Amino acid infrared spectrum classification and identification method and visualization system

Abstract

The invention provides an amino acid infrared spectrum classification and identification method which comprises the following steps: preparing amino acid tablets by using amino acid powder and a solid dispersant; placing the amino acid tablets into a Fourier infrared spectrometer, and collecting amino acid transmission spectrum data; based on the amino acid transmission spectrum data, using a classification and regression tree method and a support vector machine method to obtain a trained algorithm model for identifying the types of amino acid functional groups; based on the amino acid transmission spectrum data, using a classification and regression tree method and a support vector machine method to obtain a trained algorithm model for identifying amino acid types; and inputting to-be-determined amino acid infrared spectrum data into the algorithm models for identifying the amino acid functional groups and the amino acid types respectively to obtain a classification result of the amino acid functional groups and an amino acid classification result. According to the scheme, the vibration difference of amino acid molecular structures can be reflected on the basis of infrared signals, the method is high in recognition accuracy, a chiral reagent does not need to be combined with chiral molecules in a reaction mode, and the processing efficiency is improved on the premise that the recognition accuracy is improved.

Description

Technical Field

[0037] The present invention relates to the technical field of amino acid recognition, and in particular to an amino acid infrared spectrum classification and recognition method and a visualization system. Background Art

[0039] Amino acids are the basic units that make up proteins and play a vital role in the physiology and metabolism of organisms. The type, composition, sequence and three-dimensional structure of amino acids directly affect the function and characteristics of proteins. It is the key to understanding the structure and function of proteins, and helps to study the three-dimensional folding of proteins and their physiological effects. Secondly, amino acids are the basic units in the gene expression process, and understanding their characteristics helps to analyze gene function and protein synthesis. In addition, amino acids play an important role in protein-protein interactions and affect the molecular binding mechanism. The identification of amino acids is crucial for disease diagnosis and treatment, and helps to understand the mechanism of metabolic disorders. In drug development, amino acid sequence information guides drug design and identifies binding sites for targeted therapy. Therefore, accurate identification and determination of the types and quantities of amino acids are of great significance for understanding life processes, disease mechanisms, and drug development.

[0040] Traditional identification methods are based on the different physical and chemical properties of amino acids, including chromatography, spectroscopy and nuclear magnetic resonance spectroscopy. However, these identification methods are often complicated to operate, time-consuming, destructive and costly. In recent years, there have been sensing methods used to identify amino acids, mainly including spectroscopic sensing methods, quartz crystal balance sensing methods and electrochemical sensing methods. The quartz crystal balance sensing method can obtain recognition results in real time, but the detection instrument is expensive and has very high requirements for the experimental environment and experimental samples. Electrochemical sensors have certain advantages in portable real-time detection, but this method is irreversible for sample changes and is not conducive to the reuse of samples. Although the spectroscopic sensing method has significantly improved the recognition results, the development cost is high. Therefore, it is urgent to establish a low-cost, accurate and rapid detection method to identify amino acids. Summary of the invention

[0042] In view of this, an embodiment of the present invention provides an amino acid infrared spectrum classification and identification method and a visualization system to eliminate or improve one or more defects existing in the prior art.

[0043] To achieve the above object, the present invention provides the following solutions:

[0044] A method for classifying and identifying amino acids by infrared spectroscopy, wherein the classification and identification of amino acids by infrared spectroscopy is achieved by using a machine learning algorithm based on amino acid infrared spectroscopy, comprising the following steps:

[0045] Using amino acid powder and solid dispersant to make amino acid tablets;

[0046] The amino acid pressed tablet is placed in a Fourier transform infrared spectrometer to collect amino acid transmission spectrum data;

[0047] Based on the amino acid transmission spectrum data, the classification and regression tree method and the support vector machine method were used to obtain a trained algorithm model for identifying the types of amino acid functional groups.

[0048] Based on the amino acid transmission spectrum data, the classification and regression tree method and the support vector machine method were used to obtain a trained algorithm model for identifying amino acid types.

