Prediction method for inhibitory activity of chemical on DNA gyrase A subunit

Abstract

The invention discloses a method for predicting the inhibitory activity of a chemical to a DNA gyrase A subunit, and the method comprises the steps: obtaining the structure of the chemical and the inhibitory activity data of the chemical to the DNA gyrase A subunit from a public database, carrying out the structure cleaning and zooming, and dividing the obtained known data into a training set and a test set; constructing a graph neural network model taking the chemical SMILES codes as input; using the training set to determine an optimal hyper-parameter of the graph neural network model through cross validation so as to obtain an optimal inhibition activity prediction model; and acquiring a predicted value of the scaling value of the inhibitory activity data of the to-be-evaluated chemical to the DNA gyrase A subunit by using the inhibitory activity prediction model. According to the method, the inhibition effect of environmental chemicals on the DNA gyrase A subunit can be predicted only by inputting the molecular structure of the compound, rapid screening of molecular initial events of pollutants possibly inducing bacterial quinolone drug resistance is realized, and the method has a wide application prospect.

Description

Technical Field

[0001] The invention relates to the technical field of chemical-induced bacterial resistance assessment, and in particular to a method for predicting the inhibitory activity of chemicals on DNA gyrase A subunit. Background Art

[0002] Molecular Initiate Events (MIE) often involve the binding of pollutants to target proteins, which then triggers a molecular cascade response that leads to changes in normal physiological functions and produces irreversible harmful effects. Similarly, bacterial resistance mutations originate from the molecular recognition and interaction between environmental pollutants and many resistance-related targets, which in turn interfere with the relevant pathways that maintain normal cell functions and cause cascade effects, inducing the production of resistant bacteria. Since DNA gyrase A subunit is a key target for inducing quinolone resistance, the interaction between pollutants and DNA gyrase A subunit will also determine its ability to induce quinolone resistance. Pollutants with specific structures can inhibit DNA gyrase A subunit, thereby inhibiting DNA replication and inducing DNA damage, followed by a series of SOS and responses of DNA replication and repair systems, ultimately leading to target mutations, thereby producing quinolone resistance. Therefore, rapid screening of environmental pollutants with DNA gyrase A subunit inhibitory activity is the key to identifying high-risk pollutants with the ability to induce bacterial quinolone resistance.

[0003] Due to the large amount of chemicals added to industrial products and daily necessities, it is obviously not feasible to rely on traditional experimental methods. Therefore, computational methods are needed to speed up the screening of high-risk pollutants with the ability to induce bacterial quinolone resistance. The traditional methods for predicting the inhibitory activity of chemicals on proteins are mainly divided into two categories, namely ligand-based methods and receptor-based methods. The ligand-based method is to judge the potential effect of chemicals by comparing the structure with the known active ligand. The ligand-based method is simple and effective under the condition of high ligand similarity, but due to the limitation of the known ligand structure space, the scope of application and accuracy of this method are restricted. The receptor-based method mainly uses molecular docking to judge the interaction mode between the ligand and the target protein by calculating the three-dimensional structure, interaction energy and other parameters of the ligand-receptor complex. Although the receptor-based method fully considers the three-dimensional structural characteristics of the target protein, there are still some technical problems that have not been well solved. For example, the conformational flexibility of the protein, the accuracy of the scoring function, the solvent effect of water molecules, etc., all of which will affect the accuracy of the docking results.

[0004] Traditional methods for predicting the inhibitory activity of chemicals on proteins have problems such as low accuracy and a small scope of application. Therefore, it is urgent to propose and develop new methods for evaluating the inhibitory activity of chemicals on DNA gyrase A subunit to meet the needs of rapid screening of high-risk pollutants with the ability to induce bacterial quinolone resistance. In fact, with the rise of deep learning methods, methods for predicting chemical properties based on feature learning have gradually developed. This is because feature learning-based methods can apply deep learning algorithms to learn the characteristics of chemicals with known DNA gyrase A subunit inhibitory activity and infer the DNA gyrase A subunit inhibitory activity of unknown chemicals. In addition, relevant studies have shown that deep learning models perform better than molecular docking models and other machine learning models.

[0005] In summary, a deep learning model is needed that can fully learn the characteristics of chemicals, realize the direct mapping of chemical structure to properties, reduce the requirements for preliminary data preparation in model use, and improve the predictive ability of the model. Summary of the invention

[0006] The purpose of the present invention is to provide a method for predicting the inhibitory activity of chemicals on DNA gyrase A subunit in view of the deficiencies of the prior art.

[0007] The object of the present invention is achieved through the following technical scheme: A method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit, comprising the following steps:

[0008] S1. Obtain chemical structures and inhibitory activity data of chemicals on DNA gyrase A subunit, perform structural cleaning operations on chemical structures, perform scaling operations on inhibitory activity data to construct a data set, and randomly divide it into a training set and a test set in proportion;

[0009] S2. Constructing a graph neural network model with the chemical SMILES code as input, whose output is the predicted value of the scaled value of the inhibitory activity data of the chemical on the DNA gyrase A subunit;

[0010] S3. Use cross-validation to determine the optimal hyperparameters of the graph neural network model, and use the training set to train it to obtain the trained optimal graph neural network model, which is used as an inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit;

[0011] S4. Input the SMILES code of the chemical to be tested in the test set or the SMILES code of the chemical to be evaluated into the inhibitory activity prediction model to obtain the predicted value of the inhibitory activity data scaling value of the chemical on DNA gyrase A subunit.

