Data processing method, training method, identification method, device, equipment and medium

Abstract

The present disclosure provides a data processing method, a feature extraction model training method, a molecular attribute identification method, an apparatus, an electronic device, a storage medium, and a program product, all of which relate to the fields of data processing technology, particularly deep learning and biocomputing technology. A specific implementation scheme comprises: using the molecular structure information of the molecules to be processed to determine a target map, where the target map is used to represent the spatial structure information of multiple molecules in different dimensions; and determining a spatial structural feature representation of the molecules to be processed based on the target map.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, in particular to the field of deep learning and biocomputing technology, and specifically to data processing methods, training methods for feature extraction models, molecular attribute recognition methods, devices, electronic devices, storage media, and program products. Background Art

[0002] With the continuous development of computer technology, research on chemical materials and pharmaceuticals has gradually moved from using experimental instruments to using new machine learning techniques. New machine learning techniques applied to molecular and materials sciences, combining statistical perspectives with traditional chemical techniques, hold great research potential. Summary of the Invention

[0003] The present disclosure provides a data processing method, a feature extraction model training method, a molecular attribute recognition method, a device, an electronic device, a storage medium, and a program product.

[0004] According to one aspect of the present disclosure, a data processing method is provided, comprising: determining a target map using molecular structure information of a molecule to be processed, wherein the target map is used to characterize multiple molecular spatial structure information of different dimensions; and determining a spatial structural feature representation of the molecule to be processed based on the target map.

[0005] According to another aspect of the present disclosure, a method for training a feature extraction model is provided, comprising: determining a target sample map using molecular structure information of sample molecules, wherein the target sample map is used to characterize multiple molecular spatial structure information of different dimensions; processing the target sample map using a feature extraction model to determine a sample spatial structure feature representation of the sample molecules; determining attribute prediction information of the sample molecules based on the sample spatial structure feature representation; and training the feature extraction model using the attribute prediction information of the sample molecules and labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules.

[0006] According to another aspect of the present disclosure, a method for molecular attribute identification is provided, comprising: processing a molecule to be identified using the method described in the present disclosure to determine a spatial structural feature representation of the molecule to be identified; and determining target attribute information of the molecule to be identified based on the spatial structural feature representation.

[0007] According to another aspect of the present disclosure, a data processing device is provided, comprising: a first determination module for determining a target map using molecular structure information of a molecule to be processed, wherein the target map is used to characterize multiple molecular spatial structure information of different dimensions; and a second determination module for determining a spatial structural feature representation of the molecule to be processed based on the target map.

[0008] According to another aspect of the present disclosure, a training device for a feature extraction model is provided, comprising: a sample map determination module for determining a target sample map using molecular structure information of sample molecules, wherein the target sample map is used to characterize multiple molecular spatial structure information of different dimensions; a sample extraction module for processing the target sample map using a feature extraction model to determine a sample spatial structure feature representation of the sample molecule; a prediction module for determining attribute prediction information of the sample molecule based on the sample spatial structure feature representation; and a training module for training the feature extraction model using the attribute prediction information of the sample molecule and a label of the sample molecule to obtain a trained feature extraction model, wherein the label is used to characterize the attribute information of the sample molecule.

[0009] According to another aspect of the present disclosure, a molecular attribute identification device is provided, comprising: a feature representation determination module for processing a molecule to be identified using the data processing device described in the present disclosure to determine a spatial structural feature representation of the molecule to be identified; and an identification module for determining target attribute information of the molecule to be identified based on the spatial structural feature representation.

[0010] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method as disclosed herein.

[0011] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to execute the method of the present disclosure.

[0012] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program, wherein the computer program implements the method of the present disclosure when executed by a processor.

[0013] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0015] Figure 1 Schematically illustrates an exemplary system architecture to which the molecular property identification method and apparatus according to an embodiment of the present disclosure can be applied;

[0016] Figure 2 The following schematically shows a flow chart of a data processing method according to an embodiment of the present disclosure;

[0017] Figure 3A Schematically shows a three-dimensional structural diagram of a methylamine molecule according to an embodiment of the present disclosure;

[0018] Figure 3B Schematic diagram of the atom-chemical bond target submap of the methylamine molecule according to an embodiment of the present disclosure is schematically shown;

[0019] Figure 3C Schematic diagram of the chemical bond-bond angle target submap of the methylamine molecule according to an embodiment of the present disclosure is schematically shown;

[0020] Figure 3D Schematic diagram of the bond angle-dihedral angle target submap of the methylamine molecule according to an embodiment of the present disclosure is schematically shown;

[0021] Figure 4 Schematically shows a network structure diagram of a feature extraction model according to an embodiment of the present disclosure;

[0022] Figure 5 The schematic diagram of the molecular structure of 2-butene according to an embodiment of the present disclosure is shown;

[0023] Figure 6 A flowchart of a method for training a feature extraction model according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 7 Schematically shows a flow chart of a molecular attribute identification method according to an embodiment of the present disclosure;

[0025] Figure 8 A block diagram schematically shows a data processing device according to an embodiment of the present disclosure;

[0026] Figure 9 A block diagram schematically illustrates a training device for a feature extraction model according to an embodiment of the present disclosure;

[0027] Figure 10 A block diagram schematically shows a molecular attribute identification device according to an embodiment of the present disclosure; and

[0028] Figure 11A block diagram schematically shows an electronic device suitable for implementing a content processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0030] The present disclosure provides a data processing method, a feature extraction model training method, a molecular attribute recognition method, a device, an electronic device, a storage medium, and a program product.

[0031] According to an embodiment of the present disclosure, a data processing method is provided, comprising: determining a target map using molecular structure information of a molecule to be processed, wherein the target map is used to characterize spatial structure information of multiple molecules in different dimensions; and determining a spatial structure feature representation of the molecule to be processed based on the target map.

[0032] According to an embodiment of the present disclosure, a method for training a feature extraction model is provided, comprising: determining a target sample map using molecular structure information of sample molecules, wherein the target sample map is used to characterize multiple molecular spatial structure information of different dimensions; processing the target sample map using a feature extraction model to determine a sample spatial structure feature representation of the sample molecules; determining attribute prediction information of the sample molecules based on the sample spatial structure feature representation; and training a feature extraction model using the attribute prediction information of the sample molecules and labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules.

[0033] According to an embodiment of the present disclosure, a molecular attribute identification method is provided, comprising: processing a molecule to be identified using a method as disclosed herein to determine a spatial structural feature representation of the molecule to be identified; and determining target attribute information of the molecule to be identified based on the spatial structural feature representation.

[0034] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0035] Figure 1 An exemplary system architecture to which the molecular property identification method and apparatus according to an embodiment of the present disclosure can be applied is schematically shown.