[0049] The infrared spectrum data of the amino acid to be determined are respectively input into the algorithm model for identifying the amino acid functional group and the amino acid type, and the classification results of the amino acid functional group and the amino acid classification results are obtained.

[0050] Furthermore, the solid dispersant used is potassium bromide powder.

[0051] Furthermore, the preparation of amino acid tablets using amino acid powder and solid dispersant comprises the following steps:

[0052] Weigh the amino acid powder to be tested and potassium bromide powder in a ratio of 1:199;

[0053] Grinding: Grind the amino acid powder and potassium bromide powder separately using an agate mortar;

[0054] Filter, use a 250-mesh sieve to filter the sample to be tested, filter the ground amino acid powder and potassium bromide powder;

[0055] Mix well, put the filtered amino acid powder and potassium bromide powder into a sterile centrifuge tube, put it on a vortex mixer, and mix them well;

[0056] Tableting: Place the mixed powder in a mold to obtain an amino acid tablet with a thickness of 0.5 mm.

[0057] Furthermore, before training based on the amino acid transmission spectrum data, the collected amino acid transmission spectrum data needs to be preprocessed.

[0058] Furthermore, the preprocessing includes:

[0059] Format conversion: converting the ASF format data collected by Fourier infrared spectrometer into EXCEL format;

[0060] Dimensionality reduction, using principal component analysis to reduce the dimensionality of amino acid mid-infrared spectral data can reduce noise and eliminate redundant information. Based on the following calculation process:

[0061] Suppose there are m n-dimensional data X n×m=[x1,x2,...,x m ], where xi is an n-dimensional column vector.

[0062] Decentralization,

[0063] Calculate the covariance matrix,

[0064] Project to new coordinates, Where P n×k is the first k columns of the characteristic matrix obtained by eigenvalue decomposition of C, Y k×m It is the m k-dimensional data obtained after dimensionality reduction.

[0065] Data annotation: Annotate the infrared transmission spectrum data of each group of amino acids.

[0066] Furthermore, the algorithm model for identifying amino acid functional group categories and amino acid types using the classification and regression tree method is based on the following calculation process:

[0067] Among them, Gini(D) is the Gini index of sample set D, K means there are K categories, C k Represents the sample subset belonging to the kth class in the dataset D.

[0068] Among them, D1 and D2 represent two parts of the data set D divided according to feature A at a certain value a, and the Gini coefficient Gini(D,A) represents the uncertainty of the set D after A=a division. The larger the Gini index, the greater the uncertainty of the sample set.

[0069] Furthermore, the algorithm model for identifying amino acid functional group categories and amino acid types using the support vector machine method is based on the following calculation process:

[0070] First, the particle swarm algorithm is used to calculate the optimal penalty parameter and kernel function radius of the support vector machine. All particles are initialized, that is, their speed and position are assigned, and the individual's historical best pBest is set as the current position, and the best individual in the group is used as the current gBest.

[0071] Among them, w is the inertia weight, c1 and c2 are acceleration coefficients, and rand1 and rand2 are random numbers in [0,1].

[0072] Secondly, the support vector machine method is used to separate the hyperplane and the classification decision function. Input the training data set, x i Represents the data feature value, y iRepresents the class label, selects the appropriate kernel function K(x,z) and penalty parameter C, constructs and solves the convex quadratic programming problem, and the calculation formula is as follows: 0≤α i ≤C,i=1,2,...N

[0073] Get the optimal solution Then select α * A component of Satisfy the conditions calculate

[0074] Classification decision function

[0075] Using the radial basis kernel function K(x i ,x j ) = exp(-g||x i -x j || 2 ),g>0.

[0076] Furthermore, the algorithm model for identifying the types of amino acid functional groups specifically includes eight functional group types, namely hydrocarbon chain-containing, hydroxyl-containing, sulfur atom-containing, carboxyl-containing, amide-containing and alkaline aliphatic amino acids, aromatic amino acids and heterocyclic amino acids.

