A method for distinguishing the type of carbon assimilation pathway of methanotrophs

Abstract

The application discloses a method for distinguishing the carbon assimilation pathway type characteristics of methanotrophs. 13 CH4 stable isotope labeling and collecting single-cell Raman full spectrum data; based on the PCA-LDA discriminant model of the reference strains 13 CH4 labeled single-cell Raman full spectrum data to establish a PCA-LDA discriminant model; a plurality of single-cell Raman full spectrum data of a to-be-tested methanotroph sample are input into the previously established PCA-LDA discriminant model, the corresponding I-type or II-type prediction category of each single-cell Raman spectrum is obtained, and it is determined that the to-be-tested sample is closer to the carbon assimilation pathway characteristics of I-type methanotrophs or the carbon assimilation pathway characteristics of II-type methanotrophs. The method can distinguish different carbon assimilation pathway types of methanotrophs.

Description

Technical Field

[0001] This invention relates to the field of microbial functional detection and single-cell spectral analysis technology, specifically to a method utilizing... 13 A method for detecting and distinguishing the carbon assimilation pathway types of methanogenic bacteria using CH4 stable isotope labeling, single-cell Raman full-spectrum acquisition, and statistical discrimination models. Background Technology

[0002] Methanogenic bacteria are important functional microorganisms capable of utilizing methane as a carbon and energy source, playing a crucial role in methane biotransformation and related environmental processes. Different methanogenic bacteria exhibit distinct carbon assimilation pathways. Typical type I methanogenic bacteria primarily employ the ribulose monophosphate pathway (RuMP), while typical type II methanogenic bacteria mainly utilize the serine cycle pathway. Differentiating the characteristics of these carbon assimilation pathways helps in analyzing the utilization characteristics and functional differences of methane-derived carbon by different strains.

[0003] Existing analytical methods for methanogenic bacteria include culture analysis, nucleic acid detection, sequencing analysis, and stable isotope labeling.

[0004] However, nucleic acid methods are complex and cannot directly reflect the current phenotype of a single cell. Functional gene amplification, sequencing, and metagenomic analysis can be used for classification or functional potential assessment, but they usually require nucleic acid extraction, amplification, and analysis processes, making rapid detection at the single-cell level difficult; stable isotope probes combined with nucleic acid analysis are mainly used for activity identification. Summary of the Invention

[0005] The main objective of this invention is to provide a method based on 13 A method for distinguishing the carbon assimilation pathway types of methanogenic bacteria using CH4-labeled single-cell Raman spectroscopy. The method uses typical type I and type II methanogenic bacteria with known carbon assimilation pathway types as reference strains to establish a discrimination model for distinguishing between type I and type II reference strains. The established discrimination model is then used to analyze the carbon assimilation pathway type characteristics of the methanogenic bacteria samples under test.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for distinguishing the characteristics of carbon assimilation pathways in methanogenic bacteria, including a reference discrimination model establishment stage and a sample discrimination stage, the method comprising the following steps:

[0007] I. Reference Discriminant Model Establishment Stage

[0008] S1: Reference strain selection: Typical methane-oxidizing bacteria with known carbon assimilation pathways were selected as reference strains. The reference strains include type I reference strains and type II reference strains. Among them, the type I methane-oxidizing bacteria are those that mainly use the ribulose monophosphate pathway, and the type II methane-oxidizing bacteria are those that mainly use the serine cycle pathway.

[0009] S2: Reference strain labeling and single-cell Raman spectral acquisition: The reference strain was placed in a container... 13 Stable isotope labeling was performed in the CH4 labeled culture system, and single-cell Raman full spectrum data of each reference strain were acquired by microconfocal Raman spectrometer within a unified detection time window.

[0010] S3: Preprocessing of single-cell Raman full spectrum of reference strain: Preprocess the single-cell Raman full spectrum data of the reference strain according to a unified data processing procedure to obtain a single-cell Raman full spectrum dataset of the reference strain for establishing a discrimination model.

[0011] S4: PCA-LDA discriminant model establishment: using the aforementioned type I reference strain and type II reference strain... 13 Preprocessed single-cell Raman spectrum data obtained under CH4 labeling conditions were used as model input. First, the single-cell Raman spectrum data was reduced in dimensionality by principal component analysis. Then, a linear discriminant analysis model was established based on the obtained principal component scores and the known carbon assimilation pathway type of the reference strain, thus forming a PCA-LDA discriminant model for distinguishing between type I and type II reference spectra.

[0012] S5: Interpretation of Contribution Wavenumber Regions: Pretreated reference strains used to establish the PCA-LDA discriminant model. 13 Cluster Vectors analysis was performed on CH4-labeled single-cell Raman full spectrum data to show the Raman wavenumber regions that contributed significantly to the grouping of type I and type II reference strains. The candidate Raman wavenumber regions were then used to provide an auxiliary interpretation of the spectral differences reflected by the PCA-LDA discriminant model.

[0013] II. Sample discrimination stage

[0014] S6: Sample Labeling and Single-Cell Raman Spectrum Acquisition: The methane-oxidizing bacteria sample to be tested is labeled according to the same procedure as the reference discrimination model establishment stage. 13 The CH4 labeling conditions, detection time window, and spectral acquisition conditions were processed to obtain multiple single-cell Raman full-spectrum data of the sample to be tested.

[0015] S7: Sample Model Prediction: After preprocessing the single-cell Raman spectrum data of the sample to be tested according to the same data processing flow as the reference discrimination model establishment stage, the data is input into the pre-established PCA-LDA discrimination model, and the model outputs the type I prediction category or type II prediction category corresponding to each single-cell Raman spectrum.

[0016] S8: Pathway type characteristic determination: Count the number of single-cell Raman spectra predicted as type I and type II in the same test sample, and determine the carbon assimilation pathway type characteristic of the test sample according to the majority voting rule; when the number of single-cell Raman spectra predicted as type I is greater than the number of single-cell Raman spectra predicted as type II, the test sample is determined to be closer to the carbon assimilation pathway characteristics of type I methanogenic bacteria; when the number of single-cell Raman spectra predicted as type II is greater than the number of single-cell Raman spectra predicted as type I, the test sample is determined to be closer to the carbon assimilation pathway characteristics of type II methanogenic bacteria.

[0017] Furthermore, the Cluster Vectors analysis is used to help interpret the Raman wavenumber regions that contribute significantly to the grouping of Type I and Type II reference spectra, and is not used as an independent fixed threshold discrimination rule.

[0018] Furthermore, the type I reference strains include Methylomonas sp. LW13, Methylomonasmethanica MC09, Methylotuvimicrobium buryatense 5GB1C, and Methylobacter sp. YRD-M1; the type II reference strains include Methylosinus sp. LW4, Methylosinus sp. PW1, Methylosinustrichosporium OB3b, and Methylocystis iwaonis SD4.

