Intelligent drug research and development device, storage medium and computer equipment

Abstract

The present invention provides an intelligent drug development device, storage medium and computer equipment, the device includes: a small molecule drug design module and a macromolecular drug design module, the small molecule drug design module is used to generate new small molecule drugs based on the existing small molecule chemical library using a pre-trained small molecule simulation model, and screen out target small molecule drugs that meet the drug design requirements in the small molecule drugs using generative adversarial reinforcement learning technology; the macromolecular drug design module is used to generate target macromolecular drugs corresponding to the existing antigen protein structure using GNN network technology; the target macromolecular drug is an antibody protein with a three-dimensional spatial structure that can specifically bind to the antigen protein structure. Through the present invention, the automatic design, screening and discovery of drugs are realized, the efficiency of drug development is greatly improved, and the manpower and time costs are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of drug research and development, and in particular to an intelligent drug research and development device, a storage medium and a computer device. Background Art

[0002] The pharmaceutical industry is not only one of the fastest-growing sunrise industries in the world trade, but also a basic and strategic industry related to the national economy and people's livelihood. Traditional drug research and development generally takes 10-20 years and costs $1-2 billion, with a success rate of <5%, low R&D efficiency and high costs. The R&D process goes through hypothesis-driven problem discovery, biological experimental falsification or confirmation, biological knowledge update, and new hypothesis-driven new problem discovery, without making full use of biomedical big data. Summary of the invention

[0003] In view of the above problems, the present invention proposes an intelligent drug research and development device, storage medium and computer equipment, which integrates biomedical multimodal big data, uses multiple algorithms to realize automatic design, screening and discovery of drugs, and visualizes the results, providing a software product that is easy to operate and use.

[0004] According to a first aspect of the present invention, there is provided an intelligent drug development device, comprising:

[0005] A small molecule drug design module is used to generate new small molecule drugs based on an existing small molecule chemical library using a pre-trained small molecule simulation model, and to screen out target small molecule drugs that meet drug design requirements from the small molecule drugs using generative adversarial reinforcement learning technology;

[0006] The macromolecular drug design module is used to generate a target macromolecular drug corresponding to an existing antigen protein structure using GNN network technology; the target macromolecular drug is an antibody protein having a three-dimensional spatial structure that can specifically bind to the antigen protein structure.

[0007] Optionally, the intelligent drug development device further includes a small molecule drug prediction module;

[0008] The small molecule drug prediction module is used to characterize the drug structure characteristics of the target small molecule drug and match the target disease corresponding to the target small molecule drug according to the drug structure characteristics; the drug structure characteristics include global structure vector characteristics and protein sequence vector characteristics.

[0009] Optionally, the small molecule drug design module includes a small molecule drug generation unit and a small molecule drug detection unit;

[0010] The small molecule drug generation unit is used to use the small molecule simulation model to change part of the atomic structure of the small molecule structure in the existing small molecule chemical library to generate a new small molecule structure;

[0011] The small molecule drug detection unit is used to analyze the small molecule structure using generative adversarial reinforcement learning technology, and match the small molecule structure with the corresponding drug properties through tensor decomposition, so as to screen out small molecule structures that meet the drug design requirements as target small molecule drugs; the drug design requirements include at least one of drug novelty, drug effectiveness, and drug synthesis feasibility.

[0012] Optionally, the small molecule drug prediction module includes a microscopic characterization unit and a macroscopic characterization unit:

[0013] A microscopic characterization unit, used for exchanging vector information of atomic vector features corresponding to each atom in the target small molecule drug by using directed information transfer neural network technology, so as to update the atomic vector features corresponding to each atom in the small molecule structure;

[0014] The macroscopic characterization unit is used to characterize the global structural vector features of the target small molecule drug based on the atomic vector features corresponding to each atom in the target small molecule drug using a graph regularization algorithm, and to characterize the protein sequence vector features in the target small molecule drug based on CNN network technology.

[0015] Optionally, the small molecule drug prediction module further includes a database updating unit, a heterogeneous graph building unit and a drug prediction unit;

[0016] A database updating unit, used for updating the small molecule chemical library by adding the global structure vector features of the target small molecule drug into the existing small molecule chemical library; and updating the protein database by adding the protein sequence vector features of the target small molecule drug into the existing protein database;

[0017] A heterogeneous graph construction unit is used to generate small molecule drug relationship graphs, disease relationship graphs and protein relationship graphs based on the updated small molecule chemical library, protein database and disease relationship library using heterogeneous graph attention neural network technology;

[0018] The drug prediction unit is used to use the graph attention network technology to transfer vector information between any two of the small molecule drug relationship graph, the disease relationship graph and the protein relationship graph to obtain an updated small molecule drug-disease relationship graph, and search for the target disease corresponding to the target small molecule drug according to the small molecule drug-disease relationship graph.

