Method for constructing mirror protein interaction map and related device

Abstract

This application proposes a method, device, electronic device, and storage medium for constructing a mirror protein interaction map. The method for constructing a mirror protein interaction map includes: constructing a first graph neural network based on the original protein, and inputting the amino acid sequence of the original protein into the first graph neural network to predict the properties of the original protein and the interactions between the original protein and other original proteins; constructing a second graph neural network based on the interactions between the original proteins; inputting any amino acid sequence into the second graph neural network to predict the properties, three-dimensional structure, and interactions between the mirror protein and other proteins corresponding to the amino acid sequence; and constructing a mirror protein interaction map using the protein properties and three-dimensional structure as node information and the interactions between any two proteins as edges. This application can predict the structure and properties of mirror proteins and construct a mirror protein interaction map.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and digital medical technology, and in particular to a method for constructing a mirror protein interaction graph and related equipment. BACKGROUND

[0002] All proteins in living organisms are connected and folded by left-handed amino acids. Right-handed amino acids, which are mirror images of left-handed amino acids, do not exist in nature, but mirror proteins composed of right-handed amino acids can bind to specific receptors to inhibit or signal, thus having new drug properties. In addition, mirror proteins do not bind to most protein receptors in living organisms, so they can avoid harming living organisms and are safer.

[0003] Currently, the prediction of protein structure and properties is mainly for left-handed amino acids, and there is no tool to support the prediction of the structure and properties of mirror proteins composed of right-handed amino acids. In addition, mirror proteins cannot be generated or extracted from nature, so it is time-consuming and labor-intensive to experiment on the properties of mirror proteins to determine their properties. Therefore, how to predict the structure and properties of mirror proteins and construct a mirror protein interaction graph is a problem to be solved in the field of digital medical technology. SUMMARY

[0004] In view of the above, it is necessary to propose a method for constructing a mirror protein interaction graph and related equipment to solve the technical problem of how to predict the structure and properties of mirror proteins and construct a mirror protein interaction graph. The related equipment includes a mirror protein interaction graph construction device, an electronic device, and a storage medium.

[0005] The present application provides a method for constructing a mirror protein interaction graph, which comprises:

[0006] Collecting a plurality of amino acid sequences and original proteins corresponding to each amino acid sequence;

[0007] Constructing a first graph neural network based on the original proteins, and inputting the amino acid sequences corresponding to the original proteins into the first graph neural network to predict the properties of the original proteins and the interactions between the original proteins and other original proteins;

[0008] Constructing a second graph neural network based on the interactions between any two original proteins;

[0009] inputting any amino acid sequence into the second graph neural network, predicting the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, and the interaction between the mirror protein and other proteins, including the mirror protein and the original protein, the original protein and the mirror protein corresponding to the same amino acid sequence being mirror symmetric to each other;

[0010] constructing a mirror protein interaction graph taking the properties and three-dimensional structure of the protein as node information and the interaction between any two proteins as an edge.

[0011] In some embodiments, the first graph neural network comprises a first amino acid layer, a first protein layer, and a first prediction layer, and the constructing of the first graph neural network based on the original protein comprises:

[0012] taking any one of the amino acids in the amino acid sequence corresponding to the original protein as a target amino acid;

[0013] taking the amino acid atoms in the target amino acid as atomic nodes, taking the types and three-dimensional coordinates of the amino acid atoms as the initial node information of the corresponding atomic nodes, and constructing an amino acid subgraph of the target amino acid by taking the chemical bonds between any amino acid atoms as atomic edges;

[0014] After traversing all the amino acids in the amino acid sequence corresponding to the original protein, the amino acid subgraph of each amino acid is obtained, and all the amino acid subgraphs of the amino acids are taken as the first amino acid layer, which is used to extract the amino acid features of each amino acid;

[0015] taking the amino acids in the amino acid sequence corresponding to the original protein as amino acid nodes, taking the types, three-dimensional coordinates, and amino acid features of the amino acids as the initial node information of the corresponding amino acid nodes, and constructing the first protein layer by taking the peptide bonds between any amino acids as amino acid edges, the first protein layer being used to extract the protein features of the original protein;

[0016] the first prediction layer is used for feature processing of the protein features, to predict the properties of the original protein and the interaction between the original protein and other original proteins.

[0017] In some embodiments, the first amino acid layer is used to extract the amino acid features of each amino acid, which comprises: in the amino acid subgraph corresponding to any amino acid, updating the initial node information of each atomic node based on the edge of the atomic node and the initial node information of the other atomic nodes connected by the edge, to obtain the updated node information of each atomic node, and splicing or adding all the updated node information of the atomic nodes to obtain the amino acid features of the amino acid;

[0018] The first protein layer is used to extract the protein features of the original protein, including: updating the initial node information of the amino acid node based on the edge of each amino acid node and the initial node information of other amino acid nodes connected by the edge to obtain the updated node information of each amino acid node, and splicing or adding the updated node information of all amino acid nodes to obtain the protein features of the original protein.

[0019] In some embodiments, before inputting the amino acid sequence corresponding to the original protein into the first graph neural network and predicting the properties of the original protein and the interactions between the original protein and other original proteins, the method further includes: training the first graph neural network, specifically including:

[0020] Collecting amino acid sequences of multiple original proteins as multiple training samples, and using the properties of each original protein and the interaction between each original protein and other original proteins as sample labels of the training samples;

[0021] Inputting any training sample into the first graph neural network to obtain a prediction result, wherein the prediction result includes the predicted properties of the original protein corresponding to the training sample and the predicted interactions with other original proteins;

[0022] A first cost function is calculated based on the prediction result and the sample label, and the first cost function satisfies the relationship:

[0023]

[0024] Among them, py * is the predicted property of the original protein corresponding to the training sample, is the property of the original protein in the sample label; M is the number of other original proteins, py′ i is the predicted interaction between the original protein corresponding to the training sample and the original protein i, is the interaction between the original protein in the sample tag and the original protein i; Indicates calculation py * and The KL divergence between them, Loss1 is the value of the first cost function;

[0025] Updating the first graph neural network to reduce the value of the first cost function, completing one iterative training;

[0026] The training sample is continuously input into the first graph neural network, and multiple iteration training is performed until the number of iteration training reaches a preset number or the value of the first cost function is less than a preset value, and the training of the first graph neural network is completed.

