-lactamase, mexT (p.Asp59Ala), parS (p.Leu137Pro), oprD (p.Val359Leu), PA2020 (p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD (p.Leu229Phe), ampD (p.Val10Gly), IMP beta-lactamase, ftsI (p.Phe533Leu) and / or oprD (HIGH) in the preparation of a kit for predicting the imipenem drug sensitivity phenotype of Pseudomonas aeruginosa; preferably, in the preparation of a kit for predicting the imipenem drug sensitivity phenotype of Pseudomonas aeruginosa by metagenomic sequencing of infections.

[0031] The present application also provides the use of a reagent or component for detecting KPC beta-lactamase, parS (p.Ala13Thr), oprD (p.Ser278Pro), IMP beta-lactamase, oprD (p.Val359Leu), ftsI (p.Ala244Thr), nalD (p.Val151Leu), oprD (p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047 (p.Phe171Leu), ampD (p.Gly121Glu), ftsI (p.Arg504Cys), mpl (p.Val124Gly), PA2020 (p.Arg87Pro), parS (p.Ala13Val), PA2020 (p.Lys55Glu), mexR (p.Ala108fs), mexT (p.Gly191Arg), PA2020 (p.Gly137Asp), parS (p.Val152Ala), ftsI (p.Pro527Ser), mexR (p.Gly101Arg), parS (p.Ala149Thr), ampD (p.Ala96Thr), ftsI (p.Phe533Leu), mexT (p.Ala143Thr), parS (p.Arg383Ser), oprD (HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase and / or OXA beta-lactamase Cluster-40 in the preparation of a kit for predicting the meropenem susceptibility phenotype of Pseudomonas aeruginosa;

[0032] Preferably, it is used in the preparation of a kit for predicting the meropenem susceptibility phenotype of Pseudomonas aeruginosa from an infected metagenome;

[0033] Furthermore, the above-mentioned susceptibility phenotypes include resistant phenotypes and susceptible phenotypes;

[0034] Preferably, the above genes are detected simultaneously. If the test results are all negative, it can be inferred as sensitive, that is, sensitive detection is achieved. Similarly, the above genes can be detected separately. If any test result is positive, it can be inferred as resistant, that is, resistant detection is achieved.

[0035] Furthermore, the detection is performed at the nucleic acid level, and the nucleic acid level is obtained by sequencing technology, nucleic acid amplification technology, nucleic acid hybridization technology, electrophoresis technology, bio-mass spectrometry technology or chromatography technology; preferably, the method for obtaining the nucleic acid level includes, but is not limited to, any one of the following methods: gene sequencing method, polymerase chain reaction method, isothermal amplification reaction method, gene chip method, probe hybridization method, gel electrophoresis method, RNA blotting method, nucleic acid mass spectrometry method.

[0036] Furthermore, the sample for the detection can be from one or more of tissues, cells, body fluids, serum, plasma, whole blood, urine, semen, saliva, pleural effusion, ascites, cerebrospinal fluid, feces or synovial fluid.

[0037] The present application also provides a kit for predicting the imipenem drug susceptibility phenotype of Pseudomonas aeruginosa in an infected metagenome, including reagents capable of detecting KPC beta-lactamase, VIM beta-lactamase, oprD (p.Ser278Pro), ampR (p.Ser179Thr), GES beta-lactamase, mexT (p.Asp59Ala), parS (p.Leu137Pro), oprD (p.Val359Leu), PA2020 (p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD (p.Leu229Phe), ampD (p.Val10Gly), IMP beta-lactamase, ftsI (p.Phe533Leu) and / or oprD (HIGH).

[0038] The present application also provides a kit for predicting the meropenem drug susceptibility phenotype of Pseudomonas aeruginosa in infectious metagenomics, which includes reagents capable of detecting KPC beta-lactamase, parS (p.Ala13Thr), oprD (p.Ser278Pro), IMPbeta-lactamase, oprD (p.Val359Leu), ftsI (p.Ala244Thr), nalD (p.Val151Leu), oprD (p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047 (p.Phe171Leu), ampD (p.Gly121Glu), ftsI (p.Arg504Cys), mpl (p.Val124Gly), PA2020 (p.Arg87Pro), parS (p.Ala13Val), PA2020 (p.Lys55Glu), mexR (p.Ala108fs), mexT (p.Gly191Arg), PA2020 (p.Gly137Asp), parS (p.Val152Al a), ftsI (p.Pro527Ser), mexR (p.Gly101Arg), parS (p.Ala149Thr), ampD (p.Ala96Thr), ftsI (p.Phe533Leu), mexT (p.Ala143Thr), parS (p.Arg383Ser), oprD (HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase and / or OXA beta-lactamase Cluster-40.

