阅读说明：本技术 药物与靶蛋白反应的效果识别方法及相关装置、设备 (Method for identifying reaction effect of drug and target protein and related device and equipment ) 是由 毕研广 胡志强 于 2021-06-18 设计创作，主要内容包括：本申请公开了一种药物与靶蛋白反应的效果识别方法及相关装置、设备,其中,药物与靶蛋白反应的效果识别方法包括：获取药物的分子结构图与靶蛋白的氨基酸序列；分别对分子结构图与氨基酸序列进行特征提取,以获得药物的特征表示与靶蛋白的特征表示；将药物的特征表示和靶蛋白的特征表示进行拼接,得到药物和靶蛋白的拼接特征表示；基于拼接特征表示,确定药物与靶蛋白的反应效果类型。上述方案,能够自适应地学习药物分子和蛋白质更深层的特征,优化药物与靶蛋白的反应预测效果。(The application discloses a method for identifying the reaction effect of a drug and a target protein, and a related device and equipment, wherein the method for identifying the reaction effect of the drug and the target protein comprises the following steps: acquiring a molecular structure diagram of a medicine and an amino acid sequence of a target protein; respectively carrying out feature extraction on the molecular structure diagram and the amino acid sequence so as to obtain feature representation of the medicine and feature representation of the target protein; splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein; and determining the type of the reaction effect of the drug and the target protein based on the splicing characteristic representation. The scheme can adaptively learn deeper characteristics of the drug molecules and the proteins and optimize the reaction prediction effect of the drug and the target protein.)

1. A method for identifying an effect of a drug on a target protein, comprising:

acquiring a molecular structure diagram of the medicine and an amino acid sequence of the target protein;

respectively carrying out feature extraction on the molecular structure diagram and the amino acid sequence so as to obtain a feature representation of the medicine and a feature representation of the target protein;

splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein;

determining a type of effect of the drug's reaction with the target protein based on the splicing signature representation.

2. The method of claim 1, wherein the step of performing feature extraction on the molecular structure diagram and the amino acid sequence to obtain the feature representation of the drug and the feature representation of the target protein comprises:

extracting feature representations of a plurality of atomic nodes of the molecular structure using a deep learning network;

and performing maximum pooling on the feature representations of the atomic nodes to obtain the feature representation of the medicine.

3. The method for identifying the effect of a drug on a target protein according to claim 2, wherein the step of extracting the feature representation of the plurality of atomic nodes of the molecular structure using the deep learning network comprises:

respectively extracting initial feature representation of each atomic node in the molecular structure and adjacency relation feature representation of the atomic node and adjacent atomic nodes by utilizing a multilayer graph convolution structure of the deep learning network;

and fusing the initial feature representation of the atomic node and the adjacency feature representation, and determining a fusion result as the feature representation of the atomic node.

4. The method according to claim 3, wherein the step of fusing the initial characteristic representation of the atomic node and the adjacency characteristic representation and determining the fusion result as the characteristic representation of the atomic node comprises:

taking the fusion result as the initial characteristic representation of the atomic node, and circularly executing the step of extracting the characteristic representations of the plurality of atomic nodes of the molecular structure;

and under the condition that the number of times of executing the step of extracting the characteristic representations of the plurality of atomic nodes of the molecular structure reaches a preset value, determining the last fusion result as the characteristic representation of the atomic node.

5. The method for identifying an effect of a drug on a target protein according to any one of claims 1 to 4, wherein the step of performing feature extraction on the molecular structure diagram and the amino acid sequence to obtain a feature representation of the drug and a feature representation of the target protein comprises:

constructing a non-directional structure diagram based on the molecular structure diagram;

extracting feature representation of the medicine based on the undirected graph.

6. The method for identifying the effect of a reaction between a drug and a target protein according to any one of claims 1 to 4, wherein the step of obtaining the molecular structure diagram of the drug and the amino acid sequence of the target protein specifically comprises:

obtaining the SMILES molecular formula of the drug;

converting the SMILES molecular formula of the drug into a molecular structural diagram of the drug.

7. The method for identifying an effect of a drug on a target protein according to any one of claims 1 to 6, wherein the step of performing feature extraction on the molecular structure diagram and the amino acid sequence to obtain a feature representation of the drug and a feature representation of the target protein comprises:

obtaining a plurality of amino acid codes of the amino acid sequence;

obtaining an initial characterization of the amino acid sequence based on the plurality of amino acid codes;

extracting a hidden feature representation of the amino acid sequence by a recurrent neural network;

and fusing the initial feature representation and the hidden feature representation to obtain the feature representation of the target protein.

8. The method of claim 7, wherein the step of obtaining an initial characterization of the amino acid sequence based on the plurality of amino acid codes comprises:

extracting feature representations of a plurality of amino acids corresponding to the amino acid sequence by utilizing a multi-layer one-dimensional convolution network based on the amino acid codes;

performing maximum pooling treatment on the feature representations of the plurality of amino acids to obtain an initial feature representation of the target protein; and

the step of extracting a hidden-feature representation of the amino acid sequence by a recurrent neural network comprises:

extracting, by the recurrent neural network, a hidden feature representation of the feature representations of the plurality of amino acids.

9. The method for identifying the effect of a drug on a target protein according to claim 8, wherein the step of extracting the hidden feature representation of the plurality of amino acids by the recurrent neural network comprises:

and sequentially extracting hidden feature representations of the plurality of amino acids along the head-to-tail direction and the tail-to-head direction of the feature representations of the plurality of amino acids through the recurrent neural network respectively.

10. The method for identifying an effect of a drug on a target protein according to any one of claims 1 to 9, wherein the step of concatenating the characterization representation of the drug and the characterization representation of the target protein to obtain a concatenated characterization representation of the drug and the target protein comprises:

and performing head-to-tail splicing on the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein.

11. The method for identifying the effect of a drug on a target protein according to any one of claims 1 to 10, wherein the step of determining the type of the effect of the drug on the target protein using the splicing feature representation of the drug and the target protein comprises:

and carrying out prediction classification on the splicing characteristic representation by utilizing a multilayer neural network to obtain the reaction effect type of the drug and the target protein.