[0019] Optionally, the macromolecular drug design module includes a model building unit;

[0020] The model building unit is used to train the GNN network based on the correspondence database between the three-dimensional spatial structure of the protein and the one-dimensional amino acid sequence, and generate a macromolecular simulation model that can output the corresponding one-dimensional amino acid sequence by inputting the three-dimensional spatial structure of the protein.

[0021] Optionally, the macromolecular drug design module further includes a drug design unit;

[0022] The drug design unit is used to generate the protein three-dimensional spatial structure of the target macromolecular drug that can specifically bind to the antigen protein based on the protein three-dimensional spatial structure of the antigen protein, and predict the amino acid sequence of the target macromolecular drug using the macromolecular simulation model based on the protein three-dimensional spatial structure of the target macromolecular drug.

[0023] Optionally, the intelligent drug research and development device further includes a database integration module;

[0024] The database integration module is used to save and update the multimodal database; the multimodal database includes at least one of a small molecule chemical library, a protein database, a disease relationship library, and a correspondence database between a three-dimensional spatial structure of a protein and a one-dimensional amino acid sequence.

[0025] Optionally, the intelligent drug development device further includes an experimental data verification module;

[0026] The experimental data verification module is used to verify the drug properties of the target small molecule drug and / or the target macromolecule drug through in vivo experiments, and update the multimodal database according to the obtained experimental verification data.

[0027] Optionally, the intelligent drug development device further includes a system visualization module;

[0028] The system visualization module is used to set a terminal visualization interactive page, and the terminal interactive visualization page is used to display, search and / or download experimental verification data, multimodal database and / or drug design process of the target small molecule drug and the target macromolecular drug;

[0029] The terminal visualization interactive page is also used to search for drug functions according to the target small molecule drug and the corresponding target disease, the target macromolecule drug and the corresponding antigen protein.

[0030] According to a second aspect of the present invention, there is provided a readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0031] Generate new small molecule drugs based on the existing small molecule chemical library using the pre-trained small molecule simulation model, and screen out target small molecule drugs that meet the drug design requirements among the small molecule drugs using generative adversarial reinforcement learning technology; and,

[0032] GNN network technology is used to generate a target macromolecular drug corresponding to an existing antigen protein structure; wherein the target macromolecular drug is an antibody protein having a three-dimensional spatial structure that can specifically bind to the antigen protein structure.

[0033] According to a third aspect of the present invention, there is provided a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0034] Generate new small molecule drugs based on the existing small molecule chemical library using the pre-trained small molecule simulation model, and screen out target small molecule drugs that meet the drug design requirements among the small molecule drugs using generative adversarial reinforcement learning technology; and,

[0035] GNN network technology is used to generate a target macromolecular drug corresponding to an existing antigen protein structure; wherein the target macromolecular drug is an antibody protein having a three-dimensional spatial structure that can specifically bind to the antigen protein structure.

[0036] The present invention provides an intelligent drug development device, storage medium and computer equipment, which are provided with a small molecule drug design module and a macromolecular drug design module, and use the small molecule drug design module to generate new small molecule drugs based on the existing small molecule chemical library through a pre-trained small molecule simulation model, and use generative adversarial reinforcement learning technology to screen out the target small molecule drugs that meet the drug design requirements in the small molecule drugs; use the macromolecular drug design module to generate target macromolecular drugs corresponding to the existing antigen protein structure through GNN network technology; the target macromolecular drug is an antibody protein with a three-dimensional spatial structure that can specifically bind to the antigen protein structure. The present invention integrates biomedical multimodal big data, uses a variety of algorithms to realize the automatic design, screening and discovery of drugs, realizes the transformation of the traditional hypothesis-driven research model into a data-driven research model, and transforms the traditional wet experiment manual high-throughput drug screening work into the work of automatic iterative update by computer, which greatly improves the efficiency of drug development and reduces manpower and time costs.

[0037] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below.

[0038] Based on the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will become more aware of the above and other objects, advantages and features of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0040] Figure 1 A schematic diagram showing the structure of an intelligent drug research and development device provided by an embodiment of the present invention is shown;

[0041] Figure 2 A schematic diagram showing the structure of an intelligent drug research and development device provided by another embodiment of the present invention is shown;

[0042] Figure 3 A brief schematic diagram of a terminal visual interaction page provided by an embodiment of the present invention is shown;

[0043] Figure 4 A schematic diagram of the structure of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0044] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0045] Artificial intelligence technology was first applied to pharmaceutical research and development in the medical and health field. It has been widely applied and developed in many aspects such as health management, auxiliary diagnosis and treatment, drug mining, drug preparation and even clinical rational drug use. Artificial intelligence is expected to further provide practical solutions to some challenging problems and development difficulties in the development of the pharmaceutical industry. The amount of data in the pharmaceutical industry is extremely large. If this data is used to train artificial intelligence algorithm models and then used in the process of pharmaceutical research and development, it can effectively speed up drug research and development, reduce drug research and development costs, and increase the success rate of drug research and development.