[0077] Furthermore, the algorithm model for identifying the types of amino acids is specifically glycine, alanine, valine, leucine, isoleucine, serine, threonine, methionine, cysteine, aspartic acid, glutamic acid, asparagine, glutamine, arginine, lysine, phenylalanine, tyrosine, proline, tryptophan, and histidine, a total of 20 amino acids.

[0078] A visualization system for amino acid infrared spectrum classification and identification, comprising:

[0079] The data type and algorithm determination module is used to select the type of amino acid spectrum data to be tested and the algorithm model to be used. The spectrum data type includes transmission spectrum and absorbance spectrum. The algorithm model includes support vector machine method and classification and regression tree method.

[0080] The data input module is used to load the infrared spectrum data of amino acids tested into the system. The infrared spectrum data of amino acids needs to be stored in a table.

[0081] The algorithm model calling module is used to call the packaged algorithm model in the system to identify the spectral data input by the user.

[0082] The result display module is used to display the amino acid type to which the tested spectral data belongs.

[0083] Beneficial effects of the present invention:

[0084] The invention provides an amino acid infrared spectrum classification and recognition method and a visualization system, which can be used to detect the functional group category and amino acid category of unknown amino acid powder.

[0085] 1. By collecting infrared transmission spectrum data of amino acids, we fully utilize the infrared-active characteristics of the vibration modes of amino acid molecules, effectively expand the sample size of 20 amino acid transmission spectra in the mid-infrared band, and obtain a sample-rich amino acid transmission spectrum data set, which provides a basis for training algorithm models that can identify amino acid functional groups and amino acid types to improve classification accuracy.

[0086] 2. A low-cost amino acid identification method is proposed. The amino acid spectrum data is collected by Fourier infrared spectrometer. Compared with the traditional chromatography and nuclear magnetic resonance methods, the test equipment collects data efficiently and at a low cost. Compared with the electrochemical sensor method, it will not affect the test sample, improves the efficiency of collecting amino acid data, and effectively reduces the test cost.

[0087] 3. It is proved that the molecular recognition method combining machine learning and infrared spectroscopy is a universal method. Through the efficient extraction of known spectral features and the optimization of machine learning models, it is possible to achieve high-accuracy amino acid classification under the condition of existing sample data. This method provides a new idea for real-time and efficient molecular detection under experimental conditions with limited resources.

[0088] 4. Based on the infrared transmission spectrum data of amino acids, using the classification and regression tree method and the support vector machine method, a high-accuracy algorithm model can be established for the identification of amino acid functional group types and amino acid types at the same time. By gradually narrowing the category range, accurate identification from macro to micro can be achieved.

[0089] 5. We tried multiple machine learning models including support vector machine and classification and regression tree, and finally selected the best accuracy by comparing model performance. We maintained a classification accuracy of 99% on different test sets, and built an efficient practical example for complex spectral data processing.

[0090] 6. Traditional spectral classification relies on empirical rules and instruments, while this method uses a data-driven model to automatically capture rules. By introducing the properties of amino acid transmission spectra to impose constraints, the classification model is given more theoretical support, which can improve the training efficiency and stability of the model and provide a new modeling perspective for solving complex biochemical problems.

[0091] 7. A visual web platform was developed to intuitively display the classification results of infrared spectral data of amino acids. The platform combines the output of the machine learning model and provides an interactive graphical interface. This not only improves the comprehensibility of experimental results, but also provides convenient operating tools for non-professional experimental personnel, helping to quickly analyze and verify spectral data, and promoting the efficient development of scientific research. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of this application, and do not constitute a limitation of the present invention.

