Non-targeted screening method for characteristic mass spectrum fingerprints of pollutants in water

Abstract

The invention discloses a non-targeted screening method for characteristic mass spectrum fingerprints of pollutants in water. The non-targeted screening method comprises the following preparation steps: S1, collecting and filtering; s2, performing mass spectrometry; s3, data processing; s4, constructing a fingerprint spectrum; and S5, performing non-targeted screening. In a characteristic mass spectrum fingerprint construction process, a plurality of advanced data analysis algorithms, principal component analysis, clustering analysis and a chemometrics method are combined, and a real representative characteristic peak can be more effectively mined from massive mass spectrum data in a multi-algorithm fusion mode; meanwhile, various means such as solid-phase extraction and acid-base adjustment are combined, targeted enrichment and treatment are carried out according to pollutant properties, a comprehensive database containing known pollutants and characteristic fingerprints newly found through the method is established, and a perfect index and classification system is configured.

Description

Technical Field

[0001] The present invention relates to the technical field of screening of pollutants in water, and in particular to a non-targeted screening method for characteristic mass spectrum fingerprints of pollutants in water. Background Art

[0002] With the vigorous development of modern industries such as chemical industry, printing and dyeing, electroplating, metallurgy and the rapid increase in the scale and scope of human activities, the types of pollutants in water are becoming increasingly complex. Pollutants have changed the physical and chemical properties of water bodies, such as pH, dissolved oxygen content, and transparency.

[0003] Traditional water quality monitoring methods often target specific known pollutants and conduct quantitative and qualitative analysis through standard substances. However, these targeted analysis methods have limitations for unknown new pollutants or complex mixed pollutants. Although there are some non-targeted analysis techniques, they still need to be improved in terms of accuracy, efficiency and adaptability to complex water samples. Accordingly, this application proposes a non-targeted screening method for characteristic mass spectrometry fingerprints of pollutants in water. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water, comprising the following screening steps:

[0007] S1. Collection and filtration. The collected water is stored in a polytetrafluoroethylene test tube and kept at a low temperature of about 4°C and protected from light. The water sample is acid-base adjusted and then solid phase extraction is performed. After each extraction step, the eluent is acid-base adjusted again as needed;

[0008] S2. Mass spectrometry analysis: the water to be tested is tested through mass spectrometer host, vacuum pump system, data acquisition system and other related equipment;

[0009] S3. Data processing: convert the data format output by the mass spectrometer, such as the original binary file or the specific instrument software format, into a universal data format. The noise characteristics of the mass spectrometry data are analyzed, and the appropriate denoising method is selected. PCA and cluster analysis can be organically combined. PCA processing is performed on high-dimensional mass spectrometry data containing mass-to-charge ratio and ion intensity information. The principal component is extracted through covariance matrix calculation and eigenvalue decomposition to reduce the dimension, denoise and remove redundancy. The principal component score is then used as a new feature for cluster analysis. K-means clustering is performed and parameters are adjusted according to data characteristics to cluster similar samples.

[0010] S4, fingerprint construction, the processed data is used to construct a mass spectrum fingerprint with mass-to-charge ratio as the horizontal axis and relative intensity as the vertical axis, and the fingerprint of each sample is stored in the database, which can intuitively reflect the overall characteristics of pollutants in water samples, including the types and relative contents of pollutants;

[0011] S5. Non-targeted screening: compare the constructed mass spectral fingerprint with the known NIST mass spectral database and EPA's pollutant database, and find the parts that match the known pollutants by calculating similarity indicators such as cosine similarity and Euclidean distance.

[0012] Preferably, in step S1, the acidic regulator includes hydrochloric acid HCl, nitric acid HNO 3 and sulfuric acid H 2 SO 4 , hydrochloric acid HCl reacts quickly when adjusting pH;

[0013] Alkaline regulators include sodium hydroxide NaOH and potassium hydroxide KOH, which change the pH value of water samples.

[0014] Preferably, in step S2, the pressure of the vacuum pump system, generally a high vacuum system, needs to reach 10 -5 -10 - 6 torr or so.