12. An effect recognition apparatus for a drug to react with a target protein, comprising: the device comprises an acquisition module, a feature extraction module, a splicing module and an identification module;

the acquisition module is used for acquiring a molecular structure diagram of the medicine and an amino acid sequence of the target protein;

the characteristic extraction module is used for respectively extracting the characteristics of the molecular structure diagram and the amino acid sequence so as to obtain the characteristic representation of the medicine and the characteristic representation of the target protein;

the splicing module is used for splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein;

the identification module is used for determining the type of the reaction effect of the drug and the target protein based on the splicing feature representation.

13. An electronic device, comprising a memory and a processor coupled to each other, wherein the processor is configured to execute program instructions stored in the memory to implement the method for identifying an effect of a drug reacting with a target protein according to any one of claims 1 to 11.

14. A computer-readable storage medium having stored thereon program instructions, which when executed by a processor, implement the method for identifying an effect of a drug reacting with a target protein according to any one of claims 1 to 11.

Technical Field

The application relates to the technical field of deep learning, in particular to a method for identifying the reaction effect of a drug and a target protein and a related device and equipment.

Background

At present, a certain guarantee is provided for the health of people by a variety of medicines in the market. One typically achieves efficacy by orally administering and/or injecting the drug. Specifically, the drug interacts with specific proteins after entering the human body, thereby achieving a certain therapeutic effect. Therefore, in the stage of developing new drugs, the prediction of the reaction between the drug and the target protein is the first task, which saves the experimental cost and a large amount of manpower and material resources to a certain extent and ensures the reliability of drug development to a certain extent.

The existing reaction prediction means usually uses characteristic engineering designed by experts, such as molecular fingerprints, to predict the reaction between the drug and the target protein, but the expression capability of the method is limited, and deeper features of the drug molecules and the protein cannot be adaptively learned.

In practical applications, the amino acid sequences of drug molecules and constituent proteins are not fully equivalent to natural language, thus limiting the predicted effect of drug target protein reactions. How to model these abstract sequences into learnable mathematical models remains a challenge of current research.

Disclosure of Invention

The application at least provides a method for identifying the reaction effect of a drug and a target protein, and a related device and equipment.

The first aspect of the present application provides a method for identifying an effect of a drug reacting with a target protein, comprising: obtaining a molecular structure diagram of the medicine and an amino acid sequence of a target protein; respectively carrying out feature extraction on the molecular structure diagram and the amino acid sequence so as to obtain feature representation of the medicine and feature representation of the target protein; splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein; and determining the type of the reaction effect of the drug and the target protein based on the splicing characteristic representation.

Therefore, the method obtains the molecular structure diagram of the medicine and the amino acid sequence of the target protein, and then respectively performs feature extraction on the molecular structure diagram and the amino acid sequence to obtain feature representation of the medicine and feature representation of the target protein; splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein; based on the splicing characteristic expression, the reaction effect type of the drug and the target protein is determined, so that the characteristic splicing of the drug molecules with a two-dimensional space structure and the one-dimensional amino acid sequence can be realized, and deeper characteristics of the drug molecules and the protein can be learned respectively, so that a better prediction effect can be obtained.

Wherein, the step of respectively extracting the characteristics of the molecular structure diagram and the amino acid sequence to obtain the characteristic representation of the medicine and the characteristic representation of the target protein comprises the following steps: extracting feature representations of a plurality of atomic nodes of a molecular structure by using a deep learning network; and performing maximum pooling on the feature representations of the atomic nodes to obtain the feature representation of the medicine.

Therefore, the deep learning network is used for extracting the feature representation of a plurality of atomic nodes of the molecular structure, so that the feature representation of the molecular structure with the two-dimensional space structure is one-dimensional, the subsequent feature splicing is facilitated, and the efficiency of response prediction is improved. And performing maximum pooling treatment on the feature representations of the atomic nodes to obtain the feature representation of the medicine, so as to reduce the number of parameters in the feature representation and keep important information.

The method for extracting the feature representation of the plurality of atomic nodes of the molecular structure by utilizing the deep learning network comprises the following steps of: respectively extracting initial feature representation of each atomic node in the molecular structure and adjacency relation feature representation of the atomic node and adjacent atomic nodes by utilizing a multilayer graph convolution structure of a deep learning network; and fusing the initial characteristic representation and the adjacency characteristic representation of the atomic node, and determining a fusion result as the characteristic representation of the atomic node.

Therefore, the multi-layer graph convolution structure extraction is performed based on the initial feature representation of each atomic node and the adjacency feature representation of the atomic node and the adjacent atomic node, so as to improve the accuracy of feature extraction. And fusing the initial characteristic representation and the adjacency characteristic representation of the atomic node to obtain the characteristic representation of the atomic node, so that the characteristic representation of the atomic node can better reflect the characteristic value of the whole drug molecule.

The method comprises the following steps of fusing the initial feature representation and the adjacency feature representation of the atomic node, and determining a fusion result as the feature representation of the atomic node: taking the fusion result as the initial characteristic representation of the atomic node, and circularly executing the step of extracting the characteristic representation of a plurality of atomic nodes of the molecular structure; and under the condition that the number of times of executing the step of extracting the characteristic representations of the plurality of atomic nodes of the molecular structure reaches a preset value, determining the last fusion result as the characteristic representation of the atomic node.

Therefore, the steps of extracting the feature representation of a plurality of atomic nodes of the molecular structure are executed in a circulating mode to adapt to the drug molecules with different complexity, so that the feature representation of the atomic nodes is extracted by adopting execution times with different preset values aiming at different drug molecular structures, and the flexible operation is realized.

Wherein, the step of respectively extracting the characteristics of the molecular structure diagram and the amino acid sequence to obtain the characteristic representation of the medicine and the characteristic representation of the target protein comprises the following steps: constructing a directed structure diagram based on the molecular structure diagram, and extracting characteristic representation of the medicine based on the directed structure diagram.

Therefore, the input of the multilayer graph convolution structure which is constructed by constructing the undirected structure graph and is adapted to the deep learning network based on the molecular structure graph removes redundant information in the molecular structure graph, only retains information such as atomic nodes, connection relations of the atomic nodes and the connection relations of the atomic nodes, and the like, and improves the efficiency of characteristic representation extraction of the medicine.