[0027] In some embodiments, the input of the second graph neural network is an arbitrary amino acid sequence, and the output is the property and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, and the interaction between the mirror protein and other proteins, and the second graph neural network is constructed based on the interaction between any two original proteins, comprising:

[0028] The second graph neural network comprises a second amino acid layer, a second protein layer, an interaction layer, a second prediction layer, and a regression layer.

[0029] The structure of the second amino acid layer is the same as that of the first amino acid layer, and is used to extract the amino acid features of each amino acid in the amino acid sequence.

[0030] The structure of the second protein layer is the same as that of the first protein layer, and is used to extract the protein features of the mirror protein corresponding to the amino acid sequence.

[0031] Each original protein is taken as a protein node, the protein features extracted by the first protein layer are taken as the initial node information of the corresponding protein node, and the interaction between any proteins is taken as a protein edge, so as to construct the interaction layer, and the interaction layer is used to learn the interaction features of each protein node.

[0032] The protein features of the mirror protein are input into the regression layer to predict the three-dimensional structure of the mirror protein, and the three-dimensional structure comprises the three-dimensional coordinates of all amino acids.

[0033] All interaction features and protein features of the mirror protein are input into the second prediction layer to predict the property of the mirror protein and the interaction between the mirror protein and each protein node.

[0034] In some embodiments, after the interaction features and the protein features of the mirror protein are input into the second prediction layer to predict the property of the mirror protein and the interaction between the mirror protein and other proteins, the method further comprises:

[0035] The mirror protein is taken as a new protein node, the protein features of the mirror protein are taken as the initial node information of the new protein node, and the interaction between the mirror protein and each protein node is taken as a new protein edge, so as to update the interaction layer.

[0036] In some embodiments, before inputting the arbitrary amino acid sequence into the second graph neural network, the method further comprises: training the second graph neural network, specifically comprising:

[0037] collecting amino acid sequences of a plurality of mirror proteins as a plurality of training samples, and collecting three-dimensional structures and properties of each mirror protein and interactions with other proteins as sample labels of the training samples;

[0038] inputting any one training sample into the second graph neural network to obtain a prediction result, the prediction result comprising predicted three-dimensional structures and predicted properties of the mirror protein corresponding to the training sample, and predicted interactions with other proteins;

[0039] calculating a second cost function based on the prediction result and the sample label, the second cost function satisfying the relationship:

[0040]

[0041] wherein, pJ * is the predicted property of the mirror protein corresponding to the training sample, is the property of the mirror protein in the sample label; K is the number of the other proteins, pJ′ u is the predicted interaction between the mirror protein corresponding to the training sample and protein u, is the interaction between the mirror protein in the sample label and protein u, N is the number of amino acids in the training sample, z v is the predicted three-dimensional coordinates of amino acid v in the training sample, is the three-dimensional coordinates of amino acid v in the sample label; denotes the KL divergence between pJ * and denotes the Euclidean distance between z v and , and Loss2 is the value of the second cost function;

[0042] updating the second graph neural network to reduce the value of the second cost function to complete one iteration of training;

[0043] continuously inputting training samples into the second graph neural network, performing multiple iterations of training until the number of iterations of training reaches a preset number or the value of the second cost function is less than a preset value, and stopping the training of the second graph neural network.

[0044] The embodiments of the present application also provide a device for constructing a mirror protein interaction map, the device comprising: ​

[0045] The collection unit is configured to collect a plurality of amino acid sequences and original proteins corresponding to each of the amino acid sequences.

[0046] The first prediction unit is configured to construct a first graph neural network based on the original proteins, and input the amino acid sequence corresponding to the original protein into the first graph neural network, to predict a property of the original protein and an interaction between the original protein and other original proteins.

[0047] The first construction unit is configured to construct a second graph neural network based on the interaction between any two original proteins.

[0048] The second prediction unit is configured to input any amino acid sequence into the second graph neural network, to predict a property and a three-dimensional structure of a mirror protein corresponding to the amino acid sequence, and an interaction between the mirror protein and other proteins, the proteins including the mirror protein and the original protein, the original protein and the mirror protein corresponding to the same amino acid sequence being mirror symmetric to each other.

[0049] The second construction unit is configured to construct a mirror protein interaction graph by taking the property and the three-dimensional structure of the proteins as node information and taking the interaction between any two proteins as an edge.

[0050] The embodiments of the present application further provide an electronic device, which comprises:

[0051] The memory is configured to store at least one instruction.

[0052] The processor is configured to execute the instruction stored in the memory to implement the method for constructing the mirror protein interaction graph.

[0053] The embodiments of the present application further provide a computer readable storage medium, which stores at least one instruction, and the at least one instruction is executed by a processor in an electronic device to implement the method for constructing the mirror protein interaction graph.

[0054] In summary, the application inputs any amino acid sequence into the first graph neural network, quickly and accurately predicts the properties of the corresponding original protein, and the interaction between any two original proteins; according to the interaction between any two original proteins, an interaction layer is constructed, and then a second graph neural network is constructed, any amino acid sequence is input into the second graph neural network, and the properties of the corresponding mirror protein and the interaction between the mirror protein and other original proteins are quickly and accurately predicted; further, based on the interaction between the mirror protein and other original proteins, the interaction layer is updated to predict the interaction between any proteins, including mirror proteins and original proteins, thereby realizing the accurate and rapid prediction of the structure and properties of the mirror protein and constructing the mirror protein interaction map. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 is a flowchart of a preferred embodiment of the mirror protein interaction map construction method involved in the application.

[0056] Figure 2 is a structural schematic diagram of the first graph neural network involved in the application.

[0057] Figure 3 is a structural schematic diagram of the second graph neural network involved in the application.

[0058] Figure 4 is a functional module diagram of a preferred embodiment of the mirror protein interaction map construction device involved in the application.

[0059] Figure 5 is a structural schematic diagram of an electronic device of a preferred embodiment of the mirror protein interaction map construction method involved in the application. DETAILED DESCRIPTION

[0060] In order to enable a person skilled in the art to more clearly understand the purpose, features and advantages of the application, the application will be described in detail below in combination with the drawings and specific embodiments. It should be noted that the embodiments of the application and the features in the embodiments can be combined with each other without conflict. In the following description, a large number of specific details are set forth in order to facilitate a full understanding of the application, and the described embodiments are only part of the embodiments of the application, not all embodiments.