[0015] Preferably, in step S3, such as wavelet transform denoising, the Daubechies wavelet series and the number of decomposition layers are first selected to decompose the mass spectrometry signal. For the noise in the high-frequency sub-band, the processed sub-band is reconstructed to obtain the denoised signal after hard threshold or soft threshold processing, so as to make the data smoother and facilitate subsequent baseline correction and peak identification.

[0016] Preferably, in step S3, PCA is a statistical method that converts original high-dimensional data into a set of low-dimensional unrelated variable principal components through linear transformation. It is based on the covariance matrix of the data and aims to find the direction with the largest variance as the principal component direction, thereby retaining data information while reducing the dimension.

[0017] Preferably, in step S3, K-means clustering is performed by first specifying the number of clusters, then randomly initializing cluster centers, and calculating the distance from each sample to the cluster center, such as the Euclidean distance, to assign the sample to the class to which the nearest cluster center belongs. Then, the cluster center of each class is recalculated, and the above process is repeated until the cluster center no longer changes or the maximum number of iterations is reached.

[0018] Preferably, in step S4, standard mass spectrometry data of various known water pollutants, including organic pollutants such as polycyclic aromatic hydrocarbons, pesticides, endocrine disruptors, etc., and mass spectrometry information of inorganic pollutants, heavy metal complexes, etc., are collected to construct characteristic mass spectrometry fingerprints that are also included in the database, and a corresponding index and classification system is established, and matched with the information in the database. New fingerprints that are not matched may represent new unknown pollutants.

[0019] The present invention has the following beneficial effects:

[0020] 1. In the process of constructing characteristic mass spectrometry fingerprints, a variety of advanced data analysis algorithms are used, combining principal component analysis, cluster analysis and chemometric methods. This multi-algorithm fusion approach can more effectively mine truly representative characteristic peaks from massive mass spectrometry data, avoiding the limitations that may exist in a single algorithm and improving the accuracy and reliability of fingerprint construction.

[0021] 2. By combining solid phase extraction, acid-base adjustment and other methods, targeted enrichment and treatment are carried out according to the properties of pollutants, which improves the detection ability of low-concentration pollutants in complex water samples, so that subsequent mass spectrometry analysis can obtain pollutant information more accurately.

[0022] 3. A comprehensive database containing known pollutants and newly discovered characteristic fingerprints through this method has been established, equipped with a complete index and classification system. The continuous updating and optimization of the database, as well as the efficient matching algorithm, greatly improve the efficiency and accuracy of screening unknown pollutants. At the same time, the fingerprints of newly emerging pollutants can be included in the database in a timely manner to provide reference for subsequent research and monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a mass spectrum diagram of detecting multiple water samples in an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0025] A non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water, comprising the following screening steps;

[0026] S1. Collection and filtration. The collected water is stored in a polytetrafluoroethylene test tube and kept at a low temperature of about 4°C and protected from light. The water sample is acid-base adjusted and solid phase extraction is performed. After each extraction step, the eluent is acid-base adjusted again as needed. The acidity regulator includes hydrochloric acid HCl, nitric acid HNO 3 and sulfuric acid H2 SO 4 , hydrochloric acid HCl reacts quickly when adjusting the pH value, and alkaline regulators include sodium hydroxide NaOH and potassium hydroxide KOH, which change the pH value of water samples.

[0027] S2. Mass spectrometry analysis: The water to be tested is tested through the mass spectrometer host, vacuum pump system, data acquisition system and other related equipment. The pressure of the vacuum pump system generally needs to reach 10 -5 -10 -6 torr or so.

[0028] S3, data processing, converting the data format output by the mass spectrometer, such as the original binary file or the specific instrument software format, into a universal data format, considering the noise characteristics of the mass spectrometry data, and selecting an appropriate denoising method;

[0029] For example, for wavelet transform denoising, first select the Daubechies wavelet series and the number of decomposition layers to decompose the mass spectrometry signal. For the noise in the high-frequency sub-band, after hard threshold or soft threshold processing, the processed sub-band is reconstructed to obtain the denoised signal, making the data smoother and enabling subsequent baseline correction and peak identification. PCA and cluster analysis can be organically combined to perform PCA processing on high-dimensional mass spectrometry data containing mass-to-charge ratio and ion intensity information. After covariance matrix calculation and eigenvalue decomposition, the principal component is extracted for dimensionality reduction, denoising and redundancy removal, and the principal component score is used as a new feature for cluster analysis. K-means clustering is performed and parameters are adjusted according to data characteristics to cluster similar samples. PCA is a statistical method that converts original high-dimensional data into a set of low-dimensional unrelated principal components through linear transformation. Based on the covariance matrix of the data, the purpose is to find the direction with the largest variance as the principal component direction, thereby retaining data information while reducing the dimension.