[0039] Furthermore, the above-mentioned drug susceptibility phenotypes include drug resistance phenotypes and drug susceptibility phenotypes;

[0040] Preferably, the above genes are detected simultaneously. If the test results are all negative, it can be inferred as sensitive, that is, sensitive detection is achieved; similarly, the above genes can be detected separately. If any test result is positive, it can be inferred as drug-resistant, that is, drug-resistant detection is achieved.

[0041] Further, by way of example, the term "oprd(HIGH)" in the text refers to a type of variation in the oprD gene that has a relatively large impact on gene function, such as SNP variations that cause premature termination of the CDS, InDels that cause CDS frame shift, and nonsense SNP mutations in the initiator, etc., which have a relatively large impact on gene function; for example, oprD(p.Ser278Pro) refers to a gene mutation in oprD, and the mutation causes the Ser at position 278 in the corresponding coding region to mutate to Pro.

[0042] The present application also provides a method for predicting the susceptibility phenotypes of Pseudomonas aeruginosa to imipenem and / or meropenem, including the step of obtaining the levels of the above-mentioned important characteristic genes in a subject sample.

[0043] Preferably, the method includes the following steps:

[0044] (i) Obtain the levels of the above-mentioned important characteristics in the subject sample in the sample;

[0045] (ii) Compare the levels of the above-mentioned important characteristic genes with those of a control sample; wherein, the significant difference in the levels of the above-mentioned important characteristic genes between the subject sample and the control sample is an indication of the subject's drug susceptibility.

[0046] Alternatively, (ii) compare with a set threshold absolute amount; wherein, the subject sample level being higher than the threshold absolute amount is an indication of the subject's drug susceptibility.

[0047] Advantageous technical effects of the present application:

[0048] 1) The present application expands the application of machine learning technology in the research direction of Pseudomonas aeruginosa drug resistance. It mainly conducts correlation analysis on the deletion characteristics obtained from non-core genes related to the generation of drug resistance phenotypes, finds out important drug resistance genes with a high contribution degree to the drug resistance phenotype, and at the same time calculates the corresponding weight coefficients of these drug resistance genes, facilitating the transformation and application to clinical drug resistance detection.

[0049] 2) Through bioinformatics means, this application conducts a correlation analysis (BGWAS) based on the genomic data of a large sample size of Pseudomonas aeruginosa strains and the drug susceptibility test data of carbapenem drugs (imipenem and meropenem), constructs a machine learning model for predicting antibiotic drug susceptibility, mines and screens important genes or new features associated with the drug-resistant phenotype, and calculates the relative weight coefficients of each gene or feature to quantify the important impact degree on the drug-resistant phenotype. It can achieve the detection of drug-resistant genes of Pseudomonas aeruginosa - carbapenem and the prediction of drug-resistant phenotypes without being restricted by traditional cultivation, with characteristics such as high speed and accuracy. At the same time, it also lays a foundation for the subsequent development of a rapid detection kit (Pannel) for Pseudomonas aeruginosa - carbapenem drug-resistant detection. In addition, by grading the characteristics of oprD mutation sites and integrating important mutations into the HIGH level, the generalization ability of the machine learning model for predicting Pseudomonas aeruginosa - carbapenem drug resistance is significantly improved.

[0050] 3) When detecting drug-resistant genes, this application directly conducts drug-resistant gene detection and annotation based on the comparison method between the genomic contig sequence and the public CARD drug-resistant database, bypasses the gene prediction and the link of detecting drug-resistant genes based on the cds sequence obtained by prediction, and avoids the deviation that may be introduced in the gene prediction process.

