Rapid pathogenic bacterium identification method based on accelerated purine metabolite SERS (Surface Enhanced Raman Scattering) spectrum

Abstract

The invention discloses a rapid pathogenic bacterium identification method based on an accelerated purine metabolite SERS (Surface Enhanced Raman Scattering) spectrum, which comprises the following steps: (1) taking a pathogenic bacterium strain, pretreating, and then collecting the SERS spectrum; (2) carrying out spectrum preprocessing on the SERS spectrum acquired in the step (1), and then carrying out data division; (3) using different machine learning algorithms to establish a classification model of the pathogenic bacterium SERS spectrum, and selecting an optimal classification model by predicting accuracy; (4) separating, enriching, washing, resuspending, heating and standing a sample to be detected, and collecting an SERS spectrum of the sample to be detected; and (5) preprocessing the SERS spectrum obtained in the step (4), and inputting the SERS spectrum into the classification model established in the step (3) to obtain an identification result. The method successfully solves the problem that traditional bacteria SERS analysis needs long-time waiting, and the detection method is high in accuracy.

Description

Technical Field

[0001] The invention belongs to the field of pathogenic bacteria analysis, and specifically relates to a method for rapid identification of pathogenic bacteria based on accelerated SERS spectrum of purine metabolites. Background Art

[0002] Pathogenic bacteria infection is one of the factors that seriously threaten the life and health of the public. Early diagnosis and identification of pathogenic bacteria are key steps in the treatment of pathogenic bacteria infection. At present, the plate culture method is regarded as the "gold standard" for pathogenic bacteria identification, but this method involves multi-step bacterial culture and subsequent biochemical identification, which usually takes several days to obtain results, and it is difficult to meet the urgent need for rapid detection in practical applications. In recent years, new technologies such as molecular biology methods, immunoassays, and mass spectrometry have been continuously applied to pathogenic bacteria detection. These methods have shortened the detection time and improved the identification efficiency to a certain extent. However, they also have shortcomings such as high detection costs and complex operating methods. Therefore, it is still a top priority to develop a method for pathogenic bacteria detection that is simple, fast, cost-effective, and reliable.

[0003] Surface-enhanced Raman spectroscopy (SERS) has the advantages of fast detection speed, high sensitivity, low detection cost, portable detection and unique fingerprint detection ability, so it shows unique advantages in pathogen detection. SERS analysis technology can generally be divided into two types: labeled detection and label-free detection. Compared with labeled detection, the label-free SERS method presents an economical, efficient and simple approach that can directly detect the inherent biological fingerprints of pathogens without any labeling steps. Based on these advantages, researchers have developed a variety of label-free SERS analysis methods for pathogens. Previous studies have shown that the SERS spectra of pathogens are mainly derived from a few purine metabolites, and the generation of these metabolites at room temperature requires sufficient time (at least 10 hours), which results in the entire detection process (including sample culture, pathogen post-processing, static, SERS spectrum acquisition, etc.) usually takes 2 days to complete. Although this has been significantly improved compared with the traditional plate counting method, there is still a clear gap with the rapid detection expected by the clinic. Therefore, how to shorten the acquisition time of SERS spectra of pathogenic bacteria is one of the key challenges to promote the industrialization of this analytical technology. Summary of the invention

[0004] The purpose of the present invention is to make up for the deficiencies of the prior art, provide a method for rapid acquisition of bacterial SERS spectra, and establish a rapid identification method for pathogenic bacteria based on this strategy. The method has the advantages of rapid detection, low cost, simple operation, accuracy and reliability.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is as follows:

[0006] A method for rapid identification of pathogenic bacteria based on accelerated SERS spectra of purine metabolites, comprising the following steps:

[0007] (1) Pathogenic bacteria are enriched and cultured, then centrifuged and washed, and then the pathogens are resuspended in a buffer solution, heated and allowed to stand, and then SERS spectra are collected;

[0008] (2) performing spectral preprocessing on the SERS spectrum collected in step (1) and then performing data division;

[0009] (3) Use different machine learning algorithms to establish classification models for pathogenic bacteria SERS spectra and select the best classification model based on prediction accuracy;

[0010] (4) collecting a SERS spectrum of the sample after separating, enriching, washing, resuspending, heating and standing the sample;

[0011] (5) The SERS spectrum obtained in step (4) is input into the classification model established in step (3) after spectral preprocessing to obtain the identification result.

[0012] In step (1), the bacterial culture time is 12 to 16 hours;

[0013] In step (1), the bacterial culture solution is centrifuged for 5 to 10 minutes at a speed of 3000 to 6000 rpm;

[0014] The number of centrifugation in step (1) is 2 to 3 times, and the solution is purified water;

[0015] After the last centrifugation in step (1), the upper layer of purified water was removed and the bacterial precipitate at the bottom was suspended with 100-200 mM PBS buffer;

[0016] The volume of PBS buffer added in step (1) is no greater than 1 / 2 of the original liquid culture medium volume;

[0017] In step (1), the SERS spectra of each bacterial species are collected independently in at least three batches, with at least 30 spectra in each batch;

[0018] In step (1), the temperature of heating and standing is 55-65° C., and the heating time is 40 minutes to 1.5 hours; heating and standing at 55-65° C. effectively shortens the pretreatment time and ensures the accuracy of the detection results.

