Method and system for predicting concentration dependence of nano plastic on crop toxicity

Abstract

The invention discloses a method for predicting concentration dependence of nano-plastics on crop toxicity. The method comprises the following steps: S1, obtaining a crop sample polluted by nano-plastics, and carrying out sample treatment to obtain a prediction sample; s2, carrying out scanning electron microscope, laser confocal, Raman spectrum and Fourier infrared spectrum comprehensive experiments on the prediction sample at the same time, and collecting a scanning electron microscope spectrum, a laser confocal spectrum, a Raman spectrum and a Fourier infrared spectrum; s3, performing comprehensive operation based on the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier infrared spectrum to obtain a comprehensive spectrum; and S4, predicting the concentration dependence of the nano plastic on the crop toxicity based on machine learning and targeted metabonomics of the comprehensive spectrum. The invention further discloses a corresponding system, electronic equipment and a computer readable storage medium.

Description

Technical Field

[0001] The invention relates to the field of biology and biological computational science and technology, and in particular to a method and system for predicting the concentration dependence of nanoplastic toxicity on crops. Background Art

[0002] Plastic is a versatile and cost-effective material that is used in all industries. Plastic production is increasing year by year and is expected to reach 460 million tons in 2060. However, only 7% of plastic is recycled, while most of it is incinerated, landfilled or directly discarded in the environment. Under the influence of external conditions such as solar radiation and biodegradation, plastic breaks into microplastics, which are further decomposed into nanoplastics and enter water bodies and soil. Therefore, the almost permanent pollution of the natural environment by plastic waste has received increasing attention. Nanoplastics are below 1μm in size and can pass through cell membranes, and their toxicity has attracted people's attention. Plants are an important part of the food chain, and peanuts are one of the main sources of edible oil in the world. They are very easy to absorb toxic nanoplastic particles in the soil. After the roots absorb nanoplastics, they can be enriched in roots, stems, leaves and other parts, and accumulate in fruits, thus affecting the entire food chain. In addition, nanoplastics have been found in human blood, tissues and even brains, and studies have shown that nanoplastics can increase the risk of induced diseases. Studies have shown that when peanuts were exposed to 250 mg / kg of nanoplastics, the empty shell rate increased and the fruit set rate decreased, indicating that exposure to nanoplastics affected the growth and development of peanuts. Therefore, if the early prediction of the phytotoxicity of nanoplastics can be achieved, it will help to take relevant measures to remove nanoplastics as soon as possible, thereby improving the yield and quality of peanuts. Metabolomics is a discipline that studies biological systems by examining the changes in metabolites or their changes over time after the biological system (cells, tissues or organisms) is stimulated or disturbed. It is an emerging discipline that systematically studies the content, morphology, distribution and function of elements in organisms. Exposure to micro / nano plastics may affect the changes in the metabolome in organisms, so it is feasible to use this characteristic to predict the toxicity of micro / nano plastics. Summary of the invention

[0003] The purpose of the present invention is to propose a method and system for predicting the concentration dependence of nanoplastic toxicity to crops in view of the problems existing in the prior art. The targeted metabolomics method based on scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy comprehensive experiments and deep learning technology can predict the phytotoxicity of micro / nano plastics of different sizes. Polyvinyl chloride (PVC) is the third largest synthetic polymer plastic in the world. The vinyl chloride monomer contained in it is a carcinogen that increases the risk of liver cancer. Targeted metabolomics does not need to quantify the specific content of the metabolome, but can achieve toxicity screening by comparing the differences in the metabolome. The statistical analysis algorithm of machine learning is the basis of data processing. PCA, as the main unsupervised clustering method, can reduce the dimension of high-dimensional data. The linear classification model K nearest neighbor algorithm (KNN) and the nonlinear model support vector machine (SVM) can realize sample qualitative analysis. Considering the situation that the data is highly linearly correlated, selecting characteristic spectra can reduce the number of model parameters and improve the detection speed. Peanuts exposed to different concentrations of PVC nanoplastics (nPVC) were used as the research object. A targeted metabolomics method combining scanning electron microscopy, laser confocal microscopy, Raman spectroscopy and Fourier transform infrared spectroscopy combined with machine learning was established to predict the toxicity of nPVC to peanuts. It can also provide reference for the prediction of the toxicity of other types of micro / nano plastics.

[0004] The first aspect of the present invention is to provide a method for predicting the concentration dependence of nanoplastic toxicity to crops, comprising:

[0005] S1, obtaining crop samples contaminated by nanoplastics and processing them to obtain predicted samples;

[0006] S2, performing a comprehensive experiment of scanning electron microscopy, laser confocal microscopy, Raman spectroscopy and Fourier transform infrared spectroscopy on the predicted sample at the same time and collecting scanning electron microscopy spectrum, laser confocal microscopy spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0007] S3, performing comprehensive calculation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum;

[0008] S4, Machine learning and targeted metabolomics based on the comprehensive spectra predict the concentration dependence of nanoplastic toxicity to crops.

[0009] Preferably, the S1 comprises:

[0010] S11, quantifying the crop samples contaminated by nanoplastics;

[0011] S12, grinding the crop sample into powder;

[0012] S13, pressing the powder into thin sheets to form a prediction sample.