[0019] Furthermore, in one embodiment, the PCA-LDA (Feature Extraction and Dimensionality Reduction Combined Algorithm) discriminant model uses the aforementioned 8 reference strains in... 13 Preprocessed single-cell Raman full-spectrum data obtained under CH4 labeling conditions were used as model input to establish the discrimination rule between type I reference spectra and type II reference spectra.

[0020] Furthermore, in observation 13 When the single-cell Raman spectrum response changes induced by CH4 labeling, settings can be configured. 12 CH4 control conditions and 13 Comparison of spectra obtained under CH4 labeling conditions; the 12The CH4 control condition is used to observe the labeled response and assist in interpreting candidate Raman wavenumber regions. In one embodiment, during the establishment of the type I and type II PCA-LDA discriminant models, the model input uses... 13 Single-cell Raman full spectrum data obtained under CH4 labeling conditions.

[0021] Furthermore, the unified detection time window is determined based on the changes in the growth state of representative methanogenic bacteria and the changes in methane concentration in the culture system. In one embodiment, the mid-logarithmic growth phase is selected as the unified detection time window for single-cell Raman spectrum acquisition of the reference strain and comparative analysis of the test samples, in order to reduce the impact of differences in spectral distribution caused by different growth stages on the comparison of type characteristics.

[0022] Furthermore, the single-cell Raman full-spectrum data are preprocessed using a unified data processing workflow. The preprocessing workflow includes baseline correction and normalization used in the actual implementation process; if smoothing, wavenumber alignment, or abnormal spectrum exclusion are also used in the actual analysis process, the actual processing steps and parameters are recorded and kept consistent during the reference model establishment stage and the sample discrimination stage.

[0023] Furthermore, after the PCA-LDA discriminant model is established, it can store data processing parameters, principal component analysis projection parameters, and linear discriminant analysis classification parameters for model invocation. When determining the pathway type characteristics of the sample to be tested, after obtaining the single-cell Raman spectrum of the sample according to the same labeling conditions, detection time window, spectral acquisition conditions, and data processing flow as in the reference model establishment stage, it can be input into the pre-established PCA-LDA discriminant model for prediction.

[0024] Furthermore, the Cluster Vectors analysis is used to display the Raman wavenumber regions that contribute significantly to the grouping of type I and type II reference strains in the PCA-LDA discriminant model. In one embodiment, candidate Raman wavenumber regions associated with cytochrome C, comprising approximately 749 cm⁻¹, can be used to aid in the interpretation of the discriminant results. -1 Approximately 1152cm -1 Approximately 1587cm -1 At least a portion of the region. The candidate Raman wavenumber region is not used as an independent fixed threshold for determining whether the sample is Type I or Type II.

[0025] Furthermore, the majority voting rule refers to: for the same test sample, counting the number of single-cell Raman spectra predicted as type I and type II by the PCA-LDA discriminant model, and using the prediction category with the larger number as the result of determining the pathway type feature of the test sample.

[0026] Furthermore, to evaluate the ability of the PCA-LDA discriminant model to determine the type of reference strains that did not participate in the model establishment in the current round, the Leave One Out (LOO) method can be used for verification. In each round of verification, the entire single-cell Raman spectrum of one reference strain is used as the test set, and the single-cell Raman spectra of the other 7 reference strains are used to establish the PCA-LDA discriminant model for that round. The strain-level determination type is determined by majority voting based on the predicted category of each single-cell Raman spectrum of the test strain.

[0027] Furthermore, to evaluate the impact of the difference in the number of single-cell Raman spectra of different reference strains in the training set on the model's judgment results, a sensitivity analysis of equal random sampling of the training set can be performed. In each round of leave-one-out-of-strain validation, the same number of single-cell Raman spectra are randomly selected from each training reference strain to establish the PCA-LDA discrimination model, and random sampling and model prediction are repeated to statistically analyze the proportion of test strains whose judgment results at the strain level are consistent with their known types.

[0028] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0029] 1. This invention adopts 13 CH4 was used as a labeled carbon source, and spectral response information related to carbon utilization from methane was obtained through single-cell Raman spectroscopy. The obtained single-cell Raman spectroscopy data can serve as an analytical basis for distinguishing the characteristics of different carbon assimilation pathways.

[0030] 2. This invention uses typical type I and typical type II methanogenic bacteria with known carbon assimilation pathways as reference strains, based on which the reference strains... 13 A PCA-LDA discriminant model was established using single-cell Raman spectra obtained under CH4 labeling conditions. In this embodiment, the reference strains included four type I methanogenic bacteria and four type II methanogenic bacteria, thereby forming a discriminant model for distinguishing between type I and type II reference spectra.

[0031] 3. This invention uses the full Raman spectrum of a single cell as the classification input for the PCA-LDA discriminant model, without relying on a single Raman wavenumber region, the ratio or difference between characteristic peak shifts, or a fixed discrimination threshold to determine the type of the test sample. Through Cluster Vectors analysis, it can display the Raman wavenumber regions that contribute significantly to the grouping of type I and type II reference strains, and provide auxiliary interpretation of the spectral differences reflected by the PCA-LDA discriminant model.

[0032] 4. This invention employs a unified detection time window during the establishment of the reference model and the analysis of the test samples, and obtains single-cell Raman full-spectrum data for comparison under the same labeling conditions, spectral acquisition conditions, and data processing procedures. The unified detection time window is used to reduce the impact of spectral distribution differences caused by different growth stages on the comparison of type I and type II pathway characteristics.

[0033] 5. This invention inputs multiple single-cell Raman spectra of the sample to be tested into a pre-established PCA-LDA discriminant model to obtain the Type I or Type II prediction category corresponding to each single-cell Raman spectrum. Based on a majority voting rule, a carbon assimilation pathway type characteristic determination result is formed at the strain level or sample level for the sample to be tested. This determination method clearly distinguishes between the model prediction result of a single spectrum and the overall type determination result of the same sample.

[0034] 6. Within the range of eight typical reference strains selected in the examples, when evaluating the PCA-LDA discriminant model using the leave-one-out cross-validation method, all eight retained test strains obtained strain-level determination results consistent with their known types. Further sensitivity analysis using equal-size random sampling of the training set revealed that in all 800 repeated modeling determinations, 753 strain-level determination results were consistent with the known types of the corresponding test strains. These results demonstrate that, within the scope of the reference strains, labeling conditions, detection time windows, spectral acquisition conditions, and data processing procedures used in this invention, the method described in this invention can be used for discriminant analysis of carbon assimilation pathway type characteristics in type I and type II methanogenic bacteria. Furthermore, the difference in determination stability among different reference strains provides a basis for subsequent expansion of the reference strain set and optimization of model parameters. Attached Figure Description

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] Figure 1 Typical methanogenic bacteria in 12 CH4 control conditions and 13 Results of single-cell Raman spectral response analysis under CH4 labeling conditions. Figure 1 A shows the average Raman spectra of single cells from 8 typical reference strains; Figure 1 B is obtained based on single-cell Raman full spectrum data. 12 CH4 control group and 13 Distribution results of PCA-LDA for CH4-labeled groups; Figure 1 C represents the corresponding Cluster Vectors analysis result, used to display the results of the analysis. 12 CH4 control group and 13 The Raman wavenumber region where the difference in the spectral distribution of the CH4-labeled group contributes significantly; Figure 1 D is12 CH4 control group and 13 Comparison of the average Raman spectra of the CH4-labeled group.