The method specifically comprises the following steps of obtaining the molecular structure diagram of the medicine and the amino acid sequence of the target protein: obtaining the SMILES molecular formula of the medicine, and converting the SMILES molecular formula of the medicine into a molecular structure diagram of the medicine.

Thus, the molecular structure diagram of a drug is converted from the SMILES molecular formula of the drug molecule. The SMILES molecular formula of the medicine has uniqueness, so that the corresponding relation between the molecular structure diagram and the medicine molecules is ensured, and the accuracy of a reaction prediction result is improved.

Wherein, the step of respectively extracting the characteristics of the molecular structure diagram and the amino acid sequence to obtain the characteristic representation of the medicine and the characteristic representation of the target protein comprises the following steps: obtaining a plurality of amino acid codes of the amino acid sequence; obtaining an initial characterization of the amino acid sequence based on the plurality of amino acid codes; extracting a hidden feature representation of the amino acid sequence by a recurrent neural network; and fusing the initial feature representation and the hidden feature representation to obtain the feature representation of the target protein.

Thus, an initial characterization of the amino acid sequence is derived based on a plurality of amino acid codes; and extracting hidden feature representations of the amino acid sequences through a recurrent neural network. Fusing the initial feature representation and the hidden feature representation to obtain a feature representation of the target protein. And deeper and more comprehensive characteristic representation of the amino acid sequence is obtained, and the reliability of the reaction prediction result is improved.

Wherein the step of deriving an initial characterization of the amino acid sequence based on the plurality of amino acid codes comprises: extracting feature representations of a plurality of amino acids corresponding to the amino acid sequence by utilizing a multi-layer one-dimensional convolution network based on the amino acid coding; performing maximum pooling treatment on the feature representations of the plurality of amino acids to obtain an initial feature representation of the target protein; and the step of extracting the hidden-feature representation of the amino acid sequence by the recurrent neural network comprises: hidden feature representations of a plurality of amino acids are extracted by a recurrent neural network.

Therefore, the multi-layer one-dimensional convolution network is used for extracting the feature expressions of a plurality of amino acids corresponding to the amino acid sequence, and the feature expressions of the plurality of amino acids are subjected to maximum pooling processing to obtain the initial feature expression of the target protein, so that the number of parameters in the feature expression is reduced, important information is kept, and the practicability of the initial feature expression is improved. Hidden feature representations of a plurality of amino acids are extracted by a recurrent neural network. So as to further obtain the overall characteristic representation of a plurality of amino acids, and the initial characteristic representation of the target protein is more comprehensive and reliable.

Wherein the step of extracting hidden feature representations of the plurality of amino acids by the recurrent neural network comprises: and sequentially extracting hidden feature representations of the plurality of amino acids along the head-tail direction and the tail-head direction of the feature representations of the plurality of amino acids through a recurrent neural network respectively.

Therefore, the hidden feature representation of the feature representation of a plurality of amino acids is obtained by extracting the feature representation of the plurality of amino acids in two directions, so that the hidden feature representation is more comprehensive and reliable.

Splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein, wherein the step of splicing the characteristic representation of the drug and the characteristic representation of the target protein comprises the following steps: and splicing the characteristic representation of the drug and the characteristic representation of the target protein end to obtain the spliced characteristic representation of the drug and the target protein.

Therefore, the characteristic representation of the drug and the characteristic representation of the target protein are spliced end to adapt to the input form of the fully-connected layer in the multilayer neural network.

Wherein, the step of determining the type of the reaction effect of the drug and the target protein by using the splicing characteristic representation of the drug and the target protein comprises the following steps: and (4) performing prediction classification on the splicing characteristic representation by utilizing a multilayer neural network to obtain the reaction effect type of the drug and the target protein.

Therefore, the splicing characteristic representation is predicted and classified by utilizing the multilayer neural network so as to comprehensively predict the reaction between the drug molecules and the target protein, minimize errors and realize the accuracy of the reaction effect type.

In a second aspect, the present application provides an effect recognition apparatus for a drug reacting with a target protein, comprising: the device comprises an acquisition module, a feature extraction module, a splicing module and an identification module; the acquisition module is used for acquiring a molecular structure diagram of the medicine and an amino acid sequence of a target protein; the characteristic extraction module is used for respectively carrying out characteristic extraction on the molecular structure diagram and the amino acid sequence so as to obtain characteristic representation of the medicine and characteristic representation of the target protein; the splicing module is used for splicing the characteristic representation of the medicine and the characteristic representation of the target protein to obtain the splicing characteristic representation of the medicine and the target protein; the recognition module is used for determining the type of the reaction effect of the drug and the target protein based on the splicing characteristic representation.

A third aspect of the present application provides an electronic device, comprising a memory and a processor coupled to each other, wherein the processor is configured to execute program instructions stored in the memory to implement the method for identifying an effect of a drug reacting with a target protein in the first aspect.

A fourth aspect of the present application provides a computer-readable storage medium having stored thereon program instructions that, when executed by a processor, implement the method for identifying an effect of a drug reacting with a target protein in the first aspect described above.

According to the scheme, the SMILES expression is reconstructed into the molecular structure diagram and then further reconstructed into the undirected structure diagram, the feature representation of a plurality of atomic nodes of the molecular structure is extracted by utilizing the multi-layer diagram convolution structure so as to convert two-dimensional data into one-dimensional data, and then the feature representation of the atomic nodes is subjected to maximum pooling treatment to obtain the feature representation of the medicine, so that the feature representation of the medicine is more consistent with the structural characteristics of medicine molecules. Meanwhile, the method extracts the amino acid local characteristics through one-dimensional convolution and pooling, extracts the overall characteristics of the amino acid sequence through a circulating neural network, fuses the local characteristics and the overall characteristics to obtain the characteristic representation of the target protein, and at the moment, the characteristic representation of the target protein is also one-dimensional data, splices the characteristic representation of the drug and the characteristic representation of the target protein and predicts and classifies the characteristic representations to obtain the reaction effect type of the drug and the target protein. The whole prediction scheme can adaptively learn deeper characteristics of the drug molecules and the protein, realize the characteristic extraction of the drug molecules and the protein as a whole, increase the comprehensive degree and the reliability of the characteristic extraction of the drug molecules and the protein as a whole, improve the efficiency of the whole reaction prediction process, and enable the reaction effect type of the drug and the target protein to be more accurate.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the application.