[0036] It should be noted that Figure 1The examples shown are merely examples of system architectures to which the embodiments of the present disclosure can be applied, to help those skilled in the art understand the technical content of the present disclosure. This does not mean that the embodiments of the present disclosure cannot be applied to other devices, systems, environments, or scenarios. For example, in another embodiment, an exemplary system architecture to which the molecular property identification method and apparatus can be applied may include a terminal device, but the terminal device may implement the molecular property identification method and apparatus provided by the embodiments of the present disclosure without interacting with a server.

[0037] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0038] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software (for example only).

[0039] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0040] The server 105 may be a server that provides various services, such as a background management server (for example only) that supports the content browsed by users using the terminal devices 101, 102, and 103. The background management server may analyze and process received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0041] It should be noted that the molecular property identification method provided in the embodiment of the present disclosure can generally be executed by the terminal device 101, 102, or 103. Accordingly, the molecular property identification apparatus provided in the embodiment of the present disclosure can also be provided in the terminal device 101, 102, or 103.

[0042] Alternatively, the molecular property identification method provided in the embodiment of the present disclosure may also be generally performed by the server 105. Accordingly, the molecular property identification apparatus provided in the embodiment of the present disclosure may generally be provided in the server 105. The molecular property identification method provided in the embodiment of the present disclosure may also be performed by a server or server cluster that is different from the server 105 and that is capable of communicating with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the molecular property identification apparatus provided in the embodiment of the present disclosure may also be provided in a server or server cluster that is different from the server 105 and that is capable of communicating with the terminal devices 101, 102, 103 and / or the server 105.

[0043] For example, when a user wants to use a molecular expression of a drug molecule to obtain target attribute information of the drug molecule, such as drug toxicity, drug ligand, and protein receptor affinity, the terminal devices 101, 102, and 103 can obtain the molecular expression input by the user and then send the obtained molecular expression to the server 105. The server 105 analyzes the molecular expression to determine the molecular structure information of the drug molecule; determines a target map based on the molecular structure information; determines a spatial structural feature representation of the drug molecule based on the target map; and determines the target attribute information of the drug molecule based on the spatial structural feature representation. Alternatively, a server or server cluster capable of communicating with the terminal devices 101, 102, and 103 and / or the server 105 can analyze the molecular expression of the drug molecule and ultimately determine the target attribute information of the drug molecule.

[0044] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0045] It should be noted that the sequence numbers of the operations in the following method are only used to indicate the operation for the purpose of description, and should not be regarded as indicating the order in which the operations should be performed. Unless explicitly stated, the method does not need to be performed in the order shown.

[0046] Figure 2 The flowchart of the data processing method according to the embodiment of the present disclosure is schematically shown.

[0047] like Figure 2 As shown, the method includes operations S210 to S220.

[0048] In operation S210 , a target map is determined using the molecular structure information of the molecule to be processed, wherein the target map is used to represent a plurality of molecular spatial structure information of different dimensions.

[0049] In operation S220 , a spatial structural feature representation of the molecule to be processed is determined based on the target map.

[0050] According to an embodiment of the present disclosure, molecular structure information may include molecular property information, but is not limited thereto, and may also include molecular spatial structure information. The molecular structure information may include one or more information of atoms, chemical bonds, bond lengths, bond angles, and dihedral angles.

[0051] According to an embodiment of the present disclosure, the target map may also be referred to as a target topology map. The target map may include vertices and edges, where vertices may be used to represent an object and edges may be used to represent the association relationship between multiple objects. For example, the atoms of the molecule to be processed correspond to the vertices in the target map, and the chemical bonds between atoms may correspond to the edges in the target map, i.e., the atom-chemical bond target submap. However, it is not limited to this. The molecular spatial information may also include information such as bond length, bond angle, dihedral angle, etc., and the target map may include multiple target submaps, including, in addition to the atom-chemical bond target submap, one or more of, for example, a chemical bond-bond angle target submap and a bond angle-dihedral angle target submap.

[0052] According to embodiments of the present disclosure, the spatial structure feature representation can be referred to as a spatial structure feature vector, a spatial structure feature sequence, etc. The spatial structure feature representation can be used to characterize the spatial structure information and properties of the molecules to be processed, and then, based on the spatial structure feature representation, methods such as deep learning can be used to predict properties of the molecules to be processed, such as drug toxicity, drug ligands, and protein receptor affinity.

[0053] According to an embodiment of the present disclosure, a target map can be used to characterize multiple molecular spatial structure information of different dimensions. Multiple molecular spatial structure information can be constructed based on a variety of information including molecular structure information such as atoms, chemical bonds, bond lengths, bond angles, dihedral angles, etc. This can make the information carried in the target map sufficient and complete, and thus make the information represented by the generated spatial structure feature able to characterize more detailed spatial structure information of the molecules to be processed, solving the problem that the spatial structure feature representation does not fully capture the spatial structure information of the molecules to be processed due to the incomplete information carried by the target map.

[0054] According to an embodiment of the present disclosure, before performing operation S210 to determine the target spectrum using the molecular structure information of the molecule to be processed, the data processing method may further perform the following operations.

[0055] For example, a molecular expression of the molecule to be processed is determined, and based on the molecular expression, molecular structure information of the molecule to be processed is determined.

[0056] According to an embodiment of the present disclosure, the molecular expression may include a simplified molecular input line entry specification (SMILES), but is not limited thereto and may also include other standardized expressions that explicitly describe the molecular structure using ASCII strings.

[0057] According to the embodiments of the present disclosure, the molecular structure information of the molecules to be processed can be obtained based on molecular expressions through molecular dynamics simulation or experimental measurement, which will not be described in detail here.

[0058] Figure 3A The three-dimensional structure diagram of the methylamine molecule according to an embodiment of the present disclosure is schematically shown.

[0059] like Figure 3A As shown, methylamine, CH3-NH2, consists of carbon atoms, nitrogen atoms, and hydrogen atoms. The carbon atom is connected to each of the three hydrogen atoms via a C-H bond, the carbon atom is connected to the nitrogen atom via a C-N bond, and the nitrogen atom is connected to each of the two hydrogen atoms via an N-H bond. The C-N, C-H, and N-H bonds have different bond lengths, and the bond angles between the multiple bonds vary. Constructing an atom-to-bond target submap using atoms and bonds alone is insufficient to fully characterize the spatial structure of the molecule being processed.

[0060] Figure 3B A schematic diagram of the atom-chemical bond target submap of the methylamine molecule according to an embodiment of the present disclosure is schematically shown.

[0061] like Figure 3B As shown, an atom-bond target subgraph 330 can be determined based on atoms 310 and chemical bonds 320 of a molecule to be processed, such as a methylamine molecule. Atoms 310 can be used as nodes and chemical bonds 320 as edges to construct atom-bond target subgraph 330. Atom-bond target subgraph 330 can be used to reflect the relationships and properties between atoms.