[0037] Figure 2 The curves show the growth status of representative type I and type II reference strains of methanogenic bacteria during cultivation, as well as the changes in methane detection results in the culture system. Figure 2 A represents the type I reference strain, *Methylomonas* sp. LW13, during cultivation. 600 And the curve of CH4 concentration change; Figure 2 B represents the type II reference strain, Methylocystisiwaonis SD4, during cultivation. 600 The curves show the changes in CH4 concentration. The blue curve represents the growth status of the strain, expressed as OD. 600 The red curve represents the methane concentration in the culture system, expressed as CH4 (mmol / L). Figure 2 This is used to illustrate the growth changes and methane utilization-related changes of representative reference strains during the culture process, and to help determine a uniform detection time window for subsequent single-cell Raman spectroscopy acquisition.

[0038] Figure 3 Comparative analysis of the distribution of single-cell Raman spectra of typical methanogenic bacteria at different growth stages. (A) PCA-LDA dimensionality reduction analysis results based on single-cell Raman spectra. Different colors and symbols represent early logarithmic phase, mid-logarithmic phase, late logarithmic phase, stationary phase, and late stationary phase, respectively. Ellipses represent 95% confidence intervals. The results show that different growth stages are significantly separated in the feature space (P<0.001). (B) Cluster Vectors analysis results show the Raman wavenumber regions that contribute significantly to classification at different growth stages, indicating the differences in the position of the characteristic peaks of the main responses at different stages.

[0039] Figure 4 For typical methanogenic bacteria (Methylomonas sp. LW13) at different growth stages, 12 CH4 control conditions and 13 The results are shown in the average single-cell Raman spectra under CH4 labeling conditions. Figure 4 It is used to compare spectral responses at different growth stages and to help explain the necessity of conducting comparative analysis of type I and type II reference spectra within a unified detection time window.

[0040] Figure 5 This presents the results of a discriminant analysis of the carbon assimilation pathway characteristics of type I and type II methanogenic bacteria based on eight typical reference strains. Figure 5 A is based on 4 type I reference strains and 4 type II reference strains.13 PCA-LDA discriminant spatial distribution results obtained from single-cell Raman full spectrum under CH4 labeling conditions; Figure 5 B represents the corresponding Cluster Vectors analysis results, used to show the Raman wavenumber regions that contribute significantly to the grouping of type I and type II reference strains. Figure 5 C represents 4 type I reference strains and 4 type II reference strains. 12 CH4 control conditions and 13 The results of single-cell Raman spectra under CH4 labeling conditions are presented to help observe the differences in spectral responses between the two types of reference strains in the candidate contribution wavenumber region.

[0041] Figure 6 The results of strain-based leave-one-out cross-validation and sensitivity analysis using equal-size random sampling of the training set are presented for the PCA-LDA discriminant model. Figure 6 A is a schematic diagram of the leave-one-out cross-validation process for strains; Figure 6 B represents the proportion of single-cell Raman spectra of each strain that was left out in the eight rounds of cross-validation with one strain left out. The proportion of the strain's known type predicted by the model is represented by the 50% threshold in the figure. The threshold for determining the strain type is based on majority voting. Figure 6 C represents the proportion of strain-level determination results that are consistent with their known types in 100 repeated modeling predictions for each test strain that was left out in the sensitivity analysis of equal-size random sampling of the training set. Detailed Implementation

[0042] The technical solution of the present invention is as follows:

[0043] This invention provides a method for distinguishing the carbon assimilation pathway types of methanogenic bacteria. The method includes a reference discriminant model establishment stage and a sample discrimination stage, and may further include a validation analysis stage for evaluating the performance of the discriminant model.

[0044] I. Reference Discriminant Model Establishment Stage

[0045] Step S1: Reference strain setup

[0046] Typical methanogenic bacteria with known carbon assimilation pathways were selected as reference strains, including type I and type II reference strains. Type I methanogenic bacteria primarily utilize the ribulose monophosphate pathway, while type II methanogenic bacteria primarily utilize the serine cycle pathway. These reference strains were used to obtain single-cell Raman full-spectrum data of known types and to establish a PCA-LDA discriminant model to distinguish between type I and type II reference strains. The specific composition of the reference strains used will be described in subsequent examples.

[0047] Step S2: Determine the unified detection time window

[0048] To reduce the impact of differences in single-cell Raman spectrum distribution caused by different growth stages on type characteristic comparisons, the growth status of representative reference strains and changes in methane detection results in the culture system were monitored during cultivation. Based on the comparison results of single-cell Raman spectrum distributions at different growth stages, a unified detection time window was determined for reference model establishment and sample analysis. In this embodiment, the unified detection time window was selected as the mid-logarithmic growth phase. This time window is used to standardize sample collection conditions and is not used as an independent type determination rule.

[0049] Step S3: Reference strain isotope labeling and single-cell Raman spectroscopy acquisition

[0050] The reference strain was placed in a container 13 Stable isotope labeling was performed in the CH4 culture system, and single-cell Raman full-spectrum data were acquired within the unified detection time window. This was for observation purposes. 13 Changes in single-cell Raman spectrum response induced by CH4 labeling can be set. 12 CH4 control conditions, and 12 CH4 control conditions and 13 Comparative analysis of single-cell Raman spectra obtained under CH4 labeling conditions was conducted. The input data used to establish the type I and type II PCA-LDA discrimination model was the reference strain at... 13 Single-cell Raman full spectrum data obtained under CH4 labeling conditions; 12 The spectra obtained under the CH4 control condition were used to observe the labeled response and to help interpret the candidate Raman wavenumber regions.

[0051] Step S4: Single-cell Raman full-spectrum pretreatment

[0052] For the reference strain in 13 Single-cell Raman spectral data obtained under CH4 labeling conditions were preprocessed according to a unified data processing workflow to obtain a reference strain single-cell Raman spectral dataset for establishing the PCA-LDA discriminant model. The preprocessing workflow included baseline correction and normalization used in the actual implementation. If smoothing, wavenumber alignment, or outlier spectrum exclusion were also used in the actual analysis, the actual processing steps and parameters should be recorded and kept consistent between the reference discriminant model establishment stage and the sample discrimination stage.