[0019] The bacterial SERS spectra collected in step (1) are mainly derived from purine metabolites such as adenine, guanine, hypoxanthine, and xanthine;

[0020] In step (2), the spectral preprocessing mainly includes spectral smoothing, filtering, etc.;

[0021] The data division method in step (2) is to use the SERS spectra obtained from the last 1 to 2 acquisitions as the test set and the remaining data as the training set. This division can reflect the generalization ability of machine learning.

[0022] The classification machine learning algorithms in step (3) are mainly PLS-DA, KNN algorithm (k-NearestNeighbors, KNN), Extreme learning machine (Extreme learning machine, ELM), Naive Bayes model (NaiveBayes model, NBM), Decision Tree (Decision Tree), Random forest (Random forest, RF), Linear discriminant analysis (Linear discriminant analysis, LDA), SoftMax classifier (SoftMax classifier, SMC) and Stacked-autoencoder (Stacked-autoencoder, SAE);

[0023] The predictive ability of the model in step (3) is evaluated by the classification efficiency (Efficient rate, EFR);

[0024] The best machine learning model in step (3) is the one with the best EFR value;

[0025] The EFR calculation formula described in step (3) is as follows:

[0026]

[0027] In the above formula, TP and TN are the number of samples identified as true positive and true negative, and FP and FN are the number of samples identified as false positive and false negative.

[0028] In step (4), the sample to be tested needs to be separated and cultured in a solid culture medium first, and then the colonies are selected and transferred to a liquid culture medium for culture until the culture medium becomes turbid.

[0029] In step (4), after the sample to be tested is separated and amplified by solid and liquid culture media, the remaining bacterial treatment steps are the same as steps (1) and (2);

[0030] In step (4), at least 15 SERS spectra of the sample to be tested are collected;

[0031] In step (5), only when more than 80% of the SERS spectra of the analyte belong to a certain type of pathogenic bacteria, the analyte is determined to belong to the type of pathogenic bacteria. Preferably, the identified bacteria are one or more of Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae or Pseudomonas aeruginosa.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] The present invention successfully solves the problem of long waiting time in traditional bacterial SERS analysis, shortens the traditional bacterial SERS spectrum acquisition process from more than 10 hours to less than 1 hour, and the entire clinical pathogen infection sample detection process is expected to be completed within 17 hours, and the detection method of the present invention has high accuracy. In view of the rapidity, convenience, stability and reliability of the method of the present invention, it has shown broad prospects in promoting the application of pathogenic bacteria SERS analysis in practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Comparison of the SERS spectra of Escherichia coli obtained in step (2) at room temperature (~25°C) and 60°C at different times;

[0035] Figure 2 is the average SERS spectrum of the four pathogenic bacteria collected in step (4), wherein (a) is Escherichia coli; (b) is Staphylococcus aureus; (c) is Klebsiella pneumoniae; and (d) is Pseudomonas aeruginosa;

[0036] Figure 3 Comparison of SERS spectra of four pathogenic bacteria with simulated spectra of purine metabolites, including (a) Escherichia coli; (b) Staphylococcus aureus; (c) Klebsiella pneumoniae; (d) Pseudomonas aeruginosa;

[0037] Figure 4 This is the confusion matrix diagram for testing 31 clinical samples. DETAILED DESCRIPTION

[0038] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0039] Example 1

[0040] (1) Take Escherichia coli (E.coli), Staphylococcus aureus (S.aureus), Klebsiella pneumoniae (K.pneumonia) and Pseudomonas aeruginosa (P.aeruginosa), transfer to LB liquid medium for amplification and culture for 14 hours, and then centrifuge at a speed of 3000 rpm for 10 minutes. After centrifugation, remove the upper layer of culture medium, resuspend with the same volume of purified water, and centrifuge again. Repeat this step twice, remove the upper layer of water, and add 1 / 20 of the original volume of 100mM PBS buffer to resuspend;

[0041] (2) placing the bacterial PBS suspension treated in step (1) at 60 degrees Celsius for 1 hour, then taking out 1 mL of the suspension and collecting the bacterial SERS spectrum;

[0042] (3) Each of the above bacterial species was cultured independently for 6 times, and 6 different batches of SERS spectrum data were collected;

[0043] (4) The collected SERS spectra of pathogenic bacteria were spectrally smoothed and baseline corrected, and the last two batches of collected data were used as test set samples, and the remaining batches of data were used as training set samples.