[0012] Furthermore, the step S1 includes the following sub-steps:

[0013] S1.1. Obtain the chemical structure and the inhibitory activity data of the chemical on the DNA gyrase A subunit from public databases or literature; wherein the chemical structure is the chemical SMILES code, and the inhibitory activity data is used to indicate the inhibitory activity of the chemical on the DNA gyrase A subunit;

[0014] S1.2. Perform structure cleaning operations on chemical structures, specifically including: standardizing, removing solvents, charge correction, and deionizing chemical molecular structure cleaning operations on chemical SMILES codes to standardize and unify the form of chemical SMILES codes; wherein, the above-mentioned structure cleaning operations can perform molecular structure inspection processing to ensure that the structure of the molecule is chemically reasonable; correct drawing errors and standardize functional groups; hide hydrogen atoms; recalculate stereochemistry; remove covalently bonded impurities; ensure that the strongest acid group is ionized first; produce neutral molecules; standardize tautomers; remove salts and covalently bound metals contained in the compound;

[0015] S1.3, performing a scaling operation on the numerical range of the inhibitory activity data of the chemical on the DNA gyrase A subunit by taking the negative logarithm of the original numerical value of the inhibitory activity data of the chemical on the DNA gyrase A subunit;

[0016] S1.4, constructing a data set according to the chemical SMILES code after the structure cleaning operation in step S1.2 and the inhibitory activity data after the scaling operation in step S1.3;

[0017] S1.5. Randomly divide the dataset into training set and test set in proportion.

[0018] Furthermore, the graph neural network model includes a molecular graph embedding layer, a molecular fragment graph embedding layer, multiple molecular graph neural network encoders, multiple molecular fragment graph neural network encoders, a dual-path combiner and a predictor, wherein the molecular graph neural network encoder and the molecular fragment graph neural network encoder are both composed of a neighborhood attention module, an information aggregation module, a gating mechanism to update node feature module and a feature attention module, and the feature attention module includes a maximum pooling layer, a sum pooling layer, a multi-layer perceptron global pooling layer and an activation function; the dual-path combiner selects a graph convolutional attention layer, and the graph convolutional attention layer includes multiple attention heads and a ReLU activation function; the predictor includes n l Fully connected layer.

[0019] Furthermore, the molecular graph embedding layer calculates the molecular graph of the chemical according to the chemical SMILES code by using RDKit to extract molecular graph features from the molecular graph; wherein the molecular graph features include a×d atThe atomic feature matrix, the adjacency matrix of length b, and the b×d bt The chemical bond characteristic matrix is: a is the number of atoms in the chemical, b is twice the number of chemical bonds in the chemical, and d is at is the number of atomic features, d bt is the number of chemical bond features, and the adjacency matrix is ​​used to represent the atomic numbers of the bonding atom pairs in the molecular graph;

[0020] The molecular fragment graph embedding layer calculates the molecular graph of the chemical according to the SMILES code of the chemical by using RDKit, and extracts the molecular fragment features from the molecular graph based on the BRICS principle; wherein the molecular fragment features include a×d at The atomic feature matrix of length b f The molecular fragment adjacency matrix, b f ×d bt The chemical bond feature matrix of the molecular fragments and the molecular fragment index list of length a, where b f It is twice the number of chemical bonds remaining after the molecule is broken into molecular fragments. The molecular fragment index list is used to indicate the index of the molecular fragment where each atom in the chemical is located;

[0021] The various molecular graph features and molecular fragment features extracted from the molecular graph embedding layer and the molecular fragment graph embedding layer are first transformed into a length of d h Then, each type of feature vector is input into the molecular graph neural network encoder and the molecular fragment graph neural network encoder to obtain the molecular graph feature sequence and the molecular fragment feature sequence, respectively. h The length of is the same as the size of the hidden layer; wherein, in the molecular graph neural network encoder and the molecular fragment graph neural network encoder, the molecular graph neural network encoder and the molecular fragment graph neural network encoder update the node features of the molecular graph and the molecular fragment graph through their respective neighborhood attention modules and information aggregation modules, and then update the node feature module through a gating mechanism to balance between retaining the original information and using the new information, and in the feature attention module, the global features of the molecular graph are extracted through the maximum pooling layer and the sum pooling layer, and the attention weights are generated by the multi-layer perceptron, and the features of the nodes are weighted and adjusted, and finally the features of all nodes are aggregated into a global feature vector through the global pooling layer and the activation function, so as to obtain the molecular graph feature sequence and the molecular fragment feature sequence;

[0022] The molecular graph feature sequence and the molecular fragment feature sequence are input into the dual-path combiner. The dual-path combiner uses the graph convolution attention layer to learn the relationship between the molecular graph feature sequence and the molecular fragment feature sequence. Specifically, it adopts n hEach attention head independently calculates the attention weights of different molecular fragments, and performs nonlinear transformation through the ReLU activation function to obtain the molecule-molecular fragment interaction feature sequence;

[0023] Then, the molecule-molecule fragment interaction feature sequence and the molecular graph feature sequence output by the molecular graph neural network encoder are concatenated and input into the predictor. l After the fully connected layer, the predicted values ​​of the scaled values ​​of the inhibitory activity data of the chemicals on the DNA gyrase A subunit are output.