[0062] Figure 3C A schematic diagram of the chemical bond-bond angle target submap of the methylamine molecule according to an embodiment of the present disclosure is shown schematically.

[0063] like Figure 3C As shown, a target subgraph 350 of chemical bonds and bond angles can be determined based on the chemical bonds 320 and bond angles 340 of the molecule to be processed. The target subgraph 350 of chemical bonds and bond angles can be constructed by using chemical bonds 320 as nodes and bond angles 340 as edges to reflect the relationships and properties between chemical bonds.

[0064] Figure 3D A schematic diagram of the bond angle-dihedral angle target submap of the methylamine molecule according to an embodiment of the present disclosure is schematically shown.

[0065] like Figure 3D As shown, a bond angle-dihedral target subgraph 370 can be determined based on the bond angles 340 and dihedral angles 360 of the molecule to be processed. The bond angle-dihedral target subgraph 370 can be constructed by using the bond angles 340 as nodes and the dihedral angles 360 as edges to reflect the relationships and properties between the bond angles.

[0066] According to the embodiments of the present disclosure, the spatial structure information and the property information of the molecule to be processed can be reflected from different dimensions through multiple target sub-maps such as the atom-chemical bond target sub-map, the chemical bond-bond angle target sub-map and the bond angle-dihedral angle target sub-map, so that the spatial structure feature representation of the molecule to be processed determined based on the target map contains rich molecular spatial structure information and property information, thereby making it more accurate and fast to predict biochemical property information closely related to the spatial structure information based on the spatial structure feature representation. For example, the target map includes, for example, a bond angle-dihedral angle target sub-map, which can make it possible to use the spatial structure feature representation to characterize the information of the dihedral angle formed between three adjacent chemical bonds inside the molecule to be processed, thereby making the spatial structure feature representation more accurately capture the spatial structure information of, for example, isomers, etc., avoiding the problem that the spatial structure feature representation has limitations in capturing the spatial structure information of the molecule to be processed due to incomplete target map information, thereby improving the generalization ability of the application of the spatial structure feature representation.

[0067] According to an embodiment of the present disclosure, for operation S220, determining the spatial structural feature representation of the molecule to be processed based on the target spectrum may include: inputting the target spectrum into a feature extraction model to obtain the spatial structural feature representation of the molecule to be processed.

[0068] According to an embodiment of the present disclosure, the feature extraction model may include a graph network (GNN, Graph Neural Networks) or a fully connected network (Transformer), but is not limited thereto. It may also include a network built based on a graph network and adapted to multiple target sub-graphs, Enhanced Geo GNN.

[0069] According to an embodiment of the present disclosure, the multiple target sub-maps in the target map are each used to characterize spatial structural information of different dimensions of the molecule. The feature extraction model may include M cascaded layers of feature extraction sub-models, each of which includes N feature extraction modules corresponding one-to-one to the N target sub-maps, where N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1. The multiple feature extraction modules in the feature extraction sub-model are used to match the multiple target sub-maps to achieve the extraction of spatial structural information of multiple different dimensions.

[0070] According to an embodiment of the present disclosure, inputting a target spectrum into a feature extraction model to obtain a spatial structural feature representation of the molecule to be processed may include the following operations.

[0071] For example, for the m+1th layer feature extraction sub-model in the feature extraction model, when n is not equal to 1, the m+1th layer nth target sub-feature vector is determined based on the feature vector of the m+1th layer nth target sub-map and the m+1th layer n+1th target sub-feature vector, where m=1, ..., M-1, and n=2, ..., N. The m+1th layer 1st feature extraction module is used to determine the m+1th layer target feature vector based on the m+1th layer 2nd target sub-feature vector and the mth layer target feature vector. The mth layer target feature vector is used as the spatial structural feature representation of the molecule to be processed.

[0072] According to an embodiment of the present disclosure, the feature extraction model may include a cascade of M-layer feature extraction sub-models. The first-layer feature extraction sub-model can be used to capture the features and spatial structure information of the first-order neighbors of each of the multiple nodes in the target graph, and the second-layer feature extraction sub-model can be used to capture the features and spatial structure information of the second-order neighbors of each of the multiple nodes in the target graph. Similarly, the cascade of M-layer feature extraction sub-models can be used to capture the features and spatial structure of the n-order neighbors of each of the multiple nodes in the target graph. In this way, high-order three-dimensional spatial structure feature information can be captured and modeled, and rich and complex spatial structure information can be integrated into the spatial structure feature representation, so that the spatial structure feature representation can represent more comprehensive information, and then use the spatial structure feature representation to perform various prediction tasks, such as judging molecular toxicity, identifying targeted drugs, predicting drug combinations, etc., more accurately.

[0073] According to an embodiment of the present disclosure, the target map may include N target sub-maps. N may be an integer greater than or equal to 2, for example, N is an integer of 2, 3, 4 or a greater integer. When N is 2, the target map may include a first target sub-map and a second target sub-map. When N is 3, the target map may include a first target sub-map, a second target sub-map and a third target sub-map. For example, the first target sub-map includes an atom-chemical bond target sub-map. The second target sub-map includes a chemical bond-bond angle target sub-map. The third target sub-map includes a bond angle-dihedral angle target sub-map.

[0074] According to an embodiment of the present disclosure, the (m+1)th n-th feature extraction module includes a cascaded (m+1)th n-th input layer (Embedding), an (m+1)th n-th graph network layer, and an (m+1)th n-th residual network layer, where n is greater than 1. For example, the (m+1)th n-th feature extraction module can be used to process the second target subgraph, such as a chemical bond-bond angle target subgraph, and the (m+1)th n+1-th feature extraction module can be used to process the third target subgraph, such as a bond angle-dihedral angle target subgraph.

[0075] According to an embodiment of the present disclosure, utilizing the nth feature extraction module of the m+1th layer, based on the feature vector of the nth target subgraph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer, determining the nth target sub-feature vector of the m+1th layer includes: inputting the nth target subgraph of the m+1th layer into the nth input layer of the m+1th layer to obtain the feature vector of the nth target subgraph of the m+1th layer. Inputting the feature vector of the nth target subgraph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer into the nth graph network layer of the m+1th layer to obtain the nth initial sub-feature vector of the m+1th layer. Inputting the feature vector of the nth target subgraph of the m+1th layer and the nth initial sub-feature vector of the m+1th layer into the nth residual network layer of the m+1th layer to obtain the nth target sub-feature vector of the m+1th layer.

