Rapid detection method for bacterial drug resistance by combining bacterial metabolism fingerprint spectrum after short-time antibiotic stimulation and machine learning

Abstract

The invention discloses a method for rapidly detecting bacterial drug resistance by combining a bacterial metabolism fingerprint spectrum after short-time antibiotic stimulation with machine learning. Bacteria are incubated in a liquid culture medium containing antibiotics with break point concentration for a short time, the metabolic pathway of the bacteria can be changed, and then generation of downstream metabolites is affected. As the drug-resistant bacteria have a special drug-resistant mechanism, the change of the metabolic state of the drug-resistant bacteria is different from that of the sensitive bacteria. Analyzing the metabolite fingerprint spectrum of the bacteria after short-time stimulation of the antibiotics by adopting matrix-assisted laser desorption ionization time-of-flight mass spectrometry; after data acquisition, classified analysis is performed through machine learning, so that rapid detection of the antibiotic resistance of bacteria is realized. Compared with the prior art, the method has the advantages that bacterial drug resistance detection can be completed within 2 hours, high practicability and operability are achieved, and meanwhile the method has the advantages of being high in throughput and low in experiment cost.

Description

Technical Field

[0001] The present invention belongs to the field of medical detection, and particularly relates to a rapid detection method for bacterial drug resistance, and specifically relates to a rapid detection method for bacterial drug resistance combining a bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning. Background Art

[0002] The abuse of antibiotics has accelerated the process of bacterial resistance and has gradually become a global health problem. In addition, the prevalence of viral infections has further weakened the antibiotic management system, making the problem of bacterial resistance more serious. Rapid acquisition of bacterial resistance information is the key to addressing this problem. Although traditional antibiotic sensitivity tests can accurately determine the minimum inhibitory concentration, they take a long time, requiring at least 1 day of bacterial pre-culture and 16 hours of antibiotic sensitivity testing. Although the newer gene and genome-based detection methods have a certain degree of accuracy, they are difficult to be widely used in clinical practice due to their high cost, high technical requirements, slow speed and the influence of plasmid migration. In recent years, matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS) has been widely used for microbial identification, but has not yet been directly applied to drug resistance detection. The traditional MALDI-TOFMS microbial identification method mainly targets proteins and peptides from 2000Da to 20000Da, ignoring small molecule metabolites that can better reflect bacterial phenotypes and dynamic changes. For example, CN117612747A discloses a method and device for predicting drug sensitivity of Klebsiella pneumoniae, using isolated strains carrying drug sensitivity markers as test samples, and obtaining all mass spectrometry characteristic peaks corresponding to several test samples based on the MALDI-TOF mass spectrometry characteristic peaks corresponding to each of the test samples, generating a characteristic peak data set, and performing supervised machine learning model training after processing to obtain a drug sensitivity prediction model. However, this method relies on detecting protein characteristic peaks, does not introduce antibiotic stimulation, and cannot reflect the dynamic changes of bacteria under antibiotic pressure. In addition, this method is limited to β-lactam drugs and is difficult to generalize to sensitivity prediction of other antibiotics. Therefore, there is an urgent need for a fast, low-cost, phenotype-based and widely applicable drug resistance detection method for multiple antibiotics to meet the challenges of bacterial resistance problems in clinical practice. Summary of the invention

[0003] The present invention aims to overcome the defects in the prior art and provide a method for rapid detection of bacterial resistance by combining bacterial metabolic fingerprints after short-term antibiotic stimulation with machine learning, specifically a rapid antibiotic resistance (AMR) test method using MALDI-TOF MS to detect bacterial metabolic fingerprints after short-term antibiotic stimulation. This method overcomes the disadvantage that traditional antibiotic sensitivity tests are too time-consuming and further expands the clinical application of MALDI-TOF MS. The present invention has the advantages of simple operation, high efficiency, low cost, high throughput, etc., the required experimental consumables are easily available, and the test can be completed within 2 hours.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] The present invention provides a method for rapid detection of bacterial drug resistance by combining bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning.