[0061] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0063] The embodiments of the present application provide a method for constructing a mirror protein interaction map, which can be applied to one or more electronic devices. An electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0064] An electronic device can be any electronic product that can interact with a user, such as a personal computer, tablet computer, smartphone, personal digital assistant (PDA), game console, Internet Protocol Television (IPTV), smart wearable device, etc.

[0065] The electronic device may also include a network device and / or a client device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0066] The network where the electronic device is located includes but is not limited to the Internet, wide area network, metropolitan area network, local area network, virtual private network (VPN), etc.

[0067] like Figure 1 The figure shows a flowchart of a preferred embodiment of the method for constructing a mirror protein interaction map of the present application. Depending on different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted. The method for constructing a mirror protein interaction map provided in the embodiment of the present application can be applied to any scenario requiring accurate recommendations, and the method can be applied to products in these scenarios, such as the research and development of new drugs in the field of digital medicine, etc.

[0068] S10, collect a plurality of amino acid sequences and a corresponding original protein for each amino acid sequence.

[0069] In an optional embodiment, the amino acid sequence includes at least one amino acid, which is a left-handed amino acid, any amino acid known and existing in nature; each amino acid includes a plurality of amino acid atoms, which are connected by chemical bonds between different amino acid atoms, and each amino acid can be regarded as a three-dimensional spatial structure composed of amino acid atoms and chemical bonds. Among them, the amino acid atoms include but are not limited to carbon atoms, oxygen atoms, nitrogen atoms, sulfur atoms and the like; the chemical bond can react the interaction between the amino acid atoms, including covalent bond and non-covalent bond.

[0070] In an optional embodiment, in any amino acid sequence, different amino acids can form peptide bonds through dehydration condensation, and then combine a plurality of amino acids into at least one polypeptide chain, and further fold all polypeptide chains together to obtain the original protein corresponding to the amino acid sequence. The original protein can be regarded as a three-dimensional spatial structure composed of amino acids and peptide bonds. The original protein is any protein known and existing in nature. For example, the original protein can be a pathogen protein of different diseases, or a protein existing in the human body.

[0071] In this way, a plurality of amino acid sequences and the original protein corresponding to each amino acid sequence existing in nature are collected, which are used to predict the properties and interactions of the original protein in the subsequent steps, and a second graph neural network is constructed based on the interactions between any original proteins. The original protein is any protein existing in nature.

[0072] S11, constructing a first graph neural network based on the original protein, and inputting the amino acid sequence corresponding to the original protein into the first graph neural network to predict the properties of the original protein and the interactions between the original protein and other original proteins.

[0073] Please refer to Figure 2 , Figure 2 The structure diagram of the first graph neural network involved in the present application. In an optional embodiment, the first graph neural network includes a first amino acid layer, a first protein layer and a first prediction layer, and the first graph neural network is constructed based on the original protein, which includes:

[0074] Taking any one of the amino acid sequences corresponding to the original protein as a target amino acid;

[0075] The amino acid atoms in the target amino acid are used as atomic nodes, the types and three-dimensional coordinates of the amino acid atoms are used as initial node information of the corresponding atomic nodes, and the chemical bonds between any amino acid atoms are used as atomic edges to construct an amino acid subgraph of the target amino acid;

[0076] After traversing all amino acids in the amino acid sequence corresponding to the original protein, an amino acid subgraph of each amino acid is obtained, and the amino acid subgraphs of all amino acids are used as the first amino acid layer, and the first amino acid layer is used to extract the amino acid features of each amino acid;

[0077] The amino acids in the amino acid sequence corresponding to the original protein are used as amino acid nodes, the type, three-dimensional coordinates and amino acid characteristics of the amino acids are used as initial node information of the corresponding amino acid nodes, and the peptide bonds between any amino acids are used as amino acid edges to construct the first protein layer, which is used to extract protein characteristics of the original protein;

[0078] The first prediction layer is used to perform feature processing on the protein features and predict the properties of the original protein and the interactions between the original protein and other original proteins.

[0079] The target amino acid can be viewed as a three-dimensional structure composed of amino acid atoms and chemical bonds. The types of amino acid atoms include, but are not limited to, carbon atoms, oxygen atoms, nitrogen atoms, and sulfur atoms. The three-dimensional coordinates of the amino acid atoms are the coordinate information of the amino acid atoms in three-dimensional space. The first amino acid layer includes at least one amino acid subgraph, which corresponds one-to-one with the amino acids in the amino acid sequence corresponding to the original protein.

[0080] The original protein can be regarded as a three-dimensional structure composed of amino acids and peptide bonds. The types of amino acids include but are not limited to alanine, arginine and aspartic acid. The three-dimensional coordinates of the amino acid are the coordinate information of the amino acid in three-dimensional space.

[0081] The first prediction layer is a fully connected layer and / or a convolutional layer, which is used to extract the protein features. The specific structure of the first prediction layer is not limited in this application.

[0082] The interaction between the original protein and another original protein refers to any type of action, effect, or influence exerted by the original protein on another original protein. For example, original protein A can bind to original protein B to inhibit the activity of original protein B. The properties of the original protein may include, but are not limited to: affinity, whether the polypeptide is active, stability, drugability, etc.

[0083] In an optional embodiment, the first amino acid layer is used to extract the amino acid feature of each amino acid, including: in the amino acid subgraph corresponding to any amino acid, updating the initial node information of each atomic node based on the edge of each atomic node and the initial node information of other atomic nodes connected by the edge to obtain updated node information of each atomic node, and splicing or adding the updated node information of all atomic nodes to obtain the amino acid feature of the amino acid;

[0084] The first protein layer is used to extract the protein features of the original protein, including: updating the initial node information of the amino acid node based on the edge of each amino acid node and the initial node information of other amino acid nodes connected by the edge to obtain the updated node information of each amino acid node, and splicing or adding the updated node information of all amino acid nodes to obtain the protein features of the original protein.

[0085] It should be noted that the first graph neural network is a parameterized model. In order to ensure that the first graph neural network can accurately predict the properties of the original protein corresponding to any amino acid sequence and its interactions with other original proteins, the first graph neural network needs to be trained to constrain the accuracy of the output results of the first prediction layer.