[0046] Based on this, an embodiment of the present invention provides an intelligent drug development device, such as Figure 1 As shown, the device may include: a small molecule drug design module 110 and a macromolecule drug design module 120.

[0047] The small molecule drug design module 110 can be used to generate new small molecule drugs based on the existing small molecule chemical library using a pre-trained small molecule simulation model, and use generative adversarial reinforcement learning technology to screen out target small molecule drugs that meet drug design requirements among small molecule drugs.

[0048] The macromolecular drug design module 120 can be used to generate a target macromolecular drug corresponding to an existing antigen protein structure using GNN network technology; the target macromolecular drug is an antibody protein having a three-dimensional spatial structure that can specifically bind to the antigen protein structure.

[0049] In an embodiment of the present invention, the newly generated small molecule drugs are carried out on the basis of the existing small molecule chemical library, and new small molecule structures are generated by changing some of the atomic types or atomic structures of the original small molecules in the small molecule chemical library. Generative adversarial reinforcement learning technology is then used to generate target small molecule drugs with novelty, synthetic feasibility and biological activity. Macromolecular drug design is based on the existing antigen protein structure to generate a protein structure that can specifically bind to the antigen protein to achieve the antibody effect. The small molecule drug design and macromolecular drug design scheme provided by the embodiment of the present invention can effectively automatically design and research new drugs for diseases or viruses in a short time, and can be automatically iterated and updated by computers, which greatly improves the design efficiency and reduces the development cost.

[0050] In an optional embodiment of the present invention, Figure 2 As shown, an intelligent drug research and development device provided by an embodiment of the present invention may further include a small molecule drug prediction module 130 .

[0051] The small molecule drug prediction module 130 can be used to characterize the drug structure characteristics of the target small molecule drug and match the target disease corresponding to the target small molecule drug according to the drug structure characteristics; the drug structure characteristics include global structure vector characteristics and protein sequence vector characteristics.

[0052] After generating the target small molecule drug, it is necessary to study and analyze the diseases that the target small molecule drug can treat. The embodiment of the present invention generates the global structure vector features and protein sequence vector features of the target small molecule drug, searches for disease features associated with the global structure vector features and protein sequence vector features, and thus matches the diseases that the target small molecule drug can treat. Among them, the global structure vector features of the target small molecule drug refer to the vector features that characterize the overall structure of the small molecule drug, and the protein sequence vector features refer to the vector features that characterize the protein sequence in the small molecule drug.

[0053] In an optional embodiment of the present invention, Figure 2As shown, the small molecule drug design module 110 may include a small molecule drug generation unit 111 and a small molecule drug detection unit 112 .

[0054] The small molecule drug generation unit 111 can be used to use a small molecule simulation model to change a part of the atomic structure of a small molecule structure in an existing small molecule chemical library to generate a new small molecule structure.

[0055] The small molecule drug detection unit 112 can be used to analyze the small molecule structure using generative adversarial reinforcement learning technology, and match the small molecule structure with the corresponding drug properties through tensor decomposition, so as to screen out the small molecule structure that meets the drug design requirements as the target small molecule drug; the drug design requirements include at least one of the drug novelty, drug effectiveness and drug synthesis feasibility.

[0056] The process of designing small molecule drugs in the embodiment of the present invention is completed through a small molecule simulation model, which is a model that can represent atoms through nodes and represent the connection relationship between atoms through lines between nodes, such as covalent bonds. The small molecule simulation model can generate similar new small molecule structures by changing the type, number, position, etc. of atoms in the original small molecule structure in the small molecule chemical library, and search for molecular structures identical to the newly generated small molecule structure based on the existing small molecule patent database to ensure the novelty of the newly generated small molecule structure. The small molecule structure and the corresponding drug properties are then linked through a generative adversarial reinforcement learning algorithm to evaluate the drug activity and synthetic feasibility of the small molecule structure.