[0024] Furthermore, the step S3 includes the following sub-steps:

[0025] S3.1. Preset multiple sets of different model hyperparameters as {lr, weight-decay, batchsize, dropout, depth, feature-dimension, n-slices, r}, where lr is the learning rate, weight-decay is the weight decay parameter, batchsize is the number of samples selected for one training, dropout is the random dropout regularization term parameter, depth is the number of encoder layers, feature-dimension is the size of the hidden layer, n-slices is the number of slices, r is the decay rate, lr is used to control the progress of convergence to the local minimum, weight-decay and dropout are used to prevent model overfitting, depth, feature-dimension, n-slices and r are used to adjust the model complexity, and the size of batchsize affects the degree and speed of model optimization;

[0026] S3.2. For each set of model hyperparameters, use the data in the training set to train the graph neural network model. f The fold-cross validation was used to evaluate the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, and the trained graph neural network model corresponding to the optimal model hyperparameters was obtained, which was used as the inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit.

[0027] Furthermore, the statistical parameter is one of a coefficient of determination, a root mean square error, a mean square error, and a mean absolute error.

[0028] Furthermore, for each set of model hyperparameters, the graph neural network model is trained using the data in the training set, and n f The fold cross validation evaluates the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, including:

[0029] S3.2.1. Train the graph neural network model based on a set of preset model hyperparameters, using n f The data in the training set is randomly divided into n f subsets, each time one of the subsets is used as the validation set, and the remaining n f -1 subsets to train the graph neural network model. During the training process, the statistical parameters are calculated based on the predicted values ​​of the inhibitory activity data scaling values ​​of the chemicals on the DNA gyrase A subunit output by the graph neural network model and the corresponding true values ​​in the training set. The parameters of the graph neural network model are adjusted with the optimal statistical parameters as the optimization goal, and then the graph neural network model is verified using the data in the validation set to calculate the corresponding statistical parameters. The above process is repeated for training n f times, and finally get n f Statistical parameters and calculate n f The average value of the statistical parameters;

[0030] S3.2.2. Repeat step S3.2.1 to obtain the statistical parameter averages corresponding to all groups of model hyperparameters, and select a group of model hyperparameters corresponding to the optimal statistical parameter averages as the optimal model hyperparameters.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] (1) The present invention can make full use of existing high-throughput experimental big data, use chemical SMILES codes as inputs to the inhibitory activity prediction model, and does not require artificially defined quantifiable structural parameters as molecular descriptors, thereby saving time and computing resources for molecular descriptor calculation and descriptor selection, and has lower requirements for computational chemistry foundations when applied;

[0033] (2) Compared with existing methods, the method of the present invention has high-precision prediction performance, which is helpful for the rapid discovery of new environmental pollutants that induce bacterial resistance, and has broad application prospects in the fields of chemical risk assessment and environmental safety assessment;

[0034] (3) The structure of the inhibitory activity prediction model in the present invention is flexible, which greatly improves the predictive ability of the inhibitory activity prediction model and realizes high-precision prediction of the inhibitory activity of chemical DNA gyrase A subunit, which is conducive to the rapid discovery of new environmental pollutants that induce bacterial resistance and can also be extended to the study of different key molecular initiation events. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of a method for predicting the inhibitory activity of a chemical substance of the present invention on DNA gyrase A subunit;

[0036] Figure 2This is a schematic diagram of the structure of the graph neural network model network in Example 1 of the present invention. DETAILED DESCRIPTION

[0037] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0038] The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The singular forms "a", "the" and "the" used in the present invention and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0039] It should be understood that although the terms first, second, third, etc. may be used in the present invention to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0040] The present invention is described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the features of the following embodiments and implementations can be combined with each other.

[0041] The method for predicting the inhibitory activity of chemicals on DNA gyrase A subunit of the present invention is based on the chemical SMILES (Simplified Molecular Input Line Entry System) code, which avoids the limitations of traditional methods such as low accuracy and small application range. Compared with other similar methods, the method of the present invention has the most superior prediction ability in current research.

[0042] It should be understood that SMILES is a specification that uses an ASCII string to clearly describe a molecular structure, and the SMILES code can be used to describe the three-dimensional chemical structure of a compound.

[0043] The basic principle of the present invention is: the input of the model is the chemical SMILES; firstly, the atomic feature matrix, the adjacency matrix and the chemical bond feature matrix are obtained through the molecular graph embedding layer, and the molecular fragment features are obtained through the molecular fragment graph embedding layer, specifically including the atomic feature matrix, the molecular fragment adjacency matrix, the chemical bond feature matrix of the molecular fragments and the molecular fragment index list; then various feature vectors extracted from the molecular graph embedding layer and the molecular fragment graph embedding layer are respectively input into the molecular graph neural network encoder and the molecular fragment graph neural network encoder, and the molecular graph feature sequence and the molecular fragment feature sequence are correspondingly obtained, and the intrinsic connection between the molecule and the molecular fragment is captured by the dual-pathway combiner, and finally the predictor outputs the inhibitory activity data (IC) of the chemical on the DNA gyrase A subunit. 50 ) scaled values.

[0044] See also Figure 1 The method for predicting the inhibitory activity of a chemical substance on DNA gyrase A subunit of the present invention specifically comprises the following steps:

[0045] S1. Obtain chemical structures and data on the inhibitory activity of chemicals on DNA gyrase A subunit (IC 50 ), and performed structural cleaning operations on the chemical structures and scaling operations on the inhibitory activity data to construct a dataset, which was randomly divided into training and test sets in proportion.