[0094] Figure 1 This is a schematic flow chart of the infrared spectrum classification and identification method for amino acids in Example 1 of the present invention;

[0095] Figure 2 The infrared transmission spectrum of aliphatic amino acids with hydrocarbon chains in the side chains collected by the present invention;

[0096] Figure 3 This is the infrared transmission spectrum of aliphatic amino acids with hydroxyl groups on the side chains collected by the present invention;

[0097] Figure 4 This is the infrared spectrum of aliphatic amino acids containing sulfur atoms in the side chains collected by the present invention;

[0098] Figure 5 The infrared spectrum of aliphatic amino acids with carboxyl groups on the side chains collected by the present invention;

[0099] Figure 6 This is the infrared spectrum of aliphatic amino acids with amide groups in the side chains collected by the present invention;

[0100] Figure 7 This is the infrared spectrum of aliphatic amino acids with alkaline side chains collected by the present invention;

[0101] Figure 8 This is the infrared spectrum of aromatic amino acids collected by the present invention;

[0102] Fig. 9 This is the infrared spectrum of heterocyclic amino acids collected by the present invention;

[0103] Fig.10 A classification result diagram of the algorithm model for identifying the types of amino acid functional groups of the present invention;

[0104] Fig.11 A classification result diagram of the algorithm model for identifying amino acid types of the present invention;

[0105] Fig.12 A confusion matrix diagram of the algorithm model for identifying the types of amino acid functional groups of the present invention;

[0106] Fig.13 A confusion matrix diagram of the algorithm model for identifying amino acid types of the present invention;

[0107] Fig.14 An initialization page of the visualization system for identifying amino acid types of the present invention;

[0108] Fig.15 This is the result display page of the visualization system for identifying amino acid types of the present invention. DETAILED DESCRIPTION

[0110] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The enumerated embodiments are only used to explain the present invention and are not used to limit the scope of the present invention.

[0111] Experimental Example 1

[0112] Reference Figure 1 This embodiment provides an infrared spectrum classification and identification method for amino acids, the method comprising the following steps:

[0113] S101. Prepare amino acid tablets using amino acid powder and solid dispersant.

[0114] In this experimental example, the solid dispersant is potassium bromide powder, which hardly absorbs infrared light in the mid-infrared region.

[0115] S102, placing the amino acid pressed tablet in a Fourier transform infrared spectrometer to collect amino acid transmission spectrum data.

[0116] The Fourier transform infrared spectrometer can emit 4400cm -1 -400cm -1 The infrared light is then received, and the light transmitted through the amino acid tablet is received, and the amino acid transmission spectrum data is obtained through the digital signal processing module. The amino acid transmission spectrum can be obtained according to the following formula.

[0117] In the formula, B(v) represents the intensity of the monochromatic light source with wave number v after correction for instrument characteristics, and I(δ) represents the signal intensity detected by the detector when the optical path difference is δ.

[0118] In the formula, T is the transmittance at a certain wave number, I(v) represents the intensity of the infrared spectrum after passing through the sample, and I0(v) represents the intensity of the infrared light passing through the background. Figure 2 This is a mid-infrared transmission spectrum of some amino acids collected in the present invention.

[0119] In this experimental example, 20 tablets were made for each amino acid, and each tablet was sampled 20 times. 200 sets of transmission spectrum data were collected for each amino acid, and a total of 4000 sets of data were collected for the 20 amino acids. In a room temperature, dry test environment, the test time of each tablet needs to be shortened as much as possible. As the test time increases, the tablet will absorb carbon dioxide and moisture in the environment, and the transparency of the tablet will decrease. -1 There will be increasingly stronger transmission valleys.

[0120] S103. Based on the amino acid transmission spectrum data, a classification and regression tree method and a support vector machine method are used to obtain a trained algorithm model for identifying the types of amino acid functional groups.

[0121] Before the training based on the amino acid transmission spectrum data, the collected amino acid transmission spectrum data needs to be preprocessed. The preprocessing includes the following steps:

[0122] Format conversion: convert the ASF format data collected by the Fourier infrared spectrometer into EXCEL format. In the specific experimental example, the collected 4000 sets of mid-infrared spectrum data are organized into an EXCEL file as the data set of the recognition model. The number of rows represents the number of data, and the number of columns represents the feature dimension.

[0123] Dimensionality reduction, using principal component analysis to reduce the dimensionality of amino acid mid-infrared spectral data can reduce noise and eliminate redundant information. Based on the following calculation process:

[0124] Suppose there are m n-dimensional data X n×m =[x1,x2,...,x m ], where x i is an n-dimensional column vector.