[0086] In an optional embodiment, before inputting the amino acid sequence corresponding to the original protein into the first graph neural network and predicting the properties of the original protein and the interactions between the original protein and other original proteins, the method further includes: training the first graph neural network, specifically including:

[0087] Collecting amino acid sequences of multiple original proteins as multiple training samples, and using the properties of each original protein and the interaction between each original protein and other original proteins as sample labels of the training samples;

[0088] Inputting any training sample into the first graph neural network to obtain a prediction result, wherein the prediction result includes the predicted properties of the original protein corresponding to the training sample and the predicted interactions with other original proteins;

[0089] A first cost function is calculated based on the prediction result and the sample label, and the first cost function satisfies the relationship:

[0090]

[0091] Among them, py * is the predicted property of the original protein corresponding to the training sample, is the property of the original protein in the sample label; M is the number of other original proteins, py′i is the predicted interaction between the original protein corresponding to the training sample and the original protein i, is the interaction between the original protein in the sample tag and the original protein i; Indicates calculation py * and The KL divergence between them, Loss1 is the value of the first cost function;

[0092] Updating the first graph neural network to reduce the value of the first cost function, completing one iterative training;

[0093] Continuously input training samples into the first graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the first cost function is less than a preset value, thereby completing the training of the first graph neural network.

[0094] Among them, the training samples and sample labels of the first graph neural network can be directly obtained from a public protein database, such as the PDB protein structure database, the Uniprot database, etc.; the preset number of times and the preset value are both pre-set, and are 1000 and 0.001 respectively.

[0095] The updating of the first graph neural network to reduce the value of the first cost function can be achieved by a Monte Carlo sampling algorithm. The predicted properties py of the original protein corresponding to the training sample * is a probability vector, which includes the probability value of the original protein belonging to each property; when the probability value of any property exceeds the probability threshold, it means that the original protein has the property, wherein the probability threshold can be 0.6. Similarly, the predicted interaction py′ between the original protein corresponding to the training sample and the original protein i i It is also a probability vector, which includes the probability value of each interaction.

[0096] In an optional embodiment, after completing the training of the first graph neural network, the amino acid sequence corresponding to the original protein is input into the first graph neural network to predict the properties of the original protein and the interaction between the original protein and other original proteins.

[0097] In this way, the construction of the first graph neural network is completed. By inputting any amino acid sequence into the first graph neural network, the properties of the original protein corresponding to the amino acid sequence and the interaction between the original protein and other original proteins can be quickly and accurately obtained.

[0098] S12, construct the second graph neural network based on the interaction between any two original proteins.

[0099] In an optional embodiment, all amino acid sequences are input into the first graph neural network to obtain interactions between any original proteins. It should be noted that the original protein and the mirror image protein corresponding to the same amino acid sequence are mirror images of each other, that is, the structural differences between the original protein and the mirror image protein corresponding to the same amino acid sequence are only in the three-dimensional coordinates of the amino acids.

[0100] See Figure 3 , Figure 3 This is a schematic diagram of the structure of the second graph neural network involved in this application. In an optional embodiment, the input of the second graph neural network is an arbitrary amino acid sequence, and the output is the properties and three-dimensional structure of the mirror image protein corresponding to the amino acid sequence, as well as the interaction between the mirror image protein and other proteins. Constructing the second graph neural network based on the interaction between any two original proteins includes:

[0101] The second graph neural network includes a second amino acid layer, a second protein layer, an interaction layer, a second prediction layer and a regression layer;

[0102] The structure of the second amino acid layer is the same as that of the first amino acid layer, and is used to extract the amino acid characteristics of each amino acid in the amino acid sequence;

[0103] The structure of the second protein layer is the same as that of the first protein layer, and is used to extract protein features of the mirror image protein corresponding to the amino acid sequence;

[0104] Each original protein is regarded as a protein node, the protein features extracted by the first protein layer are used as the initial node information of the corresponding protein node, and the interactions between any proteins are used as protein edges to construct the interaction layer, which is used to learn the interaction features of each protein node;

[0105] Inputting the protein features of the mirror protein into the regression layer to predict the three-dimensional structure of the mirror protein, wherein the three-dimensional structure includes the three-dimensional coordinates of all amino acids;

[0106] All interaction features and the protein features of the mirror protein are input into the second prediction layer to predict the properties of the mirror protein and the interactions between the mirror protein and each protein node.

[0107] Wherein, the regression layer and the second prediction layer both include fully connected layers and / or convolutional layers, which are not limited in this application.

[0108] In an optional embodiment, after inputting the interaction features and the protein features of the mirror protein into the second prediction layer and predicting the properties of the mirror protein and the interactions between the mirror protein and other proteins, the method further includes:

[0109] The mirror protein is used as a newly added protein node, the protein feature of the mirror protein is used as the initial node information of the newly added protein node, and the interaction between the mirror protein and each protein node is used as a newly added protein edge to update the interaction layer.

[0110] Among them, in the updated interaction layer, all protein nodes correspond to proteins including mirror proteins and original proteins. At this time, the interaction between the mirror protein and other proteins including the mirror protein and the original protein can be predicted.

[0111] In this way, the construction of the second graph neural network is completed, and the interaction layer in the second graph neural network can be updated according to the output results of the second graph neural network, so that the second graph neural network can predict the interaction between the mirror protein and any other protein.

[0112] S13, inputting an arbitrary amino acid sequence into the second graph neural network, predicting the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins, wherein the proteins include the mirror protein and the original protein, and the original protein and the mirror protein corresponding to the same amino acid sequence are mirror-symmetric to each other.

[0113] In an optional embodiment, the second graph neural network and the first graph neural network are both parameterized models, so the second graph neural network also needs to be trained to constrain the accuracy of the prediction results of the second graph neural network.