[0030] K-means clustering, first specify the number of clusters, then randomly initialize the cluster centers, and assign the samples to the class to which the nearest cluster center belongs by calculating the distance from each sample to the cluster center, such as the Euclidean distance. Then, recalculate the cluster center of each class, and repeat the above process until the cluster center no longer changes or the maximum number of iterations is reached.

[0031] S4. Fingerprint construction: construct a mass spectrum fingerprint of the processed data with mass-to-charge ratio as the horizontal axis and relative intensity as the vertical axis. The fingerprint of each sample is stored in the database, which can intuitively reflect the overall characteristics of pollutants in water samples, including the types and relative contents of pollutants. Standard mass spectrum data of various known water pollutants will be collected, including organic pollutants such as polycyclic aromatic hydrocarbons, pesticides, endocrine disruptors, etc. The mass spectrum information of inorganic pollutants, heavy metal complexes, etc. will also be included in the database. A corresponding index and classification system will be established, and matched with the information in the database. New fingerprints that are not matched may represent new unknown pollutants.

[0032] S5. Non-targeted screening: compare the constructed mass spectral fingerprint with the known NIST mass spectral database and EPA's pollutant database, and find the parts that match the known pollutants by calculating similarity indicators such as cosine similarity and Euclidean distance.

[0033] Finally, the water to be tested is tested using the above screening method;

[0034] Embodiment 1:

[0035] Step 1: Collect 10 water samples and store them in polytetrafluoroethylene test tubes, keep them at a low temperature of about 4°C and avoid light. Adjust the acidity and base of the water samples and then perform solid phase extraction. After each extraction step, use hydrochloric acid HCl and nitric acid HNO as needed. 3 , sulfuric acid H 2 SO 4 The eluent is again acid- and base-adjusted with an acidic regulator such as hydrochloric acid (HCl) (which reacts quickly to adjust pH) or an alkaline regulator such as sodium hydroxide NaOH or potassium hydroxide KOH to change its pH value.

[0036] Step 2: Through the mass spectrometer host, the pressure must reach 10 -5 -10 -6 The vacuum pump system of the high vacuum system of about torr and the data acquisition system and other related equipment to be tested Water testing.

[0037] Step 3: Use the 10 collected water samples, each containing information on the concentration of two pollutants, to construct a 10×2 data matrix X. This simulated data has no missing values. If there are a few missing values, they can be left unprocessed. Then use the box plot method or the 3-times standard deviation principle to check for outliers. After processing, perform Z-score standardization on the data matrix X to obtain the matrix Z. For example, for pollutant 1, calculate its mean and standard deviation σ j Standardized concentration of pollutant 1 in the first water sample Similarly, we get Z. Calculate the covariance matrix of Z, where c 11is the standardized concentration variance of pollutant 1, c 12 =c 21 is the coefficient of the standardized concentration of the two pollutants 22 The difference is the variance of the standardized concentration of pollutant 2. Calculate the cumulative variance contribution rate. Here, both principal components are selected. Calculate the score of each water sample on the principal component to obtain a 10×2 principal component score matrix F. (Take the first water sample).

[0038] Using K-means clustering, the number of clusters k is determined by the elbow rule. Three 2D initial cluster centers are randomly selected in the principal component score matrix F space. The Euclidean distance of each water sample to each cluster center is calculated, and the water sample is assigned to the cluster to which the nearest cluster center belongs. At the same time, the cluster center is updated, that is, (n j is the number of water samples in cluster j, C j is a water sample set), repeat until the cluster center changes less than the set threshold y i or the maximum number of iterations is reached.