[0057] Among them, the generative adversarial reinforcement learning network is a deep learning model. The model mainly produces conclusion outputs through two modules: the mutual game learning of the generative model and the discriminative model. The generative model is used to imitate, model and learn the distribution law of real data, and the discriminative model is used to discriminate whether the input data comes from the real data distribution or from a generative model. Through the continuous competition between these two internal models, the generative ability and discriminative ability of the two models are improved. In other words, the generative adversarial reinforcement learning algorithm can be used to obtain an optimal strategy by maximizing the reward function, and then tensor decomposition is used to match the coding of small molecule structures for drug activity. When the newly generated small molecule structure has certain drug properties, it can be determined that this small molecule structure has drug activity, that is, drug effectiveness and synthetic feasibility. Small molecule drugs with drug synthesis feasibility, drug effectiveness and drug novelty are used as target small molecule drugs that meet the needs of drug design, and drug prediction is subsequently performed on the target small molecules.

[0058] In an optional embodiment of the present invention, Figure 2As shown, the small molecule drug prediction module 130 may include a microscopic characterization unit 131 , a macroscopic characterization unit 132 , a database updating unit 133 , a heterogeneous graph construction unit 134 , and a drug prediction unit 135 .

[0059] The microscopic characterization unit 131 can be used to exchange vector information of the atomic vector features corresponding to each atom in the target small molecule drug using directed information transfer neural network technology to update the atomic vector features corresponding to each atom in the small molecule structure.

[0060] In an embodiment of the present invention, the microscopic characterization unit can characterize the vector features of each atomic structure of the target small molecule drug through the directed information transfer neural network technology. Among them, the directed information transfer neural network technology is completed using the GNN neural network algorithm. GNN uses a form of neural information transfer to exchange vector feature information between the atomic nodes of the target small molecule drug, and uses a neural network to update the vector feature information of the atomic node. In each iteration of message transmission in GNN, the vector feature of the atomic node is updated according to the information aggregated from the neighborhood of the atomic node. The directed information transfer neural network does not use information related to the node, but uses information related to the directed edge. For example, the information of node 1→2 will only be propagated to nodes 3 and 4 in the next round of iteration, while in the traditional undirected information transfer neural network, the information of 1→2 will still be propagated to node 1. The advantage of the directed information transfer neural network over the undirected information transfer neural network is that the characteristics of the node will not be over-smoothed due to the repeated transmission of information, and the vector features of each atomic structure of the target small molecule drug will be generated.

[0061] The macro characterization unit 132 can be used to characterize the global structure vector features of the target small molecule drug based on the atomic vector features corresponding to each atom in the target small molecule drug using a graph regularization algorithm, and to characterize the protein sequence vector features in the target small molecule drug based on the CNN network technology.

[0062] In an embodiment of the present invention, the macro characterization unit characterizes the global structural vector features and protein sequence vector features of the target small molecule drug through a GNN+CNN model with graph warp technology. Specifically, the global structural vector features can be characterized by GNN graph neural network technology, and CNN can characterize the protein sequence vector features. Graph warping is a method of representing the connection relationship of all other nodes in the small molecule structure graph by creating a virtual node, allowing atoms to directly pass messages to each other through virtual nodes, thereby realizing the detection of global features. GNN graph neural network technology is a neural network technology that directly acts on the graph structure. By introducing the graph warping algorithm into the GNN network technology, the characterization of the global structural vector features of the target small molecule drug can be realized, and the molecular structure features can be extracted more accurately than the traditional local characterization. CNN convolutional neural network is a common deep learning model. Through pre-model training, it can be realized to characterize the protein sequence vector features in the target small molecule drug based on CNN convolutional neural network technology.

[0063] The database updating unit 133 can be used to update the small molecule chemical library by adding the global structure vector features of the target small molecule drug into the existing small molecule chemical library; and to update the protein database by adding the protein sequence vector features of the target small molecule drug into the existing protein database.

[0064] After obtaining the protein sequence vector features and global structure vector features of the target small molecule drug, the small molecule structure or global structure vector features of the target small molecule drug can be added to the existing small molecule chemical library to update the small molecule chemical library; the protein sequence or protein sequence vector features of the target small molecule drug can also be added to the existing protein database to update the protein database, and subsequent drug research and development can be carried out based on the latest updated database.

[0065] The heterogeneous graph construction unit 134 can be used to generate a small molecule drug relationship graph, a disease relationship graph and a protein relationship graph based on the updated small molecule chemical library, protein database and disease relationship library using the heterogeneous graph attention neural network technology.

[0066] The drug prediction unit 135 can be used to use the graph attention network technology to transfer vector information between any two of the small molecule drug relationship graph, the disease relationship graph and the protein relationship graph to obtain an updated small molecule drug-disease relationship graph, and search for the target disease corresponding to the target small molecule drug based on the small molecule drug-disease relationship graph.