[0046] S1.1. Obtain the chemical structure and the inhibitory activity data of the chemical on the DNA gyrase A subunit from public databases or literature; wherein the chemical structure is the chemical SMILES code, and the inhibitory activity data is used to indicate the inhibitory activity of the chemical on the DNA gyrase A subunit.

[0047] S1.2. Perform structure cleaning operations on chemical structures, including: standardization of chemical SMILES codes, desolvation, charge correction, deionization and other chemical molecular structure cleaning operations, so as to standardize and unify the form of chemical SMILES codes, which can effectively avoid interference in subsequent calculations. Among them, these structure cleaning operations can perform molecular structure inspection to ensure that the structure of the molecule is chemically reasonable; correct common drawing errors and standardize functional groups; hide hydrogen atoms; recalculate stereochemistry; remove covalently bonded impurities; ensure that the strongest acid group is ionized first; produce neutral molecules as much as possible; standardize tautomers; remove salts and covalently bound metals contained in the compound.

[0048] S1.3. By taking the negative logarithm of the original numerical value of the inhibitory activity data of the chemical on the DNA gyrase A subunit and scaling the numerical range of the inhibitory activity data of the chemical on the DNA gyrase A subunit, the numerical range of the inhibitory activity data can be reduced.

[0049] S1.4. Construct a data set based on the chemical SMILES code after the structure cleaning operation in step S1.2 and the inhibitory activity data after the scaling operation in step S1.3.

[0050] S1.5. Randomly divide the data set into training set and test set according to a certain ratio. T% of the data is used as the training set to train and optimize the graph neural network model; w% of the data is used as the test set to evaluate the performance of the final model.

[0051] S2. Construct a graph neural network model with the SMILES code of chemicals as input, and its output is the predicted value of the scaled value of the inhibitory activity data of the chemicals on the DNA gyrase A subunit.

[0052] In this embodiment, the graph neural network model includes a molecular graph embedding layer, a molecular fragment graph embedding layer, multiple molecular graph neural network encoders, multiple molecular fragment graph neural network encoders, a dual-pathway combiner and a predictor, wherein the molecular graph neural network encoder and the molecular fragment graph neural network encoder are both composed of a neighborhood attention module, an information aggregation module, a gating mechanism to update node feature module and a feature attention module, but their parameters are different, such as Figure 2 As shown in Figure 1, the feature attention module includes a maximum pooling layer, a sum pooling layer, a multi-layer perceptron global pooling layer and an activation function; the dual-path combiner uses a graph convolutional attention layer, which includes multiple attention heads and a ReLU activation function; the predictor includes n l Fully connected layer.

[0053] Specifically, the molecular graph embedding layer calculates the molecular graph of the chemical according to the chemical SMILES code by using RDKit to extract the molecular graph features from the molecular graph; wherein the molecular graph features include a×d at The atomic feature matrix, the adjacency matrix of length b, and the b×d bt The chemical bond characteristic matrix is: a is the number of atoms in the chemical, b is twice the number of chemical bonds in the chemical, and d is at is the number of atomic features, d bt is the number of chemical bond features, and the adjacency matrix is ​​used to represent the atomic numbers of the bonding atom pairs in the molecular graph.

[0054] The molecular fragment graph embedding layer uses RDKit to calculate the molecular graph of chemicals according to the SMILES code of chemicals, and extracts molecular fragment features from the molecular graph based on the BRICS principle (Bonds Representing Interaction Clusters); the molecular fragment features include a×d at The atomic feature matrix of length b f The molecular fragment adjacency matrix, bf ×d bt The chemical bond feature matrix of the molecular fragments and the molecular fragment index list of length a, where b f It is twice the number of chemical bonds remaining after the molecule is broken into molecular fragments. The molecular fragment index list is used to indicate the index of the molecular fragment where each atom in the chemical is located.

[0055] The various molecular graph features and molecular fragment features extracted from the molecular graph embedding layer and the molecular fragment graph embedding layer are first transformed into a length of d h Then, each type of feature vector is input into the molecular graph neural network encoder and the molecular fragment graph neural network encoder to obtain the molecular graph feature sequence and the molecular fragment feature sequence, respectively. h The length of is the same as the size of the hidden layer. In the molecular graph neural network encoder and the molecular fragment graph neural network encoder, the molecular graph neural network encoder and the molecular fragment graph neural network encoder update the node features of the molecular graph and the molecular fragment graph through their respective neighborhood attention modules and information aggregation modules, and then update the node feature module through the gating mechanism to balance between retaining the original information and using the new information. In the feature attention module, the global features of the molecular graph are extracted through the maximum pooling layer and the sum pooling layer, and the attention weights are generated by the multi-layer perceptron to weight the features of the nodes. Finally, the features of all nodes are aggregated into a global feature vector through the global pooling layer and the activation function, thereby obtaining the molecular graph feature sequence and the molecular fragment feature sequence.

[0056] It should be understood that feature conversion refers to compressing or amplifying feature dimensions, as well as calculating activation functions, to convert various molecular graph features and molecular fragment features into corresponding feature vectors. In addition, in deep learning models, except for the input layer and the output layer, the remaining network layers can be called hidden layers.

[0057] The molecular graph feature sequence and the molecular fragment feature sequence are input into the dual-path combiner. The dual-path combiner first adopts the interactive attention strategy to capture the intrinsic connection between molecules and molecular fragments, and uses the graph convolutional attention layer to learn the relationship between the molecular graph feature sequence and the molecular fragment feature sequence. Specifically, it adopts n h Each attention head independently calculates the attention weights of different molecular fragments, which can better capture the characteristics of molecular fragments, and obtains the molecular-molecular fragment interaction feature sequence through nonlinear transformation through the ReLU activation function. Then, the molecular-molecular fragment interaction feature sequence and the molecular graph feature sequence output by the molecular graph neural network encoder are spliced ​​and input into the predictor. l After the full connection layer, the inhibitory activity data of the chemical on DNA gyrase A subunit (IC 50) scaled values.

