A mass spectrometry peak annotation method based on co-expression, co-flow, and similar secondary mass spectrum

Abstract

The application discloses a mass spectrum peak annotation method based on co-expression, co-flowing and similar secondary mass spectrum diagrams, and belongs to the technical field of mass spectrum data annotation.The application is based on the characteristics that different mass spectrum characteristics of the same substance have co-expression, co-flowing and similar secondary mass spectrum diagrams in different samples, and a set of efficient annotation method for high-resolution mass spectrum data is established by filtering the mass differences of common adducts, multimers, isotopes and in-source cleavage fragments, so that the time-consuming and labor-consuming difficulty of the structure annotation step of unknown substances in actual metabolomics and other fields due to redundant mass spectrum information can be solved.The mass spectrum data annotation method is accurate, time-saving and labor-saving, and has important significance for accurate annotation of redundant information of unknown substances in actual samples.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mass spectrometry peak data annotation, in particular to a mass spectrometry peak annotation method based on co-expression, co-elution and similar secondary mass spectrometry graphs. BACKGROUND

[0002] Non-targeted screening technology based on high performance liquid chromatography-high resolution mass spectrometry provides a powerful tool for the identification of unknown substances. After being separated by chromatography, the substances are charged by ion source and then enter the mass spectrometer. After ionization, a single compound will produce one or more ions, including isotope ions, in-source fragmentation ions, different adducts and polymers. In the metabolomics study of E. coli by Mahieu and Patti (2017), 12797 features were reduced to 7318 by redundant peak removal, and these features were all from different features of the same analyte. If these redundant features are identified one by one, not only will it increase the workload for the identification of substances, but it may also cause the wrong interpretation of sample data, for example, the in-source fragmentation fragments may be mistakenly thought to be caused by biological or chemical transformation rather than ionization process.

[0003] Many studies group related features based on three characteristic patterns: chromatographic peak shape similarity, co-elution, and co-expression, and further annotate them with accurate mass to achieve the effect of data cleaning. Co-elution means that the peak time of the chromatographic peak is similar, and co-expression means that the abundance changes in different samples are similar. CAMERA is one of the most widely used software platforms for annotating redundant signals, which uses the co-elution principle to first create a composite spectrum with similar retention times, and then based on abundance similarity, isotopic relationship and peak shape similarity, finally uses a graph-based algorithm to divide the related features into modules to annotate the isotopic patterns and additive peaks of unknown substances. The XMSannotator R package uses co-expression and co-elution characteristics to cluster and annotate metabolites through network module analysis; other platforms such as Binner, CliqueMS, Mz.Unity, mWISE, etc. are all based on these characteristic patterns to annotate redundant peaks. However, a single characteristic pattern is not enough to distinguish different mass spectrometry features of the same substance, which may result in a high false positive rate. For example, although co-elution is one of the characteristics of redundant data of the same substance, co-elution of substances is very common in the chromatography-high resolution mass spectrometry system. For example, a sample from a sewage plant may contain tens of thousands of peaks, and the gradient of tens of minutes of chromatography cannot separate all these pollutants, and many substances have similar retention times and co-elution. The fragment characteristics of secondary mass spectrometry can reflect the structural information of the substance, and to some extent, the secondary mass spectrometry graphs of substances with similar structures are also similar, so the secondary mass spectrometry graph information is considered to have high reliability in the process of substance identification. ISFrag improves on the basis of CAMERA and adds secondary mass spectrometry fragment information to annotate in-source fragmentation. Currently, there is no literature or patent that uses co-expression, co-elution, and secondary mass spectrometry similarity to annotate multiple redundant features such as in-source fragmentation, additive, isotope, and multimer at the same time. These characteristic patterns are the most important three principles in the process of distinguishing redundant data in mass spectrometry data. SUMMARY

[0004] The purpose of the present application is to solve the problem of too much redundant data in the process of non-target substance identification, which brings a lot of workload for subsequent substance structure identification, and even may lead to incorrect sample interpretation.