[0067] Based on the updated small molecule chemical library, protein database and disease relationship library, the heterogeneous graph attention neural network technology is used to generate small molecule relationship graphs, disease relationship graphs and protein relationship graphs. Among them, the small molecule relationship graph is used to characterize the association relationship between small molecule drugs. The graph includes multiple nodes and multiple edges connecting the nodes, and the edges represent the association relationship between small molecule drugs. Similarly, the disease relationship graph and the protein relationship graph are used to characterize the association relationship between diseases and the association relationship between proteins, respectively. After generating the small molecule relationship graph, the disease relationship graph and the protein relationship graph, the network graph features of the small molecule structure, protein sequence and disease can be fused and fully extracted through the graph convolution network GCN technology, and the vector information between any two relationship graphs can be transferred using the GAT graph attention network technology to achieve the update of the association relationship between two different biological entities. Among them, the GAT graph attention network technology is a method of aggregating neighbor nodes through the attention mechanism to achieve adaptive allocation of weights of different neighbor nodes, and complete the vector feature transfer to other nodes based on the features of a certain node. For example, the protein relationship graph can be used to first update the small molecule relationship graph and the disease relationship graph, and then the vector feature information of the updated small molecule relationship graph and the disease relationship graph can be transferred to obtain an updated small molecule-disease relationship graph, and the target disease corresponding to the target small molecule drug can be found based on the small molecule-disease relationship graph.

[0068] In an optional embodiment of the present invention, Figure 2 As shown, the macromolecular drug prediction module 120 may include a model building unit 121 and a drug design unit 122 .

[0069] The model building unit 121 can be used to train the GNN network based on the correspondence database between the three-dimensional spatial structure of the protein and the one-dimensional amino acid sequence, and generate a macromolecular simulation model that can output the corresponding one-dimensional amino acid sequence by inputting the three-dimensional spatial structure of the protein.

[0070] The drug design unit 122 can be used to generate the protein three-dimensional spatial structure of the target macromolecule drug that can specifically bind to the antigen protein based on the protein three-dimensional spatial structure of the antigen protein, and predict the amino acid sequence of the target macromolecule drug using a macromolecular simulation model based on the protein three-dimensional spatial structure of the target macromolecule drug.

[0071] The macromolecular drug design in the embodiment of the present invention is generally based on the antigen protein in the existing virus, and the designed target macromolecular drug is an antibody protein with a three-dimensional spatial structure that can specifically bind to the antigen protein. The process of macromolecular drug design is specifically to represent the molecular structure in the form of a graph, and to characterize the three-dimensional spatial structure of the protein using the GNN graph neural network technology. The nodes in the graph are amino acids, and the edges are the Euclidean distances between the amino acids and the relative positions of these amino acids in the amino acid chain. The macromolecular simulation model based on the GNN graph neural network technology is trained based on the correspondence database between the three-dimensional spatial structure of the protein and the one-dimensional amino acid sequence. By minimizing the cross entropy loss between the predicted value and the actual value, the final macromolecular simulation model is trained, and the macromolecular simulation model can accurately predict the amino acid sequence of the protein by inputting the three-dimensional spatial structure of the protein to output the corresponding one-dimensional amino acid sequence. The protein three-dimensional spatial structure of the target macromolecular drug is input into the trained macromolecular simulation model to predict the amino acid sequence of the target macromolecular drug, so that the target macromolecular drug can be mass-produced according to the amino acid sequence.

[0072] In an optional embodiment of the present invention, Figure 2 As shown, an intelligent drug research and development device provided by an embodiment of the present invention may further include a database integration module 140 , an experimental data verification module 150 and a system visualization module 160 .

[0073] The database integration module 140 can be used to save and update the multimodal database; the multimodal database includes at least one of a small molecule chemical library, a protein database, a disease relationship library, and a correspondence database between a protein three-dimensional spatial structure and a one-dimensional amino acid sequence.

[0074] The experimental data verification module 150 can be used to verify the drug properties of the target small molecule drug and / or the target macromolecule drug through in vivo experiments, and update the multimodal database according to the obtained experimental verification data.

[0075] The system visualization module 160 can be used to set up a terminal visualization interaction page, which is used to display, search and / or download experimental verification data, multimodal databases and / or drug design processes of target small molecule drugs and target macromolecule drugs; the terminal visualization interaction page is also used to search for drug functions based on target small molecule drugs and corresponding target diseases, target macromolecule drugs and corresponding antigen proteins.

[0076] Among them, the database integration module can provide the integration, automatic update and viewing functions of multimodal databases, including small molecule chemical libraries, protein databases, disease relationship libraries, and the correspondence database between protein three-dimensional spatial structure and one-dimensional amino acid sequence. The molecular prediction function can be realized by using these real-time updated database information. The experimental data verification module can verify the drug activity and effectiveness of target small molecule drugs or target macromolecule drugs through in vivo experiments, and update the molecular drug functions based on the experimental verification data obtained, and add them to the multimodal database for subsequent drug development. The system visualization module supports the display of the system on the terminal page and the interaction with the user, such as Figure 3 As shown, users can view, search, call and modify the multimodal database stored in the database integration module; users can view and manipulate the small molecule design process through page interaction; search for drug functions, that is, search and find the relationship between molecular drugs and diseases; and record, save and call the drug activity and effectiveness results obtained after in vivo experiments on target small molecule drugs or target macromolecule drugs.

