Key protein identification method and device based on dynamic graph neural network

Abstract

The application provides a key protein identification method and device based on a dynamic graph neural network, the method comprising: obtaining first original protein data, and constructing a dynamic protein interaction network with a time attribute; performing random walk sampling on the dynamic protein interaction network, and constructing a training corpus; using a trained graph convolutional neural network model to extract structural features of protein nodes; inputting the structural feature data into a pre-trained long short-term memory network model, and outputting time features of the protein nodes; and identifying key proteins through pattern classification according to the structural features and the time features. The application models a protein interaction graph as a dynamic graph, and learns structural features and time features of protein nodes on the dynamic graph by combining a graph convolutional neural network model and a long short-term memory network model, so that key proteins can be identified more efficiently.

Description

Technical Field

[0001] This application belongs to the field of biological technology, and in particular relates to a method and device for identifying key proteins based on dynamic graph neural networks. Background Technology

[0002] In recent years, learning network embeddings has attracted increasing interest. Researchers represent networks as low-dimensional vectors, which can be used to perform high-precision and scalable graph mining tasks. Currently, most methods are based on static graphs. However, in the real world, networks are constantly evolving as nodes and edges are added and removed. Therefore, identifying key proteins on dynamic networks is becoming increasingly important, as key protein identification has always been a focus of research in the biological field.

[0003] Traditional protein recognition primarily relies on static PPI networks. While some methods incorporate biological characteristics of proteins, they are still based on static graphs. Given the continuous evolution and development of networks, protein recognition on static graphs faces the following main challenges:

[0004] (1) Key protein identification methods based on biological experiments are time-consuming and costly. As PPI data and protein sequence data become increasingly complete, many scholars are gradually turning to computation-based key protein identification research.

[0005] (2) Many embedding methods based on random walks are based on static graphs, while the network is dynamic. Therefore, the learning of protein features is not comprehensive enough and the dynamics of protein-protein interaction networks are ignored.

[0006] (3) Since the new random walk set involved each time is random, the network mining task results obtained by the subsequent network representation cannot be directly compared with each other.

[0007] (4) Previous studies on dynamic networks assumed that all proteins in the network were dynamic, and that conserved proteins were not filtered out. Conserved proteins could not well characterize real protein networks. Summary of the Invention

[0008] In view of this, this application aims to propose a key protein identification method and apparatus based on dynamic graph neural networks, in order to solve at least one of the above-mentioned problems.

[0009] To achieve the above objectives, the technical solution of this application is implemented as follows:

[0010] In a first aspect, this application provides a key protein identification method based on a dynamic graph neural network, the method comprising:

[0011] First raw protein data is acquired, the first raw protein data is processed, and a dynamic protein-protein interaction network with time attributes is constructed based on the processed first raw protein data. The first raw protein data includes key protein data with tags.

[0012] Random walk sampling is performed on the dynamic protein-protein interaction network to construct a training corpus;

[0013] Using a trained graph convolutional neural network model, structural features of protein nodes are extracted;

[0014] The structural feature data is input into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes.

[0015] Key proteins are identified through pattern classification based on the structural and temporal characteristics.

[0016] Secondly, based on the same inventive concept, this application also provides a key protein recognition device based on a dynamic graph neural network, comprising:

[0017] The acquisition module is configured to acquire first raw protein data, process the first raw protein data, and construct a dynamic protein-protein interaction network with time attributes based on the processed first raw protein data, wherein the first raw protein data includes key protein data with tags.

[0018] The training corpus construction module is configured to perform random walk sampling on the dynamic protein-protein interaction network to construct the training corpus.

[0019] The structural feature extraction module is configured to extract the structural features of protein nodes using a trained graph convolutional neural network model.

[0020] The temporal feature extraction module is configured to input the structural feature data into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes.

[0021] The classification and recognition module is configured to identify key proteins by pattern classification based on the structural features and the temporal features.

[0022] Thirdly, based on the same inventive concept, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the key protein identification method based on dynamic graph neural network as described in the first aspect.

[0023] Fourthly, based on the same inventive concept, this application also provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the key protein identification method based on dynamic graph neural network as described in the first aspect.