[0005] To solve the above problems, the technical scheme of the present application is as follows:

[0006] A mass spectrometry peak annotation method based on co-expression, co-elution, and secondary mass spectrometry graph similarity, which uses the characteristics of co-expression, co-elution, and secondary mass spectrometry graph similarity of each mass spectrometry feature of the same substance in each sample, and uses the mass difference of additive, multimer, isotope, and in-source fragmentation fragments to accurately identify and annotate the mass spectrometry peak, specifically including the following steps:

[0007] S1, clustering all mass spectrometry features in each sample based on abundance, and dividing feature modules through abundance similarity network analysis to determine mass spectrometry features co-expressed in each sample;

[0008] S2, clustering all mass spectrometry features in each sample based on retention time to determine co-flow detection modules, i.e. mass spectrometry feature modules with similar retention times;

[0009] S3, filtering mass spectrometry features in the same feature module in the abundance similarity network and the retention time through secondary mass spectrometry graph similarity to determine mass spectrometry features with similar structures;

[0010] S4, subtracting the accurate mass numbers of pairs of mass spectrometry features satisfying steps S1, S2 and S3, and using the mass difference characteristics of adducts, in-source fragmentation fragments and isotopic patterns to locate and annotate each mass spectrometry feature of the same substance.

[0011] Wherein, the method for obtaining all mass spectrometry features of each sample includes the following contents: after peak extraction, peak deconvolution and peak alignment of the mass spectrometry data of each sample, basic information of feature peaks is obtained, including: accurate mass number, retention time, primary mass spectrometry graph, secondary mass spectrometry graph and abundance of the same feature peak in different samples.

[0012] Further, the sample is a metabolome or an environmental sample, such as urine, blood, sewage, surface water sample, etc.

[0013] Further, each mass spectrometry feature is the result after peak extraction, peak deconvolution and peak alignment operations, and the operations such as peak extraction, peak deconvolution and peak alignment can be completed by software MSDIAL.

[0014] Further, the similarity network analysis in step S1 is based on R language package XMSannotator.

[0015] Further, in step S1, the abundance similarity network analysis is completed by weighted gene co-expression network analysis (WGCNA), which includes the following contents:

[0016] A weighted gene co-expression network is constructed based on abundance similarity, an n*n adjacency matrix is obtained, and then module detection is performed through the weighted gene co-expression network, the network interconnectivity based on topological overlap measure is calculated, and finally the closely connected co-expressed feature modules are found based on hierarchical clustering, wherein the module is defined as a closely connected feature cluster.

[0017] Preferably, in step S1, the basis for feature module division by abundance similarity network analysis is that the positive correlation threshold of the mass spectrometry feature is greater than the judgment value, and the judgment value can be selected according to the abundance change of the pollutant abundance feature in the sample, and if it cannot be judged, it can be set as a default value of 0.7.

[0018] Preferably, step S2 is completed by R language package XMSannotator to cluster each mass spectrometry feature based on retention time, and a kernel density function is used to detect and determine each module of co-flow, wherein the default value of the retention time difference is 10s.

[0019] Preferably, step S3 is realized by dot product algorithm, and the default value of similarity determination is 0.7.

[0020] Preferably, step S4 can be realized by R language, Python or EXCEL to traverse and compare the paired mass spectrometry features.

[0021] Preferably, the mass difference discrimination method of the adduct in step S4 is:

[0022] The positive ion mode of each substance is [M+H] + (The negative ion mode is [M-H] - ) adduct form must exist, wherein M is a neutral substance. For different mass spectrometry features A and B, difference is made, and if the difference value conforms to the difference value of different adduct forms and [M+H] + (or [M-H] - ), then one mass spectrometry feature is annotated as [M+H] + (or [M-H] - ), and the other mass spectrometry feature is annotated as other adduct form. For example, if A-H=B-NH4, then A is considered as [M+H] + , and B is considered as [M+NH4] + .