[0039] According to the clustering F result, the category vector y is constructed, and its elements represent the cluster category to which the first water sample i belongs. The discriminant model is established with the principal component score matrix F as the independent variable and the category vector y as the dependent variable. The intra-class scatter matrix and S are calculated. w The inter-class scatter matrix is ​​used to solve the discriminant vector w so that Maximum, with the principal component score matrix as the independent variable and the target pollutant concentration vector c as the dependent variable.

[0040] The data set was divided into 70% training set and 30% validation set for cross validation, and the appropriate number of principal components and model parameters were selected to establish the PLSR model to minimize the prediction error. The silhouette coefficient was used to evaluate the clustering quality. (a i is the average distance from water sample i to other water samples in its cluster, b i is the average distance to all water samples in the nearest neighbor cluster) Take the average value of the silhouette coefficient of all water samples, which is between [-1,1]. The closer it is to 1, the better the clustering effect. Calculate the mean square error of the PLSR model prediction (c i is the actual concentration, is the predicted concentration) and the coefficient of determination is the mean concentration of the target pollutant, R 2 In [0,1], the closer to 1, the better the prediction effect.

[0041] For newly collected water samples, they are first processed according to the above preprocessing steps, then the principal component scores are calculated by PCA, and the samples are input into the discriminant analysis model to predict the water sample category. At the same time, the PLSR model is input to predict the concentration of target pollutants, so as to realize the classification and quantitative analysis of pollutants in water and provide a basis for water quality monitoring and pollution control.

[0042] Step 4: Based on the processed data, the mass-to-charge ratio is set as the horizontal axis and the relative intensity as the vertical axis to construct a mass spectrum fingerprint. The fingerprint of each sample is stored in the database. At the same time, standard mass spectrum data of various water pollutants including organic pollutants (polycyclic aromatic hydrocarbons, pesticides, endocrine disruptors, etc.), inorganic pollutants, heavy metal complexes, etc. are collected. The constructed characteristic mass spectrum fingerprint is also included in the database and a corresponding index and classification system is established to reflect the overall characteristics of the pollutants in the water sample, such as the type and relative content. The newly constructed fingerprint is then matched with the information in the database. If there is a new fingerprint that is not matched, it may represent a new unknown pollutant.

[0043] Step 5: By calculating similarity indicators such as cosine similarity and Euclidean distance, the constructed mass spectral fingerprint is compared with the known NIST mass spectral database and EPA pollutant database to find the parts that match the known pollutants.

[0044] Example 2: An algorithm different from step 3 in Example 1 first collects and organizes the concentration information of two pollutants in 10 water samples, constructs a data set and performs minimum-maximum normalization processing, and divides it into training set, validation set and test set in a ratio of 7:2:1. Then, a three-layer feedforward neural network is constructed, and the input layer has 2 neurons corresponding to the two pollutant concentrations. The hidden layer can first try to set 5 neurons and use the ReLU activation function. The output layer has 2 neurons using a linear activation function, and the weights and biases in the network are randomly initialized in the interval [-0.1, 0.1].

[0045] Then select mean square error as the loss function and back propagation algorithm as the training algorithm, set the learning rate to 0.01, and the number of training rounds to 1000 to start training. During training, use the validation set to evaluate the model performance regularly. If the loss function value of the validation set increases, overfitting may occur, and measures such as stopping training early, adjusting the network structure, and regularization should be taken to optimize the model. Finally, use the test set to conduct a final test on the optimized model, and calculate indicators such as the test set loss function value and mean absolute error. If the performance is satisfactory, it can be used to predict the concentrations of the two pollutants in the water sample.

[0046] Example 3: Using a different algorithm from step 3 in Example 1, for 10 water samples (each containing information on the concentration of 2 pollutants), first collect and organize the data and construct them into a data set. Then perform preprocessing such as normalization on the data, and then divide the data into a training set and a test set (such as 7 water samples as a training set and 3 water samples as a test set). Then select a suitable kernel function (such as a Gaussian kernel function) to build a support vector machine model, train the model with the training set data, use an optimization algorithm (such as a sequential minimum optimization algorithm) to find the optimal hyperplane, adjust the model parameters (such as penalty parameters) during the training process to optimize the model performance and avoid overfitting, and finally use the test set to evaluate the model's detection effect on the two pollutant concentration information in the water sample.