FIG. 1 is a schematic flow chart of an embodiment of a method for identifying the effect of a reaction between a drug and a target protein according to the present application;

FIG. 2 is a schematic flow chart of another embodiment of the method for identifying the effect of the reaction of a drug with a target protein according to the present application;

FIG. 3 is a schematic diagram of one embodiment of a directed structure diagram of the present application;

FIG. 4 is a schematic diagram of one embodiment of a maximum pooling process of the present application;

FIG. 5 is a schematic diagram of one embodiment of a fully-connected layer feature connection of the present application;

FIG. 6 is a schematic block diagram of an embodiment of the effect recognition device of the present application in which a drug reacts with a target protein;

FIG. 7 is a block diagram of an embodiment of an electronic device of the present application;

FIG. 8 is a block diagram of an embodiment of a computer-readable storage medium of the present application.

Detailed Description

The following describes in detail the embodiments of the present application with reference to the drawings attached hereto.

In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, interfaces, techniques, etc. in order to provide a thorough understanding of the present application.

The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship. Further, the term "plurality" herein means two or more than two. In addition, the term "at least one" herein means any one of a plurality or any combination of at least two of a plurality, for example, including at least one of A, B, C, and may mean including any one or more elements selected from the group consisting of A, B and C.

Referring to FIG. 1, FIG. 1 is a schematic flow chart of an embodiment of a method for identifying an effect of a reaction between a drug and a target protein according to the present application. Specifically, the method may include the steps of:

step S11: obtaining the molecular structure diagram of the medicine and the amino acid sequence of the target protein.

Protein interactions within an organism form a complex network. If a protein contributes to a disease, it is necessary to block the interaction of the protein with other molecules, thereby achieving a cure or disease alleviation effect. If the protein has a disease-suppressing effect, it is necessary to activate the protein by some means. Such a protein is the target protein.

Target proteins are protein molecules that are closely related to various diseases and that can affect or control diseases. The effect of treating diseases can be achieved by combining drug molecules with target proteins. Therefore, in the process of developing new drugs, predicting the reaction of drug molecules and target proteins is of great significance in judging whether the drugs can produce certain influence or control effect on target diseases.

In one implementation scenario, the molecular structure diagram of the drug may be transformed and reconstructed from the SMILES molecular formula of the drug. Wherein SMILES (simplified Molecular Input Line entry System) is a linear symbol for inputting and representing Molecular reactions, and is also an ASCII code. The molecular structure of the SMILES molecular formula can be reconstructed into a molecular structure chart of the drug molecule by a conversion tool, wherein the molecular structure chart comprises information such as atom types and connection relations among atoms.

The protein is composed of amino acids, and the amino acid types of human proteins are twenty types, and different numbers and types of amino acids constitute different proteins. The amino acid sequence in this step includes information such as the type and positional relationship of amino acids in the protein, and may be referred to as the primary amino acid sequence structure of the protein.

Step S12: and respectively carrying out feature extraction on the molecular structure diagram and the amino acid sequence so as to obtain feature representation of the medicine and feature representation of the target protein.

In an implementation scenario, in order to improve convenience of feature extraction, a recognition network may be trained in advance, and the recognition network may include a feature extraction network for feature extraction of drug molecules, so that feature extraction may be performed on a drug molecule structure diagram for multiple times by using the feature extraction network to obtain feature representation of a drug.

In an implementation scenario, in order to accelerate feature extraction of the molecular structure, before the feature extraction of the molecular structure, the molecular structure may be further preprocessed, and specifically, the preprocessing may include classifying atoms in the molecular structure and their connection relationships and/or estimating complexity, so as to control an extraction degree in a feature extraction process and accelerate efficiency of feature extraction of the molecular structure. The classification refers to dividing based on the connection relationship between atoms and other atoms, for example: if the A atom has a connection relation with 3 other atoms, the B atom has a connection relation with 2 other atoms and the C atom has a connection relation with 2 other atoms, the B atom and the C atom are classified into one type, and the A atom is a single type. And the complexity estimation means that the connection relation among atoms is estimated based on the valence rule of each atom type to obtain the connection relation and the number among the atoms, so that the complexity estimation of the atoms is finished. For example, in general, the oxygen atom can be monovalent negative and divalent negative, and can be attached to one or two other atoms, such as two hydrogen atoms, based on the valencies of the other atoms. Through classification and/or complexity estimation of atoms in the molecular structure and the connection relation thereof, the related characteristics of the molecular structure can be known in advance before characteristic extraction, so that the extraction degree in the characteristic extraction process is controlled, and the efficiency of the characteristic extraction of the molecular structure is accelerated.

In one implementation, the primary amino acid sequence structure of the protein is characterized to extract a characterization of the target protein.

Step S13: and splicing the characteristic representation of the drug and the characteristic representation of the target protein to obtain spliced characteristic representation of the drug and the target protein.

And performing characteristic splicing on the characteristic representation of the medicine obtained in the last step and the characteristic representation of the target protein to obtain spliced characteristic representation of the medicine and the target protein.

Step S14: and determining the type of the reaction effect of the drug and the target protein based on the splicing characteristic representation.

Whether the drug is effective on the target protein is determined based on the splicing characteristic representation of the drug and the target protein, namely, the type of reaction effect of the drug and the target protein is obtained. Here, the prediction of the reaction of a drug with a target protein is essentially a binary task, which means that there are two classes in the classification task, and when the input is represented by a feature vector x, the output is represented by y ═ 0 or 1.

In one implementation scenario, the reaction of the drug with the target protein is predicted, and the final output result is one of y-0 or 1, and when y-0, it indicates that the drug in the current reaction prediction is not effective for the target protein. When y is 1, it indicates that the drug in this prediction of response is effective against the target protein.

According to the scheme, after the molecular structure diagram of the drug and the amino acid sequence of the target protein are obtained, feature extraction is respectively carried out on the molecular structure diagram of the drug and the amino acid sequence of the target protein, the feature representation of the drug and the feature representation of the target protein are extracted, and finally the reaction effect type of the drug and the target protein is determined by utilizing the splicing feature representation of the drug and the target protein, so that deeper features of the drug molecule and the protein can be learned in a self-adaptive manner, the feature extraction of the whole drug molecule and the protein is realized, and the reaction prediction effect type of the drug and the target protein is obtained.