[0023] Preferably, the mass difference discrimination method of the adduct in step S4 is:

[0024] For different mass spectrometry features A and B, if the positive ion mode is (A-H) / (B-H)=n, and the negative ion mode is (A+H) / (B+H)=n, then A is considered as the n-mer of B. Other adduct forms of the n-mer are, for example, [2M+FA-H] - , and the calculation method is similar.

[0025] Preferably, the source internal cleavage fragment discrimination method in step S4 is:

[0026] For pairs of mass spectrometry features in the same module, if the precursor m / z of one mass spectrometry feature B exists in the fragment m / z of the secondary spectrum of another mass spectrometry feature A, it is considered that B is the in-source fragmentation fragment of A.

[0027] Preferably, the discrimination mode of the isotopic pattern in step S4 is: only considering the first isotope (range: 0.997-1.006) and the second isotope (range: 1.994-2.013).

[0028] The beneficial effects of the present application are:

[0029] (1) The present application comprehensively identifies redundant data of multiple modes, including isotopes, adducts, in-source fragmentation and multimers;

[0030] (2) The present application comprehensively considers abundance correlation and similar retention time, adds secondary mass spectrum fragment information to make the credibility of annotation higher;

[0031] (3) The present application greatly reduces the workload of subsequent structure identification, and reduces the possibility of incorrect sample interpretation caused by inappropriate annotation. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a flow chart of a mass spectrometry peak annotation method based on co-expression, co-flow, and similar secondary mass spectrum of Example 1;

[0033] Figure 2 is the number of redundant peaks removed by different methods in Experimental Example 1;

[0034] Figure 3 is the standard sample annotation rate of different methods in Experimental Example 1;

[0035] Figure 4 is the abundance change of NUM1107, NUM806 and NUM807 in different samples in Experimental Example 2;

[0036] Figure 5 is the spectrum of three peaks NUM1001, NUM836 and NUM1012 in Experimental Example 2. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0038] The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the description of the application and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0039] Embodiment 1

[0040] The embodiment is a mass spectrometry peak annotation method based on co-expression, co-elution, and similar secondary mass spectrometry graphs, characterized in that each mass spectrometry characteristic of the same substance has the characteristics of co-expression, co-elution, and similar secondary mass spectrometry graphs in each sample, and the mass spectrometry peak is accurately identified and annotated by using the mass difference of adducts, multimers, isotopes, and in-source fragmentation fragments. The sample is a metabolome sample, and in this embodiment, it is urine, and specifically includes the following steps:

[0041] S1, clustering all mass spectrometry characteristics in each sample based on abundance, the mass spectrometry characteristics are the results after peak extraction, peak deconvolution, and peak alignment operations, which are completed by software MSDIAL, and the feature module is divided by abundance similarity network analysis, and the division basis of the feature module division by the abundance similarity network analysis is that the positive correlation threshold of the mass spectrometry characteristics is greater than 0.7, to determine the mass spectrometry characteristics with co-expression in each sample, wherein the abundance similarity network analysis is based on the R language package XMSannotator and is completed by weighted gene co-expression network analysis, and specifically includes the following content:

[0042] A weighted gene co-expression network is constructed according to the abundance similarity to obtain an n*n adjacency matrix, then a module detection is performed by the weighted gene co-expression network, the network interconnectivity based on the topological overlap measure is calculated, and finally the closely connected co-expression feature modules are found based on hierarchical clustering, wherein the module is defined as a closely connected feature cluster, and the division basis of the feature module division by the similarity network analysis is that the positive correlation threshold of the mass spectrometry characteristics is greater than a judgment value, and the judgment value is selected according to the change of the pollution characteristics in the sample;

[0043] S2, clustering all mass spectrometry characteristics in each sample based on retention time to determine the co-elution module, i.e. the mass spectrometry characteristics with similar retention time, and the clustering of each mass spectrometry characteristic based on the retention time is completed by the R language package XMSannotator, and the kernel density function is used to detect each module with co-elution, and in this embodiment, the retention time is 10s;

[0044] S3, the mass spectrometry characteristics in the same feature module in the abundance similarity network and the retention time are filtered according to the similarity of the secondary mass spectrometry graphs by the dot product algorithm to determine the mass spectrometry characteristics with similar structures in the same module, and the similarity judgment value is 0.7.