[0125] Decentralization,

[0126] Calculate the covariance matrix,

[0127] Project to new coordinates, Where P n×k is the first k columns of the characteristic matrix obtained by eigenvalue decomposition of C, Y k×m It is the m k-dimensional data obtained after dimensionality reduction.

[0128] In this experimental example, the infrared transmission spectrum of each group of amino acids contains 4400cm -1 -400cm -1The transmittance corresponding to 2075 wave numbers in the band. When a substance is irradiated by a beam of infrared light with a continuous wavelength, the vibration or rotation of the molecules in the substance causes a change in the dipole moment, resulting in the transition of the vibration-rotation energy level from the ground state to the excited state. The infrared light of this wavelength is absorbed by the substance, and the intensity of the transmitted light corresponding to these areas is weakened. In this experimental example, the principal component analysis method is used to reduce the original 2075 dimensions to 5 feature dimensions. The lower the data feature dimension, the less original information will be represented, and the algorithm model recognition result will be lower. However, the higher the data feature dimension, the higher the time cost and memory cost used in training the algorithm model will become. As a preferred example, the number of principal components of the infrared transmission spectrum of each group of amino acids is set to 5.

[0129] Data annotation: Annotate the infrared transmission spectrum data of each group of amino acids.

[0130] In this experimental example, the data set for training the algorithm model for identifying amino acid types is divided into 20 types: glycine, alanine, valine, leucine, isoleucine, serine, threonine, methionine, cysteine, aspartic acid, glutamic acid, asparagine, glutamine, arginine, lysine, phenylalanine, tyrosine, proline, tryptophan, and histidine. The data set for training the algorithm model for identifying amino acid functional group types is divided into eight functional group types: hydrocarbon chain-containing, hydroxyl-containing, sulfur atom-containing, carboxyl-containing, amide-containing, and alkaline aliphatic amino acids, aromatic amino acids, and heterocyclic amino acids. Figure 2 Mid-infrared transmission spectra of aliphatic amino acids containing hydrocarbon chains, including glycine, alanine, valine, leucine, and isoleucine; Figure 3 The infrared transmission spectra of aliphatic amino acids containing hydroxyl groups include serine and threonine; Figure 4 Aliphatic amino acids containing sulfur atoms, including methionine and cysteine; Figure 5 The infrared transmission spectra of aliphatic amino acids containing carboxyl groups include aspartic acid and glutamic acid; Figure 6 The infrared transmission spectra of aliphatic amino acids containing amide groups, including asparagine and glutamine; Figure 7 Alkaline aliphatic amino acids, including arginine and lysine; Figure 8 This is the infrared transmission spectrum of aromatic amino acids, including phenylalanine and tyrosine; Fig. 9 It is the mid-infrared transmission spectrum of heterocyclic amino acids, including proline, tryptophan and histidine.

[0131] In this experimental example, the classification and regression tree method specifically refers to the classification tree in the decision tree. The algorithm uses the binary recursive segmentation technique to divide the sample set into two sub-sample sets, so that each non-leaf node generated has two branches. The decision tree generated by the CART algorithm is a binary tree with a simple structure. CART can handle continuous variables and discrete variables, use the eigenvalues ​​of the training data to build the tree, and use the verification data for pruning. Based on the following calculation process:

[0132] Among them, Gini(D) is the Gini index of sample set D, K means there are K categories, C k Represents the sample subset belonging to the kth class in the dataset D.