[0024] Compared with existing technologies, the key protein identification method and apparatus based on dynamic graph neural networks described in this application have the following advantages:

[0025] The key protein identification method and apparatus based on dynamic graph neural networks described in this application model the protein interaction graph as a dynamic graph, and learn the structural and temporal features of protein nodes on the dynamic graph by combining graph convolutional neural network model and long short-term memory network model. This solves the problem of ignoring temporal features in the process of identifying key proteins, and can identify key proteins more efficiently with accurate and efficient results. Attached Figure Description

[0026] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0027] Figure 1 This is a flowchart of the key protein identification method based on dynamic graph neural networks described in the embodiments of this application;

[0028] Figure 2 This is a schematic diagram of the key protein recognition device based on a dynamic graph neural network as described in an embodiment of this application;

[0029] Figure 3 This is a schematic diagram of the electronic device structure described in an embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with specific embodiments and the accompanying drawings.

[0031] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this application should have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms "first," "second," and similar terms used in the embodiments of this application do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0032] Based on this, one or more embodiments of this application provide a key protein identification method based on a dynamic graph neural network.

[0033] The embodiments of this application are described in detail below with reference to the accompanying drawings.

[0034] Please see Figure 1 As shown, one embodiment of this application provides a key protein identification method based on a dynamic graph neural network, which includes the following steps:

[0035] S101. Obtain the first raw protein data, process the first raw protein data, and construct a dynamic protein interaction network with time attributes based on the processed first raw protein data. The first raw protein data includes key protein data with tags.

[0036] Specifically, the first set of raw protein data includes PPI protein network data from the DIP database, and gene expression profile data from the GSE3431 dataset, which contains 6777 gene expression values ​​at 36 time points. Standard key protein data integrates four databases: MIPS, SGD, DEG, and SGDP, and includes a total of 1258 key proteins.

[0037] S1011. Preprocess the protein data to remove protein data that have self-interactions and repetitive interactions.

[0038] S1012. Determine the magnitude of gene expression change V based on the mean and variance of protein gene expression values ​​in the overlapping data. u Proteins are classified into dynamic proteins and conserved proteins based on the magnitude of their changes.

[0039] The formulas for the mean and variance of gene expression values, as well as the fluctuation range, are as follows:

[0040]

[0041]

[0042]

[0043] Among them, T u (i) represents the gene expression value of protein u at time i, T′ u n represents the average gene expression value of protein u, T represents the total number of time points in the gene expression profile, and n represents the total number of time points in the gene expression profile. u The gene marker representing protein u is approximately located in [T′]. u -σ 2 , T′ u +σ 2 The number of time intervals, V u ∈[0,1],V u The closer a value is to 1, the more stable the gene expression value is, and the more likely it is to be a conserved protein. The closer a value is to 0, the more likely the protein is to be a dynamic protein.

[0044] S1013, Filtering Noise Genes: Sort all time-independent genes in ascending order according to their gene expression data. Given a threshold, if the average gene expression value is less than the threshold, the gene is considered noise and filtered out.

[0045] S1014. Screening gene expression data points: Use the threshold function Active_threshold=u+kσ×(1-F) to determine whether the dynamic protein is active at a certain time point. When the protein data is active at the corresponding time point, it is added to the time snapshot at that time.

[0046] Where u and σ are the mean and standard deviation of gene expression values, respectively. If the expression value fluctuates significantly, the corresponding σ value is higher. If the F value is small, the threshold may be greater than all of its expression values.

[0047] In other words, some proteins with large fluctuations will be filtered out. The K value represents a parameter, and in this embodiment, the k value is selected as 3.

[0048] S1015. Constructing a static protein network at a given time point: If the expression level of a gene exceeds its activity threshold at a given time point, the corresponding protein is considered active at that time point. For each time point, if two interacting proteins in the static protein network are active at the same time point, then the protein and its interaction form part of the dynamic protein interaction network at that time point, and this process is repeated.

[0049] S1016. Based on conserved proteins and dynamic proteins active at each time step, proteins are modeled as nodes, and the interactions between proteins are modeled as edges. Combining the static protein network at each time step, the final dynamic protein interaction network G is constructed. t =(V t E t ).

[0050] Among them G t It is the protein-related interaction network in the active state at time t, V t Is it Figure G? t The set of proteins contained in E t It is G t The edge set of protein-protein interactions.

[0051] S102. Random walk sampling is performed on the dynamic protein interaction network to construct the training corpus.

[0052] Specifically, in this embodiment, the protein interaction graph is sampled using a random walk node2vec method to collect training data. The specific sampling process is as follows:

[0053] S1021. Input the graph snapshots (static protein network) at all times and the corresponding adjacency matrix as parameters.