[0077] An intelligent drug development device provided by an embodiment of the present invention, by setting a small molecule drug design module, a small molecule drug prediction module, a macromolecular drug design module, a database integration module, an experimental data verification module and a system visualization module, by integrating biomedical multimodal big data, using a variety of algorithms to achieve automatic design, screening and discovery of macromolecular drugs and small molecule drugs, and timely update the database data, use the drug clinical trial results to verify and update the drug performance, and can also visualize the drug prediction results, providing a software product that is easy to operate and use. The present invention transforms the traditional wet experiment manual high-throughput drug screening work into computer automatic iterative update work through the process including data-driven problem discovery, model establishment, biological experiment verification, biological knowledge update and other technologies, and can complete the design and discovery of new drugs for difficult and serious diseases such as cancer, Alzheimer's disease, new mutant viruses, drug-resistant bacteria, etc. in a short time based on artificial intelligence algorithms, transforming the traditional hypothesis-driven research model into a data-driven research model, greatly improving the efficiency of drug development and reducing manpower and time costs.

[0078] Based on the above Figure 1 The device shown in FIG. 1 is a computer-readable storage medium. Accordingly, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0079] Use the pre-trained small molecule simulation model to generate new small molecule drugs based on the existing small molecule chemical library, and use generative adversarial reinforcement learning technology to screen out target small molecule drugs that meet the drug design requirements; and

[0080] GNN network technology is used to generate target macromolecule drugs corresponding to existing antigen protein structures; the target macromolecule drugs are antibody proteins with a three-dimensional spatial structure that can specifically bind to the antigen protein structure.

[0081] In the embodiments of the present invention, the step of generating a target small molecule drug and the step of synthesizing a target macromolecule drug may be performed simultaneously or in any order, and the present invention does not limit this.

[0082] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0083] Characterize the drug structural features of the target small molecule drug, and match the target disease corresponding to the target small molecule drug based on the drug structural features; drug structural features include global structure vector features and protein sequence vector features.

[0084] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0085] Use the small molecule simulation model to change some of the atomic structures in the small molecule structure in the existing small molecule chemical library to generate new small molecule structures;

[0086] Generative adversarial reinforcement learning technology is used to analyze small molecule structures, and the corresponding drug properties are matched to the small molecule structures through tensor decomposition, so as to screen out small molecule structures that meet the drug design requirements as target small molecule drugs; drug design requirements include at least one of drug novelty, drug efficacy, and drug synthesis feasibility.

[0087] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0088] Use directed information transfer neural network technology to exchange vector information of atomic vector features corresponding to each atom in the target small molecule drug, so as to update the atomic vector features corresponding to each atom in the small molecule structure;

[0089] Based on the atomic vector features corresponding to each atom in the target small molecule drug, the global structure vector features of the target small molecule drug are characterized by using a graph regularization algorithm, and the protein sequence vector features in the target small molecule drug are characterized based on CNN network technology.

[0090] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0091] Updating a small molecule chemical library by adding global structure vector features of target small molecule drugs into an existing small molecule chemical library; and updating a protein database by adding protein sequence vector features of target small molecule drugs into an existing protein database;

[0092] Based on the updated small molecule chemical library, protein database and disease relationship library, the heterogeneous graph attention neural network technology is used to generate small molecule drug relationship graphs, disease relationship graphs and protein relationship graphs;

[0093] The graph attention network technology is used to transfer vector information between any two of the small molecule drug relationship graph, the disease relationship graph and the protein relationship graph to obtain an updated small molecule drug-disease relationship graph. The target disease corresponding to the target small molecule drug is found according to the small molecule drug-disease relationship graph.

[0094] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0095] The GNN network is trained based on a database of the correspondence between the three-dimensional spatial structure of proteins and the one-dimensional amino acid sequence, generating a macromolecular simulation model that can output the corresponding one-dimensional amino acid sequence by inputting the three-dimensional spatial structure of proteins.

[0096] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0097] The protein three-dimensional spatial structure of the target macromolecular drug that can specifically bind to the antigen protein is generated based on the protein three-dimensional spatial structure of the antigen protein, and the amino acid sequence of the target macromolecular drug is predicted using a macromolecular simulation model based on the protein three-dimensional spatial structure of the target macromolecular drug.