[0045] S4, the accurate mass numbers of the pairs of mass spectrum characteristics satisfying step S1, step S2 and step S3 are subtracted, the mass difference characteristics of adducts, in-source fragmentation fragments and isotopic patterns are used to locate and annotate each mass spectrum characteristic of the same substance, and the pairs of mass spectrum characteristics are traversed and compared through R language, wherein:

[0046] The mass difference discrimination mode of the adduct is:

[0047] The [M+H] of each substance in the positive ion mode is set + The adduct form must exist in the negative ion mode [(M-H)-], wherein M is a neutral substance. For mass spectrum characteristics A and B, if the difference value conforms to the difference value of different adduct forms [M+H] + (or [M-H] - ), one mass spectrum characteristic is annotated as [M+H] + (or [M-H] - ), and the other mass spectrum characteristic is annotated as other adduct forms. For example, if A-H=B-NH4, it is considered that A is [M+H] + , B is [M+NH4] + , and the specific rules of the positive ion mode are shown in Table 1, and the specific rules of the negative ion mode are shown in Table 2.

[0048] Table 1: Mass difference of adducts and polymers, valence state and accurate mass in positive ion mode

[0049]

[0050]

[0051]

[0052] Table 2: Valence state and accurate mass difference of common adducts and polymers in positive ion mode

[0053] Adducts or multimers Valence state Exact mass difference [[M-H] - ]] 1 -1.00783 [[M-H2O-H] - ]]> 1 -19.01839 [M+Na-2H] - ]] 1 20.97412 [M+Cl] - ]] 1 34.96885 [M+K-2H] - ]] 1 36.94806 [M+FA-H] - ]] 1 44.99765 [[M+Hac-H] - ]]> 1 59.01330 [M+Br] - ]]> 1 78.91834 [[M+TFA-H] - ]]> 1 112.98504 [[M-C6H10O4-H] - ]]> 1 -147.06573 [[M-C6H10O5-H] - ]]> 1 -163.06065 [[M-C6H8O6-H] - ]]> 1 -177.03991 [[2M-H] - ]]> 1 -1.00783 [[2M+FA-H] - ]]> 1 44.99765 [[2M+Hac-H] - ]]> 1 59.01330 [3M-H] - ]] 1 -1.00783 [[M-2H] 2- ]]> 2 -2.01565 [[M-3H] 3- ]]> 3 -3.02348

[0054] The mass difference discrimination mode of the polymer is:

[0055] For mass spectrum characteristics A and B, if (A-H) / (B-H)=n in the positive ion mode, and (A+H) / (B+H)=n in the negative ion mode, it is considered that A is the n polymer of B. Other adduct forms of polymers, such as [2M+FA-H] - and the like, the calculation method is similar.

[0056] The discrimination mode of the in-source fragmentation fragment is:

[0057] For pairs of mass spectral features in the same module, if the precursor m / z of mass spectral feature B exists in the fragment m / z of the secondary spectrum of another mass spectral feature A, then B is considered to be the in-source cleavage fragment of A,

[0058] The discrimination method of the isotope pattern is: only the first isotope and the second isotope are considered, the determination range of the first isotope is: 0.997-1.006, and the determination range of the second isotope is: 1.994-2.013.

[0059] Example 2

[0060] The difference between this embodiment and example 1 is:

[0061] The sample is an environmental sample, which is sewage in this embodiment.