[0013] Preferably, S2 includes:

[0014] S21, placing the predicted sample in an experimental station consisting of a synchrotron radiation device to conduct comprehensive experiments of scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy;

[0015] S22, collecting scanning electron microscope spectra, laser confocal spectra, Raman spectra and Fourier transform infrared spectra during the comprehensive test;

[0016] Preferably, S3 includes:

[0017] S31, respectively obtaining the optimal characteristic wavelengths of scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0018] S32, optimally fitting the preferred characteristic wavelengths of the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum to obtain a comprehensive spectrum corresponding to the current crop variety.

[0019] Preferably, the S31 comprises: using any one of the principal component analysis method PCA and the collaborative interval partial least squares method SiPLS to obtain the preferred characteristic wavelength formed by the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum;

[0020] The principal component analysis method PCA includes:

[0021] Each spectral feature of the SEM spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum is adjusted to be centered on zero, that is, the average value of all spectral features is subtracted from each spectral feature, thereby removing the influence of the position of the spectral feature data and retaining only the information of the shape of the spectral feature data;

[0022] Calculate the eigenvectors of the covariance matrix of the adjusted spectral features, and sort all eigenvectors according to the size of the eigenvalue. The larger the eigenvalue, the more information the corresponding eigenvector can capture about the crop toxicity, i.e., the variability.

[0023] Select the first ten largest eigenvalues ​​and their corresponding eigenvectors as principal components, and define a new coordinate system based on the principal components, with the axes of the coordinate system pointing to the direction of the largest data change;

[0024] Re-projecting all spectral data included in the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum into a new coordinate system, selecting and intercepting the part with the largest spectral density as the preferred characteristic wavelength;

[0025] The collaborative interval partial least squares method SiPLS includes:

[0026] All spectral data of SEM, confocal laser, Raman and Fourier transform infrared spectra are used to generate a data set, which is divided into multiple small intervals, and the data in each interval has different importance;

[0027] Based on the collaborative interval partial least squares method, the influence of the spectral data in each interval on the crop toxicity concentration dependence prediction output result is evaluated and a value is assigned, wherein the assignment has multiple values;

[0028] Determine the top ten intervals with the greatest impact according to multiple assignments and perform mixed splicing, wherein the mixed splicing is composed of multiple splicing modes according to the arrangement and combination mode;

[0029] The spectral data intervals formed by multiple splicing methods are used to adjust the weight of each spectral data in the spectral data interval as a parameter adjustment process, and partial least squares regression analysis is performed. The optimal splicing method is determined based on the analysis results, which corresponds to the preferred characteristic wavelength.

[0030] Preferably, the S32 is performed based on a best fit competitive adaptive resampling algorithm, including:

[0031] Randomly selecting a preferred characteristic wavelength, generating a comprehensive spectrum, and evaluating the generation effect, wherein the generation effect is determined based on the degree of overlap between the generated comprehensive spectrum and historical data of typical spectra of the same crop;

[0032] After performing the above evaluations in sequence, the preferred characteristic wavelengths that fail the evaluations are eliminated;

[0033] The weights of the preferred characteristic wavelengths obtained through the evaluation are adjusted and merged to obtain the comprehensive spectrum.

[0034] Preferably, S4 includes:

[0035] S41, intuitively obtaining differences in metabolic elements in the crop plants by comparing the spectral peak areas of the comprehensive spectrum;

[0036] S42, based on the unsupervised clustering of the comprehensive spectrum, reducing the spectral data dimension in the comprehensive spectrum, converting multiple indicators into a few comprehensive indicators, and plotting a coordinate graph with the differences in metabolic elements in the crop plants corresponding to the few comprehensive indicators and the targeted metabolomics as the ordinate and abscissa respectively; the coordinate graph is used to predict the effects of different concentrations of nanoplastics on crop toxicity;

[0037] S43, a prediction model was established based on the linear model k-nearest neighbor algorithm and the nonlinear model support vector machine to distinguish and predict the concentration dependence of different concentrations of nanoplastics on the toxicity of different groups of crops;

[0038] S44, based on the competitive adaptive weighted sampling algorithm, the characteristic spectra were selected to establish a prediction model to distinguish and predict the concentration dependence of the toxicity of different concentrations of nanoplastics on different groups of crops;

[0039] S45, based on the prediction results of S42-S44, the concentration dependence of nanoplastic toxicity on crops is comprehensively obtained.

[0040] The second aspect of the present invention is to provide a concentration-dependency prediction system for the toxicity of nanoplastics to crops, comprising:

[0041] A sample acquisition module (101) is used to acquire crop samples contaminated by nanoplastics and process the samples to obtain predicted samples;

[0042] A spectrum experiment and acquisition module (102) is used to simultaneously perform a comprehensive experiment of scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy on the predicted sample and to collect the scanning electron microscopy spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0043] A comprehensive spectrum operation module (103) is used to perform comprehensive operation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum;

[0044] A dependency prediction module (104) is used to predict the concentration dependency of nanoplastic toxicity to crops based on machine learning and targeted metabolomics of the comprehensive spectrum.