[0098] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0099] Preserve and update the multimodal database; the multimodal database includes at least one of a small molecule chemical library, a protein database, a disease relationship library, and a database of correspondence between a protein three-dimensional spatial structure and a one-dimensional amino acid sequence.

[0100] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0101] The drug properties of the target small molecule drug and / or the target macromolecule drug are verified through in vivo experiments, and the multimodal database is updated based on the obtained experimental verification data.

[0102] Optionally, when the computer program stored on the computer-readable storage medium provided by the embodiment of the present invention is executed by a processor, the following steps may be implemented:

[0103] Setting a terminal visualization interactive page, the terminal visualization interactive page is used to display, search and / or download experimental verification data, multimodal databases and / or drug design processes of target small molecule drugs and target macromolecule drugs;

[0104] Among them, the terminal visualization interactive page can also be used to search for drug functions based on target small molecule drugs and corresponding target diseases, target macromolecule drugs and corresponding antigen proteins.

[0105] It should be noted that for a detailed description of the steps that can be implemented when a computer program stored on a computer-readable storage medium is executed by a processor, reference can be made to Figure 1 The corresponding description of the device shown will not be repeated here.

[0106] Based on the above Figure 1 The embodiment of the device shown in the figure, the embodiment of the present invention also provides a physical structure diagram of a computer device, such as Figure 4 As shown, the computer device may include a communication bus, a processor, a memory and a communication interface, and may also include an input / output interface and a display device, wherein each functional unit may communicate with each other through the bus. The memory stores a computer program, and the processor is used to execute the program stored in the memory, and execute the same steps implemented by the computer program stored in the computer-readable storage medium provided in the above embodiment when executed by the processor.

[0107] Those skilled in the art can clearly understand that the specific working processes of the systems, devices, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they are not further described here.

[0108] In addition, the functional units in various embodiments of the present invention may be physically independent of each other, or two or more functional units may be integrated together, or all functional units may be integrated into one processing unit. The above integrated functional units may be implemented in the form of hardware, or in the form of software or firmware.

[0109] Those skilled in the art can understand that if the integrated functional unit is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can essentially or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, which includes a number of instructions to enable a computing device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention when running the instructions. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes.

[0110] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as a computing device such as a personal computer, a server, or a network device), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of a computing device, the computing device executes all or part of the steps of the methods described in the embodiments of the present invention.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that within the spirit and principles of the present invention, the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate from the protection scope of the present invention.

Claims

1. An intelligent drug development device, characterized in that: include: A small molecule drug design module is used to generate new small molecule drugs based on an existing small molecule chemical library using a pre-trained small molecule simulation model, and to screen out target small molecule drugs that meet drug design requirements from the small molecule drugs using generative adversarial reinforcement learning technology; A macromolecular drug design module is used to generate a target macromolecular drug corresponding to an existing antigen protein structure using GNN network technology; the target macromolecular drug is an antibody protein having a three-dimensional spatial structure that can specifically bind to the antigen protein structure; The intelligent drug development device also includes a small molecule drug prediction module; The small molecule drug prediction module is used to characterize the drug structure characteristics of the target small molecule drug and match the target disease corresponding to the target small molecule drug according to the drug structure characteristics; the drug structure characteristics include global structure vector characteristics and protein sequence vector characteristics; The small molecule drug prediction module includes a microscopic characterization unit and a macroscopic characterization unit: A microscopic characterization unit, used for exchanging vector information of atomic vector features corresponding to each atom in the target small molecule drug by using directed information transfer neural network technology, so as to update the atomic vector features corresponding to each atom in the small molecule structure; The macroscopic characterization unit is used to characterize the global structural vector features of the target small molecule drug based on the atomic vector features corresponding to each atom in the target small molecule drug using a graph regularization algorithm, and to characterize the protein sequence vector features in the target small molecule drug based on CNN network technology.

2. The device according to claim 1, characterized in that The small molecule drug design module includes a small molecule drug generation unit and a small molecule drug detection unit; The small molecule drug generation unit is used to use the small molecule simulation model to change part of the atomic structure of the small molecule structure in the existing small molecule chemical library to generate a new small molecule structure; The small molecule drug detection unit is used to analyze the small molecule structure using generative adversarial reinforcement learning technology, and match the small molecule structure with the corresponding drug properties through tensor decomposition, so as to screen out small molecule structures that meet the drug design requirements as target small molecule drugs; the drug design requirements include at least one of drug novelty, drug effectiveness, and drug synthesis feasibility.