[0114] In an optional embodiment, before inputting the arbitrary amino acid sequence into the second graph neural network, the method further includes: training the second graph neural network, specifically including:

[0115] Collecting amino acid sequences of multiple mirror proteins as multiple training samples, and using the three-dimensional structure and properties of each mirror protein, as well as its interactions with other proteins, as sample labels of the training samples;

[0116] Inputting any training sample into the second graph neural network to obtain a prediction result, wherein the prediction result includes the predicted three-dimensional structure and predicted properties of the mirror protein corresponding to the training sample, as well as the predicted interaction with other proteins;

[0117] A second cost function is calculated based on the prediction result and the sample label, and the second cost function satisfies the relationship:

[0118]

[0119] Among them, pJ * is the predicted property of the mirror protein corresponding to the training sample, is the property of the mirror protein in the sample label; K is the number of other proteins, pJ′ u is the predicted interaction between the mirror protein corresponding to the training sample and protein u, is the interaction between the mirror protein and protein u in the sample label, N is the number of amino acids in the training sample, z v is the predicted three-dimensional coordinate of the amino acid v in the training sample, is the three-dimensional coordinate of amino acid v in the sample label; Indicates the calculation of pJ * and The KL divergence between Indicates the calculation of z v and The Euclidean distance of , Loss2 is the value of the second cost function;

[0120] Updating the second graph neural network to reduce the value of the second cost function, completing one iterative training;

[0121] Continuously input training samples into the second graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the second cost function is less than a preset value, thereby completing the training of the second graph neural network.

[0122] The training samples and sample labels of the second graph neural network are known mirror protein related data, which can be obtained from a public database; the preset number of times and the preset value are both pre-set, and are 1000 and 0.001 respectively; the predicted property pJ * and predicted interaction pJ′ u is a probability vector, with the predicted properties py * and the predicted interaction py′ i The updating process of the second graph neural network is similar to that of the first graph neural network, and this application will not go into details.

[0123] In this embodiment, the original protein and the mirror protein corresponding to the same amino acid sequence are mirror-symmetrical to each other. For example, the three-dimensional coordinates of amino acid A in the original protein corresponding to any amino acid sequence are (1, 1, 1), and (0, 0, 0) is used as the mirror symmetry point. Then, the position of amino acid A in the mirror protein corresponding to the amino acid sequence is the coordinates (-1, -1, -1). By performing mirror symmetry on all amino acids in the original protein, the corresponding mirror protein can be obtained.

[0124] In an optional embodiment, after completing the training of the second graph neural network, an arbitrary amino acid sequence is input into the second graph neural network, and the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins are output, and the proteins include the mirror protein and the original protein.

[0125] In this way, after completing the training of the second graph neural network, inputting any amino acid sequence into the first graph neural network can quickly and accurately obtain the three-dimensional structure and properties of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins.

[0126] S14, using the properties and three-dimensional structure of the protein as node information and the interaction between any two proteins as edges to construct a mirror protein interaction map.

[0127] In an optional embodiment, the properties and three-dimensional structure of the mirror image protein corresponding to any amino acid sequence, as well as the interaction between the mirror image protein and other proteins can be obtained through the mirror image protein interaction map, and the other proteins include mirror image proteins and original proteins.

[0128] In this way, based on the mirror image protein interaction map, in vivo experimental simulation of the mirror image protein can be performed, which can be used for efficacy simulation and safety simulation in the process of developing new drugs.

[0129] It can be seen from the above technical solution that the present application inputs an arbitrary amino acid sequence into the first graph neural network to quickly and accurately predict the properties of the corresponding original protein and the interaction between any two original proteins; constructs an interaction layer based on the interaction between any two original proteins, and then constructs a second graph neural network, inputs an arbitrary amino acid sequence into the second graph neural network, and quickly and accurately predicts the properties of the corresponding mirror protein and the interaction between the mirror protein and other original proteins; further, based on the interaction between the mirror protein and other original proteins, the interaction layer is updated to predict the interaction between arbitrary proteins, including mirror proteins and original proteins, thereby achieving accurate and rapid prediction of the structure and properties of the mirror protein and constructing a mirror protein interaction map.

[0130] See Figure 4 , Figure 4 This is a functional block diagram of a preferred embodiment of the apparatus for constructing a mirror protein interaction map of the present application. The apparatus 11 for constructing a mirror protein interaction map includes an acquisition unit 110, a first prediction unit 111, a first construction unit 112, a second prediction unit 113, and a second construction unit 114. The modules / units referred to herein are a series of computer-readable instruction segments that can be executed by the processor 13 and perform a fixed function, and are stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0131] In an optional embodiment, the collection unit 110 is configured to collect multiple amino acid sequences and the original protein corresponding to each amino acid sequence.

[0132] In an optional embodiment, the first prediction unit 111 is used to construct a first graph neural network based on the original protein, and input the amino acid sequence corresponding to the original protein into the first graph neural network to predict the properties of the original protein and the interaction between the original protein and other original proteins.

[0133] In an optional embodiment, the first graph neural network includes a first amino acid layer, a first protein layer, and a first prediction layer, and constructing the first graph neural network based on the original protein includes:

[0134] Taking any amino acid in the amino acid sequence corresponding to the original protein as the target amino acid;

[0135] The amino acid atoms in the target amino acid are used as atomic nodes, the types and three-dimensional coordinates of the amino acid atoms are used as initial node information of the corresponding atomic nodes, and the chemical bonds between any amino acid atoms are used as atomic edges to construct an amino acid subgraph of the target amino acid;

[0136] After traversing all amino acids in the amino acid sequence corresponding to the original protein, an amino acid subgraph of each amino acid is obtained, and the amino acid subgraphs of all amino acids are used as the first amino acid layer, and the first amino acid layer is used to extract the amino acid features of each amino acid;

[0137] The amino acids in the amino acid sequence corresponding to the original protein are used as amino acid nodes, the type, three-dimensional coordinates and amino acid characteristics of the amino acids are used as initial node information of the corresponding amino acid nodes, and the peptide bonds between any amino acids are used as amino acid edges to construct the first protein layer, which is used to extract protein characteristics of the original protein;

[0138] The first prediction layer is used to perform feature processing on the protein features and predict the properties of the original protein and the interactions between the original protein and other original proteins.

[0139] In an optional embodiment, the first amino acid layer is used to extract the amino acid feature of each amino acid, including: in the amino acid subgraph corresponding to any amino acid, updating the initial node information of each atomic node based on the edge of each atomic node and the initial node information of other atomic nodes connected by the edge to obtain updated node information of each atomic node, and splicing or adding the updated node information of all atomic nodes to obtain the amino acid feature of the amino acid;

[0140] The first protein layer is used to extract the protein features of the original protein, including: updating the initial node information of the amino acid node based on the edge of each amino acid node and the initial node information of other amino acid nodes connected by the edge to obtain the updated node information of each amino acid node, and splicing or adding the updated node information of all amino acid nodes to obtain the protein features of the original protein.