[0045] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is configured to read the instructions and execute the method described in the first aspect.

[0046] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor and execute the method described in the first aspect.

[0047] Beneficial effects of the method and system of the present invention:

[0048] Micro / nano plastics in soil inevitably affect plants, so the prediction of phytotoxicity of micro / nano plastics in the present invention can provide a handle for the management of micro / nano plastics in soil. Peanuts were used as the research object, and a targeted metabolomics method based on comprehensive spectroscopy and machine learning was developed to predict the toxicity of polyvinyl chloride nanoplastics (nPVC) to peanuts. First, peanuts were exposed to different concentrations (2500ng / g and 2500μg / g) of nPVC, and peanut leaves were collected after 35 days of cultivation; secondly, the changes in the metabolome in peanut leaves after exposure to nPVC were studied using comprehensive spectroscopy; then, machine learning methods were used to distinguish peanut samples exposed to different concentrations of nPVC. The integrated spectra formed by scanning electron microscopy, laser confocal spectroscopy, Raman spectroscopy and Fourier transform infrared spectroscopy were subjected to principal component analysis (PCA) unsupervised clustering, and it was found that the 2500μg / g group could be well clustered, while there was no significant difference between the 2500ng / g group and the control group, indicating that the toxicity of 2500ng / g nPVC exposure to plants is much lower than that of 2500μg / gnPVC. For the comprehensive spectrum, the linear model k-nearest neighbor algorithm (kNN) and the nonlinear model support vector machine (SVM) were used to establish a prediction model, and the accuracy of distinguishing different groups can reach 94.12%. In order to improve the operation speed and reduce the amount of model calculation, the competitive adaptive weighted sampling algorithm (CARS) was used to select characteristic spectra to establish a prediction model, and the accuracy of distinguishing different groups was 89.51%. Compared with the full spectrum model, although the prediction accuracy of the characteristic spectrum prediction model decreased by 4.61%, the model input parameters were reduced by 99.38%, so it also has good potential. Studies have shown that non-target metabolomics based on comprehensive spectra and machine learning can accurately predict the degree of interference of different concentrations of nPVC on the peanut metabolome, thereby reflecting the concentration dependence of nPVC toxicity on peanuts. The method can also be used to predict the concentration dependence of the toxicity of other micro / nano plastics. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 A flow chart of a method for predicting the concentration dependence of nanoplastic toxicity to crops provided according to an embodiment of the present invention;

[0051] Figure 2 This is a schematic flow chart of step S1 of a method for predicting the concentration dependence of nanoplastic toxicity to crops provided in an embodiment of the present invention;

[0052] Figure 3 This is a schematic flow chart of step S2 of a method for predicting the concentration dependence of nanoplastic toxicity to crops provided in an embodiment of the present invention;

[0053] Figure 4 This is a schematic flow chart of step S3 of a method for predicting the concentration dependence of nanoplastic toxicity to crops provided in an embodiment of the present invention;

[0054] Figure 5 This is a schematic flow chart of step S4 of a method for predicting the concentration dependence of nanoplastic toxicity to crops provided in an embodiment of the present invention;

[0055] Figure 6 A schematic diagram of the architecture of a system for predicting the concentration dependence of nanoplastic toxicity to crops provided according to an embodiment of the present invention;

[0056] Figure 7 A structural diagram of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0059] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] Embodiment 1

[0061] like Figure 1As shown, this embodiment provides a method for predicting the concentration dependence of nanoplastic toxicity to crops, comprising:

[0062] S1, obtaining crop samples contaminated by nanoplastics and processing them to obtain predicted samples;

[0063] like Figure 2 As shown, as a preferred embodiment, the S1 includes:

[0064] S11, quantifying the crop samples contaminated by nanoplastics;

[0065] S12, grinding the crop sample into powder;

[0066] S13, pressing the powder into thin sheets to form a prediction sample.

[0067] S2, performing a comprehensive experiment of scanning electron microscopy, laser confocal microscopy, Raman spectroscopy and Fourier transform infrared spectroscopy on the predicted sample at the same time and collecting scanning electron microscopy spectrum, laser confocal microscopy spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0068] like Figure 3 As shown, as a preferred embodiment, the S2 includes:

[0069] S21, placing the predicted sample in an experimental station consisting of a synchrotron radiation device to conduct comprehensive experiments of scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy;

[0070] S22, collecting scanning electron microscope spectra, laser confocal spectra, Raman spectra and Fourier transform infrared spectra during the comprehensive test;

[0071] S3, performing comprehensive calculation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum.

[0072] like Figure 4 As shown, as a preferred implementation, S3 includes:

[0073] S31, respectively obtaining the optimal characteristic wavelengths of scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0074] S32, optimally fitting the preferred characteristic wavelengths of the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum to obtain a comprehensive spectrum corresponding to the current crop variety.

[0075] As a preferred embodiment, S31 includes: using any one of the principal component analysis method PCA and the collaborative interval partial least squares method SiPLS to obtain the preferred characteristic wavelength formed by the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum.