3. The device according to claim 1, characterized in that The small molecule drug prediction module also includes a database updating unit, a heterogeneous graph building unit and a drug prediction unit; A database updating unit, used for updating the small molecule chemical library by adding the global structure vector features of the target small molecule drug into the existing small molecule chemical library; and updating the protein database by adding the protein sequence vector features of the target small molecule drug into the existing protein database; A heterogeneous graph construction unit is used to generate small molecule drug relationship graphs, disease relationship graphs and protein relationship graphs based on the updated small molecule chemical library, protein database and disease relationship library using heterogeneous graph attention neural network technology; The drug prediction unit is used to use the graph attention network technology to transfer vector information between any two of the small molecule drug relationship graph, the disease relationship graph and the protein relationship graph to obtain an updated small molecule drug-disease relationship graph, and search for the target disease corresponding to the target small molecule drug according to the small molecule drug-disease relationship graph.

4. The device according to claim 1, characterized in that The macromolecular drug design module includes a model building unit; The model building unit is used to train the GNN network based on the correspondence database between the three-dimensional spatial structure of the protein and the one-dimensional amino acid sequence, and generate a macromolecular simulation model that can output the corresponding one-dimensional amino acid sequence by inputting the three-dimensional spatial structure of the protein.

5. The device according to claim 4, characterized in that The macromolecular drug design module also includes a drug design unit; The drug design unit is used to generate the protein three-dimensional spatial structure of the target macromolecular drug that can specifically bind to the antigen protein based on the protein three-dimensional spatial structure of the antigen protein, and predict the amino acid sequence of the target macromolecular drug using the macromolecular simulation model based on the protein three-dimensional spatial structure of the target macromolecular drug.

6. The device according to claim 1, characterized in that The intelligent drug development device also includes a database integration module; The database integration module is used to save and update the multimodal database; the multimodal database includes at least one of a small molecule chemical library, a protein database, a disease relationship library, and a correspondence database between a three-dimensional spatial structure of a protein and a one-dimensional amino acid sequence.

7. The device according to claim 1, characterized in that The intelligent drug development device also includes an experimental data verification module; The experimental data verification module is used to verify the drug properties of the target small molecule drug and / or the target macromolecule drug through in vivo experiments, and update the multimodal database according to the obtained experimental verification data.

8. The device according to any one of claims 1 to 7, characterized in that: The intelligent drug development device also includes a system visualization module; The system visualization module is used to set a terminal visualization interactive page, and the terminal interactive visualization page is used to display, search and / or download experimental verification data, multimodal database and / or drug design process of the target small molecule drug and the target macromolecular drug; The terminal visualization interactive page is also used to search for drug functions according to the target small molecule drug and the corresponding target disease, the target macromolecule drug and the corresponding antigen protein.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the following steps are implemented: Generate new small molecule drugs based on the existing small molecule chemical library using the pre-trained small molecule simulation model, and use generative adversarial reinforcement learning technology to screen out target small molecule drugs that meet the drug design requirements among the small molecule drugs; as well as, Generate a target macromolecular drug corresponding to an existing antigen protein structure using GNN network technology; wherein the target macromolecular drug is an antibody protein having a three-dimensional spatial structure capable of specifically binding to the antigen protein structure; Characterize the drug structure characteristics of the target small molecule drug, and match the target disease corresponding to the target small molecule drug according to the drug structure characteristics; drug structure characteristics include global structure vector characteristics and protein sequence vector characteristics; The directed information transfer neural network technology is used to exchange vector information of the atomic vector features corresponding to each atom in the target small molecule drug, so as to update the atomic vector features corresponding to each atom in the small molecule structure; Based on the atomic vector features corresponding to each atom in the target small molecule drug, the global structure vector features of the target small molecule drug are characterized by using a graph regularization algorithm, and the protein sequence vector features in the target small molecule drug are characterized based on CNN network technology.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the following steps are implemented: Generate new small molecule drugs based on the existing small molecule chemical library using the pre-trained small molecule simulation model, and use generative adversarial reinforcement learning technology to screen out target small molecule drugs that meet the drug design requirements among the small molecule drugs; as well as, Generate a target macromolecular drug corresponding to an existing antigen protein structure using GNN network technology; wherein the target macromolecular drug is an antibody protein having a three-dimensional spatial structure capable of specifically binding to the antigen protein structure; Characterize the drug structure characteristics of the target small molecule drug, and match the target disease corresponding to the target small molecule drug according to the drug structure characteristics; drug structure characteristics include global structure vector characteristics and protein sequence vector characteristics; Use directed information transfer neural network technology to exchange vector information of atomic vector features corresponding to each atom in the target small molecule drug, so as to update the atomic vector features corresponding to each atom in the small molecule structure; Based on the atomic vector features corresponding to each atom in the target small molecule drug, the global structure vector features of the target small molecule drug are characterized by using a graph regularization algorithm, and the protein sequence vector features in the target small molecule drug are characterized based on CNN network technology.

Images