[0051] 4) This application determines important drug-resistant genes or features with predictive significance. By directly performing metagenomic sequencing (mNGS) on clinical specimens to identify Pseudomonas aeruginosa and the presence of drug-resistant genes, based on the detection or non-detection of these important drug-resistant genes or features, the drug susceptibility results of carbapenem drugs (imipenem and meropenem) are directly predicted. This application can effectively predict the drug susceptibility results of carbapenem drugs (imipenem, meropenem), and the prediction accuracy is extremely high: Sampling verification of clinical specimens shows that the overall drug susceptibility prediction accuracy (PPV) of carbapenem can reach more than 89%, and the proportion of sample cases with clear drug susceptibility prediction results is more than 60%. This is not easy in practice, and the effect exceeds the conventional expectation. Description of the Drawings

[0052] Figure 1 Mechanism of Pseudomonas aeruginosa drug resistance;

[0053] Figure 2 Technical route for constructing the Pseudomonas aeruginosa drug resistance model;

[0054] Figure 3 Bubble chart of the influence degree of different drug-resistant genes;

[0055] Figure 4 Heat map of the detection distribution of oprD HIGH mutations in imipenem;

[0056] Figure 5, Heat map of the detection distribution of meropenem oprD HIGH mutations;

[0057] Figure 6 , CV error rate and AUC change curves of the model under different numbers of characteristic genes (P. aeruginosa - imipenem);

[0058] Figure 7 , Characteristic genes of P. aeruginosa - imipenem drug - resistant phenotype;

[0059] Figure 8 , Characteristic genes of P. aeruginosa - meropenem drug - resistant phenotype;

[0060] Figure 9 , Verification of the reliability of P. aeruginosa - imipenem characteristic genes by ROC curve;

[0061] Figure 10 , Verification of the reliability of P. aeruginosa - meropenem characteristic genes by ROC curve;

[0062] Figure 11 , Change curve graph of the performance (AUC value) of the drug - resistant prediction models of imipenem and meropenem under simulated different sequencing data volumes;

[0063] Figure 12 , For the training set and the validation set, simulate the performance (AUC value) of the imipenem and meropenem prediction models under 15X genome data volume. Detailed implementation manners

[0064] The following will describe the implementation schemes of the present application in detail in combination with embodiments. However, those skilled in the art will understand that the following embodiments are only used to illustrate the present application and should not be regarded as limiting the scope of the present application. For those conditions not specified in the embodiments, they are carried out according to conventional conditions or the conditions recommended by the manufacturer. For the reagents or instruments not specified by the manufacturer, they are all conventional products that can be obtained through market purchase.

[0065] Definition of some terms

[0066] Unless otherwise defined hereinafter, the meanings of all technical terms and scientific terms used in the specific implementation manners of the present application are intended to be the same as those generally understood by those skilled in the art. Although the following terms are believed to be well - understood by those skilled in the art, the following definitions are still set forth to better explain the present application.

[0067] As used in this application, the terms "comprising", "including", "having", "containing" or "involving" are inclusive or open-ended and do not exclude other unrecited elements or method steps. The term "consisting of" is considered a preferred embodiment of the term "comprising". If a group is defined hereinafter as including at least a certain number of embodiments, this should also be understood to disclose a group preferably consisting only of these embodiments.

[0068] The indefinite or definite articles used in reference to singular nouns, such as "a" or "an", "the", include the plural forms of such nouns.

[0069] The term "about" in this application means an accuracy range that those skilled in the art can understand and still ensure the technical effect of the feature being discussed. This term generally means ±10% deviation from the indicated value, preferably ±5%.

[0070] In addition, the terms first, second, third, (a), (b), (c), and the like in the specification and claims are used to distinguish similar elements and are not necessarily for describing order or time sequence. It should be understood that the terms so applied can be interchanged in appropriate circumstances, and the embodiments described in this application can be implemented in an order different from that described or illustrated in this application.

[0071] The present application will be described below in conjunction with specific embodiments.

[0072] Example 1. Screening for the phenotypic drug resistance characteristics of Pseudomonas aeruginosa - carbapenem

[0073] Figure 2 The overall technical roadmap screened for this application is described in detail as follows:

[0074] Step 1: Search and download the genomic data of Pseudomonas aeruginosa strains and their corresponding antibiotic susceptibility test result data from public databases.