[0058] S3. Use cross-validation to determine the optimal hyperparameters of the graph neural network model, and use the training set to train it to obtain the trained optimal graph neural network model, which is used as the inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit.

[0059] S3.1. Based on experience, we preset multiple sets of different model hyperparameters as {lr, weight-decay, batchsize, dropout, depth, feature-dimension, n-slices, r}, where lr is the learning rate, weight-decay is the weight decay parameter, batchsize is the number of samples selected for one training, dropout is the random dropout regularization term parameter, depth is the number of encoder layers, feature-dimension is the size of the hidden layer, n-slices is the number of slices, r is the decay rate, lr is used to control the progress of convergence to the local minimum, weight-decay and dropout are used to prevent the model from overfitting, depth, feature-dimension, n-slices and r are used to adjust the model complexity, and the size of batchsize affects the degree and speed of model optimization.

[0060] S3.2. For each set of model hyperparameters, use the data in the training set to train the graph neural network model. f The fold-cross validation was used to evaluate the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, and the trained graph neural network model corresponding to the optimal model hyperparameters was obtained, which was used as the inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit.

[0061] Furthermore, the statistical parameter is the coefficient of determination (R 2 ), root mean square error (RMSE), mean square error (MSE), mean absolute error (MAE), and their calculation formulas are:

[0062]

[0063] Where n represents the number of samples, y i represents the true value of the scaling value of the inhibitory activity data of the chemical corresponding to the i-th sample on the DNA gyrase A subunit, represents the predicted value of the scaling value of the inhibitory activity data of the chemical corresponding to the i-th sample on the DNA gyrase A subunit, Represents the average of all true values.

[0064] Furthermore, for each set of model hyperparameters, the graph neural network model is trained using the data in the training set. f The fold cross validation evaluates the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, including:

[0065] S3.2.1. Train the graph neural network model based on a set of preset model hyperparameters, using n f The data in the training set is randomly divided into n f subsets, each time one of the subsets is used as the validation set, and the remaining n f -1 subsets to train the graph neural network model. During the training process, the statistical parameters are calculated based on the predicted values ​​of the inhibitory activity data scaling values ​​of the chemicals on the DNA gyrase A subunit output by the graph neural network model and the corresponding true values ​​in the training set. The parameters of the graph neural network model are adjusted with the optimal statistical parameters as the optimization goal, and then the graph neural network model is verified using the data in the validation set to calculate the corresponding statistical parameters. The above process is repeated for training n f times, and finally get n f Statistical parameters, take n f The average value of the statistical parameters evaluates the predictive ability of the graph neural network model when different model hyperparameters are selected.

[0066] S3.2.2. Repeat step S3.2.1 to obtain the statistical parameter averages corresponding to all groups of model hyperparameters, and select a group of model hyperparameters corresponding to the optimal statistical parameter averages as the optimal model hyperparameters.

[0067] It should be understood that when different statistical parameters are selected, the requirements for the optimal statistical parameters are also different. For example, when the statistical parameter is RMSE, the smaller the RMSE, the better the statistical parameter; when the statistical parameter is R 2 When R 2 The larger the value, the better the statistical parameter.

[0068] S4. Input the SMILES code of the chemical to be tested in the test set or the SMILES code of the chemical to be evaluated into the inhibitory activity prediction model to obtain the predicted value of the inhibitory activity data scaling value of the chemical on the DNA gyrase A subunit. The larger the predicted value, the stronger the inhibitory activity of the chemical on the DNA gyrase A subunit, and vice versa.

[0069] In summary, the present invention utilizes the SMILES code of chemicals to avoid the application limitations of traditional prediction models, and greatly improves the prediction ability of the model, thereby achieving high-precision prediction of the inhibitory activity of chemicals on DNA gyrase A subunit; in addition, the present invention does not rely on molecular descriptors, but directly establishes a connection between the SMILES code of chemicals and the inhibitory activity to achieve rapid screening of key molecular initiation events of pollutant-induced bacterial resistance, elucidate the structural basis of pollutant-induced bacterial resistance, and has broad application prospects in the field of chemical health risk assessment.

[0070] The technical solution of the present invention is further described below by means of specific embodiments in combination with the accompanying drawings. It should be noted that the following specific embodiments are only used as examples, and the protection scope of the present invention is not limited thereto.

[0071] Example 1

[0072] See also Figure 1-Figure 2 This embodiment implements a method for predicting the inhibitory activity of chemicals on DNA gyrase A subunit based on a graph neural network model, which specifically includes the following steps:

[0073] (1) Acquisition and preprocessing of chemical structures and data on the inhibitory activity of chemicals on DNA gyrase A subunit.

[0074] First, the Uniprot ID of bacterial DNA gyrase A subunit was obtained from the BindingDB database. Based on these Uniprot IDs, the inhibitory activity data of chemical DNA gyrase A subunit were found in the BindingDB database and the ChEMBL database, which contained the chemical SMILES code and inhibitory activity data (IC 50 ) numerical value. The inhibitory activity data indicates the strength of the compound's inhibitory activity on the enzyme.