[0054] S1022. Traverse each graph snapshot at each time step, and perform node2vec random walk sampling on each graph snapshot with a fixed walk length and a fixed number of walks as the training corpus for positive sampling. The positive sampling result is used as the context node of the protein at a certain time step.

[0055] S1023. Traverse each graph and negative sample each graph snapshot at each time step with a fixed walk length and number of walks. The negative sampling result is a node that does not belong to the protein context at a certain time step.

[0056] S1024. Add the positive and negative sampling sequences of the protein nodes to the graph information. The purpose of the positive and negative sampling sequences is to calculate the final loss value.

[0057] In some implementations, a random walk node2vec sampling method is used to sample nodes in a dynamic protein-protein interaction network, obtain node visit sequences, and construct training corpus.

[0058] Through the above steps, sequence information associated with protein nodes can be effectively collected and added to the graph information at the corresponding time points, providing initial training data for subsequent model building. At the same time, this sampling process helps to avoid the data sparsity problem in protein-protein interaction networks and enhances the model's context awareness ability.

[0059] S103. Using the trained graph convolutional neural network model, extract the structural features of protein nodes.

[0060] Specifically, the node structure features are obtained by using the static encoder GCN on the multiple time snapshots obtained in step S102. The specific steps are as follows:

[0061] S1031. Use one-hot vectors as the initial features for each node in each time snapshot.

[0062] S1032. Calculate the t-plot G at each time step. t The corresponding adjacency matrix A, plus the adjacency matrix after self-connection Then perform a normalization operation:

[0063]

[0064] Among them, diagonal array have I n This represents a self-join, where A is the adjacency matrix. It involves adding a self-connected adjacency matrix. It is regularized This represents the diagonal element in the i-th row and i-th column. It is the weight value from node i to node j in the adjacency matrix after adding self-connection.

[0065] S1033. Calculate the graph convolution result for each layer sequentially for all time snapshots:

[0066]

[0067] Where H l W represents the node vector of the l-th layer. l H represents the parameters of the corresponding layer. l+1 This represents the output of the (l+1)th layer, i.e., the current layer, and f(.) is the graph convolution operation.

[0068] S1034. Obtain the structural features {z} of the nodes in each time snapshot. 1 ,z 2 ,…,z T}, z T This represents the structural characteristics of the node at time T.

[0069] S104. Input the structural feature data into the pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes.

[0070] Specifically, in S104, the temporal evolution characteristics of nodes are learned using a long short-term memory network model. The specific steps are as follows:

[0071] S1041, The output of step 3 is {z} 1 ,z 2 ,…,z T}, z T This represents the representation of all nodes in the protein network at time T. Now, we extract the representation of each node at each time step to construct... The features of each node are sorted in chronological order. As input to the Long Short-Term Memory (LSTM) network.

[0072] The process of learning the temporal evolution characteristics of a single node follows these steps:

[0073] S1042. Use the forget gate in LSTM to determine which information from previous cell states should be discarded, considering the previous hidden state h. t-1 and new input How the forget gate in LSTM controls the discarding of cell states C t-1 Information about the corresponding element.

[0074] The Gate of Oblivion can be described as:

[0075]

[0076] Among them, W f and U f Indicates the parameter, b f The deviation term is represented by σ(), which is a logical sigmoid function that maps the input to a range between 0 and 1. t-1 It is the hidden state of the previous layer. It is the feature representation of node v at time t.

[0077] S1043. Use the LSTM input gate to determine the new input. Which information needs to be stored in the new cell state?

[0078]

[0079] Among them, W i and U i It is a parameter, b i Indicates the deviation term. h is the feature representation of node v at time t. t-1 It is the hidden state of the previous layer.

[0080] S1044, Input Information Candidate values ​​are generated after processing by a neural network.

[0081]

[0082] Among them, W c and U c It is a parameter, b c Indicates the deviation term. h is the feature representation of node v at time t. t-1 It is the hidden state of the previous layer.

[0083] S1045, via combined unit state C t-1 and new candidate cell state Generate a new cell state C t :

[0084]

[0085] Among them, f t Information discarded through the forget gate, i t It is the information that needs to be stored in the new unit state, and ⊙ is the XOR operation.

[0086] S1046. Calculate the output at this position using the output gate:

[0087]

[0088] Among them, W o and U o It is a parameter, b o Indicates the deviation term. h is the feature representation of node v at time t. t-1 It is the hidden state of the previous layer.