[0062] Step S4 traverses and compares pairs of mass spectral features through Python.

[0063] Example 3

[0064] The difference between this embodiment and example 1 is:

[0065] Step S4 traverses and compares pairs of mass spectral features through EXCEL.

[0066] Experimental Example 1

[0067] The object of this experimental example is a standard sample prepared by dissolving 126 standard samples in Fisher water (Emerson, USA, HPLC grade) with concentration gradients of 10, 50, 100, 200, 250 and 500 ug / L. Peak extraction, peak deconvolution and peak alignment are performed by MSDIAL. The specific parameters are as follows: Data collection: MS1 tolerance: 0.01 Da, MS2 tolerance: 0.025 Da, Peak detection: Minimum peak height: 1000 amplitude, Mass slice width: 0.1 Da; Alignment: Retention time tolerance: 0.05 min, MS1 tolerance: 0.015 Da, Adduct: All possible adduct forms are checked in positive and negative modes.

[0068] In the alignment result, in positive ion mode: 3241 peaks; in negative ion mode: 1159 peaks. Using MSDIAL, a total of 1401 H peaks (positive) and 568 H peaks (negative) are identified.

[0069] Firstly, the redundant peak network was constructed, and the peaks in positive ion mode and negative ion mode were edited into expression matrix data.csv imported into R language, the first column was mz(Da), and the second column was time(s), representing the accurate mass number and retention time, respectively.

[0070] The specific command line is as follows:

[0071] multilevelannotation(dataA, max.mz.diff = 60, max.rt.diff = 60, cormethod = "pearson", num_nodes = 3, queryadductlist = c("all"), mode = "pos", outloc = "D:\\gongzuomulu\\xmsannotator", adduct_weights = NA, num_sets = 3000, allsteps = FALSE, corthresh = 0.7, NOPS_check = TRUE, customIDs = NA, missing.value = NA, deepsplit = 2, networktype = "unsigned", minclustsize = 10, module.merge.dissimilarity = 0.2, filter.by = c("M+H"), redundancy_check = TRUE, min_ions_perchem = 1, biofluid.location = NA, origin = NA, status = NA, boostIDs = NA, max_isp = 5, MplusH.abundance.ratio.check = FALSE, customDB = NA, HMDBselect = "union", mass_defect_window = 0.01, mass_defect_mode = "pos", pathwaycheckmode = "pm")

[0072] The parameters can be adjusted according to the XMSannotator user manual. The first column in stage1.csv in the output file is the clustering result of XMSannotator according to the abundance similarity between different sewage samples of the same feature and the retention time between different features. For the same module, the substances are edited into a file including secondary mass spectrum fragment information, and the secondary spectrum similarity of the features in the same module is calculated.

[0073] The experiment example adopts Python to calculate the correlation between each two spectra, and the basic principle is DP similarity, to determine the substances with similar structures.

[0074] For the features similar to the pair of spectra, the mass difference of the pair of mass spectrometry features is calculated by Python software, and the mass difference of the common adduct, multimer, in-source fragmentation fragment, and isotope pattern is used to locate different mass spectrometry features of the same substance. Table 1 and Table 2 are lists of common adducts and multimers respectively. The first isotope range is 0.997-1.006, and the second isotope range is 1.994-2.013. Whether it is an in-source fragmentation fragment is determined by whether the mass number of the fragment is included in the secondary spectrum of the precursor. Whether it is a different spectrum feature of the same substance is determined by traversing the difference. And the redundant peaks are annotated in combination with abundance similarity, retention time similarity and secondary spectrum similarity. In the experiment example, the threshold for abundance similarity is 0.7, and the threshold for secondary spectrum similarity is 0.7.