[0047] The screening effect was compared by using indicators such as data processing dimension change (pieces), model interpretability (0-100 points), small sample accuracy (%), model training speed (minutes), and prediction accuracy (%). Comparative Example 1 used a conventional algorithm, and Comparative Example 2 used another algorithm of the conventional algorithm.

[0048] It should also be noted that, for the conventional algorithm in Example 1, the workflow of pollutants in water includes sampling at appropriate locations and recording relevant information, properly storing and transporting water samples, conducting preliminary treatment of water samples (such as filtration, adjusting pH), enriching target pollutants when necessary (such as liquid-liquid extraction, solid phase extraction), and then analyzing and detecting through physical property testing and various chemical analysis methods (acid-base titration, gravimetric analysis, spectrophotometry, chromatography, electrochemical analysis, etc.), and finally processing the data and evaluating the results, comparing them with water quality standards to determine the pollution status and impact, and ensuring quality control.

[0049] Another algorithm in comparative example 2 first obtains the sample to be tested from the water sample to be tested, and pre-treats the water sample to be tested, including sequentially eluting the unloaded solid phase extraction column and the adsorbed organic matter with a specific solvent, activating the solid phase extraction column with methanol and ultrapure water, blowing the eluent with nitrogen to dryness and dissolving it in an organic solvent, etc.;

[0050] Next, the sample to be tested is subjected to two-dimensional gas chromatography analysis to obtain an intermediate analyte, which is then divided into two parts. One part is subjected to mass spectrometry analysis after bombardment with an EI ion source to obtain a first spectrum, and the other part is subjected to ECD detection to obtain a second spectrum. Then, the mass spectrometry difference method is used to screen the first spectrum, and the organic chloride, organic bromide or nitro substitution contained in the sample to be tested is determined based on the specific mass-to-charge ratio signal and compared with the second spectrum (determine whether the signal-to-noise ratio, relative abundance and peak time of the signal meet the conditions). After that, the screened pollutants can be compared with the NIST spectral library to further identify the material components. Through specific examples, it is demonstrated that this method can accurately screen out pollutants such as chloroform, monobromoacetonitrile, nitrobenzene, etc. according to the above process in actual application, and can be combined with the NIST spectral library for subsequent identification:

[0051] Table 1: Comparison of various performances of 10 water samples containing 2 kinds of pollutant concentration information in various embodiments and comparative examples:

[0052]

[0053] In summary, the water pollutant screening method prepared by the present invention, compared with the traditional screening method and another screening method on the market, has changes in data processing dimensions (pieces), model interpretability (0-100 points), small sample accuracy (%), model training speed (minutes), and prediction accuracy (%). For example, Example 1 reduces the data dimension from 2 to 1-2 through principal component analysis, while Examples 2, 3 and Comparative Examples 1 and 2 all maintain the original 2 dimensions, indicating that Example 1 has unique advantages in data dimensionality reduction and can simplify the data structure.

[0054] Furthermore, the mass spectrogram is prepared by changing the amount of water sample prepared in the calculation process of each embodiment 1, as shown in the following example: Figure 1 As shown, the horizontal axis (m / z) represents the mass-to-charge ratio, and the vertical axis (I,%) represents the ion intensity (expressed as a percentage). Example 1 has a very high peak at around m / z=220, and its ion intensity is close to 100%. This is the most significant peak in the entire mass spectrum and may represent the most important component in the sample or the ion with the highest abundance.

[0055] There are multiple smaller peaks between m / z = 120-180 and 240-260. The ion intensities of these peaks are relatively low, but still detectable, indicating that the ions corresponding to these mass-to-charge ratios also have a certain abundance in the sample.