[0133] Among them, D1 and D2 represent two parts divided on the data set D according to feature A at a certain value a, and the Gini coefficient Gini (D, A) represents the uncertainty of the set D after A = a segmentation. The larger the Gini index, the greater the uncertainty of the sample set. First, the Gini gain is calculated based on the current feature. The feature with the smallest Gini gain is selected as the partition feature. From this feature, find the classification category with the smallest Gini index as the optimal partition point. The current sample is divided into two categories, one is that the category of the partition feature is equal to the optimal partition point, and the other is that it is not equal. The above partitioning work is recursively performed for these two categories, and the number of samples in the direct node is less than the predetermined threshold, or the Gini index of the sample set is less than the predetermined threshold (the samples basically belong to the same category), or there are no more features. In this experimental example, on the classification tree for identifying the type of amino acid functional group, a branch is established using the third principal component and the fifth principal component. If the third principal component is less than -352.05 and the fifth principal component is greater than or equal to 176.331, it can be judged that the mid-infrared transmission spectrum data belongs to a hydroxyl-containing aliphatic amino acid. In the classification tree for identifying amino acid types, starting from the root node, the two parts are divided according to whether the fifth principal component is greater than or equal to 176.331. It can be determined whether it belongs to proline or tryptophan, indicating that the information entropy of the fifth principal component is relatively high.

[0134] In this experimental example, the support vector machine method is used to establish an algorithm model for identifying amino acid functional group categories and amino acid types, based on the following calculation process:

[0135] First, the particle swarm algorithm is used to calculate the optimal penalty parameter and kernel function radius of the support vector machine. Initialize all particles, that is, assign values ​​to their speed and position, and set the individual's historical optimal pBest to the current position, and the optimal individual in the group as the current gBest; in each generation of evolution, calculate the fitness function value of each particle; if the current fitness function value is better than the historical optimal value, update pBest; if the current fitness function value is better than the global historical optimal value, update gBest; update the speed and position of the dth dimension of each particle i according to the following formula. Finally, determine whether the termination condition is met. If not, re-determine the fitness value of the particle, and continue to update the position and speed of the particle until the termination condition is met. Otherwise, end the algorithm and obtain the optimal solution.

[0136] Among them, w is the inertia weight, c1 and c2 are acceleration coefficients, and rand1 and rand2 are random numbers in [0,1].

[0137] Secondly, the support vector machine method is used to separate the hyperplane and the classification decision function. Input the training data set, x i Represents the data feature value, y i Represents the class label, selects the appropriate kernel function K(x,z) and penalty parameter C, constructs and solves the convex quadratic programming problem, and the calculation formula is as follows:

[0138] Get the optimal solution Then select α * A component of Satisfy the conditions calculate

[0139] Classification decision function

[0140] Using the radial basis kernel function K(x i ,x j ) = exp(-g||x i -x j || 2 ),g>0.

[0141] In the specific implementation process, the initial inertia weight is 0.5, the cognitive component is 2, the social component is 2, the number of particles is 30, the maximum number of iterations is 10, and the hyperparameters C and g are 527.81 and 0.7221, respectively, obtained through the particle swarm optimization algorithm.

[0142] In this experimental example, the labeled amino acid mid-infrared transmission spectrum data are divided into a training set and a test set in a ratio of 8:2. In this experimental example, the aliphatic amino acids containing hydrocarbon chains include 5 amino acids, namely glycine, alanine, valine, leucine and isoleucine. 200 sets of data were collected for each amino acid, and the training set is 800 sets of data randomly selected from the transmission spectrum data of these 5 amino acids.

[0143] S103. Based on the amino acid transmission spectrum data, a classification and regression tree method and a support vector machine method are used to obtain a trained algorithm model for identifying amino acid types.

[0144] In this experimental example, the labeled amino acid mid-infrared transmission spectrum data is divided into a training set and a test set in a ratio of 8:2. In this experimental example, a total of 4000 sets of data are collected by Fourier infrared spectrometer, so 160 sets of data are randomly selected from each type, and a total of 3200 sets of data are input into the recognition model for model training. The trained model can identify the amino acid type through the test transmission spectrum data.

[0145] The infrared spectrum data of the amino acid to be determined are respectively input into the algorithm model for identifying the amino acid functional group and the amino acid type, and the classification results of the amino acid functional group and the amino acid classification results are obtained.