[0006] The present invention combines the bacterial metabolic fingerprint spectrum after short-term antibiotic stimulation with machine learning to perform rapid testing of bacterial resistance. The specific steps are as follows:

[0007] S1: prepare antibiotic solution;

[0008] S2: inoculating bacteria into a culture container, adding the antibiotic solution prepared in step S1, and incubating the bacteria for a short time to obtain bacteria incubated with antibiotics;

[0009] S3: centrifuging and washing the incubated bacteria obtained in step S2, and extracting metabolites thereof;

[0010] S4: dropping the bacterial metabolites obtained in step S3 onto a target plate of matrix-assisted laser desorption ionization time-of-flight mass spectrometry, and covering the plate with a matrix after natural drying to obtain a sample;

[0011] S5: analyzing the sample in step S4 using matrix-assisted laser desorption ionization time-of-flight mass spectrometry to obtain a metabolic fingerprint of the incubated bacteria;

[0012] S6: performing data preprocessing on the metabolic fingerprint spectrum obtained in step S5 to obtain preprocessed data;

[0013] S7: Analyze the pre-processed data obtained in step S6 using a classification algorithm based on machine learning to directly evaluate the antibiotic resistance of bacteria.

[0014] Furthermore, the final antibiotic concentration in step S2 (the concentration of the antibiotic after the bacterial suspension and the antibiotic solution are combined) is the antibiotic breakpoint concentration specified by EUCAST.

[0015] Furthermore, the final concentration of bacteria in step S2 (the concentration of bacteria after the bacterial suspension and the antibiotic solution are combined) ranges from 5×10 5 ~8×10 5 CFU / mL.

[0016] Furthermore, the bacterial incubation time in step S2 is controlled between 10 minutes and 2 hours.

[0017] Furthermore, the matrix in step S4 includes one or more of common organic matrix and inorganic nanomaterial matrix.

[0018] Furthermore, the organic matrix can be selected from α-cyano-4-hydroxycinnamic acid, sinapinic acid, ferulic acid, and the like.

[0019] Furthermore, the inorganic nanomaterial matrix can be selected from TiO2 nanoparticles and the like.

[0020] Furthermore, the data preprocessing step of the metabolic fingerprint spectrum in step S6 includes operations such as background subtraction, baseline smoothing, binning and peak extraction.

[0021] Furthermore, the machine learning model involved in step S7 includes one or more models such as logistic regression, support vector machine, multi-layer perceptron, random forest and neural network.

[0022] Furthermore, in the classification analysis based on machine learning in step S7, the test result of the best performing model among multiple machine learning models can be taken as the test result of the final sample.

[0023] Compared with the prior art, the present invention has the following advantages:

[0024] 1) The innovative use of MALDI-TOF MS metabolic fingerprint analysis for AMR testing can quickly detect changes in bacterial metabolites under the action of antibiotics, thereby identifying their resistance tags.

[0025] 2) Use machine learning models to analyze MALDI-TOF metabolic fingerprint data and deeply explore the potential associations in mass spectrometry information.

[0026] 3) The experimental results show that compared with the clinical gold standard - the microbroth dilution method, the accuracy of this method can reach more than 80%, and it can be used as a rapid screening method for bacterial resistance.

[0027] 4) The entire test process can be completed within 2 hours, and the test speed is fast.

[0028] 5) This method expands the application of MALDI-TOF MS in clinical microbiology laboratories, and realizes rapid bacterial identification and drug resistance testing on the same platform. It has the characteristics of simple operation, high throughput and low cost, and has the potential for large-scale application. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 Schematic diagram of the basic process of the rapid AMR testing method of the present invention.

[0030] Figure 2 The MALDI-TOF raw metabolic fingerprints of Escherichia coli resistant and sensitive to ceftriaxone sodium after short-time incubation with antibiotics.