[0141] In an optional embodiment, before inputting the amino acid sequence corresponding to the original protein into the first graph neural network and predicting the properties of the original protein and the interactions between the original protein and other original proteins, the method further includes: training the first graph neural network, specifically including:

[0142] Collecting amino acid sequences of multiple original proteins as multiple training samples, and using the properties of each original protein and the interaction between each original protein and other original proteins as sample labels of the training samples;

[0143] Inputting any training sample into the first graph neural network to obtain a prediction result, wherein the prediction result includes the predicted properties of the original protein corresponding to the training sample and the predicted interactions with other original proteins;

[0144] A first cost function is calculated based on the prediction result and the sample label, and the first cost function satisfies the relationship:

[0145]

[0146] Among them, py * is the predicted property of the original protein corresponding to the training sample, is the property of the original protein in the sample label; M is the number of other original proteins, py′ i is the predicted interaction between the original protein corresponding to the training sample and the original protein i, is the interaction between the original protein in the sample tag and the original protein i; Indicates calculation py * and The KL divergence between them, Loss1 is the value of the first cost function;

[0147] Updating the first graph neural network to reduce the value of the first cost function, completing one iterative training;

[0148] Continuously input training samples into the first graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the first cost function is less than a preset value, thereby completing the training of the first graph neural network.

[0149] In an optional embodiment, the first construction unit 112 is configured to construct a second graph neural network based on the interaction between any two original proteins.

[0150] In an optional embodiment, the input of the second graph neural network is an arbitrary amino acid sequence, and the output is the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins. The construction of the second graph neural network based on the interaction between any two original proteins includes:

[0151] The second graph neural network includes a second amino acid layer, a second protein layer, an interaction layer, a second prediction layer and a regression layer;

[0152] The structure of the second amino acid layer is the same as that of the first amino acid layer, and is used to extract the amino acid characteristics of each amino acid in the amino acid sequence;

[0153] the second protein layer is the same as the first protein layer, and is configured to extract protein features of a mirror protein corresponding to the amino acid sequence;

[0154] each original protein is taken as a protein node, the protein features extracted by the first protein layer are taken as initial node information of the corresponding protein node, and an interaction between any proteins is taken as a protein edge, so as to construct the interaction layer, and the interaction layer is configured to learn interaction features of each protein node;

[0155] the protein features of the mirror protein are input into the regression layer, and a three-dimensional structure of the mirror protein is predicted, the three-dimensional structure including three-dimensional coordinates of all amino acids;

[0156] all interaction features and the protein features of the mirror protein are input into the second prediction layer, and properties of the mirror protein and interactions between the mirror protein and each protein node are predicted.

[0157] In an optional embodiment, after the interaction features and the protein features of the mirror protein are input into the second prediction layer, and the properties of the mirror protein and the interactions between the mirror protein and other proteins are predicted, the method further includes:

[0158] the mirror protein is taken as a new protein node, the protein features of the mirror protein are taken as initial node information of the new protein node, and an interaction between the mirror protein and each protein node is taken as a new protein edge, so as to update the interaction layer.

[0159] In an optional embodiment, the second prediction unit 113 is configured to input any amino acid sequence into the second graph neural network, and predict properties and a three-dimensional structure of a mirror protein corresponding to the amino acid sequence, and interactions between the mirror protein and other proteins, the proteins including the mirror protein and original proteins, and the original protein and the mirror protein corresponding to the same amino acid sequence are mirror symmetric to each other.

[0160] In an optional embodiment, before the any amino acid sequence is input into the second graph neural network, the method further includes training the second graph neural network, and specifically includes:

[0161] a plurality of amino acid sequences of mirror proteins are collected as a plurality of training samples, and three-dimensional structures and properties of each mirror protein and interactions between the mirror protein and other proteins are taken as sample labels of the training samples;

[0162] Inputting any training sample into the second graph neural network to obtain a prediction result, wherein the prediction result includes the predicted three-dimensional structure and predicted properties of the mirror protein corresponding to the training sample, as well as the predicted interaction with other proteins;

[0163] A second cost function is calculated based on the prediction result and the sample label, and the second cost function satisfies the relationship:

[0164]

[0165] Among them, pJ * is the predicted property of the mirror protein corresponding to the training sample, is the property of the mirror protein in the sample label; K is the number of other proteins, pJ′ u is the predicted interaction between the mirror protein corresponding to the training sample and protein u, is the interaction between the mirror protein and protein u in the sample label, N is the number of amino acids in the training sample, z v is the predicted three-dimensional coordinate of the amino acid v in the training sample, is the three-dimensional coordinate of amino acid v in the sample label; Indicates the calculation of pJ * and The KL divergence between Indicates the calculation of z v and The Euclidean distance of , Loss2 is the value of the second cost function;

[0166] Updating the second graph neural network to reduce the value of the second cost function, completing one iterative training;

[0167] Continuously input training samples into the second graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the second cost function is less than a preset value, thereby completing the training of the second graph neural network.

[0168] In an optional embodiment, the second construction unit 114 is configured to construct a mirror protein interaction map by using the properties and three-dimensional structures of the proteins as node information and the interactions between any two proteins as edges.

[0169] It can be seen from the above technical solution that the present application inputs an arbitrary amino acid sequence into the first graph neural network to quickly and accurately predict the properties of the corresponding original protein and the interaction between any two original proteins; constructs an interaction layer based on the interaction between any two original proteins, and then constructs a second graph neural network, inputs an arbitrary amino acid sequence into the second graph neural network, and quickly and accurately predicts the properties of the corresponding mirror protein and the interaction between the mirror protein and other original proteins; further, based on the interaction between the mirror protein and other original proteins, the interaction layer is updated to predict the interaction between arbitrary proteins, including mirror proteins and original proteins, thereby achieving accurate and rapid prediction of the structure and properties of the mirror protein and constructing a mirror protein interaction map.

[0170] See Figure 5 , is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Electronic device 1 includes memory 12 and processor 13. Memory 12 is used to store computer-readable instructions, and processor 13 is used to execute the computer-readable instructions stored in the memory to implement the method for constructing a mirror protein interaction map described in any of the above embodiments.

[0171] In an optional embodiment, the electronic device 1 further includes a bus, a computer program stored in the memory 12 and executable on the processor 13 , such as a program for constructing a mirror protein interaction map.