Referring to FIG. 2, FIG. 2 is a schematic flow chart of another embodiment of the method for identifying the effect of a drug reacting with a target protein according to the present application. Specifically, the method may include the steps of:

s21: obtaining the molecular structure diagram of the medicine and the amino acid sequence of the target protein.

And acquiring the SMILES molecular formula of the medicine, and converting the SMILES molecular formula of the medicine into a molecular structure diagram of the medicine. In other implementation scenarios, other encoding form molecular formulas capable of performing the conversion of the molecular structure diagram through a conventional conversion tool may also be obtained, and the subsequent feature extraction step is performed after the molecular formulas of other encoding forms of the drug are converted into the molecular structure diagram, which is not limited in this step.

Proteins are polymers formed by the linkage of twenty different amino acids. In a specific embodiment, an amino acid sequence including information on the type and positional relationship of amino acids is obtained. I.e., the primary amino acid sequence structure of the protein.

S22: and extracting the feature representation of a plurality of atomic nodes of the molecular structure by using a deep learning network, and performing maximum pooling treatment on the feature representation of the atomic nodes to obtain the feature representation of the medicine.

The molecular structure diagram is further constructed into a directed structure diagram by utilizing the chemical bond connection condition in the molecular structure diagram so as to extract the characteristic representation of the medicine based on the directed structure diagram. The undirected structure chart is a data structure chart which is constructed by using an atom node, atom connection relation and atom connection mode on the basis of a molecular structure chart, and the data structure chart comprises node characteristics and adjacent relation characteristic representation.

Referring to fig. 3, fig. 3 is a schematic diagram of an embodiment of a non-directional structural diagram of the present application.

The undirected structure 10 of the present embodiment comprises a plurality of nodes 11 and a plurality of connecting edges 12. There is no direction on connecting edges 12, and connecting edges 12 in undirected structure 10 are all unordered pairs of multiple nodes 11. Each node 11 is connected to other nodes 11 by connecting edges 12.

The undirected structure is constructed by the molecular structure of the drug, the characteristics of each atom in the molecular structure of the drug and the characteristics of the adjacency relation among the atoms are reserved, redundant information in the molecular structure is deleted, the input form of the multilayer graph convolution structure of the deep learning network is adapted, and the efficiency of the characteristic representation extraction of the drug is improved.

The deep learning network in this step may include various types such as a convolutional neural network, a cyclic neural network, or a transform network model. In this embodiment, the initial feature representation of each atomic node in the undirected structure diagram and the adjacency relation feature representation of the atomic node and the adjacent atomic node may be extracted through the multi-layer graph convolution structure of the convolutional neural network. And fusing the initial characteristic representation and the adjacency characteristic representation of the atomic node, and determining a fusion result as the characteristic representation of the atomic node.

In the step, a multidirectional structure diagram with a two-dimensional space structure is subjected to feature extraction through a multilayer diagram convolution structure, and finally feature representation of one-dimensional sequence data-atomic nodes is obtained. The step realizes that abstract two-dimensional data is converted into learnable one-dimensional data, so that the method is suitable for a natural language processing method, and the flexibility of reaction prediction of the drug target protein is improved.

The deep learning network is a general name of a type of pattern analysis method, and mainly relates to three types of methods in terms of specific research contents: (1) a neural network system based on convolution operations, i.e. a Convolutional Neural Network (CNN). (2) self-Coding neural networks based on multi-layer neurons include self-Coding (Auto encoder) and Sparse Coding (Sparse Coding). (3) And pre-training in a multilayer self-coding neural network mode, and further optimizing a Deep Belief Network (DBN) of the neural network weight by combining the identification information.

The multilayer graph convolution structure belongs to a neural network system based on convolution operation, is used for converting a graph of a non-European space into an European space, and provides a convolution kernel capable of processing a variable-length neighbor node so as to extract features on the graph. The graph is a graph formed by a plurality of nodes and edges connecting the two nodes and is used for depicting the relationship between different nodes.

The fusion method of the initial feature representation and the adjacency feature representation of the atomic node may perform fusion in a manner of direct splicing or weighted fusion, and in this embodiment, each atomic node and its adjacent node have a certain weight and contribution. Therefore, feature extraction of preset iteration times can be performed on each atomic node through a multilayer graph convolution structure, and weighting fusion is performed on each atomic node and the feature values of adjacent nodes to obtain new nodes correspondingly. The structure between the new nodes is the same as that between the corresponding atomic nodes before weighting, but the eigenvalues of the new nodes change. After obtaining the new node and the new adjacent node characteristics, the multilayer graph convolution structure can be recycled to carry out weighted fusion on the new node and the new adjacent node characteristics so as to enrich the characteristics of each atomic node. And simultaneously judging whether the times of the steps of circularly executing and extracting the feature representation of the plurality of atomic nodes of the molecular structure reach a preset value, namely presetting iteration times, and if the times reach the preset value, determining the last fusion result as the feature representation of the atomic nodes. If the preset value is not reached, the step of extracting the characteristic representation of the plurality of atomic nodes of the molecular structure is executed in a loop. The specific value of the preset value may be set based on actual conditions, and is not limited herein.

The eigenvalues of each atomic node consist of a string of vectors, initialized with 0's and 1's. After the multilayer graph convolution structure is subjected to weighted fusion of preset iteration times, the representation state of each atomic node changes, and the characteristic value is richer. After the characteristics of the molecular structure are extracted by the multilayer graph convolution structure with preset iteration times, the characteristic propagation and characteristic multiplexing of the molecular structure are enhanced, and the reliability of the characteristic representation of a plurality of atomic nodes is enhanced.

In a specific implementation scenario, the number of preset iterations is determined by the complexity of the structure of the drug molecule. The complexity of the structure of a drug molecule is proportional to the number of iterations of the multi-layer graph convolution structure. The number of iterations of the multi-layer graph convolution structure may be set before each response prediction, for example, 2, 3, or 5, and is not limited herein.