[0075] Download from the NCBI NDARO database: Open the website https: / / www.ncbi.nlm.nih.gov / pathogens / isolates, enter "Pseudomonas aeruginosa" in the search bar to retrieve information on Pseudomonas aeruginosa. Then, in the Matched Isolates sub-window, click "Choose columns" and select "AST pheotypes" to display the information in this column. Next, download the table data of the entire window, organize the Pseudomonas aeruginosa strains with drug susceptibility test results data, and batch download the genomic sequences from the NCBI genome database (ftp: / / ftp.ncbi.nlm.nih.gov / genomes) according to the Assembly ID information.

[0076] Download from the PATRIC database: Open the website https: / / patricbrc.org, click the BACTERIA button in the BROWSE column of the search window. First, select "AMR Phenotypes", enter "Pseudomonas aeruginosa" in the KEYWORDS column for screening, and at the same time filter out the entries with "Evidence" column as "Computational Method" and only retain the entries with "Laboratory Method" to obtain the drug susceptibility information of Pseudomonas aeruginosa strains and download the data table. Then, select "Genome", add the "Assembly Accession" column information and download the data table. Find the genomic PATRIC ID or Assembly ID of the strains with drug susceptibility test results data according to the Genome ID in the two downloaded tables, and then batch download the genomic sequences from the PATRIC or NCBI genome database (ftp: / / ftp.ncbi.nlm.nih.gov / genomes).

[0077] Merge the genomes collected and downloaded from the NCBI Pathogen Detection and PATRIC databases, and filter out redundant genomes. Finally, obtain a total of 2275 Pseudomonas aeruginosa genomes and their drug susceptibility test results data. At the same time, randomly divide the strains into two subsets in a random manner to be used as the model training set and validation set respectively. Among them, the number of strains against imipenem and meropenem is as follows in the table:

[0078]

[0079] Step 2: Perform a comparison with the CARD resistance database and detect and annotate resistance genes (ARGs) based on the downloaded Pseudomonas aeruginosa genomic contig sequences. Use RGI (v5.2.1) to detect resistance genes, filter out hits with an identity less than 80% or a reference gene coverage less than 80%, and then select the best hit (first hit) for the aligned region on each contig as the final alignment result for that contig region, and add the annotation information of the resistance genes. Count the detection of resistance genes in each strain, and finally summarize them into a 0-1 matrix table, where 0 indicates that no resistance gene is detected, and 1 indicates that the resistance gene is detected.

[0080] Step 3: Break the downloaded Pseudomonas aeruginosa genomic contig sequences into 100-bp sequences, and use BWA (v0.7.17-r1198-dirty) to align the sequence fragments to the Pseudomonas aeruginosa reference genome (GCF_000006765.1) to obtain the initial alignment result in BAM format. Then, use the Pisces software (5.2.5.20) to detect variant sites and obtain a VCF text file for describing variant results such as SNPs and InDels. Then, input the VCF file into the SnpEff software (V5.0) to perform functional annotation on the detected variant sites. The gene list related to carbapenem resistance is obtained by referring to the literature as follows:

[0081]

[0082]

[0083] Calculate the overall PPV of each gene in the above classifications based on the variant impact levels (HIGH, MODERATE, LOW) given by the SnpEff software. For a certain type, especially the HIGH type variant, if the gene with PPV >= 0.9 or above is determined as a gene highly related to drug resistance, the model will be directly constructed based on the major categories of variant impact levels during the subsequent modeling process. The PPV of each gene variant impact level is as Figure 3 shown.

[0084] At the same time, in order to further improve the sensitivity of the model and recall more samples carrying key variant features. For genes with PPV > 0.9, additional templates will be used for variant detection. Here, only the HIGH of the oprD gene satisfies PPV greater than 0.9. Therefore, ATCC_27853, F23197, FRD1, LESB58, MTB-1, PA-VAP-4, and UCBPP-PA14 are added as templates to perform variant detection on oprD (see Figure 3in oprD of PAO1+). In addition, mutation detection based on the HIGH dimension further improves the generalization ability of the model, as shown in Figures 4 - 5 , many mutation sites only appear in the training set, while some mutation sites only appear in the validation set. These samples can be identified if detected based on the HIGH dimension.