[0031] Figure 3 Confusion matrix for typing Escherichia coli resistant or sensitive to ceftriaxone sodium. DETAILED DESCRIPTION

[0032] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several modifications and improvements may be made without departing from the concept of the present invention. These all fall within the scope of protection of the present invention.

[0034] Any features such as preparation methods, materials, structures or composition ratios that are not clearly described in this technical solution shall be deemed as common technical features disclosed in the prior art.

[0035] The present invention discloses a method for rapid detection of bacterial drug resistance by combining bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning. By incubating bacteria in a liquid culture medium containing breakpoint concentration antibiotics for a short time, the metabolic pathway of the bacteria will change, thereby affecting the generation of downstream metabolites. Since drug-resistant bacteria have a special drug resistance mechanism, the change in their metabolic state is different from that of sensitive bacteria. Matrix-assisted laser desorption ionization time-of-flight mass spectrometry is used to analyze the metabolite fingerprint of bacteria after short-term antibiotic stimulation; after data acquisition, classification analysis is performed through machine learning, thereby realizing rapid detection of bacterial antibiotic resistance. Compared with existing methods, this method can complete bacterial resistance detection within 2 hours, has high practicality and operability, and has the characteristics of high throughput and low experimental cost.

[0036] like Figure 1 As shown, the present invention provides a rapid detection method for bacterial drug resistance (rapid AMR test method) combining bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning, comprising the following steps:

[0037] S1: prepare antibiotic solution;

[0038] S2: inoculating bacteria into a culture container, adding the antibiotic solution prepared in step S1, and incubating the bacteria for a short time to obtain bacteria incubated with antibiotics;

[0039] S3: centrifuging and washing the bacteria incubated with the antibiotic obtained in step S2, and extracting bacterial metabolites;

[0040] S4: dropping the bacterial metabolites obtained in step S3 onto a target plate of matrix-assisted laser desorption ionization time-of-flight mass spectrometry, and covering the plate with a matrix after natural drying to obtain a sample;

[0041] S5: analyzing the sample in step S4 using matrix-assisted laser desorption ionization time-of-flight mass spectrometry to obtain a metabolic fingerprint of the incubated bacteria;

[0042] S6: performing data preprocessing on the metabolic fingerprint spectrum obtained in step S5 to obtain preprocessed data;

[0043] S7: Analyze the pre-processed data obtained in step S6 using a classification algorithm based on machine learning to directly evaluate the antibiotic resistance of bacteria.

[0044] Unless otherwise specified, the raw materials, reagents or devices used in the following examples can be obtained from conventional commercial sources or by existing known methods.

[0045] Example 1

[0046] This embodiment provides a rapid test method for the resistance of Escherichia coli to ceftriaxone sodium based on the determination of bacterial metabolic fingerprints by MALDI-TOF MS after short-term antibiotic stimulation and combined with machine learning. The specific steps are as follows:

[0047] Medium preparation: According to the instructions of the purchased broth medium, weigh 21 g of MHB medium powder (purchased from Thermo Fisher Scientific), dissolve it in 1 L of deionized water, and sterilize it by high pressure.

[0048] Preparation of ceftriaxone sodium antibiotic solution: According to the EUCAST standard, prepare an antibiotic solution with a concentration twice the breakpoint, i.e. 2 mg / L, where this is the concentration corrected according to the potency of the antibiotic.

[0049] Concentration determination: The concentration of Escherichia coli culture medium was determined using an ultraviolet spectrophotometer.

[0050] Dilution: Dilute the E. coli culture to 1×10 6 ~1.6×10 6 CFU / mL to obtain bacterial suspension.

[0051] Add samples: Take 5 mL of antibiotic solution and 5 mL of diluted bacterial suspension, add them to the centrifuge tube to obtain a mixed solution. The concentration of bacteria in the mixed solution ranges from 5×10 5 ~8×10 5 CFU / mL, and the antibiotic concentration in the mixed solution is the antibiotic breakpoint concentration specified by the EUCAST standard.