[0076] As a preferred embodiment, the principal component analysis method PCA includes:

[0077] Each spectral feature of the SEM spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum is adjusted to be centered on zero, that is, the average value of all spectral features is subtracted from each spectral feature, thereby removing the influence of the position of the spectral feature data and retaining only the information of the shape of the spectral feature data;

[0078] Calculate the eigenvectors of the covariance matrix of the adjusted spectral features, and sort all eigenvectors according to the size of the eigenvalue. The larger the eigenvalue, the more information the corresponding eigenvector can capture about the crop toxicity, i.e., the variability.

[0079] Select the first ten largest eigenvalues ​​and their corresponding eigenvectors as principal components, and define a new coordinate system based on the principal components, with the axes of the coordinate system pointing to the direction of the largest data change;

[0080] All spectral data contained in the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum are re-projected into a new coordinate system, and the part with the largest spectral density is selected and intercepted as the preferred characteristic wavelength.

[0081] As a preferred embodiment, the collaborative interval partial least squares method SiPLS includes:

[0082] All spectral data of SEM, confocal laser, Raman and Fourier transform infrared spectra are used to generate a data set, which is divided into multiple small intervals, and the data in each interval has different importance;

[0083] Based on the collaborative interval partial least squares method, the influence of the spectral data in each interval on the crop toxicity concentration dependence prediction output result is evaluated and a value is assigned, wherein the assignment has multiple values;

[0084] Determine the top ten intervals with the greatest impact according to multiple assignments and perform mixed splicing, wherein the mixed splicing is composed of multiple splicing modes according to the arrangement and combination mode;

[0085] The spectral data intervals formed by multiple splicing methods are used to adjust the weight of each spectral data in the spectral data interval as a parameter adjustment process, and partial least squares regression analysis is performed. The optimal splicing method is determined based on the analysis results, which corresponds to the preferred characteristic wavelength.

[0086] As a preferred implementation, the S32 is performed based on a best fit competitive adaptive resampling algorithm, including:

[0087] Randomly selecting a preferred characteristic wavelength, generating a comprehensive spectrum, and evaluating the generation effect, wherein the generation effect is determined based on the degree of overlap between the generated comprehensive spectrum and historical data of typical spectra of the same crop;

[0088] After performing the above evaluations in sequence, the preferred characteristic wavelengths that fail the evaluations are eliminated;

[0089] The weights of the preferred characteristic wavelengths obtained through the evaluation are adjusted and merged to obtain the comprehensive spectrum.

[0090] S4, Machine learning and targeted metabolomics based on the comprehensive spectra predict the concentration dependence of nanoplastic toxicity to crops.

[0091] like Figure 5 As shown, as a preferred embodiment, the S4 includes:

[0092] S41, intuitively obtaining differences in metabolic elements in the crop plants by comparing the spectral peak areas of the comprehensive spectrum;

[0093] S42, based on the unsupervised clustering of the comprehensive spectrum, reducing the spectral data dimension in the comprehensive spectrum, converting multiple indicators into a few comprehensive indicators, and plotting a coordinate graph with the differences in metabolic elements in the crop plants corresponding to the few comprehensive indicators and the targeted metabolomics as the ordinate and abscissa respectively; the coordinate graph is used to predict the effects of different concentrations of nanoplastics on crop toxicity;

[0094] S43, a prediction model was established based on the linear model k-nearest neighbor algorithm (kNN) and the nonlinear model support vector machine (SVM) to distinguish and predict the concentration dependence of different concentrations of nanoplastics on the toxicity of different groups of crops;

[0095] S44, based on the competitive adaptive weighted sampling algorithm (CARS), a prediction model was established by selecting characteristic spectra to distinguish and predict the concentration dependence of the toxicity of different concentrations of nanoplastics on different groups of crops;

[0096] S45, based on the prediction results of S42-S44, the concentration dependence of nanoplastic toxicity on crops is comprehensively obtained.

[0097] Embodiment 2

[0098] like Figure 6 As shown, this embodiment provides a concentration-dependency prediction system for nanoplastic toxicity to crops, comprising:

[0099] The sample acquisition module 101 is used to acquire crop samples contaminated by nanoplastics and process the samples to obtain predicted samples;

[0100] The spectrum experiment and acquisition module 102 is used to simultaneously perform scanning electron microscope, laser confocal, Raman spectrum and Fourier infrared spectrum comprehensive experiments on the predicted samples and to collect scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier infrared spectrum;

[0101] A comprehensive spectrum operation module 103 is used to perform comprehensive operation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum;

[0102] The dependency prediction module 104 is used to predict the concentration dependence of nanoplastic toxicity on crops based on machine learning and targeted metabolomics of the comprehensive spectrum.

[0103] Specific test example:

[0104] 1 Experimental methods

[0105] 1.1 Sample preparation

[0106] The peanut seeds of Jihua 54 variety were selected as experimental objects. PVC micron plastics were purchased from Shanghai Aladdin. The PVC micron plastics were placed in a planetary ball mill for ball milling, cleaned, and dried to obtain nano-scale PVC plastics. The particle size distribution range was (1594±150) nm, and the concentrations of nPVC aqueous solutions of 0, 2500 ng / g, and 2500 μg / g were prepared (control group, 500 ng / g group, and 500 μg / g group, respectively). Culture conditions: daytime temperature 28℃, light 300-350 lux, culture time 10h; night temperature 15℃, light intensity 0 lux, culture 8h. 250 peanut seeds were placed in each culture dish in triplicate. Before seed germination, the nPVC aqueous solution was maintained at 5 mL to facilitate seed germination. After germination, nutrient solution was added to continue culture for 35 days. The leaves of peanut seedlings were collected, freeze-dried, and ground into powder.