[0085] Step 4: Based on the training samples, use the lasso regression model to perform association analysis on genotype and antibiotic resistance phenotype data to screen out important characteristic genes related to drug resistance. Taking imipenem as an example, other drugs are similar. According to the antibiotic drug classification information corresponding to each drug resistance gene recorded in the CARD library, based on the gene detection matrix table obtained in Steps 2 and 3 and the oprD gene with 3 mutation levels, select the sub-matrix table of imipenem-related drug resistance genes, and filter out genes with a low detection frequency (preferably, the detection frequency is less than 3) and a low PPV (preferably, the PPV is less than 0.9), and then use the filtered table data to perform association analysis with the imipenem drug susceptibility results. The format of the sub-matrix table data (denoted as X) is as follows (only part of the data is presented):

[0086]

[0087]

[0088] The format of the imipenem drug susceptibility result data (denoted as Y) is as follows:

[0089] Sample AST 11 R 1 R 23 R 28 R 287.1477 R 287.2972 R ... ...

[0090] Taking the above two data sets (X and Y) as inputs, use the glmnet package in R language to perform association analysis on genotype and drug resistance phenotype data, and perform 10-fold cross-validation to screen out important characteristic genes related to imipenem resistance. Part of the running program code is as follows:

[0091] library(glmnet)

[0092] cv.model<-

[0093] cv.glmnet(X,Y,family="binomial",nlambda=100,alpha=1,standardize=F,nfolds=10,type.measure="class")

[0094] coefficients<-coef(cv.model,s=cv.model$lambda.min)

[0095] library(pROC)

[0096] predict<-predict(cv.model,trainx,s=cv.model$lambda.min,type="response")

[0097] roc.predict<-roc(Y,as.numeric(predict)) ...

[0099] After running the above program, the genes associated with carbapenem resistance were analyzed. Given that the number of genes obtained in the initial analysis is often large, it is necessary to rank these genes according to their importance and finally select the important genes with the highest ranking.

[0100] The important gene selection process is as follows:

[0101] Based on the ranked genes obtained from the preliminary analysis, a gradient was set to select different numbers of gene combinations. The Lasso regression model was constructed according to the above program code to obtain the model AUC value, CV error rate, and AUC-error difference, and then a curve was drawn. The number of horizontal coordinates corresponding to the first inflection point drop in the process of gradual increase of AUC-error value, or the number of horizontal coordinates corresponding to the first maximum value of AUC or the first minimum value of error was used to select the number of genes as the final selection (see Figure 6 ).

[0102] Here, 14 important gene variation features were finally screened out for imipenem based on the machine learning model.

[0103] In summary, the important genes related to carbapenem resistance of Pseudomonas aeruginosa and their weight coefficients are finally screened out as follows:

[0104]

[0105] It can be seen that for imipenem, the important genes / gene characteristics associated with the drug-resistant phenotype screened include (see Figure 7): KPC beta-lactamase, VIM beta-lactamase, oprD(p.Ser278Pro), ampR(p.Ser179Thr), GES beta-lactamase, mexT(p.Asp59Ala), parS(p.Leu137Pro), oprD(p.Val359Leu), PA2020(p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD(p.Leu229Phe), ampD(p.Val10Gly), IMP beta-lactamase, ftsI(p.Phe533Leu) and oprD(HIGH).

[0106] For meropenem, the important genes / gene features related to the drug-resistant phenotype screened out include (see Figure 8 ): KPC beta-lactamase, parS(p.Ala13Thr), oprD(p.Ser278Pro), IMP beta-lactamase, oprD(p.Val359Leu), ftsI(p.Ala244Thr), nalD(p.Val151Leu), oprD(p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047(p.Phe171Leu), ampD(p.Gly121Glu), ftsI(p.Arg504Cys), mpl(p.Val124Gly), PA2020(p.Arg87Pro), parS(p.Ala13Val), PA2020(p.Lys55Glu), mexR(p.Ala108fs), mexT(p.Gly191Arg), PA2020(p.Gly137Asp), parS(p.Val152Ala), ftsI(p.Pro527Ser), mexR(p.Gly101Arg), parS(p.Ala149Thr), ampD(p.Ala96Thr), ftsI(p.Phe533Leu), mexT(p.Ala143Thr), parS(p.Arg383Ser), oprD(HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase and OXA beta-lactamase Cluster-40.