[0053] Step S5: PCA-LDA Discriminant Model Establishment

[0054] Using type I reference strains and type II reference strains in 13Preprocessed single-cell Raman full-spectrum data obtained under CH4 labeling conditions were used as model input. First, principal component analysis (PCA) was used to reduce the dimensionality of the single-cell Raman full-spectrum data, obtaining principal component scores reflecting the differences in the main spectra. Then, using the principal component scores and the known carbon assimilation pathway type of the reference strain as input, a linear discriminant analysis (PCA-LDA) model was established, thus forming a PCA-LDA discriminant model to distinguish between Type I and Type II reference spectra. This PCA-LDA discriminant model uses single-cell Raman full-spectrum data as classification input and does not rely on the peak position of a single Raman wavenumber region, the ratio or difference between multiple characteristic peak offsets, or a fixed Raman peak threshold as the basis for type determination.

[0055] Step S6: Interpretation of Contributing Wavenumber Regions

[0056] Based on the pretreated reference strain 13CH4 labeled single-cell Raman spectrum data obtained in step S4 for establishing the PCA-LDA discriminant model, Cluster Vectors analysis was performed to display the Raman wavenumber regions that significantly contributed to the grouping of type I and type II reference strains. In this embodiment, candidate Raman wavenumber regions related to cytochrome C can be used to further interpret the spectral differences reflected by the PCA-LDA discriminant model. These candidate Raman wavenumber regions may include approximately 749 cm⁻¹. -1 Approximately 1152cm -1 Approximately 1587cm -1 At least a portion of the region. The Cluster Vectors analysis and candidate Raman wavenumber region interpretation are used to help explain the spectral differences between the Type I and Type II reference spectra, and are not used as an independent fixed threshold rule for determining the type of the sample to be tested.

[0057] The "Raman wavenumber regions that contribute relatively significantly" refer to Raman shift regions in the Cluster Vectors curve that exhibit higher absolute amplitudes, local peaks, or significant changes compared to the adjacent wavenumber background. The presence of relatively high amplitudes, significant local peaks, or local changes in certain wavenumber regions within the Cluster Vectors curve indicates that these regions contribute more significantly to the differences between Type I and Type II groupings.

[0058] In a specific embodiment, combined with approximately 749 cm -1 Approximately 1152cm -1 Approximately 1587cm -1 At least some candidate Raman wavenumber regions in the model can be used to help interpret the spectral differences between the Type I and Type II reference spectra reflected by the PCA-LDA discriminant model.

[0059] The "supplementary explanation" includes:

[0060] (1) Explain which Raman wavenumber regions the differences in PCA-LDA grouping mainly originate from. In a specific embodiment, for example, Figure 5 A observed that the distributions of type I and type II are different. Figure 5 B uses Cluster Vectors to help illustrate which wavenumber regions this difference is primarily related to.

[0061] (2) Explain whether these contributing wavenumber regions are related to candidate biologically relevant peaks.

[0062] In a specific embodiment, for example, approximately 749 cm -1 Approximately 1152cm -1 Approximately 1587cm -1 Regions such as these can serve as candidate Raman wavenumber regions associated with cytochrome c, and can be used to help explain whether spectral differences may have biological significance.

[0063] Cluster Vectors only help explain which regions contribute more significantly in the full-spectrum model. The final classification is still based on the PCA-LDA model's prediction of the single-cell Raman spectrum, and determined by majority vote at the strain or sample level.

[0064] II. Sample discrimination stage

[0065] Step S7: Labeling, spectral acquisition, and model prediction of the sample to be tested.

[0066] The methane-oxidizing bacteria samples to be tested were processed in accordance with the steps established in the reference discrimination model. 13 The CH4 labeling conditions, unified detection time window, single-cell Raman spectral acquisition conditions, and data processing workflow were used to obtain multiple pre-processed single-cell Raman spectral data of the sample to be tested. Each single-cell Raman spectral data of the sample to be tested was input into a pre-established PCA-LDA discriminant model, which outputs the type I or type II prediction category corresponding to each single-cell Raman spectrum.

[0067] Step S8: Determining the characteristics of majority voting and approach type

[0068] For the same sample, the number of single-cell Raman spectra predicted as type I and type II by the PCA-LDA discriminant model is counted separately. When the number of single-cell Raman spectra predicted as type I is greater than the number of single-cell Raman spectra predicted as type II, the sample is determined to be closer to the carbon assimilation pathway characteristics of type I methanogenic bacteria; when the number of single-cell Raman spectra predicted as type II is greater than the number of single-cell Raman spectra predicted as type I, the sample is determined to be closer to the carbon assimilation pathway characteristics of type II methanogenic bacteria. When the number of single-cell Raman spectra predicted as type I and type II is equal, the number of single-cell Raman spectra included in the analysis can be increased and a majority vote can be performed again, or the sample can be recorded as not being able to determine its pathway type characteristics under the current model and current sampling quantity. This situation is a supplementary explanation of the judgment rules, and no situation of equal numbers occurred in the verification results shown in this embodiment.

[0069] III. Model Validation Analysis Explanation

[0070] To evaluate the ability of the PCA-LDA discriminant model to determine the pathway type characteristics of reference strains that did not participate in the establishment of the current model, the leave-one-out cross-validation method can be used for verification.

[0071] In each round of leave-one-out cross-validation, the complete single-cell Raman spectrum data of one reference strain is used as the test set, while the single-cell Raman spectrum data of the remaining reference strains are used as the training set. The PCA-LDA discriminant model for that round is built using only the training set. Then, the complete single-cell Raman spectrum data of the retained test strains are input into the model to obtain the type I or type II prediction category for each test spectrum. The strain-level determination type is then determined based on majority voting. Subsequently, the obtained strain-level determination type is compared with the known type of the test strain.

[0072] To evaluate the impact of differences in the number of single-cell Raman spectra of different reference strains in the training set on the model's judgment results, a sensitivity analysis of equal-sampling random sampling of the training set can be further performed. In this analysis, in each round of cross-validation by strain, the same number of single-cell Raman full-spectrum data are randomly selected from each training reference strain to rebuild the PCA-LDA discrimination model. The random sampling, model building, test strain prediction, and majority voting are repeated to statistically analyze the consistency between the strain-level judgment results and their known types under different training spectrum composition conditions.

[0073] The leave-one-out cross-validation by strain and the sensitivity analysis using equal-size random sampling of the training set are used to evaluate the performance and stability of the discriminant model within the reference strain range in this embodiment, and are not necessary steps in the judgment process for the test samples. The specific sample size, calculation settings, and results of the validation analysis will be described in subsequent embodiments.

[0074] In this invention, the term "single-cell Raman spectrum" mainly refers to the Raman spectral curve obtained from a single cell, which can be used for graphical display, spectral response description, and spectral quantity statistics.