[0107] 1.2 Data Acquisition

[0108] 100 mg of peanut leaf powder was weighed and pressed into 6 mm thin slices. Scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy comprehensive experiments were carried out at the 6W2B experimental station of the synchrotron radiation facility (BSRF). After collecting scanning electron microscopy spectra, laser confocal spectra, Raman spectra and Fourier transform infrared spectra, comprehensive calculations were performed to obtain a comprehensive spectrum.

[0109] The method for obtaining the comprehensive spectrum is as follows:

[0110] (1) respectively obtaining the optimal characteristic wavelengths of scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum;

[0111] In this embodiment, the preferred characteristic wavelength formed by the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum is obtained by any one of the two methods, including:

[0112] 1. Principal Component Analysis (PCA)

[0113] 1. Adjust each spectral feature of the SEM spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to be centered on zero, that is, subtract the average value of all spectral features from each spectral feature, thereby removing the influence of the spectral feature data position and retaining only the information of the spectral feature data shape;

[0114] 2. Calculate the eigenvectors of the covariance matrix of the adjusted spectral features, and sort all eigenvectors according to the size of the eigenvalue. The larger the eigenvalue, the more information the corresponding eigenvector can capture about crop toxicity, i.e., variability.

[0115] 3. Select the first ten largest eigenvalues ​​and their corresponding eigenvectors as principal components, and define a new coordinate system based on the principal components, with the coordinate axes pointing to the direction where the data changes the most;

[0116] 4. Reproject all spectral data contained in the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum into a new coordinate system, select the part with the largest spectral density and intercept it as the preferred characteristic wavelength.

[0117] Code part:

[0118]

[0119]

[0120]

[0121] (II) Synergistic Interval Partial Least Squares (SiPLS)

[0122] 1. Generate a data set from all the spectral data of SEM, confocal laser, Raman and Fourier transform infrared spectra, and divide the data set into multiple small intervals, with different importance assigned to the data in each interval;

[0123] 2. Based on the collaborative interval partial least squares method, the influence of the spectral data in each interval on the crop toxicity concentration dependence prediction output result is evaluated and assigned a value, wherein the assignment has multiple values;

[0124] 3. Determine the top ten intervals with the greatest impact according to multiple assignments and perform mixed splicing, wherein the mixed splicing is composed of multiple splicing methods according to the arrangement and combination method;

[0125] 4. According to the spectral data interval formed by multiple splicing methods, the weight of each spectral data in the spectral data interval is adjusted as a parameter adjustment process, and partial least squares regression analysis is performed. Based on the analysis results, the optimal splicing method is determined, which corresponds to the preferred characteristic wavelength.

[0126] Code section

[0127]

[0128]

[0129]

[0130] (2) The optimal characteristic wavelengths of the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum are optimally fitted to obtain the comprehensive spectrum corresponding to the current crop variety;

[0131] The best fit is based on a competitive adaptive resampling algorithm, including:

[0132] Randomly selecting a preferred characteristic wavelength, generating a comprehensive spectrum, and evaluating the generation effect, wherein the generation effect is determined based on the degree of overlap between the generated comprehensive spectrum and historical data of typical spectra of the same crop;

[0133] After performing the above evaluations in sequence, the preferred characteristic wavelengths that fail the evaluations are eliminated;

[0134] The weights of the preferred characteristic wavelengths obtained through the evaluation are adjusted and merged to obtain the comprehensive spectrum.

[0135] Code section

[0136] importnumpyasnp

[0137] importmatplotlib.pyplotasplt

[0138] fromsklearn.feature_selectionimportSelectKBest,f_regression

[0139] #Assume that X is the simulated spectral data and y is the target variable of the corresponding simulation

[0140] np.random.seed(0)

[0141] X = np.random.rand(100,200) #100 samples, 200 wavelength points

[0142] y = np.random.rand(100) # target value of 100 samples

[0143] #Use SelectKBest for feature selection, here we use the f_regression method

[0144] #f_regression is a feature selection function for regression tasks, it is not the CARS algorithm

[0145] selector=SelectKBest(score_func=f_regression,k=10)

[0146] X_selected=selector.fit_transform(X,y)

[0147] #Draw the original data and the data after feature selection

[0148] plt.figure(figsize=(14,6))

[0149] #Draw the original data

[0150] plt.subplot(1,2,1)

[0151] plt.imshow(XT,aspect='auto')

[0152] plt.colorbar()

[0153] plt.title('OriginalSpectraData')

[0154] ————————————————

[0155] #Draw the data after feature selection

[0156] plt.subplot(1,2,2)

[0157] selected_features=selector.get_support(indices=True)

[0158] selected_spectra=X[:,selected_features]

[0159] plt.imshow(selected_spectra.T,aspect='auto')

[0160] plt.colorbar()

[0161] plt.title('SelectedFeatures')

[0162] plt.show()

[0163] In this embodiment, the operating energy of the synchrotron radiation facility (BSRF) is 12.5 GeV, and the beam intensity is 750-1250 mA. A polychromatic beam with an energy of 50-90 keV is used as the incident X-ray. The silicon drift detector detects the fluorescence signal, and the spot size of the collection beam is 250 μm × 250 μm. The spectrum of the sample can be obtained in one run. In this embodiment, 900 comprehensive spectra are obtained for each group of samples, and a total of 540 spectral data are obtained for the three groups of samples. All samples are divided into training sets and validation sets in a ratio of 7:3.