[0075] As shown in Figure 2 , the method removes 71 redundant peaks (6.13%) in negative ion mode; and removes 180 redundant peaks (5.55%) in positive ion mode. The annotation rate is tested by the accurate mass number of the standard substance, and the result is shown in the figure: a total of 126 standard substances, 59 are annotated in negative ion mode (excluding 57 repeated annotations); the number of MSDIAL annotations is 40 (excluding 43 repeated annotations). In positive ion mode, 71 are annotated (excluding 39 repeated annotations), and 53 are annotated by MSDIAL (excluding 67 repeated annotations).

[0076] The total standard substance annotation rate of MSDIAL is 65, and the annotation rate is 51.59%; the total annotation number of the experiment example is 90, and the annotation rate is 71.43%. As shown in Figure 2 , compared with MSDIAL, the method has a higher annotation rate (true positive rate), but the number of removed redundant peaks is low. In the process of removing redundancy, the method deletes the redundancy by more stringent standards (co-elution, spectrum similarity and abundance similarity), greatly retains the target substance, thereby increasing the credibility of the process of removing redundant peaks, and finally the negative ion mode redundant peak annotation result is shown in Table 3, and the positive ion mode redundant peak annotation result is shown in Table 4.

[0077] Table 3 Experiment Example 1 Negative Ion Mode Redundant Peak Annotation Results

[0078]

[0079]

[0080] Table 4 Experiment Example 1 Positive Ion Mode Redundant Peak Annotation Results

[0081] Redundant peak forms Number [[2M+H] + ]]> 2 [[2M+NH4] + ]]> 1 [M+H+Na] 2+ ]]> 4 [M+H-H2O] + ]] 2 [M+NH4] + ]]> 4 In-source fragmentation 18 Isotopes 18 Same substance 38 Redundant peak forms Number

[0082] Experimental Example 2

[0083] The object of this experimental example is to collect municipal sewage water samples from 15 different cities across the country, and perform peak extraction, peak deconvolution and peak alignment via MSDIAL. The specific parameters are as follows: Data collection: MS1 tolerance: 0.01 Da, MS2 tolerance: 0.025 Da, Peak detection: Minimum peak height: 10000 amplitude, Massslice width: 0.1 Da; Alignment: Retention time tolerance: 0.05 min, MS1 tolerance: 0.015 Da, Adduct: Check all possible adduct forms in positive and negative modes.

[0084] Export the alignment results, and in total 1326 peaks in positive ion mode and 294 peaks in negative ion mode are obtained from the 15 sewage plants, and each peak is named as NUM1~1620.

[0085] The step of removing redundant peaks and the parameters are the same as in Experimental Example 1.

[0086] In negative mode, 16 redundant peaks (5.44%) were removed, and 280 redundant peaks (21.12%) were removed. To illustrate the superiority of the method, the accurate mass of three peaks NUM806, NUM807 and NUM1107 were 232.1902, 232.1904 and 232.1904, respectively, and the retention time were 21.863, 21.011 and 22.505, respectively. The accurate mass of the three peaks were similar, and the spectrum similarity were both high. The DP between each other were: NUM806-NUM807: 0.92582, NUM806-NUM1107: 0.92582, NUM807-NUM1107: 1. Through xMSannotator clustering, it was found that although the difference of the accurate mass of the three peaks was less than 0.0002 Da, the abundance of the three peaks changed differently in different samples, and the retention time was quite different (more than 1 min). NUM807 and NUM1107 might be isomers of each other, and NUM806 might be a substance with similar structure to the two, with a slight structural change. This illustrates the importance of abundance and retention time in the identification of redundant peaks. In addition, NUM1001, NUM836 and NUM1012 belong to the same abundance and retention time module (illustrating that the abundance is similar and the retention time is similar), but the DP similarity between the three peaks were: NUM1001-NUM836: 0.719092, NUM1001-NUM1012: 0.544331, NUM836-NUM1012: 0.480384. NUM1012 has only the same two small fragments with the highest abundance as the other two peaks, and the fragments with larger mass number are quite different. Using CFM-ID to predict the structure shows that NUM836 and NUM1001 might be Demethylphylloquinone, and NUM1012 might be (25S)-26-Hydroxy-24-methylenecycloartan-3-one. Therefore, only abundance similarity and retention time similarity are not enough to prove that different peaks belong to different mass spectrum characteristics of the same substance. The final negative ion mode redundant peak annotation results are shown in Table 5, and the positive ion mode redundant peak annotation results are shown in Table 6.