[0089] S1047. Generate a new hidden state:

[0090] h t =o t ⊙tanh(C t )

[0091] Here, ⊙ represents the XOR operation.

[0092] S1048, Finally, the evolutionary characteristics of each node are learned.

[0093] S1049. After learning one round of structural and temporal features, forward propagation is performed using a loss function to calculate the loss value, where the loss function is:

[0094]

[0095] Among them, L v Here, u represents the set of all positively sampled nodes of node v during the walk, u′ is the set of all negatively sampled nodes of node v, and w is the loss function value. nIt is a custom parameter that represents the proportion of negative sampling results selected.

[0096] S10410, Backpropagation recalculates node features. When the training rounds are reached, the final node features are obtained. The learning process of all nodes is carried out in the same way as above. Multiple nodes learn at the same time, and finally the features of all nodes are learned.

[0097] S105. Identify key proteins through pattern classification based on structural and temporal characteristics.

[0098] Specifically, after obtaining the structural and temporal features of the protein, SVM is used to identify key proteins. The specific steps are as follows:

[0099] S1051. The protein data with acquired features and their corresponding labels are used as raw data to divide the data into training and testing sets, with the corresponding labels coming from key proteins.

[0100] All protein data were processed, with critical proteins assigned a value of 1 and non-critical proteins assigned a value of 0.

[0101] S1052. Use a single-class support vector machine (SVM) to train the protein data in the training set.

[0102] S1053. Test the test set to calculate the accuracy of key protein identification.

[0103] This application models the protein interaction graph as a dynamic graph and combines a graph convolutional neural network model and a long short-term memory network model to learn the structural and temporal features of protein nodes on the dynamic graph. This solves the problem of ignoring temporal features in the process of identifying key proteins, and can identify key proteins more efficiently. The key protein identification method based on dynamic graph neural network described in this application can achieve accurate identification of key proteins with high accuracy and efficiency.

[0104] It should be noted that the above description describes some embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the above embodiments and still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0105] Based on the same inventive concept, and corresponding to any of the above embodiments, the embodiments of this application also provide a key protein identification device based on a dynamic graph neural network.

[0106] like Figure 2 As shown, the key protein recognition device based on a dynamic graph neural network includes:

[0107] The acquisition module 11 is configured to acquire first raw protein data, process the first raw protein data, and construct a dynamic protein interaction network with time attributes based on the processed first raw protein data, wherein the first raw protein data includes key protein data with tags.

[0108] Training corpus construction module 12 is configured to perform random walk sampling on the dynamic protein interaction network to construct training corpus;

[0109] The structural feature extraction module 13 is configured to extract the structural features of protein nodes using a trained graph convolutional neural network model.

[0110] The temporal feature extraction module 14 is configured to input structural feature data into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes.

[0111] The classification and identification module 15 is configured to identify key proteins by pattern classification based on structural features and the time features.

[0112] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware.

[0113] The apparatus of the above embodiments is used to implement the corresponding key protein identification method based on dynamic graph neural network in any of the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be described again here.

[0114] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, embodiments of this application also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the key protein identification method based on dynamic graph neural network as described in any of the above embodiments.

[0115] Figure 3This embodiment illustrates a more specific hardware structure of an electronic device, which may include a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected internally via the bus 1050.

[0116] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0117] The memory 1020 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 1020 and is called and executed by the processor 1010.

[0118] The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components within the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touchscreens, microphones, various sensors, etc., while output devices may include displays, speakers, vibrators, indicator lights, etc.

[0119] The communication interface 1040 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0120] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0121] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0122] The electronic devices described above are used to implement the corresponding key protein identification methods based on dynamic graph neural networks in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0123] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, this application also provides a non-transitory computer-readable storage medium storing computer instructions for causing the computer to execute the key protein identification method based on dynamic graph neural network as described in any of the above embodiments.

[0124] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0125] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to execute the key protein identification method based on dynamic graph neural network as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0126] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this application (including the claims) is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of the embodiments of this application as described above, which are not provided in the details for the sake of brevity.

[0127] Additionally, to simplify the description and discussion, and to avoid obscuring the embodiments of this application, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the embodiments of this application, and this also takes into account the fact that the details of the implementation of these block diagram apparatuses are highly dependent on the platform on which the embodiments of this application will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) have been set forth to describe exemplary embodiments of this application, it will be apparent to those skilled in the art that the embodiments of this application can be implemented without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0128] Although this application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0129] The embodiments of this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of this application should be included within the protection scope of this application.