[0164] The data analysis method preferably used in this embodiment is to preliminarily use PCA to analyze the acquired SRXRF spectrum. The first three principal component variables are used as the input of the classifier and compared with other model results. The linear model KNN and the nonlinear model SVM were tested and compared to distinguish the control group, the 500ng / g group and the 500μg / g group. The optimal parameters of the model are determined by a grid optimization algorithm. The model is evaluated by comparing the accuracy (Accuracy) of the training set, the cross-validation set and the validation set. The calculation method is shown in formula (1): Accuracy = N1 / N 2 ×100% (1);

[0165] Among them, N 1 and N 2 are the number of correct predictions and the total number of samples, respectively. The confusion matrix and receiver operating characteristic curve (ROC) were calculated to further evaluate the model effect. The closer the area (AUC) value between the ROC curve and the coordinate axis is to 1, the better the effect. Selecting characteristic spectra that affect the classification results of the model from the high-dimensional linearly correlated SRXRF spectra can reduce the calculation time and optimize the model. The competitive adaptive reweighted sampling algorithm (CARS) is used to eliminate redundant channels, thereby establishing a characteristic spectrum optimization model. Data processing and model development are carried out based on the PyCharm environment and the Scikit-learn algorithm library in the Window system.

[0166] 2 Results and analysis

[0167] 2.1 SRXRF spectrum comparison analysis of different groups shows the original spectra and mean spectra of the control group, 2500ng / g group, and 2500μg / g group with channel numbers of 1000-5000. The spectral trends of the three groups of samples are basically the same, and the peak positions are consistent. From the mean spectrum, it can be seen that in the channel number of 1000-3500, with the increase of nPVC concentration, the spectral intensity is higher, and in the channel number of 3500-5000, the higher the concentration, the lower the spectral intensity. Considering that the quality of peanut tablets is the same, by comparing the spectral peak area, the difference in metabolic elements in peanut plants can be intuitively obtained. As can be seen from the figure, the element contents corresponding to the peaks 1333, 1367, 1589, and 1641 in the 2500μg / g group are significantly higher than those in the 2500ng / g group and the control group. After fitting, they correspond to potassium (K), calcium (Ca), manganese (Mn) and iron (Fe) elements. Although K only accounts for 7.5% of the total elements in peanuts, it affects peanut photosynthesis, nutrient absorption and transport, and seed development and maturity. Ca affects the integrity of the cell wall, enzyme activity, and ion balance, and in severe cases affects the degree of peanut fruiting. Mn promotes the changes in peanut enzymes, metabolism, and the synthesis of nutrients. Fe deficiency causes yellowing and bleaching of leaves, affects photosynthesis, and causes peanuts to die slowly. Therefore, the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum, and Fourier transform infrared spectrum of peanut leaves were analyzed, indicating that nPVC exposure can interfere with the steady-state balance of the metabolome in peanut leaves, and high-concentration nPVC exposure can directly affect the yield of peanuts, showing its toxicity to peanuts.

[0168] 2.2 PCA cluster analysis

[0169] PCA is an unsupervised clustering method that can reduce data dimensions and transform multiple indicators into a few comprehensive indicators. The first three variables accounted for 49.42%, 23.80% and 18.04% of the total variables, respectively, with a cumulative contribution rate of 91.25%, which can express most of the information in the original spectrum. The three-dimensional score map composed of PC1-PC3, in which green, yellow and purple represent the control group, 2500ng / g group and 2500μg / g group, respectively. These three categories are arranged roughly from the lower left corner to the upper right corner, with the control group at the lower left and the 2500μg / g group at the upper right. Among them, the control group and the 2500ng / g group have some data overlap, while the 2500μg / g group is independently clustered together. This may be due to the lower concentration of nPVC in the 2500ng / g group, which has less interference with peanut growth. The nPVC in the 2500μg / g group interfered with the steady-state balance of the metabolome in peanut plants, indicating that the toxicity of nPVC to peanuts is concentration-dependent.

[0170] 2.3 The comprehensive spectrum formed by scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum was used to predict the toxicity of nPVC to peanuts. The linear model KNN and the nonlinear model SVM were used to establish a full spectrum classification model to distinguish samples from different groups. In order to correspond to the aforementioned principal components, the first three principal components were used to input the model for comparison. Using 4000 channels as input, the classification effect of KNN reached the best, and the accuracy of the model reached 99.12%. Using the first three principal components as input, the accuracy of the model was slightly lower than that of the full spectrum model, and the accuracy of the best model was 97.21%. This situation may be caused by the loss of variables that affect the classification in the first three principal components. Although the accuracy of the model has decreased, the first three comprehensive principal components have improved the calculation speed of the model.