[0056] In Example 1, the water sample is subjected to mass spectrometry analysis after pretreatment steps such as collection, filtration, acid-base adjustment, and solid phase extraction. This mass spectrum may be the result obtained after a certain water sample has undergone these treatments. By analyzing the peaks in the mass spectrum, the pollutant components in the water sample can be identified. The high abundance peak at m / z=220 may correspond to a specific pollutant or its fragment ion. By comparing with the known NIST mass spectrum database and the EPA's pollutant database, the parts that match the known pollutants can be found, and then the types of pollutants in the water sample can be determined. At the same time, combined with the data processing methods in Example 1, such as principal component analysis (PCA) and cluster analysis, the pollutants in the water sample can be further classified and quantitatively analyzed, providing a basis for water quality monitoring and pollution control.

[0057] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water, characterized in that: The screening steps include: S1. Collection and filtration. The collected water is stored in a polytetrafluoroethylene test tube and kept at a low temperature of about 4°C and protected from light. The water sample is acid-base adjusted and then solid phase extraction is performed. After each extraction step, the eluent is acid-base adjusted again as needed. S2. Mass spectrometry analysis: the water to be tested is tested through mass spectrometer host, vacuum pump system, data acquisition system and other related equipment; S3. Data processing: convert the data format output by the mass spectrometer, such as the original binary file or the specific instrument software format, into a universal data format. The noise characteristics of the mass spectrometry data are analyzed, and the appropriate denoising method is selected. PCA and cluster analysis can be organically combined. PCA processing is performed on high-dimensional mass spectrometry data containing mass-to-charge ratio and ion intensity information. The principal component is extracted through covariance matrix calculation and eigenvalue decomposition to reduce the dimension, denoise and remove redundancy. The principal component score is then used as a new feature for cluster analysis. K-means clustering is performed and parameters are adjusted according to data characteristics to cluster similar samples. S4, fingerprint construction, the processed data is used to construct a mass spectrum fingerprint with mass-to-charge ratio as the horizontal axis and relative intensity as the vertical axis, and the fingerprint of each sample is stored in the database, which can intuitively reflect the overall characteristics of pollutants in water samples, including the types and relative contents of pollutants; S5. Non-targeted screening: compare the constructed mass spectral fingerprint with the known NIST mass spectral database and EPA's pollutant database, and find the parts that match the known pollutants by calculating similarity indicators such as cosine similarity and Euclidean distance.

2. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1 is characterized in that: In step S1, the acidic regulator includes hydrochloric acid HCl, nitric acid HNO3 and sulfuric acid H2SO4, and the hydrochloric acid HCl reacts quickly when adjusting the pH value; Alkaline regulators include sodium hydroxide NaOH and potassium hydroxide KOH, which change the pH value of water samples.

3. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1 is characterized in that: In step S2, the pressure of the vacuum pump system, generally a high vacuum system, needs to reach 10 -5 -10 -6 torr or so.

4. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1, characterized in that: In step S3, such as wavelet transform denoising, first select the Daubechies wavelet series and the number of decomposition layers to decompose the mass spectrometry signal. For the noise in the high-frequency sub-band, after hard threshold or soft threshold processing, the processed sub-band is reconstructed to obtain the denoised signal, making the data smoother and facilitating subsequent baseline correction and peak identification.

5. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1, characterized in that: In step S3, PCA is a statistical method that converts original high-dimensional data into a set of low-dimensional unrelated variable principal components through linear transformation. It is based on the covariance matrix of the data and aims to find the direction with the largest variance as the principal component direction, thereby retaining the data information while reducing the dimension.

6. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1, characterized in that: In step S3, K-means clustering first specifies the number of clusters, then randomly initializes the cluster centers, and assigns the samples to the class to which the nearest cluster center belongs by calculating the distance from each sample to the cluster center, such as the Euclidean distance. Then, the cluster center of each class is recalculated, and the above process is repeated until the cluster center no longer changes or the maximum number of iterations is reached.

7. The non-targeted screening method for characteristic mass spectrometric fingerprints of pollutants in water according to claim 1, characterized in that: In step S4, standard mass spectrometry data of various known water pollutants are collected, including organic pollutants such as polycyclic aromatic hydrocarbons, pesticides, endocrine disruptors, etc. The mass spectrometry information of inorganic pollutants, heavy metal complexes, etc. The constructed characteristic mass spectrometry fingerprints are also included in the database, and a corresponding index and classification system is established, and matched with the information in the database. New fingerprints that are not matched may represent new unknown pollutants.

Images