[0146] In this experimental example, 800 groups of data in the test set are input into the trained algorithm model for identifying amino acid types. Fig.10 The classification tree method for identifying amino acid functional groups has a model recognition accuracy of 99.16%, and the support vector machine method for identifying amino acid functional groups has a model recognition accuracy of 99.8%. Fig.11 The classification tree model has an accuracy rate of 98.6% in identifying amino acid types, and the support vector machine model has an accuracy rate of 99.6%. The support vector machine model training time takes about 2 minutes, and the classification tree model training time takes less than 1 minute.

[0147] In some embodiments of the present invention, the amino acid powder to be tested and the solid dispersant powder are compressed into amino acid tablets, comprising the following steps:

[0148] Weigh the amino acid powder to be tested and potassium bromide powder in a ratio of 1:199;

[0149] In the specific implementation process, the amino acid powder and solid dispersant powder to be tested were obtained at a mass ratio of 1:199, with 0.01g of amino acid powder and 1.99g of solid dispersant powder. By observing the spectra at different ratios, it was found that the transmittance of the absorption peak in the spectrum at a ratio of 1:199 was almost between 10% and 80%, and the peak position and peak value could be clearly seen in the absorption spectrum, which met the requirements of the standard spectrum. Therefore, 20 kinds of amino acids were selected to make tablets at a ratio of 1:199.

[0150] Grind. Grind the amino acid powder and potassium bromide powder separately using an agate mortar.

[0151] During the specific implementation process, apply uniform force in the agate mortar in the same direction to avoid crystal rotation and affect the spectral quality.

[0152] Filter the sample using a 250 mesh sieve to filter the ground amino acid powder and potassium bromide powder. Ensure that the particle size is below 2.5 μm to avoid being affected by scattered light during testing.

[0153] Mix well, put the filtered amino acid powder and potassium bromide powder into a sterile centrifuge tube, place it on a vortex mixer, and mix the two thoroughly.

[0154] In this experimental example, the powder was poured into a 5 ml centrifuge tube, which was placed on a vortex mixer, the switch was turned on, and waited for 2 minutes.

[0155] Tableting: Place the mixed powder in a mold to obtain an amino acid tablet with a thickness of 0.5 mm.

[0156] As a preferred example, 0.1683 g of the mixture of the amino acid powder and the solid dispersant powder is placed in a mold and pressed at a pressure of 12 MPa at room temperature for 3 minutes to obtain an amino acid tablet with a thickness of 0.5 mm. If the tablet thickness is greater than 0.5 mm, interference fringes often appear in the spectrum.

[0157] Experimental Example 2

[0158] The visualization system for amino acid infrared spectrum classification and identification in this experimental example includes:

[0159] The data type and algorithm determination module is used to select the test range of the amino acid spectrum data to be tested and the algorithm model to be used; the spectrum data test range includes amino acid types and functional group types; the algorithm model includes support vector machine method and classification and regression tree method, see Fig.14 ;

[0160] The data input module is used to load the infrared spectrum data of amino acids tested into the system; the infrared spectrum data of amino acids needs to be stored in a table;

[0161] The algorithm model calling module is used to call the algorithm model packaged in the system to identify the spectral data input by the user;

[0162] The result display module is used to display the amino acid types to which the tested spectral data belongs, such as Fig.15 shown.

[0163] Experimental Example 3

[0164] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute an amino acid infrared spectrum classification and identification method and a visualization system in Example 1 and Experimental Example 2.

[0165] As an optional implementation, the memory is a readable storage medium.

[0166] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for classifying and identifying amino acids by infrared spectroscopy, wherein the classification and identification of amino acids by infrared spectroscopy is achieved by using a machine learning algorithm based on amino acid infrared spectroscopy, and comprises the following steps: Using amino acid powder and solid dispersant to make amino acid tablets; placing the amino acid pressed tablet into a Fourier transform infrared spectrometer to collect amino acid transmission spectrum data; Based on the amino acid transmission spectrum data, the classification and regression tree method and the support vector machine method were used to obtain a trained algorithm model for identifying the types of amino acid functional groups. Based on the amino acid transmission spectrum data, the classification and regression tree method and the support vector machine method were used to obtain a trained algorithm model for identifying amino acid types. The infrared spectrum data of the amino acid to be determined are respectively input into the algorithm model for identifying the amino acid functional group and the amino acid type, and the classification results of the amino acid functional group and the amino acid classification results are obtained.

2. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The solid dispersant used is potassium bromide powder.

3. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The method of using amino acid powder and solid dispersant to prepare amino acid tablets comprises the following steps: Weigh the amino acid powder to be tested and potassium bromide powder in a ratio of 1:199; Grinding: Grind the amino acid powder and potassium bromide powder separately using an agate mortar; Filter, use a 250-mesh sieve to filter the sample to be tested, filter the ground amino acid powder and potassium bromide powder; Mix well, and place the filtered amino acid powder and potassium bromide powder in a sterile centrifuge tube. Place on a vortex mixer to mix the two thoroughly; Tableting: Place the mixed powder in a mold to obtain an amino acid tablet with a thickness of 0.5 mm.

4. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: Before the training based on the amino acid transmission spectrum data, the collected amino acid transmission spectrum data needs to be preprocessed, and the preprocessing includes: Format conversion: converting the ASF format data collected by Fourier infrared spectrometer into EXCEL format; Dimensionality reduction: using principal component analysis to reduce the dimensionality of amino acid mid-infrared spectral data can reduce noise and eliminate redundant information; Data annotation: Annotate the infrared transmission spectrum data of each group of amino acids.

5. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The algorithm model for identifying amino acid functional group categories and amino acid types using the classification and regression tree method is based on the following calculation process: Among them, Gini(D) is the Gini index of sample set D, K means there are K categories, C k Represents the sample subset belonging to the kth class in the dataset D. Among them, D1 and D2 represent the two parts divided on the data set D according to the feature A at a certain value a, and the Gini coefficient Gini(D,A) represents the uncertainty of the set D after A=a division. The larger the Gini index is, the greater the uncertainty of the sample set is.

6. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The algorithm model for identifying amino acid functional group categories and amino acid types using the support vector machine method is based on the following calculation process: First, the particle swarm algorithm is used to calculate the optimal penalty parameter and kernel function radius of the support vector machine. All particles are initialized, that is, their speed and position are assigned, and the individual's historical best pBest is set as the current position, and the best individual in the group is used as the current gBest. Among them, w is the inertia weight, c1 and c2 are acceleration coefficients, and rand1 and rand2 are random numbers in [0,1]. Secondly, the support vector machine method is used to separate the hyperplane and the classification decision function. Input the training data set, x i Represents the data feature value, y i Represents the class label, selects the appropriate kernel function K(x,z) and penalty parameter C, constructs and solves the convex quadratic programming problem, and the calculation formula is as follows: 0≤α i ≤C,i=1,2,...N Get the optimal solution Then select α * A component of Satisfy the conditions calculate Classification decision function Using the radial basis kernel function K(x i ,x j ) = exp(-g||x i -x j || 2 ),g>0.

7. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The algorithm model for identifying the types of amino acid functional groups includes eight functional group types: hydrocarbon chain-containing, hydroxyl-containing, sulfur atom-containing, carboxyl-containing, amide-containing and alkaline aliphatic amino acids, aromatic amino acids and heterocyclic amino acids.

8. A method for amino acid infrared spectrum classification and identification according to claim 1, characterized in that: The algorithm model for identifying amino acid types includes 20 amino acids: glycine, alanine, valine, leucine, isoleucine, serine, threonine, methionine, cysteine, aspartic acid, glutamic acid, asparagine, glutamine, arginine, lysine, phenylalanine, tyrosine, proline, tryptophan, and histidine.

9. A visualization system for amino acid infrared spectrum classification and identification, the visualization system comprising the following modules: a data type and algorithm determination module, for selecting a range to be tested and an algorithm model to be used; A data input module is used to load the infrared spectrum data of amino acids tested into the system; The algorithm model calling module is used to call the algorithm model packaged in the system to identify the spectral data input by the user; The result display module is used to display the amino acid type to which the tested spectral data belongs.

Images