[0171] 2.4 Prediction of nPVC toxicity to peanuts using characteristic spectra of comprehensive spectra

[0172] In order to improve the operation speed and reduce the amount of model calculation, in the present invention, the CARS algorithm is further selected to calculate and select characteristic wavelengths, and the selected characteristic wavelengths are 332, 452, 591, 640 and 806. After peak area fitting, they correspond to K, Ti, Mn, Fe and Cu respectively. Among them, K, Mn and Fe have shown differences in the spectral curve, and Ti can promote peanut growth, increase yield and increase cell water content. Although Cu does not exist in peanut seeds, its roots, stems and leaves are richly distributed, participating in the activation and catalysis of enzymes in peanuts, affecting leaf yellowing and respiration. The selected characteristic spectrum is used as an input parameter to build a characteristic spectrum classification model. The optimal accuracy of the characteristic spectrum model is 92.51%. Compared with the full spectrum model, the accuracy of the characteristic spectrum model has decreased, which is reduced by about 4.61%. However, the model input parameter is reduced from 4000 to 25, a decrease of 99.38%, which is of great significance for achieving rapid detection.

[0173] Table 2 Accuracy of machine learning model based on characteristic spectrum

[0174] The confusion matrix and ROC curve of the characteristic spectrum CARS-SVM model were calculated. In the confusion matrix, the 2500μg / g group can be accurately predicted, while the control group and the 2500ng / g group are misclassified. 60 samples of the control group are misclassified as the 2500ng / g group, and 25 samples of the 2500ng / g group are misclassified as the control group. In the ROC curve, the AUC values ​​of the three groups are 0.93, 0.93, and 1.00, respectively, which is consistent with the PCA unsupervised classification.

[0175] 3 Conclusion

[0176] For the peanut plants in the control group, 2500ng / g group and 2500μg / g group, the comprehensive spectrum of peanut leaves was acquired, and the three groups of samples were discriminated and classified by combining the machine learning model. First, the original spectral curve was analyzed, and it was found that the peanut leaves exposed to nPVC had significant differences in K, Ca, Mn and Fe. PCA analysis of the comprehensive spectrum showed that the 2500μg / g group was significantly different from the control group and the 2500ng / g group. A full spectrum prediction model was established, and the optimal accuracy reached 98.12%. Combined with the CARS algorithm, the characteristic elements of K, Ti, Mn, Fe and Cu were selected, and the characteristic spectrum model was established, and the model accuracy reached 92.51%. The AUC of the control group, 500ng / g group and 2500μg / g group reached 0.93, 0.93 and 1.00, respectively. Although the accuracy of this method is lower than that of the full spectrum model, the number of model input parameters has decreased by 99.38%, so that rapid detection can be achieved. This study shows that the targeted metabolomics based on comprehensive spectroscopy and machine learning formed by scanning electron microscopy, laser confocal spectroscopy, Raman spectroscopy and Fourier transform infrared spectroscopy can accurately predict the degree of interference of different concentrations of nPVC on the peanut metabolome, thereby reflecting the concentration dependence of nPVC toxicity on peanuts. This method can also be used to predict the concentration dependence of the toxicity of other micro / nano plastics.

[0177] The present invention also provides a memory storing a plurality of instructions, wherein the instructions are used to implement the method as in the first embodiment.

[0178] like Figure 7 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301, the memory 302 stores a plurality of instructions, and the instructions can be loaded and executed by the processor so that the processor can execute the method as in the first embodiment.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the concentration dependence of nanoplastic toxicity to crops, characterized in that: include: S1, obtaining crop samples contaminated by nanoplastics and processing them to obtain predicted samples; S2, performing a comprehensive experiment of scanning electron microscopy, laser confocal microscopy, Raman spectroscopy and Fourier transform infrared spectroscopy on the predicted sample at the same time and collecting scanning electron microscopy spectrum, laser confocal microscopy spectrum, Raman spectrum and Fourier transform infrared spectrum; S3, performing comprehensive calculation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum; S4, Machine learning and targeted metabolomics based on the comprehensive spectra predict the concentration dependence of nanoplastic toxicity to crops.

2. The method for predicting the concentration dependence of the toxicity of nanoplastics to crops according to claim 1, characterized in that: The S1 includes: S11, quantifying the crop samples contaminated by nanoplastics; S12, grinding the crop sample into powder; S13, pressing the powder into thin sheets to form a prediction sample.

3. The method for predicting the concentration dependence of the toxicity of nanoplastics to crops according to claim 2, characterized in that: The S2 includes: S21, placing the predicted sample in an experimental station composed of a synchrotron radiation device to conduct comprehensive experiments of scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy; S22, collecting scanning electron microscope spectra, laser confocal spectra, Raman spectra and Fourier transform infrared spectra during the comprehensive test process.

4. The method for predicting the concentration dependence of the toxicity of nanoplastics to crops according to claim 3, characterized in that: The S3 includes: S31, respectively obtaining the optimal characteristic wavelengths of scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum; S32, optimally fitting the preferred characteristic wavelengths of the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum to obtain a comprehensive spectrum corresponding to the current crop variety.