[0172] Figure 5 Only the electronic device 1 having the memory 12 and the processor 13 is shown. It can be understood by those skilled in the art that Figure 5 The structure shown does not constitute a limitation on the electronic device 1 , and the electronic device 1 may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0173] Combine Figure 1 The memory 12 in the electronic device 1 stores a plurality of computer-readable instructions to implement a method for constructing a mirror protein interaction map, and the processor 13 can execute the plurality of instructions to implement:

[0174] Collect multiple amino acid sequences and the original protein corresponding to each amino acid sequence;

[0175] constructing a first graph neural network based on the original protein, and inputting the amino acid sequence corresponding to the original protein into the first graph neural network to predict the properties of the original protein and the interaction between the original protein and other original proteins;

[0176] Construct a second graph neural network based on the interaction between any two original proteins;

[0177] Inputting an arbitrary amino acid sequence into the second graph neural network, predicting the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins, wherein the proteins include the mirror protein and the original protein, and the original protein and the mirror protein corresponding to the same amino acid sequence are mirror-symmetric to each other;

[0178] The properties and three-dimensional structures of the proteins are used as node information, and the interactions between any two proteins are used as edges to construct a mirror protein interaction map.

[0179] Specifically, the specific implementation method of the processor 13 for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0180] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. The electronic device 1 may have a bus structure or a star structure. The electronic device 1 may also include more or less other hardware or software than shown in the figure, or a different arrangement of components. For example, the electronic device 1 may also include input and output devices, network access devices, etc.

[0181] It should be noted that the electronic device 1 is only an example, and other existing or future electronic products that are suitable for this application should also be included in the scope of protection of this application and incorporated herein by reference.

[0182] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium can be non-volatile or volatile. The readable storage medium includes a flash memory, a mobile hard disk, a multimedia card, a card-type memory (for example: SD or DX memory, etc.), a magnetic memory, a disk, an optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 1. The memory 12 can not only be used to store application software and various types of data installed on the electronic device 1, such as the code of the program for constructing the mirror protein interaction map, but can also be used to temporarily store data that has been output or is to be output.

[0183] In some embodiments, the processor 13 may be comprised of an integrated circuit, such as a single packaged integrated circuit or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and combinations of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, connecting the various components of the entire electronic device 1 using various interfaces and circuits. It executes or runs programs or modules stored in the memory 12 (e.g., a program for constructing a mirrored protein interaction map), and calls data stored in the memory 12 to perform various functions of the electronic device 1 and process data.

[0184] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-mentioned embodiments of the method for constructing the mirror protein interaction map, for example Figure 1 Steps shown.

[0185] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into an acquisition unit 110, a first prediction unit 111, a first construction unit 112, a second prediction unit 113, and a second construction unit 114.

[0186] The above-mentioned integrated unit implemented in the form of a software functional module can be stored in a computer-readable storage medium. The above-mentioned software functional module is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, computer device, or network device, etc.) or a processor to execute the portion of the method for constructing a mirror protein interaction map described in various embodiments of the present application.

[0187] If the modules / units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment methods, and can also instruct the relevant hardware devices to complete them through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments.

[0188] The computer program includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory, or other memory.

[0189] Furthermore, the computer-readable storage medium may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0190] The blockchain referred to in this application is a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0191] The bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 The diagram is represented by only one arrow, but it does not mean that there is only one bus or one type of bus. The bus is configured to implement connection and communication between the memory 12 and at least one processor 13, etc.

[0192] The embodiment of the present application further provides a computer readable storage medium (not shown in the figure), which stores computer readable instructions, and the computer readable instructions are executed by a processor in an electronic device to implement the mirror protein interaction map construction method in any of the above embodiments.

[0193] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the above described device embodiment is merely illustrative, and for example, the division of the modules is merely a logical function division, and there can be another division manner in actual implementation.

[0194] The modules described as separated components can or can not be physically separated, and the components displayed as modules can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. According to actual needs, some or all of the modules can be selected to achieve the purpose of the embodiment.

[0195] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware, or in the form of hardware plus software functional modules.

[0196] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. The plurality of units or devices stated in the specification can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any specific order.

[0197] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A method for constructing a mirror protein interaction map, characterized in that: The method comprises: Collect multiple amino acid sequences and the original protein corresponding to each amino acid sequence; constructing a first graph neural network based on the original protein, and inputting the amino acid sequence corresponding to the original protein into the first graph neural network to predict the properties of the original protein and the interaction between the original protein and other original proteins; Construct a second graph neural network based on the interaction between any two original proteins; Inputting an arbitrary amino acid sequence into the second graph neural network, predicting the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins, wherein the proteins include the mirror protein and the original protein, and the original protein and the mirror protein corresponding to the same amino acid sequence are mirror-symmetric to each other; The properties and three-dimensional structures of the proteins are used as node information, and the interactions between any two proteins are used as edges to construct a mirror protein interaction map.

2. The method for constructing a mirror protein interaction map according to claim 1, wherein: The first graph neural network includes a first amino acid layer, a first protein layer, and a first prediction layer. Constructing the first graph neural network based on the original protein includes: Taking any amino acid in the amino acid sequence corresponding to the original protein as the target amino acid; The amino acid atoms in the target amino acid are used as atomic nodes, the types and three-dimensional coordinates of the amino acid atoms are used as initial node information of the corresponding atomic nodes, and the chemical bonds between any amino acid atoms are used as atomic edges to construct an amino acid subgraph of the target amino acid; After traversing all amino acids in the amino acid sequence corresponding to the original protein, an amino acid subgraph of each amino acid is obtained, and the amino acid subgraphs of all amino acids are used as the first amino acid layer, and the first amino acid layer is used to extract the amino acid features of each amino acid; The amino acids in the amino acid sequence corresponding to the original protein are used as amino acid nodes, the type, three-dimensional coordinates and amino acid characteristics of the amino acids are used as initial node information of the corresponding amino acid nodes, and the peptide bonds between any amino acids are used as amino acid edges to construct the first protein layer, which is used to extract protein characteristics of the original protein; The first prediction layer is used to perform feature processing on the protein features and predict the properties of the original protein and the interactions between the original protein and other original proteins.