After the feature representations of the plurality of atomic nodes are obtained, maximal pooling is performed on the feature representations to obtain the feature representation of the drug. The characteristics of the drug finally obtained in this step are represented as a one-dimensional data column. The pooling process reduces the number of parameters in the feature representation, but retains important information. The maximum pooling process is to take a point with the maximum value in the local acceptance domain in the whole data category. In a specific implementation scenario, if each atomic node is a length-10 vector, n length-10 vectors are combined together to form a 10 × n matrix. When the matrix is processed by maximum pooling, each column of the matrix takes the maximum eigenvalue to obtain the overall eigenvalue of the matrix.

Referring to fig. 4, fig. 4 is a schematic diagram of an embodiment of a maximum pooling process according to the present application.

The first table 21 is a matrix of 10 x 7. When performing maximum pooling for each column of the first table 21, one follows to choose the maximum value in each column as the pooling result for that column, and finally generates a1 × 7 matrix-the second table 22. Wherein the data of each column in the second table 22 is derived from the corresponding column in the first table 21. For example, "16" in the first column of the second table 22 is the result of maximum pooling of the data in the first column of the first table 21, i.e., selecting the maximum value "16" of the column.

S23: acquiring a plurality of amino acid codes of an amino acid sequence, acquiring initial feature representation of the amino acid sequence based on the amino acid codes, extracting hidden feature representation of the amino acid sequence through a circulating neural network, and fusing the initial feature representation and the hidden feature representation to obtain feature representation of the target protein.

After the primary amino acid sequence structure of the target protein is obtained, the position of the target protein is coded to obtain a plurality of amino acid codes of the amino acid sequence. The position code is used for coding the amino acid type and position in the primary amino acid sequence structure so as to obtain a plurality of amino acid codes, and the plurality of amino acid codes comprise the information of the type and position of each amino acid in the amino acid sequence.

In one implementation scenario, in order to improve the convenience of position coding, a coding network may be trained in advance, and the coding network may include an amino acid coding sub-network for position coding, so that the amino acid coding sub-network may be used to position code the primary amino acid sequence structure of the target protein, and obtain a plurality of amino acid codes corresponding to the primary amino acid sequence structure of the target protein.

And extracting the characteristic representation of a plurality of amino acids corresponding to the amino acid sequence by utilizing a multi-layer one-dimensional convolution network based on the plurality of amino acid codes. Since the amino acid codes are a row sequence, in this step, the feature representation of a plurality of amino acids can be extracted by using a one-dimensional convolution network. The convolution kernel of the one-dimensional convolution is a one-dimensional convolution kernel, which is applied to perform convolution processing on the one-dimensional sequence data.

The different amino acids themselves have a different composition structure, resulting in a plurality of amino acid codes of different lengths. So that the characteristic representations of a plurality of amino acids after one-dimensional convolution processing are also different. After the feature representations of the plurality of amino acids are pooled, they are subjected to a maximum pooling process to obtain an initial feature representation of the target protein.

The plurality of amino acid codes are input into a recurrent neural network to extract a hidden feature representation of the plurality of amino acids. Specifically, hidden feature representations of a plurality of amino acids are sequentially extracted in the head-to-tail direction and the tail-to-head direction of the feature representations of the plurality of amino acids through a recurrent neural network, respectively. The hidden feature representation includes the entire feature of the protein.

The recurrent neural network comprises an input unit, an output unit and a hiding unit, wherein the hiding unit is used for finishing the most main work in the recurrent neural network. Specifically, the features of the input data are abstracted to another dimensional space to show more abstract features of the input data. The hidden feature representation of the plurality of amino acids in the step is the feature value processed by the hidden unit of the recurrent neural network.

And performing feature fusion on the initial feature representation and the hidden feature representation to obtain a feature representation of the target protein. The steps of obtaining the initial feature representation and the hidden feature representation may be performed simultaneously, or the hidden feature representation may be obtained first and then the initial feature representation is obtained, which is not limited herein.

Wherein, the steps S23 and S22 can be performed simultaneously or in reverse order, and the steps S23 and S22 are independent of each other.

S24: and splicing the characteristic representation of the drug and the characteristic representation of the target protein end to obtain the spliced characteristic representation of the drug and the target protein.

And splicing the characteristic representation of the drug and the characteristic representation of the target protein end to obtain the spliced characteristic representation of the drug and the target protein. The characteristic representation of the medicine and the characteristic representation of the target protein are connected in series to form a group of 1 x n tiled structures, so that the influence of characteristic positions on classification in regression prediction classification is reduced.

The head-to-tail splicing comprises splicing the head expressed by the characteristics of the medicine and the tail expressed by the characteristics of the target protein or splicing the tail expressed by the characteristics of the medicine and the head expressed by the characteristics of the target protein to obtain a group of tiled structures, and the specific splicing sequence is not limited.

S25: and (4) performing prediction classification on the splicing characteristic representation by utilizing a multilayer neural network to obtain the reaction effect type of the drug and the target protein.

And carrying out regression prediction classification on the splicing characteristic representation by utilizing a multilayer neural network to obtain the classification confidence coefficient of the splicing characteristic representation of the drug and the target protein. Judging the type of the reaction effect of the drug and the target protein according to the classification confidence. Wherein the reaction effect types comprise: the reaction of the drug with the target protein is effective and the reaction of the drug with the target protein is ineffective.

In a specific implementation scenario, the multilayer neural network may include a full link layer and a softmax layer, so that feature connection may be performed on the spliced feature representation by using the full link layer, and probability normalization may be performed by using the softmax layer to obtain a classification confidence of the spliced feature representation, so that a reaction effect type corresponding to the classification confidence may be used as a reaction prediction result of the drug and the target protein.

The full connection layer is used for assembling all local features obtained in the convolution layer and the pooling layer into an overall feature through the weight matrix again, and therefore regression prediction classification of the splicing feature representation is achieved.

Referring to fig. 5, fig. 5 is a schematic diagram of an embodiment of a full link layer feature connection according to the present application.

The nodes x1, x2, x3., and xn in this embodiment are input layers of the full connection layer, i.e. the splicing feature. Nodes a1, a2, a3..

Each node a1, a2, a3., and an in the input layer is fully connected to each node x1, x2, x3., and xn in the input layer to integrate all local features in the input layer and output the global feature.