[0075] Subsequently, the chemical SMILES codes in the DNA gyrase A subunit inhibitory activity data were standardized; desolvation; charge correction; deionization and other chemical molecular structure cleaning operations were performed to avoid interference in subsequent calculations. These SMILES code cleaning operations can perform molecular structure inspection to ensure that the structure of the molecule is chemically reasonable; correct common drawing errors and standardize functional groups; hide hydrogen atoms; recalculate stereochemistry; remove covalently bonded impurities; ensure that the strongest acid group is ionized first; produce neutral molecules as much as possible; standardize tautomers; and remove salts and covalently bound metals contained in the compound. By using the half inhibitory concentration (IC50) of each chemical 50 ) to scale the inhibitory activity range.

[0076] Finally, the preprocessed chemical SMILES codes and the inhibitory activity data of the chemicals on DNA gyrase A subunit were randomly divided into training set and test set in a ratio of 8:2.

[0077] (2) Construction, training and hyperparameter search of graph neural network models.

[0078] The constructed graph neural network model includes a molecular graph embedding layer, a molecular fragment graph embedding layer, three molecular graph neural network encoders, three molecular fragment graph neural network encoders, a dual-pathway combiner and a predictor, such as Figure 2 shown.

[0079] The chemical SMILES codes in the training set are used as input to the graph neural network model.

[0080] RDKit is used in the molecular graph embedding layer to extract an a×46 atomic feature matrix, an adjacency matrix of length b (used to represent the atomic numbers of bonding atom pairs in molecules), and a b×10 chemical bond feature matrix from the molecular graph, where a is the number of atoms in the chemical, b is twice the number of chemical bonds in the chemical, 46 is the number of atomic features, and 10 is the number of chemical bond features.

[0081] RDKit is used in the molecular fragment embedding layer to extract molecular fragment features from the molecular graph. The extracted molecular fragment features include: an a×46 atomic feature matrix with a length of b f The molecular fragment adjacency matrix, b f ×10 chemical bond feature matrix of molecular fragments and a molecular fragment index list of length a (indicating the index of the molecular fragment where each atom in the chemical is located), where b f It is twice the number of chemical bonds remaining after the molecule is broken into molecular fragments.

[0082] The above-mentioned types of molecular features and molecular fragment features are respectively input into the molecular graph neural network encoder and the molecular fragment graph neural network encoder, and the node features of the molecular graph and the molecular fragment graph are updated through neighborhood attention and information aggregation. Then, the node feature module is updated through a gating mechanism to balance between retaining the original information and using new information. In the feature attention module, the global features of the graph are extracted through the maximum pooling layer and the sum pooling layer, and the attention weights are generated by the multi-layer perceptron. The features of the nodes are weighted and adjusted. Finally, the features of all nodes are aggregated into a global feature vector through the global pooling layer and the activation function, thereby obtaining the molecular graph feature sequence and the molecular fragment feature sequence.

[0083] The dual-path combiner first uses an interactive attention strategy to capture the intrinsic connection between molecules and molecular fragments. It uses a graph attention convolution layer to learn the relationship between the molecular graph feature sequence and the molecular fragment feature sequence, calculates the attention weights of different molecular fragments, and uses the ReLU activation function to perform nonlinear transformation to obtain the molecule-molecular fragment interactive feature sequence. The graph attention convolution layer uses four attention heads to perform independent calculations, which can better capture the characteristics of molecular fragments. The above molecular graph feature sequence and molecule-molecular fragment interactive feature sequence are then spliced ​​and input into the predictor.

[0084] The predictor consists of 1 fully connected layer, and the output is the predicted value s of the IC50 scaled value of the chemical for DNA gyrase A subunit out .

[0085] During training, the adaptive momentum optimizer Adam is used to update the network parameters based on the gradient, and the learning rate lr is 0.0013. In addition, in order to control the progress of convergence to the local minimum, batchsize is added to the model, and the parameter sets the batch size to 64. In order to improve the generalization ability of the model and prevent the model from overfitting, the weight decay parameter weight-decay and the random dropout regularization parameter dropout are set. The weight decay parameter weight-decay is set to 0.00001, and the random dropout regularization parameter dropout is set to 0.1. In addition, the number of layers depth of the encoder, the size of the hidden layer feature-dimension, the number of slices n-slices, and the decay rate r are used to adjust the complexity of the model. The depth, feature-dimension, n-slices and r are 4, 256, 4 and 4 respectively. A 10-fold cross-validation is used, and the training is repeated 10 times. Each time, the validation set and the training set are randomly selected according to a certain ratio for training. Each training is performed for a maximum of 500 iterations. When the validation set R of the current iteration is 2 Less than the optimal validation set R 2 And the current number of iterations is greater than the optimal validation set R 2 The training is terminated when the number of iterations is greater than 50. The prediction ability of the model when different hyperparameters are selected is evaluated based on the average value of the statistical parameters of the model on the validation set during 10 trainings. Finally, the model is retrained on the entire training set based on the optimal hyperparameter combination, and the test set R is calculated. 2 , RMSE, MSE, and MAE statistical parameters are used to characterize the prediction ability of the model.