5. The method for predicting the concentration dependence of the toxicity of nanoplastics to crops according to claim 4, characterized in that: The S31 includes: using any one of the principal component analysis method PCA and the collaborative interval partial least squares method SiPLS to obtain the preferred characteristic wavelength formed by the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum; The principal component analysis method PCA includes: Each spectral feature of the SEM spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum is adjusted to be centered on zero, that is, the average value of all spectral features is subtracted from each spectral feature, thereby removing the influence of the position of the spectral feature data and retaining only the information of the shape of the spectral feature data; Calculate the eigenvectors of the covariance matrix of the adjusted spectral features, and sort all eigenvectors according to the size of the eigenvalue. The larger the eigenvalue, the more information the corresponding eigenvector can capture about the crop toxicity, i.e., the variability. Select the first ten largest eigenvalues ​​and their corresponding eigenvectors as principal components, and define a new coordinate system based on the principal components, with the axes of the coordinate system pointing to the direction of the largest data change; Re-projecting all spectral data included in the scanning electron microscope spectrum, the laser confocal spectrum, the Raman spectrum and the Fourier transform infrared spectrum into a new coordinate system, selecting and intercepting the part with the largest spectral density as the preferred characteristic wavelength; The collaborative interval partial least squares method SiPLS includes: All spectral data of SEM, confocal laser, Raman and Fourier transform infrared spectra are used to generate a data set, which is divided into multiple small intervals, and the data in each interval has different importance; Based on the collaborative interval partial least squares method, the influence of the spectral data in each interval on the crop toxicity concentration dependence prediction output result is evaluated and a value is assigned, wherein the assignment has multiple values; Determine the top ten intervals with the greatest impact according to multiple assignments and perform mixed splicing, wherein the mixed splicing is composed of multiple splicing modes according to the arrangement and combination mode; The spectral data intervals formed by multiple splicing methods are used to adjust the weight of each spectral data in the spectral data interval as a parameter adjustment process, and partial least squares regression analysis is performed. The optimal splicing method is determined based on the analysis results, and the corresponding one is the preferred characteristic wavelength.

6. The method for predicting the concentration dependence of nanoplastic toxicity to crops according to claim 5, characterized in that: The S32 is performed based on the optimal fit based competitive adaptive resampling algorithm, including: Randomly selecting a preferred characteristic wavelength, generating a comprehensive spectrum, and evaluating the generation effect, wherein the generation effect is determined based on the degree of overlap between the generated comprehensive spectrum and historical data of typical spectra of the same crop; After performing the above evaluations in sequence, the preferred characteristic wavelengths that fail the evaluations are eliminated; The weights of the preferred characteristic wavelengths obtained through the evaluation are adjusted and merged to obtain the comprehensive spectrum.

7. The method for predicting the concentration dependence of the toxicity of nanoplastics to crops according to claim 6, characterized in that: The S4 includes: S41, intuitively obtaining differences in metabolic elements in the crop plants by comparing the spectral peak areas of the comprehensive spectrum; S42, based on the unsupervised clustering of the comprehensive spectrum, reducing the spectral data dimension in the comprehensive spectrum, converting multiple indicators into a few comprehensive indicators, and plotting a coordinate graph with the differences in metabolic elements in the crop plants corresponding to the few comprehensive indicators and the targeted metabolomics as the ordinate and abscissa respectively; the coordinate graph is used to predict the effects of different concentrations of nanoplastics on crop toxicity; S43, a prediction model was established based on the linear model k-nearest neighbor algorithm and the nonlinear model support vector machine to distinguish and predict the concentration dependence of different concentrations of nanoplastics on the toxicity of different groups of crops; S44, based on the competitive adaptive weighted sampling algorithm, the characteristic spectra were selected to establish a prediction model to distinguish and predict the concentration dependence of the toxicity of different concentrations of nanoplastics on different groups of crops; S45, based on the prediction results of S42-S44, the concentration dependence of nanoplastic toxicity on crops is comprehensively obtained.

8. A system for predicting the concentration dependence of the toxicity of nanoplastics on crops, used for implementing the method according to any one of claims 1 to 7, characterized in that: include: A sample acquisition module (101) is used to acquire crop samples contaminated by nanoplastics and process the samples to obtain predicted samples; A spectrum experiment and acquisition module (102) is used to simultaneously perform scanning electron microscopy, laser confocal, Raman spectroscopy and Fourier transform infrared spectroscopy comprehensive experiments on the predicted samples and to collect scanning electron microscopy, laser confocal, Raman and Fourier transform infrared spectra; A comprehensive spectrum operation module (103) is used to perform comprehensive operation based on the scanning electron microscope spectrum, laser confocal spectrum, Raman spectrum and Fourier transform infrared spectrum to obtain a comprehensive spectrum; A dependency prediction module (104) is used to predict the concentration dependency of nanoplastic toxicity to crops based on machine learning and targeted metabolomics of the comprehensive spectrum.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is used to read the instructions and execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor to execute the method according to any one of claims 1 to 7.

Images