[0052] Incubation: Incubate the centrifuge tube with the mixed solution in a 37°C constant temperature oscillator for 30 minutes to obtain an incubated bacterial suspension.

[0053] Washing: Centrifuge the incubated bacterial suspension, discard the supernatant, collect the precipitate, and wash the precipitate.

[0054] Metabolite extraction: The washed bacteria were transferred to 30 μL 80% (v / v) ethanol aqueous solution, heated at 80° C. for 2 min, and then 10 μL deionized water was added. The solution was centrifuged in a handheld centrifuge to obtain a metabolite extract.

[0055] Preparation of matrix solution: Weigh 1 mg of TiO2 nanoparticles (25 nm) and dissolve them in 1 mL of 50% (v / v) ethanol aqueous solution.

[0056] Spotting: Take 1.5 μL of metabolite extract twice and drop it on the target plate. After natural drying, cover it with 2 μL of matrix solution.

[0057] MALDI-TOF MS mass spectrometry detection: The instrument was set in reflectron positive ion mode, ion acceleration voltage 20 kV, delayed extraction time 120 ns, and laser intensity 85%.

[0058] MALDI-TOF MS analysis: The original metabolic fingerprints of sensitive and resistant Escherichia coli after stimulation with ceftriaxone sodium are shown in Figure 2 .

[0059] MALDI-TOF MS data preprocessing: Preprocess MALDI-TOF MS data to achieve background subtraction, baseline smoothing, binning, and peak enhancement.

[0060] Machine learning analysis: The above analysis was performed on multiple strains of resistant or non-resistant E. coli. The support vector machine model was used to classify the drug resistance of the metabolic fingerprint of pretreated E. coli after stimulation with ceftriaxone sodium. The confusion matrix is ​​shown in Figure 3 .

[0061] Figure 2 The metabolic fingerprint peaks in the strains indicated that there were significant differences in the metabolites of resistant and sensitive strains after antibiotic stimulation. Figure 3The confusion matrix in the figure shows that 27 out of 30 fingerprints of resistant bacteria were correctly classified; 27 out of 35 fingerprints of sensitive bacteria were correctly classified, and the overall typing accuracy was 83.1%.

[0062] This example demonstrates the rapid drug resistance test of Escherichia coli under the stimulation of ceftriaxone sodium. This method is not only applicable to the bacteria-antibiotic pair of Escherichia coli-cefotaxime sodium, but can also be used for rapid AMR testing of other bacteria and antibiotics, such as Escherichia coli-ciprofloxacin, Staphylococcus aureus-ciprofloxacin, etc.

[0063] The description of this embodiment is intended to help those skilled in the art understand and use the present invention. The skilled person may modify the specific implementation according to the description herein and apply it to other situations without creative work. Therefore, the protection scope of the present invention is not limited to the above-mentioned embodiments, and all improvements and adjustments based on the principles of the present invention are included in the protection scope of the present invention.

[0064] In order to highlight the innovation of the present invention, the present invention is now compared with the prior art.

[0065] Comparative Example 1

[0066] CN117612747A discloses a method and device for predicting drug sensitivity of Klebsiella pneumoniae. The method directly detects the peptide / protein fingerprint peak of Klebsiella pneumoniae by MALDI-TOF mass spectrometry, uses a supervised machine learning model for training, and finally obtains a drug sensitivity prediction model.

[0067] Unlike the comparative technology, the present invention focuses on detecting changes in bacterial metabolites after short-term incubation with antibiotics. Compared with proteins and peptides, metabolites can more directly reflect the phenotypic changes of bacteria and are more dynamic. At the same time, the present invention combines the stimulation of antibiotic incubation to better reflect the phenotypic characteristics of drug-resistant bacteria. Therefore, the present invention can not only detect the drug resistance of various bacteria, but also is applicable to a variety of antibiotics, with a wider range of applications.