[0075] The term "single-cell Raman full spectrum" emphasizes the complete single-cell Raman spectrum acquired within a preset wavenumber range, and is used to distinguish it from single-peak or local characteristic peak analysis.

[0076] The term "single-cell Raman full spectrum data" refers to the Raman wavenumber-intensity data set formed after the full spectrum is digitized, which is the actual input for the PCA-LDA discriminant model and Cluster Vectors analysis.

[0077] Principal component analysis (PCA-LDA) and other necessary spectral processing procedures, such as Cluster Vectors, were performed in the IRootLab toolkit (https: / / code.google.com / p / irootlab / ) within MATLAB (version 2012a). The IRootLab version was IRootLab 0.13.11.14bj.

[0078] PCA-LDA and Cluster Vectors Analysis Parameter Setting Instructions:

[0079] In a specific embodiment, PCA-LDA analysis uses a uniformly preprocessed... 13 CH4-labeled single-cell Raman spectral data were used as input. The wavenumber range of the single-cell Raman spectrum was 500–3200 cm⁻¹. -1 Spectral preprocessing includes baseline correction and normalization; the same preprocessing procedure is used for reference strain data and test sample data.

[0080] PCA-LDA analysis is a tandem analysis process. PCA is used to reduce the dimensionality of single-cell Raman spectral data and extract principal component scores that reflect the differences in major spectra. LDA uses the principal component scores and the known type I or type II label of the reference strain as input to establish a discriminant model to distinguish between type I and type II reference spectra. The single-cell Raman spectral data of the sample to be tested is preprocessed in the same way and then input into the pre-established PCA-LDA discriminant model. The model outputs the type I or type II prediction category corresponding to each single-cell Raman spectrum.

[0081] During model building and model invocation, the spectral range, preprocessing procedures, PCA-LDA model parameters, and category label settings remain consistent. This invention does not use a fixed LD1 value or fixed projection coordinates as a universal discrimination threshold across models, but rather uses the predicted category output by the pre-built PCA-LDA discrimination model as the determination result at the single-cell spectral level.

[0082] Cluster Vectors analysis was performed using the same preprocessed single-cell Raman spectral data of the reference strain and the same type / type II grouping labels as the PCA-LDA discriminant model. The Cluster Vectors analysis was used to display the Raman wavenumber regions that contributed relatively significantly to the grouping differences between type I and type II reference strains.

[0083] Cluster Vectors analysis results are used to help interpret the spectral differences between the Type I and Type II reference spectra reflected by the PCA-LDA discriminant model, and are not used as an independent fixed threshold rule for determining whether a sample is Type I or Type II.

[0084] Example 1

[0085] 1. Strain source and type settings

[0086] To verify the feasibility of the technical solution of the present invention, typical methanogenic bacteria with known carbon assimilation pathways were selected as reference strains, and a control system of type I and type II methanogenic bacteria was established.

[0087] The selected reference strains include:

[0088] (1) Type I methanogenic bacteria (RuMP cycle)

[0089] ①Methylomonas sp. LW13

[0090] ②Methylomonas methanica MC09

[0091] ③Methylotuvimicrobium buryatense 5GB1C

[0092] ④Methylobacter sp. YRD-M1

[0093] (2) Type II methanogenic bacteria (serine cycle)

[0094] ①Methylosinus sp. LW4

[0095] ②Methylosinus sp. PW1

[0096] ③Methylosinus trichosporium OB3b

[0097] ④Methylocystis iwaonis SD4

[0098] The above 8 typical reference strains were used to obtain single-cell Raman full spectrum data with known carbon assimilation pathway types, and were used for subsequent PCA-LDA discriminant model establishment, leave-one-out cross-validation by strain, and sensitivity analysis by random sampling of the training set.

[0099] 2. Cultivation of pure culture strains and sample preparation

[0100] Methanogenic bacteria were inoculated into a closed culture system containing inorganic salts and cultured using methane as the sole carbon source.

[0101] The cultivation conditions are as follows:

[0102] Culture medium: NMS1 inorganic salt medium (the formulation and preparation method of this medium are existing technologies); Culture container: 100mL serum bottle; Liquid volume: 20mL; Temperature: 30℃;

[0103] NMS1 Inorganic Salt Culture Medium Preparation Method

[0104] (1) Add the following substances sequentially to 800 mL of ultrapure water:

[0105]

[0106] (2) After making up to 1000 mL with ultrapure water, transfer it to a sealed blue-capped bottle.

[0107] (3) High temperature and high pressure sterilization (parameters: 121℃, 20 min).

[0108] (4) In the laminar flow hood, add 50 mL / L of autoclaved phosphate solution and 2 mL / L of trace element solution sterilized by sterile filter head to the sterilized solution.

[0109] Oscillation conditions: 180–220 rpm; methane concentration: 20% (V / V).

[0110] During the cultivation process, the methane content in the system is monitored regularly.

[0111] 3. Monitoring of changes in methane utilization and determination of a unified detection time window.

[0112] During the pure culture of the reference strain, the amount of methane remaining in the culture system or the change in methane detection results were measured every 6 hours within the range of 0–42 h, and the culture process was divided into stages based on the growth status of the strain.

[0113] Based on the changes in the growth status of the strain and the methane detection results during the culture process, the culture process can be divided into early logarithmic growth phase, middle logarithmic growth phase, late logarithmic growth phase, stationary phase, and late stationary phase. Furthermore, combining the comparison results of single-cell Raman spectrum distributions at different growth stages, this embodiment selects the middle logarithmic growth phase as the unified detection time window used for comparative analysis of the reference strain and analysis of the samples to be tested.

[0114] The unified detection time window is used to reduce the impact of differences in single-cell Raman spectrum distribution caused by different growth stages on the comparison of the characteristics of type I and type II carbon assimilation pathways, rather than as an independent type determination rule.

[0115] 4. 13 CH4 isotope labeled culture

[0116] During the sample processing within the unified detection time window, the reference strains were placed in... 13 CH4 labeling conditions and 12 The two groups were treated under CH4 control conditions. Except for the difference in methane isotope composition, the other culture conditions were kept the same for both groups.

[0117] In this embodiment, 13 CH4 marker group and 12 In the CH4 control group, the amount of methane added was 20% (v / v) of the gas phase volume of the culture system, and the treatment time was 24 h. After treatment, bacterial samples were collected for single-cell Raman spectroscopy acquisition.

[0118] in, 12 Single-cell Raman spectra obtained under CH4 control conditions were used for observation. 13 The spectral response changes induced by CH4 labeling and the auxiliary interpretation of candidate Raman wavenumber regions; data used to establish type I and type II PCA-LDA discriminant models and to perform strain-based leave-one-out cross-validation and sensitivity analysis using equal-size random sampling of the training set. 13 Single-cell Raman full spectrum data obtained under CH4 labeling conditions.