According to the scheme, the SMILES expression is reconstructed into the molecular structure diagram and then further reconstructed into the undirected structure diagram, the feature representation of a plurality of atomic nodes of the molecular structure is extracted by utilizing the multi-layer diagram convolution structure so as to convert two-dimensional data into one-dimensional data, and then the feature representation of the atomic nodes is subjected to maximum pooling treatment to obtain the feature representation of the medicine, so that the feature representation of the medicine is more consistent with the structural characteristics of medicine molecules. Meanwhile, the method extracts the amino acid local characteristics through one-dimensional convolution and pooling, extracts the overall characteristics of the amino acid sequence through a circulating neural network, fuses the local characteristics and the overall characteristics to obtain the characteristic representation of the target protein, and at the moment, the characteristic representation of the target protein is also one-dimensional data, splices the characteristic representation of the drug and the characteristic representation of the target protein and predicts and classifies the characteristic representations to obtain the reaction effect type of the drug and the target protein. The whole prediction scheme can adaptively learn deeper characteristics of the drug molecules and the proteins, realize the integral characteristic extraction of the drug molecules and the proteins, and optimize the prediction effect of the drug and the target protein.

It will be understood by those skilled in the art that in the method of the present invention, the order of writing the steps does not imply a strict order of execution and any limitations on the implementation, and the specific order of execution of the steps should be determined by their function and possible inherent logic.

Referring to fig. 6, fig. 6 is a schematic diagram of an embodiment of an effect recognition device for drug reaction with a target protein according to the present application.

The effect recognition device 60 for the reaction of the drug and the target protein comprises an acquisition module 61, a feature extraction module 62, a splicing module 63 and a recognition module 64, wherein the acquisition module 61 is used for acquiring a molecular structure diagram of the drug and an amino acid sequence of the target protein; the feature extraction module 62 is configured to perform feature extraction on the molecular structure diagram and the amino acid sequence, respectively, to obtain feature representation of the drug and feature representation of the target protein; the splicing module 63 is configured to splice the feature representation of the drug and the feature representation of the target protein to obtain a spliced feature representation of the drug and the target protein; the recognition module 64 is used for determining the type of the reaction effect of the drug with the target protein based on the splicing feature representation.

According to the scheme, after the molecular structure diagram of the drug and the amino acid sequence of the target protein are obtained, feature extraction is respectively carried out on the molecular structure diagram of the drug and the amino acid sequence of the target protein, so that feature representation of the drug and the feature representation of the target protein are extracted, the reaction effect type of the drug and the target protein is determined based on splicing feature representation of the drug and the target protein, deeper features of the drug molecule and the protein can be learned in a self-adaptive mode, feature extraction of the drug molecule and the protein is achieved, and the reaction prediction effect type of the drug and the target protein is obtained.

In some disclosed embodiments, the step of feature extraction module 62 for performing feature extraction on the molecular structure diagram and the amino acid sequence to obtain a feature representation of the drug and a feature representation of the target protein respectively comprises: extracting feature representations of a plurality of atomic nodes of a molecular structure by using a deep learning network; and performing maximum pooling on the feature representations of the atomic nodes to obtain the feature representation of the medicine.

Different from the embodiment, the feature representation of a plurality of atomic nodes of the molecular structure is extracted by specifically utilizing the deep learning network, so that the feature representation of the molecular structure with a two-dimensional space structure is one-dimensional, the subsequent feature splicing is facilitated, and the efficiency of response prediction is improved. And performing maximum pooling treatment on the feature representations of the atomic nodes to obtain the feature representation of the medicine, so as to reduce the number of parameters in the feature representation and keep important information.

In some disclosed embodiments, the step of extracting feature representations of a plurality of atomic nodes of a molecular structure using a deep learning network by the feature extraction module 62 further comprises: respectively extracting initial feature representation of each atomic node in the molecular structure and adjacency relation feature representation of the atomic node and adjacent atomic nodes by utilizing a multilayer graph convolution structure of a deep learning network; and fusing the initial characteristic representation and the adjacency characteristic representation of the atomic node, and determining a fusion result as the characteristic representation of the atomic node.

Different from the foregoing embodiment, the multi-layer graph convolution structure extraction is performed based on the initial feature representation of each atomic node and the adjacency feature representation of the atomic node and the adjacent atomic node, so as to improve the accuracy of feature extraction. And fusing the initial characteristic representation and the adjacency characteristic representation of the atomic node to obtain the characteristic representation of the atomic node, so that the characteristic representation of the atomic node can better reflect the characteristic value of the whole drug molecule.

In some disclosed embodiments, the step of merging the initial feature representation and the adjacency feature representation of the atomic node and determining the merging result as the feature representation of the atomic node further includes: taking the fusion result as the initial characteristic representation of the atomic node, and circularly executing the step of extracting the characteristic representation of a plurality of atomic nodes of the molecular structure; and under the condition that the number of times of executing the step of extracting the characteristic representations of the plurality of atomic nodes of the molecular structure reaches a preset value, determining the last fusion result as the characteristic representation of the atomic node.

Different from the foregoing embodiment, the steps of extracting the feature representation of a plurality of atomic nodes of a molecular structure are performed in a loop to adapt to drug molecules with different complexity, so that the extraction of the feature representation of the atomic nodes is realized by adopting execution times with different preset values for different drug molecular structures, and the flexible operation is realized.

In some disclosed embodiments, the step of characterizing the molecular structure map and the amino acid sequence to obtain a characterization of the drug and a characterization of the target protein, respectively, comprises: constructing a directed structure diagram based on the molecular structure diagram, and extracting characteristic representation of the medicine based on the directed structure diagram.

Different from the embodiment, the redundant information in the molecular structure diagram is removed through constructing the undirected structure diagram based on the molecular structure diagram to adapt to the input of the multilayer diagram convolution structure of the deep learning network, only the information such as the atom nodes, the connection relation of the atom nodes and the like is reserved, and the efficiency of the feature representation extraction of the medicine is improved.

In some disclosed embodiments, the step of obtaining the molecular structure diagram of the drug and the amino acid sequence of the target protein specifically comprises: and acquiring the SMILES molecular formula of the medicine, and converting the SMILES molecular formula of the medicine into a molecular structure diagram of the medicine.

In distinction to the previous examples, the molecular structure diagram of the substance is converted from the SMILES molecular formula of the drug molecule. The SMILES molecular formula of the medicine has uniqueness, so that the corresponding relation between the molecular structure diagram and the medicine molecules is ensured, and the accuracy of a reaction prediction result is improved.