[0107] Step 5. ROC analysis to determine the performance of the classification model constructed based on the screened feature genes

[0108] Define the Score metric ( where arg_W represents the weight coefficient value of the detected corresponding gene), and use this as the positive and negative judgment metric. For imipenem, based on the important gene weight coefficient matrix screened by the above model, combined with the actual detection situation of the sample drug-resistant genes, calculate the Score value of each sample, and then perform ROC curve analysis. The AUC value of the training set (n = 695) for model A is 0.9056. Then further perform ROC analysis on the validation set (n = 177), and the obtained AUC value is 0.8869 (see Figure 9 ). The relatively high AUC values of the training set and validation set models indicate that the method of the present application has good performance, that is, the screening method and model of the present application are accurate and effective.

[0109] Similarly, for meropenem, the AUC values of the training set (n = 1121) and the validation set (n = 282) are 0.924 and 0.8878 respectively ( Figure 10 ).

[0110] Therefore, based on the model of the present application, according to the important feature genes related to carbapenem-resistant Pseudomonas aeruginosa screened out, when detecting these feature genes and combining the gene weight coefficients obtained by the model, the drug sensitivity results of the corresponding antibiotics can be predicted.

[0111] For imipenem, the target resistance genes / gene signatures include KPC beta-lactamase, VIM beta-lactamase, oprD (p.Ser278Pro), ampR (p.Ser179Thr), GES beta-lactamase, mexT (p.Asp59Ala), parS (p.Leu137Pro), oprD (p.Val359Leu), PA2020 (p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD (p.Leu229Phe), ampD (p.Val10Gly), IMP beta-lactamase, ftsI (p.Phe533Leu), oprD (HIGH); considering factors such as comprehensive gene weights, gene occurrence frequencies, and all possible mechanisms of resistance generation, in practice, resistance phenotype prediction can be carried out through KPC beta-lactamase, VIM beta-lactamase, oprD (p.Ser278Pro), ampR (p.Ser179Thr), GES beta-lactamase, mexT (p.Asp59Ala), parS (p.Leu137Pro), oprD (p.Val359Leu), PA2020 (p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD (p.Leu229Phe), ampD (p.Val10Gly), IMP beta-lactamase, ftsI (p.Phe533Leu), and oprD (HIGH) that occur frequently and have high weights and mainly mediate resistance generation.

[0112] For meropenem, the target genes / gene features include: KPC beta-lactamase, parS (p.Ala13Thr), oprD (p.Ser278Pro), IMP beta-lactamase, oprD (p.Val359Leu), ftsI (p.Ala244Thr), nalD (p.Val151Leu), oprD (p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047 (p.Phe171Leu), ampD (p.Gly121Glu), ftsI (p.Arg504Cys), mpl (p.Val124Gly), PA2020 (p.Arg87Pro), parS (p.Ala13Val), PA2020 (p.Lys55Glu), mexR (p.Ala108fs), mexT (p.Gly191Arg), PA2020 (p.Gly137Asp), parS (p.Val152Ala), ftsI (p.Pro527Ser), mexR (p.Gly101Arg), parS (p.Ala149Thr), ampD (p.Ala96Thr), ftsI (p.Phe533Leu), mexT (p.Ala143Thr), parS (p.Arg383Ser), oprD (HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase, OXA beta-lactamase Cluster-40.Considering factors such as comprehensive gene weights, gene occurrence frequencies, and all possible mechanisms of drug resistance generation, it can be known that in practice, for KPC beta-lactamase, parS (p.Ala13Thr), oprD (p.Ser278Pro), IMP beta-lactamase, oprD (p.Val359Leu), ftsI (p.Ala244Thr), nalD (p.Val151Leu), oprD (p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047 (p.Phe171Leu), ampD (p.Gly121Glu), ftsI (p.Arg504Cys), mpl (p.Val124Gly), PA2020 (p.Arg87Pro), parS (p.Ala13Val), PA2020 (p.Lys55Glu), mexR (p.Ala108fs), mexT (p.Gly191Arg), PA2020 (p.Gly137As p), parS (p.Val152Ala), ftsI (p.Pro527Ser), mexR (p.Gly101Arg), parS (p.Ala149Thr), ampD (p.Ala96Thr), ftsI (p.Phe533Leu), mexT (p.Ala143Thr), parS (p.Arg383Ser), oprD (HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase, and OXA beta-lactamase Cluster-40, which have a high occurrence frequency and high weight and mainly mediate drug resistance generation, drug resistance phenotype prediction is carried out.