[0119] 5. Single-cell Raman spectroscopy acquisition

[0120] Will be completed 12 CH4 control treatment or 13After CH4-labeled bacterial cells were washed with sterile water by centrifugation, they were dropped onto an aluminum foil surface and allowed to air dry. Single-cell Raman spectra were acquired using a LabRAM Aramis confocal micro-Raman system (HORIBA Jobin-Yvon, Japan). This system was equipped with a 532 nm Nd:YAG excitation source and a 300 grooves / mm diffraction grating, and a 100x air objective lens (Olympus, Japan) with a numerical aperture of 0.9 was used for focusing and acquiring the bacterial Raman signals.

[0121] Before each measurement, the silicon wafer was measured at 520.6 cm⁻¹. -1 The instrument was calibrated using the Raman peak. At least 20 single cells were randomly selected from each reference strain to collect the full Raman spectrum, with the spectral scanning range being 500–3200 cm⁻¹. -1 The single-point spectral acquisition time is 15s.

[0122] In this embodiment, the model is used for establishing and validating the discriminant model between type I and type II PCA-LDA. 13 The actual number of CH4-labeled single-cell Raman spectra included is shown in Table 1. Subsequent model analyses used preprocessed single-cell Raman full-spectrum data as input.

[0123] 6. Data Analysis and Model Validation Calculation Methods

[0124] The data used for model building and validation analysis consisted of eight typical reference strains. 13 The single-cell Raman spectrum data obtained under CH4 labeling conditions after uniform preprocessing consisted of 193 spectra, of which 100 spectra corresponded to type I reference strains and 93 spectra corresponded to type II reference strains.

[0125] Each single-cell Raman spectrum consists of the spectral intensity values ​​of that single cell at multiple Raman wavenumber positions. During model analysis, each single-cell Raman spectrum is used as an independent input sample.

[0126] A PCA-LDA method was used to establish a discrimination model between type I and type II reference spectra. PCA was used to reduce the dimensionality of the large number of wavenumber intensity variables contained in the single-cell Raman spectrum to a comprehensive coordinate system that can reflect the main spectral changes. LDA was used to learn the classification rules between the two reference spectra based on the known type I or type II labels of the reference strain, and output the type I or type II predicted category for the test single-cell Raman spectrum input to the model.

[0127] In the leave-one-out cross-validation, in each round, the complete single-cell Raman spectrum of one reference strain is used as the test set, and the single-cell Raman spectra of the remaining seven reference strains are used as the training set. The left-out test strains do not participate in the model building for that round, but are input into the PCA-LDA discriminant model as objects to be judged after the model is built.

[0128] For each selected test strain, the number of Raman spectra of all its single-cell samples predicted by the model to be type I and type II is counted. When the number of spectra predicted as type I is large, the strain-level determination type of the test strain is determined to be type I; when the number of spectra predicted as type II is large, the strain-level determination type of the test strain is determined to be type II.

[0129] To evaluate the impact of the number of spectra from different reference strains in the training set and the specific composition of the training spectra on the judgment results, a sensitivity analysis of equal-sampling random sampling of the training set was further conducted. In each round of leave-one-out cross-validation, the same number of single-cell Raman spectra were randomly selected from each training reference strain for re-establishing the model. In this embodiment, 22 spectra were randomly selected from each training reference strain, and a total of 154 training spectra were used in each modeling session; the modeling and judgment were repeated 100 times for each retained test strain.

[0130] 7. Experimental Results

[0131] 7.1 13 Results of single-cell Raman spectrum response analysis of typical reference strains under CH4 labeling conditions

[0132] For observation 13 The single-cell Raman spectrum response of typical methanogenic bacteria reference strains under CH4 labeling conditions was analyzed in this example. Eight typical reference strains were subjected to different conditions. 12 CH4 control conditions and 13 CH4 labeling conditions were applied, and single-cell Raman full-spectrum data were collected.

[0133] like Figure 1 As shown in Figure A, single-cell Raman average spectra suitable for comparative analysis were obtained for all eight typical reference strains. Figure 1B shows the results obtained based on single-cell Raman full-spectrum data. 12 CH4 control group and 13 The PCA-LDA distribution results of the CH4-labeled group are used to observe the overall spectrum distribution under different isotope treatment conditions. Figure 1 C displays the corresponding ClusterVectors analysis results, used to show the... 12 CH4 control group and 13 The Raman wavenumber region where differences in the spectral distribution of the CH4-labeled group contribute significantly. Figure 1 D showed 12 CH4 control group and13 Comparison of average Raman spectra of CH4-labeled groups.

[0134] The above results demonstrate that, under the culture, isotope treatment, and Raman sampling conditions employed in this embodiment, samples suitable for comparison can be obtained. 12 CH4 control conditions and 13 Single-cell Raman full-spectrum data of spectral response differences under CH4 labeling conditions. Based on these results, subsequent discriminant analysis of type I and type II carbon assimilation pathways was conducted. 13 The single-cell Raman spectrum obtained under CH4 labeling conditions was used as the model input data.

[0135] 7.2 Growth status of representative reference strains and changes in methane content in the culture system

[0136] To determine a uniform detection time window for comparative analysis of different reference strains, this example selected type I reference strain *Methylomonas* sp. LW13 and type II reference strain *Methylocystis iwaonis* SD4 as representative strains. Their growth status and changes in methane content during cultivation were observed, and the results are as follows: Figure 2 As shown.

[0137] Figure 2 A shows the growth status of Methylomonas sp. LW13 and the changes in methane detection results. Figure 2 B presents the corresponding results for Methylocystis iwaonis SD4. Based on the phased changes exhibited by the representative reference strain during cultivation, the cultivation process was divided into early logarithmic growth phase, middle logarithmic growth phase, late logarithmic growth phase, stationary phase, and late stationary phase.

[0138] This embodiment further combines the comparative analysis results of single-cell Raman spectra at different growth stages, selecting the mid-logarithmic growth phase as the unified detection time window for subsequent reference strain model establishment and sample analysis. This time window is used to control the consistency of growth stage conditions when comparing spectra of different samples, and is not used as an independent criterion for type I or type II determination.

[0139] 7.3 Effects of different growth stages on the distribution of single-cell Raman spectra

[0140] To observe the effect of growth stage changes on the distribution of single-cell Raman spectra, this example compares and analyzes the single-cell Raman spectra obtained from the typical reference strain Methylomonas sp. LW13 in the early logarithmic growth phase, middle logarithmic growth phase, late logarithmic growth phase, stationary phase, and late stationary phase.

[0141] like Figure 3 As shown in Figure A, the positions of single-cell Raman spectra obtained at different growth stages differ in the PCA-LDA distribution map. For example... Figure 3 As shown in B, Cluster Vectors analysis is used to display the Raman wavenumber regions that contribute significantly to the differences in spectral distribution at different growth stages. Figure 4 This demonstrates typical reference strains at different growth stages. 12 CH4 control conditions and 13 Average single-cell Raman spectra obtained under CH4 labeling conditions.