[0076] According to an embodiment of the present disclosure, the Nth target subgraph of the m+1th layer is input into the Nth input layer of the m+1th layer using the Nth feature extraction module to obtain the feature vector of the Nth target subgraph of the m+1th layer. The feature vector of the Nth target subgraph of the m+1th layer is input into the Nth graph network layer of the m+1th layer to obtain the Nth initial sub-feature vector of the m+1th layer. The feature vector of the Nth target subgraph of the m+1th layer and the Nth initial sub-feature vector of the m+1th layer are input into the Nth residual network layer of the m+1th layer to obtain the Nth target sub-feature vector of the m+1th layer.

[0077] According to an embodiment of the present disclosure, the (m+1)th layer 1 feature extraction module includes a cascaded (m+1)th layer 1 graph network layer and an (m+1)th layer 1 residual network layer. For example, the (m+1)th layer 1 feature extraction module can be used to process a first target subgraph, such as an atom-chemical bond target subgraph.

[0078] According to an embodiment of the present disclosure, determining the target feature vector of the m+1th layer using the first feature extraction module of the m+1th layer based on the second target sub-feature vector of the m+1th layer and the target feature vector of the mth layer includes: obtaining the first initial sub-feature vector of the m+1th layer based on the second target sub-feature vector of the m+1th layer and the target feature vector of the mth layer using the first residual network layer of the m+1th layer based on the first initial sub-feature vector of the m+1th layer and the target feature vector of the mth layer, and obtaining the target feature vector of the m+1th layer based on the first initial sub-feature vector of the m+1th layer and the target feature vector of the mth layer using the first residual network layer of the m+1th layer.

[0079] According to an embodiment of the present disclosure, the first layer, the first feature extraction module can be a cascaded first layer, the first input layer, the first graph network layer and the first residual network layer. Using the first layer, the first feature extraction module, the first layer, the first input layer can be used to obtain the feature vector of the first target sub-graph based on the first target sub-graph. Using the first layer, the first graph network layer can be used to obtain the first layer, the first initial sub-feature vector based on the first layer, the second target sub-feature vector and the feature vector of the first target sub-graph. Using the first layer, the first residual network layer can be used to obtain the first layer, the first target feature vector based on the first initial sub-feature vector and the feature vector of the first target sub-graph.

[0080] According to the embodiments of the present disclosure, multiple feature extraction modules within a feature extraction sub-model can be used to process multiple target sub-graphs of different dimensions, fusing spatial structural information from these different dimensions to obtain target sub-feature vectors. Furthermore, by combining multiple layers of feature extraction sub-models, higher-order three-dimensional spatial structural feature information can be captured, allowing the spatial structural feature representation to extract complete spatial structural information.

[0081] Figure 4 The network structure diagram of the feature extraction model according to the embodiment of the present disclosure is schematically shown.

[0082] like Figure 4 As shown, the feature extraction model includes two layers of feature extraction sub-models, such as the first layer feature extraction sub-model 410 and the second layer feature extraction sub-model 420. Each layer of feature extraction sub-model includes three feature extraction modules, such as the first layer feature extraction module, the second layer feature extraction module, and the third layer feature extraction module.

[0083] For the first-layer feature extraction submodel 410 in the feature extraction model, the bond angle-dihedral angle target subgraph 450 is input into the first-layer third input layer 1 4131 of the first-layer third feature extraction module to obtain a feature vector for the first-layer third target subgraph. The feature vector for the first-layer third target subgraph is input into the first-layer third graph network layer GNN1 4132 to obtain a first-layer third initial sub-feature vector. The feature vector for the first-layer third target subgraph and the first-layer third initial sub-feature vector are input into the first-layer third residual network layer 4133 to obtain a first-layer third target sub-feature vector.

[0084] The chemical bond-bond angle target subgraph 440 is input into the first-layer second input layer 1 4121 of the first-layer second feature extraction module to obtain a feature vector of the first-layer second target subgraph. The feature vector of the first-layer second target subgraph and the first-layer third target subfeature vector are input into the first-layer second graph network layer GNN1 4122 to obtain a first-layer second initial subfeature vector. The feature vector of the first-layer second target subgraph and the first-layer second initial subfeature vector are input into the first-layer second residual network layer 4123 to obtain a first-layer second target subfeature vector.

[0085] The atom-chemical bond target subgraph 430 is input into the first input layer 1 4111 of the first feature extraction module of the first layer to obtain a feature vector of the first target subgraph of the first layer. The feature vector of the first target subgraph of the first layer and the second target sub-feature vector of the first layer are input into the first graph network layer GNN1 4112 of the first layer to obtain a first initial sub-feature vector of the first layer. The feature vector of the first target subgraph of the first layer and the first initial sub-feature vector of the first layer are input into the first residual network layer 4113 of the first layer to obtain a first target feature vector of the first layer.

[0086] For the second-layer feature extraction submodel in the feature extraction model, the bond angle-dihedral angle target subgraph 450 is input into the second-layer third input layer 2 4231 of the second-layer third feature extraction module to obtain a feature vector for the second-layer third target subgraph. The feature vector for the second-layer third target subgraph is input into the second-layer third graph network layer GNN2 4232 to obtain a second-layer third initial sub-feature vector. The feature vector for the second-layer third target subgraph and the second-layer third initial sub-feature vector are input into the second-layer third residual network layer 4233 to obtain a second-layer third target sub-feature vector.

[0087] The chemical bond-bond angle target subgraph 440 is input into the second input layer 2 4221 of the second feature extraction module to obtain a feature vector of the second target subgraph. The feature vector of the second target subgraph and the third target sub-feature vector are input into the second graph network layer GNN2 4222 to obtain a second initial sub-feature vector. The feature vector of the second target subgraph and the second initial sub-feature vector are input into the second residual network layer 4223 to obtain a second target sub-feature vector.

[0088] The first target feature vector of the first layer and the second target sub-feature vector of the second layer are input into the first graph network layer GNN2 4212 of the second layer to obtain the first initial sub-feature vector of the second layer. The first initial sub-feature vector of the second layer and the first initial sub-feature vector of the second layer are input into the first residual network layer 4213 of the second layer to obtain the first target feature vector of the second layer. The first target feature vector of the second layer is used as the spatial structure feature representation 460.

[0089] It should be noted that Figure 4 The illustrated embodiment is merely an exemplary embodiment. m can be 1, but is not limited to this, and can also be any integer such as 2 or 3. However, the more layers of the feature extraction sub-model, the greater the data processing volume, and the more complete the spatial structure information captured by the spatial structure feature representation. The number of m can be determined based on actual needs. In addition, n can be 2 or 3, but is not limited to this, and can also be other integers greater than 1. This can be determined based on the number of target sub-maps, and will not be further explained here.

[0090] Figure 5 The schematic diagram of the molecular structure of 2-butene according to an embodiment of the present disclosure is shown schematically.