[0113] Example 2: Detection and identification of carbapenem-resistant genes and prediction of drug resistance phenotypes of Pseudomonas aeruginosa based on metagenomic sequencing technology

[0114] Step 1: Based on the genome of Pseudomonas aeruginosa strains, important carbapenem-related drug resistance genes are screened by a machine learning model, and the weight coefficients of the genes are calculated (see Example 1 for details).

[0115] Step 2: Based on the genomes of 2275 Pseudomonas aeruginosa strains in the training set, NGS sequencing reads are simulated, and carbapenem-resistant genes are detected and the process is corrected by a reads-based alignment method.

[0116] Use the ART_Illumina software (Version 2.5.8) to simulate 75bp short reads (NGS sequencing platform) for the test and verification of the reads-based drug resistance gene detection process. Simulate different gradient data volumes such as 0.05X, 0.1X, 0.2X, 0.3X, 0.4X, 0.5X, 0.6X, 0.7X, 0.8X, 0.9X, 1X, 2X, 3X, 5X, 10X, 30X, etc. (parameter settings: -ss NS50 -l75 -f 5 -nf 0 -rs 1), and then perform the detection and screening of carbapenem drug resistance genes. For the detailed method, please refer to the applicant's early patent CN202111680866.8. After obtaining the drug resistance gene detection results of each simulated specimen, according to the weight coefficients of important genes and corresponding gene families, define and calculate the Score index value of the sample. The calculation formula is as follows:

[0117]

[0118] In the formula, genevariation_wi represents the weight coefficient of the drug resistance characteristic variation characteristic, and genefamily_wi represents the weight coefficient of the corresponding gene family. If only a certain type of characteristic is detected, such as gene variation, the drug resistance model is mainly driven by variation.

[0119] For the simulated tests at different data volumes based on the training set strains, based on the drug resistance gene detection results, the actual drug susceptibility results of the training set strains, and the sample Score index value, perform ROC curve analysis to obtain the AUC values of the model performance of meropenem and imipenem at different data volumes, and then draw a curve of the change of the model performance AUC value as Figure 11 . When the data volume is 15X, the model performance has been stable. At this time, the coverage of Pseudomonas aeruginosa genome is greater than 95%. The AUC values of the model performance of meropenem and imipenem at this time are as Figure 12 . Finally, determine the reporting rules ("resistant" or "sensitive") and cutoff thresholds for Pseudomonas aeruginosa against meropenem and imipenem as shown in the following table. For the detailed method, please refer to the method in the applicant's early patent CN202111680866.8.

[0120]

[0121] Example 3: Detection and verification of meropenem and imipenem drug resistance genes in clinical samples

[0122] In this embodiment, 105 clinical samples containing Pseudomonas aeruginosa identified by clinical culture were collected. After extracting nucleic acids from all clinical samples, a metagenomic next-generation library (with an insert fragment length of 200 - 400 bp) was constructed and sequenced on the second-generation (Illumina nextseq CN500 SE75) instrument, followed by bioinformatics analysis.

[0123] Based on the data analysis process for predicting antibiotic resistance phenotypes by aligning gene sequencing reads, the pathogenic bacteria and their carried resistance genes in the downloaded data were identified, and then the Score value and drug sensitivity results were calculated for prediction and judgment.

[0124] Finally, the detection and identification results of Pseudomonas aeruginosa and its resistance genes and the drug sensitivity prediction results of each specimen were obtained. The results of 43 clinical metagenomic samples are shown in the following table:

[0125]

[0126] Note: ND indicates not detected, and "-" indicates not predicted.

[0127] After statistics, the drug sensitivity prediction accuracy rate and the proportion of reportable sample numbers are shown in the following table:

[0128]

[0129] The above results indicate that the combination of resistance characteristic genes determined in this application can effectively and accurately identify Pseudomonas aeruginosa and its carried resistance genes in clinical samples, and can effectively predict the drug sensitivity results of imipenem and meropenem, and can be used to assist clinical detection and diagnosis of drug-resistant Pseudomonas aeruginosa infections.