[0142] The above results indicate that changes in growth stage may affect the overall distribution of single-cell Raman spectra. Therefore, in the subsequent establishment of type I and type II PCA-LDA discrimination models based on reference strains and the model prediction of the test samples, this embodiment uses a unified detection time window to obtain single-cell Raman full-spectrum data for comparison.

[0143] 7.4 Results of Type I and Type II PCA-LDA Discriminant Analysis Based on Eight Typical Reference Strains

[0144] To establish a reference discriminant model for the carbon assimilation pathway characteristics of type I and type II methanogenic bacteria, this embodiment uses eight typical reference strains with known carbon assimilation pathway types, including four type I reference strains and four type II reference strains. The model input data consists of the data from the aforementioned reference strains. 13 The single-cell Raman spectrum data obtained under CH4 labeling conditions after uniform preprocessing consisted of 193 spectra, of which 100 spectra corresponded to type I reference strains and 93 spectra corresponded to type II reference strains.

[0145] like Figure 5 As shown in Figure A, after establishing the PCA-LDA discriminant model based on the single-cell Raman spectra of the above 8 typical reference strains, the type I and type II reference spectra exhibit different distributions in the discriminant space of this reference model. This result is used to illustrate that the reference strains used in this embodiment, 13 Within the range of CH4 labeling conditions, spectral acquisition conditions, and data processing procedures, single-cell Raman full spectrum can be used to establish a discrimination model that distinguishes between type I and type II reference spectra.

[0146] Figure 5 B shows the corresponding Cluster Vectors analysis results, which are used to display the Raman wavenumber regions that contribute significantly to the grouping of type I and type II reference strains. Figure 5 C shows 4 type I reference strains and 4 type II reference strains in 12 Comparison conditions and 13Single-cell Raman spectra under CH4 labeling conditions were used to assist in observing the differences in spectral responses between two types of reference strains in the candidate contribution wavenumber region.

[0147] In this embodiment, the classification input for distinguishing between type I and type II pathways is: 13 Single-cell Raman full-spectrum data obtained under CH4 labeling conditions; Cluster Vectors analysis and candidate Raman wavenumber region observation are used to help interpret the spectral differences reflected by the model, and are not used as independent fixed threshold classification rules. Figure 5 The complete reference model shown is used to demonstrate the distribution of eight known type reference strains; whether strains that did not participate in the model establishment of the current round can obtain strain-level determination results consistent with the known types is evaluated by subsequent leave-one-out cross-validation.

[0148] 7.5 Results of leave-one-out cross-validation of the PCA-LDA discriminant model by strain

[0149] To evaluate the ability of the established PCA-LDA discriminant model to determine the carbon assimilation pathway type characteristics of reference strains that did not participate in the model establishment in this round, this embodiment uses a leave-one-out cross-validation method for validation, the process of which is as follows: Figure 6 As shown in Figure A.

[0150] This embodiment includes a total of 8 typical reference strains, therefore a total of 8 rounds of validation were performed. In each round of validation, the entire sample of one reference strain was used. 13 CH4-labeled single-cell Raman spectra were used as the test set, along with the remaining 7 reference strains. 13 CH4-labeled single-cell Raman spectra are used as the training set; the PCA-LDA discrimination model for this round is established using only the training set, and then all single-cell Raman spectra of the retained test strains are input into the model to obtain the type I or type II prediction category corresponding to each test spectrum, and the strain-level determination type of the test strain is determined based on the majority vote.

[0151] The results of eight rounds of cross-validation with one strain retained are as follows: Figure 6As shown in B and Table 2, for type I reference strains, all 24 single-cell Raman spectra of *Methylomonas* sp. LW13 were predicted to be type I; all 29 single-cell Raman spectra of *Methylomonasmethanica* MC09 were predicted to be type I; of the 22 single-cell Raman spectra of *Methylotuvimicrobiumburyatense* 5GB1C, 14 were predicted to be type I and 8 were predicted to be type II; and of the 25 single-cell Raman spectra of *Methylobacter* sp. YRD-M1, 14 were predicted to be type I and 11 were predicted to be type II. Based on the majority vote, the strain-level classification of the above four type I reference strains is all type I.

[0152] For the type II reference strains, 23 out of 24 single-cell Raman spectra of *Methylosinus* sp. LW4 were predicted to be type II; 17 out of 22 single-cell Raman spectra of *Methylosinus* sp. PW1 were predicted to be type II; 23 out of 24 single-cell Raman spectra of *Methylosinus trichosporium* OB3b were predicted to be type II; and 21 out of 23 single-cell Raman spectra of *Methylocystis iwaonis* SD4 were predicted to be type II. Based on the majority vote, the strain-level determination of the above four type II reference strains is all type II.

[0153] Therefore, within the range of the eight typical reference strains selected in this embodiment, the strain-level determination types of the eight test strains obtained by eight rounds of cross-validation with one strain left over are all consistent with their known types.

[0154] 7.6 Sensitivity Analysis Results Based on Equal-Quantity Random Sampling

[0155] Since the number of single-cell Raman spectra of different reference strains included in the PCA-LDA discriminant model varies in this embodiment, in order to examine the impact of changes in the number of spectra and specific spectra composition of different strains in the training set on the strain-level judgment results, this embodiment further conducts sensitivity analysis of equal-quantity random sampling of the training set.

[0156] In each round of leave-one-out cross-validation, the selected test strains were still predicted using all their single-cell Raman spectra. For the remaining seven training reference strains, 22 single-cell Raman spectra were randomly selected from each strain to rebuild the PCA-LDA discriminant model. Therefore, the number of training spectra used in each remodeling was 154. For each selected test strain, random sampling, model building, test strain prediction, and majority voting were repeated 100 times, and the proportion of times the strain-level judgment result matched the known type of the test strain was counted.

[0157] The results are as follows Figure 6 As shown in C and Table 3. For type I reference strains, *Methylomonas* sp. LW13 and *Methylomonas methanica* MC09 were consistent with their known type in 100 out of 100 replicate modeling tests; *Methylotuvimicrobium buryatense* 5GB1C was consistent with its known type in 71 out of 100 replicate modeling tests; and *Methylobacter* sp. YRD-M1 was consistent with its known type in 82 out of 100 replicate modeling tests. For type II reference strains, *Methylosinus* sp. LW4, *Methylosinus* sp. PW1, *Methylosinustrichosporium* OB3b, and *Methylocystis iwaonis* SD4 were consistent with their known type in 100 out of 100 replicate modeling tests.

[0158] As shown in Table 4, out of all 800 repeated modeling decisions, 753 results at the strain level were consistent with the known type of the corresponding test strain. Among them, 353 out of 400 repeated modeling decisions for type I reference strains were consistent, and 400 out of 400 repeated modeling decisions for type II reference strains were consistent.