[0091] like Figure 5 As shown, 2-butene includes two stereo structures, such as cis-2-butene and trans-2-butene, which are isomers of each other.

[0092] According to an embodiment of the present disclosure, a target map including an atom-chemical bond target submap, a chemical bond-bond angle target submap and a bond angle-dihedral angle target submap is used to determine the spatial structure feature representation of the molecule to be processed. This allows the spatial structure feature representation to capture a more complete three-dimensional spatial structure of the molecule to be processed, for example, by utilizing the dihedral angle information formed between three adjacent chemical bonds of the molecule to be processed, so that the spatial structure feature representation can characterize fine and important spatial structure information such as isomer information, thereby improving the generalization performance of the spatial structure feature representation.

[0093] Figure 6A flowchart of a method for training a feature extraction model according to an embodiment of the present disclosure is schematically shown.

[0094] like Figure 6 As shown, the method includes operations S610 to S640.

[0095] In operation S610 , a target sample map is determined using molecular structure information of sample molecules, wherein the target sample map is used to represent multiple molecular spatial structure information of different dimensions.

[0096] In operation S620, the target sample map is processed using a feature extraction model to determine a sample spatial structure feature representation of the sample molecules.

[0097] In operation S630 , attribute prediction information of the sample molecules is determined based on the sample spatial structure feature representation.

[0098] In operation S640 , a feature extraction model is trained using the attribute prediction information of the sample molecules and the labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules.

[0099] According to an embodiment of the present disclosure, a pre-trained molecular property prediction model may be used to process a sample spatial structure feature representation to obtain property prediction information of the sample molecules.

[0100] According to the embodiments of the present disclosure, the attribute prediction information and the label may not be limited. For example, it may be attribute information such as solubility, toxicity, absorption, metabolism, etc. of the sample molecule, but it is not limited to this. It can be determined according to actual conditions, as long as the attribute prediction information and the label are of the same category.

[0101] According to an embodiment of the present disclosure, a feature extraction model is trained using a target sample spectrum for characterizing multiple molecular spatial structure information of different dimensions, so that the trained feature extraction model can be adapted to extract features from target spectra with multiple spatial structure information of different dimensions, thereby obtaining a spatial structure feature representation with complete information.

[0102] According to an embodiment of the present disclosure, multiple target sub-sample maps in the target sample map are each used to characterize spatial structural information of molecules in different dimensions. The feature extraction model may include M-layer cascaded feature extraction sub-models, and each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to N target sub-sample maps, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

[0103] According to an embodiment of the present disclosure, inputting a target sample spectrum into a feature extraction model to obtain a spatial structural feature representation of the molecule to be processed may include the following operations.

[0104] For example, for the feature extraction sub-model at the m+1th layer in the feature extraction model, when n is not equal to 1, the feature extraction module at the m+1th layer, n, is used to determine the target sub-feature vector at the m+1th layer, n, based on the feature vector of the target sub-sample map at the m+1th layer and the target sub-feature vector at the m+1th layer, n+1, where m=1, ..., M-1, and n=2, ..., N. The first feature extraction module at the m+1th layer is used to determine the target feature vector at the m+1th layer, m+1, based on the second target sub-feature vector at the m+1th layer and the target feature vector at the mth layer. The target feature vector at the mth layer is used as the spatial structural feature representation of the molecule to be processed.

[0105] According to an embodiment of the present disclosure, the target sample map may include N target sub-sample maps. N may be an integer greater than or equal to 2, for example, N is an integer of 2, 3, 4 or a greater integer. When N is 2, the target sample map may include a first target sub-sample map and a second target sub-sample map. When N is 3, the target sample map may include a first target sub-sample map, a second target sub-sample map, and a third target sub-sample map. For example, the first target sub-sample map includes an atom-chemical bond target sub-sample map. The second target sub-sample map includes a chemical bond-bond angle target sub-sample map. The third target sub-sample map includes a bond angle-dihedral angle target sub-sample map.

[0106] According to an embodiment of the present disclosure, the (m+1)th n-th feature extraction module includes a cascaded (m+1)th n-th input layer, an (m+1)th n-th graph network layer, and an (m+1)th n-th residual network layer, where n is greater than 1. For example, the (m+1)th n-th feature extraction module can be used to process a second target subsample map, such as a chemical bond-bond angle target subsample map, and the (m+1)th n+1-th feature extraction module can be used to process a third target subsample map, such as a bond angle-dihedral angle target subsample map.

[0107] According to an embodiment of the present disclosure, using the nth feature extraction module of the m+1th layer, based on the feature vector of the nth target subsample map of the m+1th layer and the feature vector of the nth target subsample of the m+1th layer, determining the feature vector of the nth target subsample of the m+1th layer includes: inputting the nth target subsample map of the m+1th layer into the nth input layer of the m+1th layer to obtain the feature vector of the nth target subsample map of the m+1th layer. Inputting the feature vector of the nth target subsample map of the m+1th layer and the feature vector of the nth target subsample of the m+1th layer into the nth graph network layer of the m+1th layer to obtain the feature vector of the nth initial subsample of the m+1th layer. Inputting the feature vector of the nth target subsample map of the m+1th layer and the feature vector of the nth initial subsample of the m+1th layer into the nth residual network layer of the m+1th layer to obtain the feature vector of the nth target subsample of the m+1th layer.

[0108] According to an embodiment of the present disclosure, the (m+1)th layer 1 feature extraction module includes a cascaded (m+1)th layer 1 graph network layer and an (m+1)th layer 1 residual network layer. For example, the (m+1)th layer 1 feature extraction module can be used to process a first target subsample map, such as an atom-chemical bond target subsample map.

[0109] According to an embodiment of the present disclosure, using the first feature extraction module of the m+1th layer to determine the target sample feature vector of the m+1th layer based on the second target subsample feature vector of the m+1th layer and the target sample feature vector of the mth layer includes: using the first graph network layer of the m+1th layer to obtain the first initial subsample feature vector of the m+1th layer based on the second target subsample feature vector of the m+1th layer and the target sample feature vector of the mth layer. Using the first residual network layer of the m+1th layer to obtain the target sample feature vector of the m+1th layer based on the first initial subsample feature vector of the m+1th layer and the target sample feature vector of the mth layer.

[0110] Figure 7 The flowchart of molecular attribute identification according to an embodiment of the present disclosure is schematically shown.

[0111] like Figure 7 As shown, the method includes operations S710 to S720.

[0112] In operation S710, the molecule to be identified is processed using a data processing method to determine a spatial structural feature representation of the molecule to be identified.

[0113] In operation S720 , target attribute information of the molecule to be identified is determined based on the spatial structure feature representation.