[0087] Table 5 Redundant peak annotation results of negative ion mode in experimental example 2

[0088] Redundant peak forms Number [2M+H]+ 1 [M+CH3OH+H]+ 1 [M+H-H2O]+ 12 [M+NH4]+ 24 In-source fragmentation 17 Isotopes 1 Same substance 224 [2M+H]+ 1

[0089] Table 6 Redundant peak annotation results of positive ion mode in experimental example 2

[0090] Redundant peak forms Number In-source fragmentation 4 Isotopes 4 Same substance 8

Claims

1. A method for annotating mass spectrometry peaks based on co-expression, co- flow-out, and similarity of secondary mass spectra, characterized in that, The mass spectrometry characteristics of the same substance in each sample have the characteristics of co-expression, co-elution and similar secondary mass spectrum, and the mass spectrometry peaks are accurately identified and annotated by using the mass difference of adducts, multimers, isotopes and in-source fragmentation fragments, and the specific steps include the following steps: S1, clustering all mass spectrometry characteristics in each sample based on abundance, and performing feature module division through abundance similarity network analysis to determine the co-expressed mass spectrometry characteristics in each sample; the sample is a metabolome or an environmental sample; the abundance similarity network analysis is realized based on R language package XMSannotator; the abundance similarity network analysis is completed through weighted gene co-expression network analysis, and the specific contents include the following: a weighted gene co-expression network is constructed according to the abundance similarity, an n*n adjacency matrix is obtained, then module detection is performed through the weighted gene co-expression network, the network interconnectivity based on topological overlap measure is calculated, and finally the closely connected co-expressed feature modules are found based on hierarchical clustering, wherein the module is defined as a closely connected feature cluster; the division basis of the feature module division through the abundance similarity network analysis is that the positive correlation threshold of the mass spectrometry characteristics is greater than a judgment value, the judgment value can be selected according to the abundance change of the pollution characteristics in the sample, or can be a default value; S2, clustering all mass spectrometry characteristics in each sample based on retention time to determine the detected co-elution module, i.e. the mass spectrometry feature module with similar retention time; step S2 completes the clustering of each mass spectrometry characteristic based on retention time through R language package XMSannotator, and the kernel density function is used for detecting each module for determining the detected co-elution; S3, filtering the mass spectrometry characteristics in the same feature module in the abundance similarity network and the retention time through the secondary mass spectrum similarity to determine the mass spectrometry characteristics with similar structures; S4, subtracting the accurate mass numbers of the pairs of mass spectrometry characteristics satisfying steps S1, S2 and S3, and positioning and annotating each mass spectrometry characteristic of the same substance by using the mass difference characteristics of adducts, in-source fragmentation fragments and isotope patterns.

2. The method for annotating MS peaks based on co-expression, co- flow-out, and similar MS spectra of the second-order mass spectra according to claim 1, wherein, The each mass spectrometry characteristic is the result after peak extraction, peak deconvolution and peak alignment operation, and can be completed by software MSDIAL.

3. The mass spectral peak annotation method based on co-expression, co-eluenting, and similarity of secondary mass spectra as described in claim 1, characterized in that, The step S3 is realized by a dot product algorithm.

4. The mass spectral peak annotation method based on co-expression, co-eluenting, and similarity of secondary mass spectra as described in claim 1, characterized in that, The step S4 can be traversed and compared for the pairs of mass spectrometry characteristics through R language, Python or EXCEL.

Images