3. The method for constructing a mirror protein interaction map according to claim 2, wherein: The first amino acid layer is used to extract the amino acid feature of each amino acid, including: in the amino acid subgraph corresponding to any amino acid, based on the edge of each atomic node and the initial node information of other atomic nodes connected by the edge, updating the initial node information of the atomic node to obtain the updated node information of each atomic node, and splicing or adding the updated node information of all atomic nodes to obtain the amino acid feature of the amino acid; The first protein layer is used to extract the protein features of the original protein, including: updating the initial node information of the amino acid node based on the edge of each amino acid node and the initial node information of other amino acid nodes connected by the edge to obtain the updated node information of each amino acid node, and splicing or adding the updated node information of all amino acid nodes to obtain the protein features of the original protein.

4. The method for constructing a mirror protein interaction map according to claim 1, wherein: Before inputting the amino acid sequence corresponding to the original protein into the first graph neural network and predicting the properties of the original protein and the interactions between the original protein and other original proteins, the method further includes: training the first graph neural network, specifically including: Collecting amino acid sequences of multiple original proteins as multiple training samples, and using the properties of each original protein and the interaction between each original protein and other original proteins as sample labels of the training samples; Inputting any training sample into the first graph neural network to obtain a prediction result, wherein the prediction result includes the predicted properties of the original protein corresponding to the training sample and the predicted interactions with other original proteins; A first cost function is calculated based on the prediction result and the sample label, and the first cost function satisfies the relationship: Among them, py * is the predicted property of the original protein corresponding to the training sample, is the property of the original protein in the sample label; M is the number of other original proteins, py′ i is the predicted interaction between the original protein corresponding to the training sample and the original protein i, is the interaction between the original protein in the sample tag and the original protein i; Indicates calculation py * and The KL divergence between them, Loss1 is the value of the first cost function; Updating the first graph neural network to reduce the value of the first cost function, completing one iterative training; Continuously input training samples into the first graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the first cost function is less than a preset value, thereby completing the training of the first graph neural network.

5. The method for constructing a mirror protein interaction map according to claim 2, wherein: The input of the second graph neural network is an arbitrary amino acid sequence, and the output is the properties and three-dimensional structure of the mirror protein corresponding to the amino acid sequence, as well as the interaction between the mirror protein and other proteins. The construction of the second graph neural network based on the interaction between any two original proteins includes: The second graph neural network includes a second amino acid layer, a second protein layer, an interaction layer, a second prediction layer and a regression layer; The structure of the second amino acid layer is the same as that of the first amino acid layer, and is used to extract the amino acid characteristics of each amino acid in the amino acid sequence; The structure of the second protein layer is the same as that of the first protein layer, and is used to extract protein features of the mirror image protein corresponding to the amino acid sequence; Each original protein is regarded as a protein node, the protein features extracted by the first protein layer are used as the initial node information of the corresponding protein node, and the interactions between any proteins are used as protein edges to construct the interaction layer, which is used to learn the interaction features of each protein node; Inputting the protein features of the mirror protein into the regression layer to predict the three-dimensional structure of the mirror protein, wherein the three-dimensional structure includes the three-dimensional coordinates of all amino acids; All interaction features and the protein features of the mirror protein are input into the second prediction layer to predict the properties of the mirror protein and the interactions between the mirror protein and each protein node.

6. The method for constructing a mirror protein interaction map according to claim 5, wherein: After inputting the interaction features and the protein features of the mirror protein into the second prediction layer to predict the properties of the mirror protein and the interactions between the mirror protein and other proteins, the method further comprises: The mirror protein is used as a newly added protein node, the protein feature of the mirror protein is used as the initial node information of the newly added protein node, and the interaction between the mirror protein and each protein node is used as a newly added protein edge to update the interaction layer.

7. The method for constructing a mirror protein interaction map according to claim 1, wherein: Before inputting the arbitrary amino acid sequence into the second graph neural network, the method further includes: training the second graph neural network, specifically including: Collecting amino acid sequences of multiple mirror proteins as multiple training samples, and using the three-dimensional structure and properties of each mirror protein, as well as its interactions with other proteins, as sample labels of the training samples; Inputting any training sample into the second graph neural network to obtain a prediction result, wherein the prediction result includes the predicted three-dimensional structure and predicted properties of the mirror protein corresponding to the training sample, as well as the predicted interaction with other proteins; A second cost function is calculated based on the prediction result and the sample label, and the second cost function satisfies the relationship: Among them, pJ * is the predicted property of the mirror protein corresponding to the training sample, is the property of the mirror protein in the sample label; K is the number of other proteins, pJ′ u is the predicted interaction between the mirror protein corresponding to the training sample and protein u, is the interaction between the mirror protein and protein u in the sample label, N is the number of amino acids in the training sample, z v is the predicted three-dimensional coordinate of the amino acid v in the training sample, is the three-dimensional coordinate of amino acid v in the sample label; Indicates the calculation of pJ * and The KL divergence between Indicates the calculation of z v and The Euclidean distance of , Loss2 is the value of the second cost function; Updating the second graph neural network to reduce the value of the second cost function, completing one iterative training; Continuously input training samples into the second graph neural network and perform multiple iterative training until the number of iterative training reaches a preset number or the value of the second cost function is less than a preset value, thereby completing the training of the second graph neural network.

8. A device for constructing a mirror protein interaction map, characterized in that: The device comprises: A collection unit, used for collecting multiple amino acid sequences and the original protein corresponding to each amino acid sequence; a first prediction unit, configured to construct a first graph neural network based on the original protein, input an amino acid sequence corresponding to the original protein into the first graph neural network, and predict properties of the original protein and interactions between the original protein and other original proteins; A first building unit is used to construct a second graph neural network based on the interaction between any two original proteins; a second prediction unit, configured to input an arbitrary amino acid sequence into the second graph neural network, and predict the properties and three-dimensional structure of a mirror image protein corresponding to the amino acid sequence, as well as the interaction between the mirror image protein and other proteins, wherein the proteins include the mirror image protein and the original protein, and the original protein and the mirror image protein corresponding to the same amino acid sequence are mirror-symmetric to each other; The second construction unit is used to construct a mirror protein interaction map by taking the properties and three-dimensional structure of the protein as node information and the interaction between any two proteins as edges.

9. An electronic device, characterized in that: The electronic device comprises: a memory storing computer-readable instructions; and A processor executes the computer-readable instructions stored in the memory to implement the method for constructing a mirror protein interaction map according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the method for constructing a mirror protein interaction map according to any one of claims 1 to 7 is implemented.

Images