In some disclosed embodiments, the step of feature extraction module 62 for performing feature extraction on the molecular structure diagram and the amino acid sequence to obtain a feature representation of the drug and a feature representation of the target protein respectively comprises: obtaining a plurality of amino acid codes of the amino acid sequence; obtaining an initial characterization of the amino acid sequence based on the plurality of amino acid codes; extracting a hidden feature representation of the amino acid sequence by a recurrent neural network; and fusing the initial feature representation and the hidden feature representation to obtain the feature representation of the target protein.

In distinction to the preceding examples, an initial characterization of the amino acid sequence is obtained based on a plurality of amino acid codes; and extracting hidden feature representations of the amino acid sequences through a recurrent neural network. Fusing the initial feature representation and the hidden feature representation to obtain a feature representation of the target protein. And deeper and more comprehensive characteristic representation of the amino acid sequence is obtained, and the reliability of the reaction prediction result is improved.

In some disclosed embodiments, the step of deriving an initial characterization of the amino acid sequence based on the plurality of amino acid codes comprises: extracting feature representations of a plurality of amino acids corresponding to the amino acid sequence by utilizing a multi-layer one-dimensional convolution network based on the amino acid coding; performing maximum pooling treatment on the feature representations of the plurality of amino acids to obtain an initial feature representation of the target protein; and the step of extracting the hidden-feature representation of the amino acid sequence by the recurrent neural network comprises: hidden feature representations of a plurality of amino acids are extracted by a recurrent neural network.

Different from the previous embodiment, the method utilizes a multilayer one-dimensional convolution network to extract the feature representation of a plurality of amino acids corresponding to the amino acid sequence, and performs maximum pooling treatment on the feature representation of the plurality of amino acids to obtain the initial feature representation of the target protein, so as to reduce the number of parameters in the feature representation, retain important information and improve the practicability of the initial feature representation. Hidden feature representations of a plurality of amino acids are extracted by a recurrent neural network. So as to further obtain the overall characteristic representation of a plurality of amino acids, and the initial characteristic representation of the target protein is more comprehensive and reliable.

In some disclosed embodiments, the step of extracting hidden feature representations of a plurality of amino acids by a recurrent neural network comprises: and sequentially extracting hidden feature representations of the plurality of amino acids along the head-tail direction and the tail-head direction of the feature representations of the plurality of amino acids through a recurrent neural network respectively.

In distinction to the previous embodiments, the hidden feature representation of the plurality of amino acids is obtained by extracting the feature representation of the plurality of amino acids in both directions, making the hidden feature representation more comprehensive and reliable.

In some disclosed embodiments, the step of splicing the signature representation of the drug and the signature representation of the target protein by the splicing module 63 to obtain a spliced signature representation of the drug and the target protein comprises: and splicing the characteristic representation of the drug and the characteristic representation of the target protein end to obtain the spliced characteristic representation of the drug and the target protein.

Unlike the previous examples, the drug signature and the target protein signature are spliced end-to-end to accommodate the input modality of the fully-connected layer in the multilayer neural network.

In some disclosed embodiments, the step of determining the type of reaction effect of the drug with the target protein using the splice signature representation of the drug and the target protein comprises: and (4) performing prediction classification on the splicing characteristic representation by utilizing a multilayer neural network to obtain the reaction effect type of the drug and the target protein.

Different from the embodiment, the splicing characteristic representation is predicted and classified by utilizing the multilayer neural network so as to comprehensively predict the reaction between the drug molecules and the target protein, minimize errors and realize the accuracy of the reaction effect type.

Referring to fig. 7, fig. 7 is a schematic diagram of a frame of an electronic device according to an embodiment of the present application.

The electronic device 70 comprises a memory 71 and a processor 72 coupled to each other, and the processor 72 is configured to execute the program instructions stored in the memory 71 to implement the steps of any of the above embodiments of the method for identifying the effect of a drug reacting with a target protein. In one particular implementation scenario, the electronic device 70 may include, but is not limited to: a microcomputer, a server, and the electronic device 70 may also include a mobile device such as a notebook computer, a tablet computer, and the like, which is not limited herein.

Specifically, the processor 72 is configured to control itself and the memory 71 to implement the steps of any of the above-mentioned embodiments of the method for identifying the effect of a drug reacting with a target protein. The processor 72 may also be referred to as a CPU (Central Processing Unit). The processor 72 may be an integrated circuit chip having signal processing capabilities. The Processor 72 may also be a general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. Additionally, the processor 72 may be collectively implemented by an integrated circuit chip.

By the scheme, deeper characteristics of the drug molecules and the protein can be learned in a self-adaptive manner, the characteristic extraction of the drug molecules and the protein as a whole is realized, the comprehensive degree and the reliability of the characteristic extraction of the drug molecules and the protein as a whole are increased, the efficiency of the whole reaction prediction process is improved, and the reaction effect type of the drug and the target protein is more accurate.

Referring to fig. 8, fig. 8 is a block diagram illustrating an embodiment of a computer-readable storage medium according to the present application.

The computer readable storage medium 80 stores program instructions 801 that can be executed by the processor, the program instructions 801 being for implementing the steps of any of the above-described method embodiments of identifying an effect of a drug on a target protein.

By the scheme, deeper characteristics of the drug molecules and the protein can be learned in a self-adaptive manner, the characteristic extraction of the drug molecules and the protein as a whole is realized, the comprehensive degree and the reliability of the characteristic extraction of the drug molecules and the protein as a whole are increased, the efficiency of the whole reaction prediction process is improved, and the reaction effect type of the drug and the target protein is more accurate.

In some embodiments, functions of or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the method described in the above method embodiments, and specific implementation thereof may refer to the description of the above method embodiments, and for brevity, will not be described again here.

The foregoing description of the various embodiments is intended to highlight various differences between the embodiments, and the same or similar parts may be referred to each other, and for brevity, will not be described again herein.

In the several embodiments provided in the present application, it should be understood that the disclosed method and apparatus may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a module or a unit is merely one type of logical division, and an actual implementation may have another division, for example, a unit or a component may be combined or integrated with another system, or some features may be omitted, or not implemented. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some interfaces, and may be in an electrical, mechanical or other form.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.