[0159] The above results demonstrate that, within the data range of this embodiment, when the composition of the training spectra changes randomly, most reference strains can maintain strain-level determination results consistent with their known types. Meanwhile, the consistency rates for *Methylotuvimicrobiumburyatense* 5GB1C and *Methylobacter sp.* YRD-M1 are 71% and 82%, respectively, lower than the other reference strains, suggesting that the current discrimination model's determination results for some type I reference strains are more susceptible to changes in the composition of the training spectra. The sensitivity analysis is used to evaluate the stability of the model's determination results within the current range of reference strains; it does not increase the number of independent validation strains, nor does it replace subsequent independent validation using newly added strains.

[0160] Table 1. Number of spectra for eight typical reference strains and their discrimination models

[0161]

[0162] Table 2. Results of leave-one-out cross-validation by strain

[0163]

[0164] Table 3. Results of Repeatability Determination Based on Equal-Quantity Random Sampling

[0165]

[0166] Table 4. Summary of results of strain-based leave-one-out cross-validation and random sampling sensitivity analysis

[0167]

[0168] The above description is merely a preferred embodiment of the present invention, and therefore should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made in accordance with the scope of the patent and the contents of the specification should still fall within the scope of the present invention.

Claims

1. A method for distinguishing the characteristics of carbon assimilation pathways in methanogenic bacteria, characterized in that, Includes the following steps: Step 1, refer to the discriminant model establishment steps, including sub-steps S1 to S5: S1: Reference strain selection: Typical methane-oxidizing bacteria with known carbon assimilation pathways were selected as reference strains. These reference strains included type I and type II reference strains; the type I reference strains included... Methylomonas sp. LW13、 Methylomonas methanica MC09 Methylotuvimicrobium buryatense 5GB1C and Methylobacter sp. YRD-M1; the type II reference strain includes Methylosinus sp. LW4、 Methylosinus sp. PW1、 Methylosinus trichosporium OB3b and Methylocystis iwaonis SD4; S2: Reference strain labeling and single-cell Raman spectral acquisition: The reference strain was placed in a container... 13 Stable isotope labeling was performed in the CH4 labeled culture system, and single-cell Raman full spectrum data of each reference strain were acquired by microconfocal Raman spectrometer within a unified detection time window. S3: Preprocessing of single-cell Raman full spectrum of reference strain: Preprocess the single-cell Raman full spectrum data of the reference strain according to a unified data processing procedure to obtain a single-cell Raman full spectrum dataset of the reference strain for establishing a discrimination model. S4: PCA-LDA discriminant model establishment: Using the preprocessed single-cell Raman spectrum data of the reference strain obtained in step S3 as the model input, the single-cell Raman spectrum data is first reduced in dimensionality by principal component analysis. Then, a linear discriminant analysis model is established based on the obtained principal component scores and the known carbon assimilation pathway type of the reference strain, thereby forming a PCA-LDA discriminant model for distinguishing between type I and type II reference spectrum types. S5: Interpretation of Contributing Wavenumber Regions: Based on the pretreated reference strain obtained in step S3 for establishing the PCA-LDA discriminant model. 13 Cluster Vectors analysis was performed on CH4-labeled single-cell Raman full spectrum data to show the Raman wavenumber regions that contributed significantly to the grouping of type I and type II reference strains. The candidate Raman wavenumber regions were then used to provide an auxiliary interpretation of the spectral differences reflected by the PCA-LDA discriminant model. Step 2, the sample discrimination step, includes sub-steps S6 to S8: S6: Sample labeling and single-cell Raman spectroscopy acquisition: The methane-oxidizing bacteria sample is labeled according to the same procedure as in step S2. 13 The CH4 labeling conditions, detection time window, and spectral acquisition conditions were processed to obtain multiple single-cell Raman full-spectrum data of the sample to be tested. S7: Sample Model Prediction: After preprocessing the single-cell Raman spectrum data of the sample to be tested according to the same data processing flow as step S3, the data is input into the PCA-LDA discrimination model established in step S4. The model outputs the type I prediction category or type II prediction category corresponding to each single-cell Raman spectrum. S8: Determination of pathway type characteristics: Count the number of single-cell Raman spectra predicted as type I and type II in the same sample, and determine the carbon assimilation pathway type characteristics of the sample according to the majority voting rule. When the number of single-cell Raman spectra predicted as type I is greater than the number of single-cell Raman spectra predicted as type II, the sample to be tested is determined to be closer to the carbon assimilation pathway characteristics of type I methanogenic bacteria. When the number of single-cell Raman spectra predicted as type II is greater than the number of single-cell Raman spectra predicted as type I, the sample to be tested is determined to be closer to the carbon assimilation pathway characteristics of type II methanogenic bacteria.

2. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that; The Cluster Vectors analysis is used to help interpret the Raman wavenumber regions that contribute significantly to the grouping of Type I and Type II reference spectra, and is not used as an independent fixed threshold discrimination rule.

3. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that, The PCA-LDA discriminant model uses the eight reference strains in claim 1 in... 13 Preprocessed single-cell Raman full-spectrum data obtained under CH4 labeling conditions were used as model input to establish the discrimination rule between type I reference spectra and type II reference spectra.

4. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that, In observation 13 When CH4 labeling causes changes in the single-cell Raman spectrum response, set 12 CH4 control conditions and 13 Comparison of spectra obtained under CH4 labeling conditions; the 12 The CH4 control condition is used to observe the labeled response and to help interpret candidate Raman wavenumber regions.

5. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that, The mid-logarithmic growth phase was selected as the unified detection time window for single-cell Raman spectrum acquisition of the reference strain and comparative analysis of the test samples.

6. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that, In step S3, the preprocessing process includes baseline correction and normalization processes used in the actual implementation.

7. The method according to claim 1, characterized in that, In step S5, the spectral differences reflected by the PCA-LDA discriminant model are interpreted by combining candidate Raman wavenumber regions related to cytochrome C.

8. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 7, characterized in that, In step S5, the Raman wavenumber region that contributes significantly includes 749 cm⁻¹. -1 1152cm -1 and 1587cm -1 The area.

9. The method for distinguishing the carbon assimilation pathway types of methanogenic bacteria according to claim 1, characterized in that, In step S8, the majority voting rule refers to: When the number of single-cell Raman spectra predicted as type I by the PCA-LDA discrimination model in the same test sample is greater than the number of single-cell Raman spectra predicted as type II, the test sample is determined to be closer to the carbon assimilation pathway characteristics of type I methanogenic bacteria. When the number of single-cell Raman spectra predicted as type II by the PCA-LDA discrimination model in the same test sample is greater than the number of single-cell Raman spectra predicted as type I, the test sample is determined to be closer to the carbon assimilation pathway characteristics of type II methanogenic bacteria.

Images