[0114] According to an embodiment of the present disclosure, the target attribute information of the molecule to be identified may include absorption, distribution, metabolism, excretion, and toxicity, but is not limited thereto. The target attribute information of the molecule to be identified may also include attribute information of drug combination (DDI, Drug-Drug Interaciton) and attribute information of diffusion tensor imaging (DTI, Diffusion Tensor Imaging).

[0115] According to the embodiments of the present disclosure, the use of spatial structure feature representations that carry multi-dimensional and fine spatial structure information to predict target attribute information of molecules to be identified, such as drug molecules, can be more accurate and effective, thereby improving the efficiency of drug research and development, reducing costs, and expanding the scope of research and development.

[0116] Figure 8 The block diagram schematically shows a data processing device according to an embodiment of the present disclosure.

[0117] like Figure 8 As shown, the data processing device 800 may include a first determining module 810 and a second determining module 820 .

[0118] The first determination module 810 is configured to determine a target map using the molecular structure information of the molecule to be processed, wherein the target map is used to represent multiple molecular spatial structure information of different dimensions.

[0119] The second determination module 820 is used to determine the spatial structural feature representation of the molecule to be processed based on the target map.

[0120] According to an embodiment of the present disclosure, the second determining module may include an extracting submodule.

[0121] The extraction submodule is used to input the target spectrum into the feature extraction model to obtain the spatial structural feature representation of the molecule to be processed.

[0122] According to an embodiment of the present disclosure, the target graph includes N target sub-graphs; the feature extraction model includes M cascaded layers of feature extraction sub-models, each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to the N target sub-graphs, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

[0123] According to an embodiment of the present disclosure, the extraction submodule may include a first extraction unit, a second extraction unit, and a first determination unit.

[0124] The first extraction unit is used to determine the nth target sub-feature vector of the m+1th layer based on the feature vector of the nth target sub-graph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer using the nth feature extraction module of the m+1th layer when n is not equal to 1, wherein m=1, ..., M-1, and n=2, ..., N.

[0125] The second extraction unit is used to determine the m+1th layer target feature vector based on the m+1th layer second target sub-feature vector and the mth layer target feature vector by using the m+1th layer first feature extraction module.

[0126] The first determining unit is configured to use the target feature vector of the Mth layer as a spatial structural feature representation of the molecule to be processed.

[0127] According to an embodiment of the present disclosure, the m+1th layer nth feature extraction module includes a cascaded m+1th layer nth input layer, an m+1th layer nth graph network layer, and an m+1th layer nth residual network layer, where n is greater than 1.

[0128] According to an embodiment of the present disclosure, the first extraction unit may include a first input subunit, a first extraction subunit, and a first fusion subunit.

[0129] The first input subunit is used to input the nth target sub-graph of the m+1th layer into the nth input layer of the m+1th layer to obtain the feature vector of the nth target sub-graph of the m+1th layer.

[0130] The first extraction subunit is used to input the feature vector of the nth target sub-graph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer into the nth graph network layer of the m+1th layer to obtain the nth initial sub-feature vector of the m+1th layer.

[0131] The first fusion subunit is used to input the feature vector of the nth target sub-graph of the m+1th layer and the nth initial sub-feature vector of the m+1th layer into the nth residual network layer of the m+1th layer to obtain the nth target sub-feature vector of the m+1th layer.

[0132] According to an embodiment of the present disclosure, the (m+1)th layer 1st feature extraction module includes a cascaded (m+1)th layer 1st graph network layer and an (m+1)th layer 1st residual network layer.

[0133] According to an embodiment of the present disclosure, the second extraction unit may include a second extraction subunit and a second fusion subunit.

[0134] The second extraction subunit is used to obtain the m+1th layer first initial sub-feature vector based on the m+1th layer second target sub-feature vector and the mth layer target feature vector by using the m+1th layer first graph network layer.

[0135] The second fusion subunit is used to obtain the m+1th layer target feature vector based on the m+1th layer first initial sub-feature vector and the mth layer target feature vector using the m+1th layer first residual network layer.

[0136] According to an embodiment of the present disclosure, the target map includes an atom-chemical bond target submap, a chemical bond-bond angle target submap, and a bond angle-dihedral angle target submap.

[0137] According to an embodiment of the present disclosure, the first determination module may include a first map determination unit, a second map determination unit, and a third map determination unit.

[0138] The first spectrum determination unit is used to determine an atom-chemical bond target sub-spectrum based on the atoms of the molecule to be processed and the chemical bonds of the molecule to be processed.

[0139] The second spectrum determination unit is used to determine a chemical bond-bond angle target sub-spectrum based on the chemical bonds and bond angles of the molecules to be processed.

[0140] The third spectrum determination unit is used to determine the bond angle-dihedral angle target sub-spectrum based on the bond angle of the molecule to be processed and the dihedral angle of the molecule to be processed.

[0141] According to an embodiment of the present disclosure, the data processing apparatus may further include, before the first determination module: an expression determination module and a structure determination module.

[0142] The expression determination module is used to determine the molecular expression of the molecule to be processed.

[0143] The structure determination module is used to determine the molecular structure information of the molecule to be processed based on the molecular expression.

[0144] Figure 9 A block diagram of a training device for a feature extraction model according to an embodiment of the present disclosure is schematically shown.

[0145] like Figure 9 As shown, the training device 900 for the feature extraction model may include a sample map determination module 910 , a sample extraction module 920 , a prediction module 930 and a training module 940 .

[0146] The sample map determination module 910 is used to determine a target sample map using the molecular structure information of the sample molecules, wherein the target sample map is used to represent multiple molecular spatial structure information of different dimensions.

[0147] The sample extraction module 920 is used to process the target sample map using a feature extraction model to determine the sample spatial structure feature representation of the sample molecules.

[0148] The prediction module 930 is used to determine the attribute prediction information of the sample molecules based on the sample spatial structure feature representation.

[0149] The training module 940 is used to train a feature extraction model using the attribute prediction information of the sample molecules and the labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules.

[0150] Figure 10 The block diagram of the molecular attribute identification device according to an embodiment of the present disclosure is schematically shown.

[0151] like Figure 10 As shown, the molecular attribute identification apparatus 1000 may include a feature representation determination module 1010 and an identification module 1020 .

[0152] The feature representation determination module 1010 is used to process the molecules to be identified using the data processing device of the present disclosure to determine the spatial structure feature representation of the molecules to be identified.

[0153] The identification module 1020 is used to determine target attribute information of the molecule to be identified based on the spatial structure feature representation.

[0154] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0155] According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method as in the embodiment of the present disclosure.

[0156] According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute a method according to an embodiment of the present disclosure.

[0157] According to an embodiment of the present disclosure, a computer program product includes a computer program. When the computer program is executed by a processor, the method according to the embodiment of the present disclosure is implemented.