[0130] Therefore, for imipenem, a carbapenem drug, considering factors such as gene weights, gene family weights, gene occurrence frequencies, and all possible mechanisms of drug resistance generation, by simultaneously detecting the genes with high occurrence frequencies and high weights that mainly mediate drug resistance generation, namely KPC beta-lactamase, VIM beta-lactamase, oprD (p.Ser278Pro), ampR (p.Ser179Thr), GES beta-lactamase, mexT (p.Asp59Ala), parS (p.Leu137Pro), oprD (p.Val359Leu), PA2020 (p.Gln134*), OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD (p.Leu229Phe), ampD (p.Val10Gly), IMP beta-lactamase, ftsI (p.Phe533Leu), and oprD (HIGH), if the test results are all negative, it can be inferred that it is sensitive, that is, drug susceptibility is achieved; similarly, these genes can be detected separately, and if the test results of any one or more drug resistance characteristics are positive, it can be inferred that it is drug resistant.

[0131] For meropenem, a carbapenem antibiotic, considering factors such as gene weights, gene family weights, gene occurrence frequencies, and all possible mechanisms of drug resistance, simultaneous detection of the genes KPC beta-lactamase, parS(p.Ala13Thr), oprD(p.Ser278Pro), IMP beta-lactamase, oprD(p.Val359Leu), ftsI(p.Ala244Thr), nalD(p.Val151Leu), oprD(p.Gly316Asp), OXA beta-lactamase Cluster-121, PA3047(p.Phe171Leu), ampD(p.Gly121Glu), ftsI(p.Arg504Cys), mpl(p.Val124Gly), PA2020(p.Arg87Pro), parS(p.Ala13Val), PA2020(p.Lys55Glu), mexR(p.Ala108fs), mexT(p.Gly191Arg), PA2020(p.Gly137Asp), parS(p.Val152Al a), ftsI(p.Pro527Ser), mexR(p.Gly101Arg), parS(p.Ala149Thr), ampD(p.Ala96Thr), ftsI(p.Phe533Leu), mexT(p.Ala143Thr), parS(p.Arg383Ser), oprD(HIGH), OXA beta-lactamase Cluster-52, VIM beta-lactamase, and OXA beta-lactamase Cluster-40, which are highly frequent and have high weights in mediating drug resistance, can be performed. If the detection result of any one or more drug resistance characteristics is positive, drug resistance can be inferred; if all detection results are negative, susceptibility can be inferred, that is, drug susceptibility testing is achieved.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. Use of a reagent for detecting a set of important characteristic genes, namely KPC beta-lactamase, VIM beta-lactamase, oprD p.Ser278Pro, ampR p.Ser179Thr, GES beta-lactamase, mexT p.Asp59Ala, parS p.Leu137Pro, oprD p.Val359Leu, PA2020 p.Gln134*, OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD p.Leu229Phe, ampD p.Val10Gly, IMP beta-lactamase, ftsI p.Phe533Leu and a high-level variant of Oprd in the preparation of a kit for predicting the imipenem susceptibility phenotype of Pseudomonas aeruginosa; the high-level variant of Oprd includes SNP mutations that cause premature termination of the CDS, InDels that cause CDS frame shift, and nonsense SNP mutations in the initiator; the susceptibility phenotype includes a drug-resistant phenotype and a drug-sensitive phenotype.

2. A kit for predicting the imipenem drug susceptibility phenotype of Pseudomonas aeruginosa, characterized in that, The kit is composed of a reagent capable of detecting a set of characteristic genes, namely KPC beta-lactamase, VIM beta-lactamase, oprD p.Ser278Pro, ampR p.Ser179Thr, GES beta-lactamase, mexT p.Asp59Ala, parS p.Leu137Pro, oprD p.Val359Leu, PA2020 p.Gln134*, OXA beta-lactamase Cluster-40, PER beta-lactamase, oprD p.Leu229Phe, ampD p.Val10Gly, IMP beta-lactamase, ftsI p.Phe533Leu and a high-level variant of Oprd; the high-level variant of Oprd includes SNP mutations that cause premature termination of the CDS, InDels that cause CDS frame shift, and nonsense SNP mutations in the initiator; the susceptibility phenotype includes a drug-resistant phenotype and a drug-sensitive phenotype.

Images