[0044] (5) Using PLS-DA, KNN algorithm, extreme learning machine, naive Bayes method, decision tree, random forest, linear discriminant analysis, SoftMax classifier and stacked autoencoder algorithms, the classification model of the above four pathogens was established, and the classification efficiency EFR of the test set was calculated to select the best classification algorithm. Table 1 shows the calculated EFR of different algorithms for the test set. From the results, it can be seen that the SoftMax classifier has the best classification effect with an accuracy of 100%. Therefore, the SoftMax classifier is selected as the best classification algorithm.

[0045] (6) Take 31 clinical pathogenic bacteria infection samples (19 Escherichia coli, 7 Staphylococcus aureus, 4 Klebsiella pneumoniae, 1 Pseudomonas aeruginosa), culture them in blood plate culture medium for 12 hours, then select colonies and transfer them to 15 mL LB liquid culture medium for further culture for 2 hours. Then, the centrifugation, washing, buffer resuspension, heating and standing of the culture solution of the bacteria to be tested are completed in the same manner as steps (1) and (2);

[0046] (7) collecting SERS spectra of the tested bacteria after treatment, with no less than 15 spectra for each sample;

[0047] (8) The SERS spectrum obtained in step (7) is input into the classification model established in step (3) after spectrum smoothing and baseline correction to obtain the identification result with an accuracy of 100%.

[0048] Table 1 Recognition accuracy of different machine learning algorithms for four pathogens

[0049]

[0050] Comparative Example 1

[0051] The other operations are the same as those in Example 1, except that in step (2), the static temperature of the treated E. coli PBS suspension is set at 25 to 60° C., and the SERS spectra of E. coli at different time periods are obtained.

Claims

1. A method for rapid identification of pathogenic bacteria based on accelerated SERS spectra of purine metabolites, characterized in that: The steps include: (1) Pathogenic bacteria are enriched and cultured, then centrifuged and washed, and then the pathogens are resuspended in a buffer solution, heated and allowed to stand, and then SERS spectra are collected; (2) performing spectral preprocessing on the SERS spectrum collected in step (1) and then performing data division; (3) Use different machine learning algorithms to establish classification models for pathogenic bacteria SERS spectra and select the best classification model based on prediction accuracy; (4) collecting a SERS spectrum of the sample after separating, enriching, washing, resuspending, heating and standing the sample; (5) The SERS spectrum obtained in step (4) is input into the classification model established in step (3) after spectral preprocessing to obtain the identification result.

2. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: In step (1), the enrichment culture time is 12 to 16 hours; In step (1), the culture solution is centrifuged for 5 to 10 minutes at a speed of 3000 to 6000 rpm; The number of centrifugation is 2 to 3 times. After each centrifugation, add purified water to resuspend and then centrifuge again for washing; After the last centrifugation in step (1), the upper layer of purified water was removed and the bacterial precipitate at the bottom was suspended with 100-200 mM PBS buffer; The volume of PBS buffer added should not be greater than 1 / 2 of the original liquid culture volume.

3. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: The heating and standing temperature in step (1) is 55-65° C., and the heating time is 40 minutes to 1.5 hours.

4. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: In step (1), the SERS spectra of each bacterial species are collected independently in at least three batches, with at least 30 spectra in each batch; The collected bacterial SERS spectra are mainly derived from purine metabolites such as adenine, guanine, hypoxanthine, and xanthine.

5. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: The spectral preprocessing in step (2) mainly includes spectral smoothing and filtering. The data division method is to use the SERS spectra obtained from the last 1 to 2 acquisitions as the test set and the remaining data as the training set.

6. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: The classification machine learning algorithm in step (5) includes one or more of PLS-DA, KNN algorithm (k-Nearest Neighbors, KNN), Extreme learning machine (Extreme learning machine, ELM), Naive Bayes model (Naive Bayes model, NBM), Decision Tree (Decision Tree), Random forest (Random forest, RF), Linear discriminant analysis (Linear discriminant analysis, LDA), SoftMax classifier (SoftMax classifier, SMC) and Stacked-autoencoder (Stacked-autoencoder, SAE).

7. The method for rapid identification of pathogenic bacteria according to claim 1, characterized in that: The predictive ability of the model in step (3) is evaluated by the classification efficiency (Efficient rate, EFR); The best machine learning model is the one with the best EFR value.

8. The method for detecting pathogenic bacteria according to claim 1, characterized in that: In step (4), the sample to be tested needs to be separated and cultured in a solid culture medium first, and then the colonies are selected and transferred to a liquid culture medium for culture until the culture medium becomes turbid. In step (4), after the sample to be tested is separated and amplified by solid and liquid culture media, the remaining bacterial treatment steps are the same as steps (1) and (2).

9. The method for detecting pathogenic bacteria according to claim 1, characterized in that: In step (4), at least 15 SERS spectra of the sample to be tested are collected.

10. The method for detecting pathogenic bacteria according to claim 1, characterized in that: In step (5), when more than 85% of the SERS spectra of the analyte belong to a certain type of pathogenic bacteria, the analyte is determined to belong to that type of pathogenic bacteria.

Images