[0158] Figure 11 A schematic block diagram of an example electronic device 1100 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0159] like Figure 11 As shown, the device 1100 includes a computing unit 1101, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1102 or a computer program loaded from a storage unit 1108 into a random access memory (RAM) 1103. Various programs and data required for the operation of the device 1100 can also be stored in the RAM 1103. The computing unit 1101, the ROM 1102, and the RAM 1103 are connected to each other via a bus 1104. An input / output (I / O) interface 1105 is also connected to the bus 1104.

[0160] Various components in device 1100 are connected to I / O interface 1105, including an input unit 1106, such as a keyboard and mouse; an output unit 1107, such as various types of displays and speakers; a storage unit 1108, such as a magnetic disk and optical disk; and a communication unit 1109, such as a network card, a modem, a wireless communication transceiver, etc. Communication unit 1109 allows device 1100 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0161] The computing unit 1101 can be various general and / or special processing components with processing and computing power. Some examples of the computing unit 1101 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units for running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1101 performs the various methods and processes described above, such as data processing methods, training methods for feature extraction models, or molecular attribute identification methods. For example, in some embodiments, data processing methods, training methods for feature extraction models, or molecular attribute identification methods can be implemented as computer software programs, which are tangibly included in machine-readable media, such as storage units 1108. In some embodiments, part or all of the computer program can be loaded and / or installed on device 1100 via ROM 1102 and / or communication unit 1109. When the computer program is loaded into RAM 1103 and executed by the computing unit 1101, one or more steps of the data processing methods, training methods for feature extraction models, or molecular attribute identification methods described above can be executed. Alternatively, in other embodiments, the computing unit 1101 may be configured to execute a data processing method, a feature extraction model training method, or a molecular property identification method in any other appropriate manner (eg, by means of firmware).

[0162] Various embodiments of the systems and techniques described above can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0163] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0164] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0165] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0166] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0167] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0168] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0169] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A data processing method, comprising: Determining a target map using molecular structure information of the molecule to be processed, wherein the target map is used to represent multiple molecular spatial structure information of different dimensions, and the target map includes an atom-chemical bond target submap, a chemical bond-bond angle target submap, and a bond angle-dihedral angle target submap; and Based on the target map, determining the spatial structural feature representation of the molecule to be processed; The method of determining the target spectrum using the molecular structure information of the molecule to be processed includes: Determining the atom-chemical bond target submap based on the atoms of the molecule to be processed and the chemical bonds of the molecule to be processed; Determining the chemical bond-bond angle target submap based on the chemical bonds of the molecule to be processed and the bond angles of the molecule to be processed; and Determining the bond angle-dihedral angle target submap based on the bond angle of the molecule to be processed and the dihedral angle of the molecule to be processed; Determining the spatial structural feature representation of the molecule to be processed based on the target map includes: The target map is input into a feature extraction model to obtain a spatial structural feature representation of the molecule to be processed; wherein the target map includes N target sub-maps; the feature extraction model includes M cascaded layers of feature extraction sub-models, each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to the N target sub-maps, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

2. The method according to claim 1, wherein Inputting the target spectrum into a feature extraction model to obtain a spatial structural feature representation of the molecule to be processed comprises: For the m+1th layer feature extraction submodel in the feature extraction model, when n is not equal to 1, using the m+1th layer nth feature extraction module, based on the feature vector of the m+1th layer nth target subgraph and the m+1th layer n+1th target subfeature vector, determine the m+1th layer nth target subfeature vector, where m=1, ..., M-1, and n=2, ..., N; Determine, using the first feature extraction module of the m+1th layer, a target feature vector of the m+1th layer based on the second target sub-feature vector of the m+1th layer and the target feature vector of the mth layer; and The M-th layer target feature vector is used as the spatial structural feature representation of the molecule to be processed.

3. The method according to claim 2, wherein: The m+1th layer nth feature extraction module includes a cascaded m+1th layer nth input layer, an m+1th layer nth graph network layer, and an m+1th layer nth residual network layer, wherein n is greater than 1; The method of using the nth feature extraction module of the m+1th layer to determine the nth target sub-feature vector of the m+1th layer based on the feature vector of the nth target sub-graph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer comprises: Input the nth target subgraph of the m+1th layer into the nth input layer of the m+1th layer to obtain the feature vector of the nth target subgraph of the m+1th layer; Inputting the feature vector of the nth target subgraph in the m+1th layer and the n+1th target sub-feature vector in the m+1th layer into the nth graph network layer in the m+1th layer to obtain the nth initial sub-feature vector in the m+1th layer; and The feature vector of the nth target sub-graph of the m+1th layer and the nth initial sub-feature vector of the m+1th layer are input into the nth residual network layer of the m+1th layer to obtain the nth target sub-feature vector of the m+1th layer.

4. The method according to claim 2, wherein: The m+1th layer 1st feature extraction module includes a cascaded m+1th layer 1st graph network layer and an m+1th layer 1st residual network layer; The determining of the target feature vector of the m+1th layer by using the first feature extraction module of the m+1th layer based on the second target sub-feature vector of the m+1th layer and the target feature vector of the mth layer includes: Using the m+1th layer 1 graph network layer, based on the m+1th layer 2 target sub-feature vector and the mth layer target feature vector, obtain the m+1th layer 1 initial sub-feature vector; and The m+1th layer target feature vector is obtained by utilizing the m+1th layer first residual network layer based on the m+1th layer first initial sub-feature vector and the mth layer target feature vector.

5. The method according to any one of claims 1 to 4, further comprising, before determining the target spectrum using the molecular structure information of the molecule to be processed: Determining the molecular expression of the molecule to be processed; and Based on the molecular expression, the molecular structure information of the molecule to be processed is determined.

6. A method for training a feature extraction model, comprising: Determine a target sample map using the molecular structure information of the sample molecules, wherein the target sample map is used to characterize multiple molecular spatial structure information of different dimensions, and the target sample map includes an atom-chemical bond target sub-sample map, a chemical bond-bond angle target sub-sample map, and a bond angle-dihedral angle target sub-sample map; Processing the target sample map using a feature extraction model to determine a sample spatial structural feature representation of the sample molecules; Determining attribute prediction information of the sample molecules based on the sample spatial structure feature representation; and Training the feature extraction model using the attribute prediction information of the sample molecules and the labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules; The method of determining the target sample profile using the molecular structure information of the sample molecules includes: determining the atom-chemical bond target subsample map based on the atoms of the sample molecules and the chemical bonds of the sample molecules; Determining the chemical bond-bond angle target subsample spectrum based on the chemical bonds of the sample molecules and the bond angles of the sample molecules; and Determining the bond angle-dihedral angle target subsample map based on the bond angle of the sample molecule and the dihedral angle of the sample molecule; The step of processing the target sample map using a feature extraction model to determine the sample spatial structure feature representation of the sample molecules includes: The target sample map is input into a feature extraction model to obtain a sample space structural feature representation of the sample molecule; wherein the target sample map includes N target sub-sample maps; the feature extraction model includes M cascaded layers of feature extraction sub-models, each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to the N target sub-sample maps, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

7. A method for identifying molecular properties, comprising: Processing the molecule to be identified using the method according to any one of claims 1 to 5 to determine the spatial structural feature representation of the molecule to be identified; as well as Based on the spatial structure feature representation, target attribute information of the molecule to be identified is determined.