[0086] In order to further avoid overfitting of the model and improve the generalization ability of the model, the hyperparameters used are searched and optimized within a certain range and a certain step size:

[0087] Select the learning rate parameter lr to get R corresponding to different learning rates2 , lr is selected as 0.0013;

[0088] Select the weight decay parameter weight-decay and get the R corresponding to different weight decay parameters 2 , select λ as 0.00001;

[0089] Select the batch size and get the R corresponding to different batch sizes. 2 , select batchsize as 64;

[0090] Select the random loss regularization term parameter dropout, and get the R corresponding to different random loss regularization term parameters 2 , select dropout as 0.1;

[0091] Select the number of encoder layers depth, and get the R corresponding to the number of encoder layers 2 , select depth as 4;

[0092] Select the feature-dimension of the hidden layer and get the R corresponding to the size of different hidden layers. 2 , select feature-dimension as 256;

[0093] Select the number of slices n-slices and get the R corresponding to different numbers of slices 2 , select n-slices as 4;

[0094] Select the attenuation rate r and get the R corresponding to different attenuation rates 2 , select r as 4.

[0095] (3) Prediction of DNA gyrase A subunit inhibitory activity of the chemical to be evaluated.

[0096] US8889671,32 was selected from the test set as the chemical to be evaluated to predict its inhibitory activity on DNA gyrase A subunit. The chemical structure of US8889671,32 is shown in the following formula:

[0097]

[0098] US8889671,32 is a fluorinated organic compound that is widely present in the environment. As the chemical to be tested in this embodiment, the SMILES code of US8889671,32 is queried through the PubChem molecular database, specifically "C[C@@H]1O[C@H](C)CN2c3c(cc4c(N5C[C@@H](CF)OC5=O)noc4c3Cl)CC3(C(=O)NC(=O)NC3=O)[C@@H]12", and standardization; solvent removal; charge correction; deionization and other chemical molecular structure cleaning operations are performed to avoid interference during subsequent calculations, wherein these SMILES coding cleaning operations can perform molecular structure inspection processing to ensure that the structure of the molecule is chemically reasonable; correct common drawing errors and standardize functional groups; hide hydrogen atoms; recalculate stereochemistry; remove covalently bonded impurities; ensure that the strongest acid group is ionized first; produce neutral molecules as much as possible; standardize tautomers; remove salts and covalently bound metals contained in the compound. The processed SMILES code is used as the input information of the model to calculate its p(IC 50 ) has a predicted value of 6.38.

[0099] The p(IC 50 ) is 6.20, and the predicted result is consistent with the facts.

[0100] In summary, the present invention establishes a graph neural network model, which can predict the inhibitory activity of chemicals on DNA gyrase A subunit based only on the SMILES code of the chemicals.

[0101] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit, characterized in that: The following steps are involved: S1. Obtain chemical structures and inhibitory activity data of chemicals on DNA gyrase A subunit, perform structural cleaning operations on chemical structures, perform scaling operations on inhibitory activity data to construct a data set, and randomly divide it into a training set and a test set in proportion; S2. Constructing a graph neural network model with the chemical SMILES code as input, whose output is the predicted value of the scaled value of the inhibitory activity data of the chemical on the DNA gyrase A subunit; S3. Use cross-validation to determine the optimal hyperparameters of the graph neural network model, and use the training set to train it to obtain the trained optimal graph neural network model, which is used as an inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit; S4. Input the SMILES code of the chemical to be tested in the test set or the SMILES code of the chemical to be evaluated into the inhibitory activity prediction model to obtain the predicted value of the inhibitory activity data scaling value of the chemical on DNA gyrase A subunit.

2. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 1, characterized in that: The step S1 comprises the following sub-steps: S1.

1. Obtain the chemical structure and the inhibitory activity data of the chemical on the DNA gyrase A subunit from public databases or literature; wherein the chemical structure is the chemical SMILES code, and the inhibitory activity data is used to indicate the inhibitory activity of the chemical on the DNA gyrase A subunit; S1.

2. Perform structure cleaning operations on chemical structures, specifically including: standardizing, removing solvents, charge correction, and deionizing chemical molecular structure cleaning operations on chemical SMILES codes to standardize and unify the form of chemical SMILES codes; wherein, the above-mentioned structure cleaning operations can perform molecular structure inspection processing to ensure that the structure of the molecule is chemically reasonable; correct drawing errors and standardize functional groups; hide hydrogen atoms; recalculate stereochemistry; remove covalently bonded impurities; ensure that the strongest acid group is ionized first; produce neutral molecules; standardize tautomers; remove salts and covalently bound metals contained in the compound; S1.3, performing a scaling operation on the numerical range of the data on the inhibitory activity of the chemical on the DNA gyrase A subunit by taking the negative logarithm of the original numerical value of the data on the inhibitory activity of the chemical on the DNA gyrase A subunit; S1.4, constructing a data set based on the chemical SMILES code after the structure cleaning operation in step S1.2 and the inhibitory activity data after the scaling operation in step S1.3; S1.

5. Randomly divide the dataset into training set and test set in proportion.

3. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 1, characterized in that: The graph neural network model includes a molecular graph embedding layer, a molecular fragment graph embedding layer, multiple molecular graph neural network encoders, multiple molecular fragment graph neural network encoders, a dual-path combiner and a predictor, wherein the molecular graph neural network encoder and the molecular fragment graph neural network encoder are both composed of a neighborhood attention module, an information aggregation module, a gating mechanism to update node feature module and a feature attention module, and the feature attention module includes a maximum pooling layer, a sum pooling layer, a multi-layer perceptron global pooling layer and an activation function; the dual-path combiner selects a graph convolutional attention layer, and the graph convolutional attention layer includes multiple attention heads and a ReLU activation function; the predictor includes n l Fully connected layer.

4. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 3, characterized in that: The molecular graph embedding layer calculates the molecular graph of the chemical according to the chemical SMILES code by using RDKit to extract the molecular graph features from the molecular graph; wherein the molecular graph features include a×d at The atomic feature matrix, the adjacency matrix of length b, and the b×d bt The chemical bond characteristic matrix is: a is the number of atoms in the chemical, b is twice the number of chemical bonds in the chemical, and d is at is the number of atomic features, d bt is the number of chemical bond features, and the adjacency matrix is ​​used to represent the atomic numbers of the bonding atom pairs in the molecular graph; The molecular fragment graph embedding layer calculates the molecular graph of the chemical according to the SMILES code of the chemical by using RDKit, and extracts the molecular fragment features from the molecular graph based on the BRICS principle; wherein the molecular fragment features include a×d at The atomic feature matrix of length b f The molecular fragment adjacency matrix, b f ×d bt The chemical bond feature matrix of the molecular fragments and the molecular fragment index list of length a, where b f It is twice the number of chemical bonds remaining after the molecule is broken into molecular fragments. The molecular fragment index list is used to indicate the index of the molecular fragment where each atom in the chemical is located; The various molecular graph features and molecular fragment features extracted from the molecular graph embedding layer and the molecular fragment graph embedding layer are first transformed into a length of d h Then, each type of feature vector is input into the molecular graph neural network encoder and the molecular fragment graph neural network encoder to obtain the molecular graph feature sequence and the molecular fragment feature sequence, respectively. h The length of is the same as the size of the hidden layer; wherein, in the molecular graph neural network encoder and the molecular fragment graph neural network encoder, the molecular graph neural network encoder and the molecular fragment graph neural network encoder update the node features of the molecular graph and the molecular fragment graph through their respective neighborhood attention modules and information aggregation modules, and then update the node feature module through a gating mechanism to balance between retaining the original information and using the new information, and in the feature attention module, the global features of the molecular graph are extracted through the maximum pooling layer and the sum pooling layer, and the attention weights are generated by the multi-layer perceptron, and the features of the nodes are weighted and adjusted, and finally the features of all nodes are aggregated into a global feature vector through the global pooling layer and the activation function, so as to obtain the molecular graph feature sequence and the molecular fragment feature sequence; The molecular graph feature sequence and the molecular fragment feature sequence are input into the dual-path combiner. The dual-path combiner uses the graph convolution attention layer to learn the relationship between the molecular graph feature sequence and the molecular fragment feature sequence. Specifically, it adopts n h Each attention head independently calculates the attention weights of different molecular fragments, and performs nonlinear transformation through the ReLU activation function to obtain the molecule-molecular fragment interaction feature sequence; Then, the molecule-molecule fragment interaction feature sequence and the molecular graph feature sequence output by the molecular graph neural network encoder are concatenated and input into the predictor. l After the fully connected layer, the predicted values ​​of the scaled values ​​of the inhibitory activity data of the chemicals on the DNA gyrase A subunit are output.

5. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 1, characterized in that: The step S3 comprises the following sub-steps: S3.

1. Preset multiple sets of different model hyperparameters as {lr, weight-decay, batchsize, dropout, depth, feature-dimension, n-slices, r}, where lr is the learning rate, weight-decay is the weight decay parameter, batchsize is the number of samples selected for one training, dropout is the random dropout regularization term parameter, depth is the number of encoder layers, feature-dimension is the size of the hidden layer, n-slices is the number of slices, r is the decay rate, lr is used to control the progress of convergence to the local minimum, weight-decay and dropout are used to prevent model overfitting, depth, feature-dimension, n-slices and r are used to adjust the model complexity, and the size of batchsize affects the degree and speed of model optimization; S3.

2. For each set of model hyperparameters, use the data in the training set to train the graph neural network model. f The fold-cross validation was used to evaluate the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, and the trained graph neural network model corresponding to the optimal model hyperparameters was obtained, which was used as the inhibitory activity prediction model to predict the inhibitory activity data scaling value of chemicals on DNA gyrase A subunit.

6. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 5, characterized in that: The statistical parameter is one of a coefficient of determination, a root mean square error, a mean square error, and a mean absolute error.

7. The method for predicting the inhibitory activity of a chemical on DNA gyrase A subunit according to claim 5, characterized in that: For each set of model hyperparameters, the graph neural network model is trained using the data in the training set. f The fold cross validation evaluates the graph neural network model corresponding to the set of model hyperparameters based on statistical parameters to determine the optimal model hyperparameters, including: S3.2.

1. Train the graph neural network model based on a set of preset model hyperparameters, using n f The data in the training set is randomly divided into n f subsets, each time one of the subsets is used as the validation set, and the remaining n f -1 subsets to train the graph neural network model. During the training process, the statistical parameters are calculated based on the predicted values ​​of the inhibitory activity data scaling values ​​of the chemicals on the DNA gyrase A subunit output by the graph neural network model and the corresponding true values ​​in the training set. The parameters of the graph neural network model are adjusted with the optimal statistical parameters as the optimization goal, and then the graph neural network model is verified using the data in the validation set to calculate the corresponding statistical parameters. The above process is repeated for training n f times, and finally get n f Statistical parameters and calculate n f The average value of the statistical parameters; S3.2.

2. Repeat step S3.2.1 to obtain the statistical parameter averages corresponding to all groups of model hyperparameters, and select a group of model hyperparameters corresponding to the optimal statistical parameter averages as the optimal model hyperparameters.

Images