[0068] Comparative Example 2

[0069] CN117265149A discloses a method, primer set and kit for detecting drug resistance of Staphylococcus aureus, wherein the detection method includes primer design, PCR reaction, shrimp alkaline phosphatase reaction, single base extension reaction, resin desalting and MALDI-TOF mass spectrometry detection, and can simultaneously detect the drug resistance of 9 mutation sites of 6 genes corresponding to 3 antibiotics.

[0070] This comparison scheme focuses on the detection of drug-resistant genes in drug-resistant bacteria, while the present invention focuses on determining drug resistance by detecting changes in bacterial metabolites after short-term incubation with antibiotics. Metabolites are closer to a direct reflection of bacterial phenotypes than gene detection, and the changes are more dynamic. At the same time, the present invention combines antibiotic incubation stimulation to better reflect the phenotypic characteristics of drug-resistant bacteria. Therefore, the present invention can not only be applied to a single bacterium and antibiotic, but can also be widely used in rapid resistance detection of multiple bacteria and antibiotics, with a wider range of applications.

Claims

1. A method for rapid detection of bacterial resistance by combining bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning, characterized in that: The following steps are involved: S1: prepare antibiotic solution; S2: inoculating bacteria into a culture container, adding the antibiotic solution prepared in step S1, and incubating the bacteria for a short time to obtain bacteria incubated with antibiotics; S3: centrifuging and washing the bacteria incubated with the antibiotic obtained in step S2, and extracting bacterial metabolites; S4: dropping the bacterial metabolites obtained in step S3 onto a target plate of matrix-assisted laser desorption ionization time-of-flight mass spectrometry, and covering the plate with a matrix after natural drying to obtain a sample; S5: Analyze the sample in step S4 using matrix-assisted laser desorption ionization time-of-flight mass spectrometry to obtain a metabolic fingerprint of the incubated bacteria; S6: performing data preprocessing on the metabolic fingerprint spectrum obtained in step S5 to obtain preprocessed data; S7: Analyze the pre-processed data obtained in step S6 using a classification algorithm based on machine learning to directly evaluate the antibiotic resistance of bacteria.

2. According to claim 1, a method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation with machine learning, characterized in that: The final concentration of the antibiotic in step S2 is the antibiotic breakpoint concentration specified by EUCAST.

3. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: After the antibiotic solution was added in step S2, the final concentration of bacteria ranged from 5 × 10 5 ~8×10 5 CFU / mL.

4. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: In step S2, bacteria are inoculated in a culture container to obtain a bacterial culture solution, the bacterial culture solution is diluted to obtain a diluted bacterial suspension, the antibiotic solution prepared in step S1 and the diluted bacterial suspension are mixed to obtain a mixed solution, and the bacteria in the mixed solution are incubated for a short time to obtain bacteria incubated with antibiotics.

5. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: In step S3, extracting bacterial metabolites includes the following process: The washed bacteria were transferred into an ethanol-water solution, heated, and then water was added to obtain a metabolite extract.

6. A method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: The bacterial incubation time in step S2 is controlled between 10 minutes and 2 hours.

7. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: The matrix used in step S4 includes one or more of an organic matrix and an inorganic nanomaterial matrix.

8. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: The data preprocessing steps in step S6 include background subtraction, baseline smoothing, binning and peak lifting operations.

9. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: The machine learning model involved in step S7 includes one or more of logistic regression, support vector machine, multi-layer perceptron, random forest and neural network model.

10. The method for rapid detection of bacterial resistance combining bacterial metabolic fingerprint after short-term antibiotic stimulation and machine learning according to claim 1, characterized in that: In step S7, classification analysis based on machine learning is performed, and the best result among the test results of multiple machine learning models is taken as the test result of the final sample.

Images