8. A data processing device comprising: a first determination module, configured to determine a target map using molecular structure information of the molecule to be processed, wherein the target map is used to represent multiple molecular spatial structure information of different dimensions, and the target map includes an atom-chemical bond target submap, a chemical bond-bond angle target submap, and a bond angle-dihedral angle target submap; and A second determination module is used to determine the spatial structural feature representation of the molecule to be processed based on the target map; The first determining module includes: A first spectrum determination unit is used to determine the atom-chemical bond target sub-spectrum based on the atoms of the molecule to be processed and the chemical bonds of the molecule to be processed; A second spectrum determination unit is configured to determine the chemical bond-bond angle target sub-spectrum based on the chemical bonds of the molecule to be processed and the bond angles of the molecule to be processed; and A third spectrum determination unit is used to determine the bond angle-dihedral angle target sub-spectrum based on the bond angle of the molecule to be processed and the dihedral angle of the molecule to be processed; The second determining module includes: An extraction submodule is used to input the target map into a feature extraction model to obtain a spatial structural feature representation of the molecule to be processed; wherein, the target map includes N target sub-maps; the feature extraction model includes M cascaded layers of feature extraction sub-models, each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to the N target sub-maps, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

9. The device according to claim 8, wherein The extraction submodule includes: a first extraction unit, configured to determine, for the m+1th layer feature extraction submodel in the feature extraction model, a feature vector of the nth target sub-graph in the m+1th layer and a feature vector of the n+1th target sub-graph in the m+1th layer, using the nth feature extraction module in the m+1th layer when n is not equal to 1, where m=1, ..., M-1, and n=2, ..., N; A second extraction unit is configured to determine an m+1th layer target feature vector based on the m+1th layer second target sub-feature vector and the mth layer target feature vector using the m+1th layer first feature extraction module; and The first determining unit is configured to use the target feature vector of the Mth layer as the spatial structural feature representation of the molecule to be processed.

10. The device according to claim 9, wherein The m+1th layer nth feature extraction module includes a cascaded m+1th layer nth input layer, an m+1th layer nth graph network layer, and an m+1th layer nth residual network layer, wherein n is greater than 1; The first extraction unit includes: The first input subunit is used to input the nth target subgraph of the m+1th layer into the nth input layer of the m+1th layer to obtain the feature vector of the nth target subgraph of the m+1th layer; A first extraction subunit is configured to input the feature vector of the nth target subgraph of the m+1th layer and the n+1th target sub-feature vector of the m+1th layer into the nth graph network layer of the m+1th layer to obtain the nth initial sub-feature vector of the m+1th layer; and The first fusion subunit is used to input the feature vector of the nth target sub-graph of the m+1th layer and the nth initial sub-feature vector of the m+1th layer into the nth residual network layer of the m+1th layer to obtain the nth target sub-feature vector of the m+1th layer.

11. The device according to claim 9, wherein The m+1th layer 1st feature extraction module includes a cascaded m+1th layer 1st graph network layer and an m+1th layer 1st residual network layer; The second extraction unit includes: A second extraction subunit is configured to obtain an m+1th layer first initial sub-feature vector based on the m+1th layer second target sub-feature vector and the mth layer target feature vector by using the m+1th layer first graph network layer; as well as The second fusion subunit is used to obtain the m+1th layer target feature vector based on the m+1th layer first initial sub-feature vector and the mth layer target feature vector by using the m+1th layer first residual network layer.

12. The apparatus according to any one of claims 8 to 11, further comprising, before the first determining module: an expression determination module, configured to determine the molecular expression of the molecule to be processed; and The structure determination module is used to determine the molecular structure information of the molecule to be processed based on the molecular expression.

13. A training device for a feature extraction model, comprising: A sample map determination module is used to determine a target sample map using the molecular structure information of the sample molecules, wherein the target sample map is used to characterize multiple molecular spatial structure information of different dimensions, and the target sample map includes an atom-chemical bond target sub-sample map, a chemical bond-bond angle target sub-sample map, and a bond angle-dihedral angle target sub-sample map; A sample extraction module, configured to process the target sample map using a feature extraction model to determine a sample spatial structural feature representation of the sample molecule; A prediction module, configured to determine attribute prediction information of the sample molecules based on the sample spatial structure feature representation; and a training module, configured to train the feature extraction model using the attribute prediction information of the sample molecules and the labels of the sample molecules to obtain a trained feature extraction model, wherein the labels are used to characterize the attribute information of the sample molecules; The sample map determination module is used to: determine the atom-chemical bond target subsample map based on the atoms of the sample molecules and the chemical bonds of the sample molecules; determine the chemical bond-bond angle target subsample map based on the chemical bonds of the sample molecules and the bond angles of the sample molecules; and determine the bond angle-dihedral angle target subsample map based on the bond angles of the sample molecules and the dihedral angles of the sample molecules; The sample extraction module is used to: input the target sample spectrum into the feature extraction model to obtain the sample space structure feature representation of the sample molecule; wherein, the target sample spectrum includes N target sub-sample spectra; the feature extraction model includes M layers of cascaded feature extraction sub-models, each feature extraction sub-model includes N feature extraction modules corresponding one-to-one to the N target sub-sample spectra, wherein N is an integer greater than or equal to 2, and M is an integer greater than or equal to 1.

14. A molecular property recognition device, comprising: a feature representation determination module, configured to process a molecule to be identified using the data processing apparatus according to any one of claims 8 to 12, and determine a spatial structural feature representation of the molecule to be identified; as well as An identification module is used to determine target attribute information of the molecule to be identified based on the spatial structure feature representation.

15. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the data processing method described in any one of claims 1 to 5, the training method of the feature extraction model described in claim 6, or the molecular attribute identification method described in claim 7.

16. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to enable the computer to execute the data processing method according to any one of claims 1 to 5, the feature extraction model training method according to claim 6, or the molecular attribute identification method according to claim 7.

17. A computer program product, comprising a computer program, which, when executed by a processor, implements the data processing method according to any one of claims 1 to 5, the feature extraction model training method according to claim 6, or the molecular attribute identification method according to claim 7.

Images