Claims

1. A method for identifying key proteins based on dynamic graph neural networks, characterized in that, The method includes: First raw protein data is acquired, and data processing is performed on the first raw protein data. Based on the processed first raw protein data, a dynamic protein-protein interaction network with time attributes is constructed. The first raw protein data includes tagged key protein data, including: Based on the magnitude of changes in gene expression values, proteins are classified into dynamic proteins and conserved proteins. A threshold function is used to determine whether the dynamic protein is in an active state at a certain time point; In response to the dynamic protein being in an active state, a static protein network at a certain moment is constructed; Based on the conserved proteins and the dynamic proteins with active states at each time step, a dynamic protein interaction network is constructed based on the static protein network at each time step. Random walk sampling is performed on the dynamic protein-protein interaction network to construct a training corpus; Using a trained graph convolutional neural network model, structural features of protein nodes are extracted; The structural feature data is input into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes. Key proteins are identified through pattern classification based on the structural and temporal characteristics.

2. The key protein identification method based on dynamic graph neural network according to claim 1, characterized in that, The classification of proteins into dynamic proteins and conserved proteins based on the magnitude of changes in gene expression values ​​includes: The first raw protein data also includes PPI protein network data and gene expression profile data; Select overlapping data from the PPI protein network data and the gene expression profile data; The mean and variance of protein gene expression values ​​in the overlapping data are calculated to determine the range of variation in gene expression values. Based on the magnitude of the change, the proteins are classified into dynamic proteins and conserved proteins.

3. The key protein identification method based on dynamic graph neural network according to claim 1, characterized in that, The step of performing random walk sampling on the dynamic protein-protein interaction network to construct training corpus includes: The node2vec random walk sampling method is used to sample nodes in the dynamic protein interaction network to obtain node visit sequences and construct training corpus.

4. The key protein identification method based on dynamic graph neural network according to claim 3, characterized in that, The process of extracting structural features of protein nodes using a trained graph convolutional neural network model includes: The dynamic protein is input into the graph convolutional neural network model, and the structural features of the protein nodes are learned through the graph convolutional neural network model; The structural features are used as the output of the graph convolutional neural network model.

5. The key protein identification method based on dynamic graph neural network according to claim 4, characterized in that, The step of inputting the structural feature data into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and outputting the temporal features of protein nodes includes: The structural feature data is sorted in chronological order and then used as input to the Long Short-Term Memory network model. The temporal evolution features of protein nodes are obtained using the Long Short-Term Memory network model, and the model is trained to obtain the final temporal features.

6. The key protein identification method based on dynamic graph neural network according to claim 5, characterized in that, The identification of key proteins through pattern classification based on the structural features and the temporal features includes: The protein data with the obtained structural features and temporal features, as well as the corresponding tag data, are used as the second original protein data, wherein key proteins are used as the tag data; The second raw protein data is divided into a training set and a test set; The protein data in the training set are trained using the single-class support vector machine method to obtain an SVM classification model; The SVM classification model was tested using a test set to classify and identify key proteins.

7. A key protein recognition device based on a dynamic graph neural network, characterized in that, include: The acquisition module is configured to acquire first raw protein data, process the first raw protein data, and construct a dynamic protein-protein interaction network with time attributes based on the processed first raw protein data. The first raw protein data includes tagged key protein data, including: Based on the magnitude of changes in gene expression values, proteins are classified into dynamic proteins and conserved proteins. A threshold function is used to determine whether the dynamic protein is in an active state at a certain time point; In response to the dynamic protein being in an active state, a static protein network at a certain moment is constructed; Based on the conserved proteins and the dynamic proteins with active states at each time step, a dynamic protein interaction network is constructed based on the static protein network at each time step. The training corpus construction module is configured to perform random walk sampling on the dynamic protein-protein interaction network to construct the training corpus. The structural feature extraction module is configured to extract the structural features of protein nodes using a trained graph convolutional neural network model. The temporal feature extraction module is configured to input the structural feature data into a pre-trained long short-term memory network model to learn the temporal evolution features of protein nodes and output the temporal features of protein nodes. The classification and recognition module is configured to identify key proteins by pattern classification based on the structural features and the temporal features.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the key protein identification method based on dynamic graph neural networks as described in any one of claims 1-6.

9. A non-transitory computer-readable storage medium, characterized in that, in, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to execute the key protein identification method based on dynamic graph neural networks as described in any one of claims 1-6.

Images