Protein complex identification method, system and equipment and storage medium

By constructing a weighted protein interaction network using GO and gene expression data, and applying graph convolution and Markov clustering, the method addresses low recognition accuracy in existing algorithms, achieving improved protein complex detection.

CN119091960APending Publication Date: 2024-12-06SHANXI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411250191.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing protein complex recognition algorithms fail to fully consider the overall structural information of complex networks, leading to low recognition accuracy and relying solely on topological connectivity.

Method used

A method that constructs a weighted protein interaction network combining gene ontology (GO) information and gene expression data, uses graph convolution to calculate high-order similarity matrices, and applies Markov clustering with a fitness function to iteratively refine protein clusters.

Benefits of technology

Improves protein complex recognition accuracy by integrating biological properties and high-order structural information, enhancing the efficiency and applicability of protein complex detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119091960A_ABST
    Figure CN119091960A_ABST
Patent Text Reader

Abstract

The invention provides a protein complex identification method, system and device and a storage medium, and belongs to the field of biological data intelligent analysis, and the method comprises the steps: obtaining required protein interaction network data, and GO information and gene expression data corresponding to protein; aiming at the protein interaction network, constructing a weighted protein interaction network G in combination with GO information and gene expression data of the protein; g is expressed as an adjacent matrix A, A is normalized to obtain an adjacent similarity matrix SA, the SA is used as a feature matrix for graph convolution, and a high-order similar matrix SK of each layer of graph convolution is calculated; combining each layer of SK to form a total similar matrix S, and obtaining an initial protein cluster set through MCL; updating and iterating the proteins in the initial protein cluster set by using a fitness function; and carrying out screening operation on the obtained protein clusters so as to obtain the final protein complex. The invention aims at fusing high-order structure information and biological information so as to improve the accuracy of protein complex detection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of biological data intelligent analysis, and specifically relates to a protein complex recognition method, system, device and storage medium. Background Art

[0002] Proteins are not only the core part of life activities, but also the main participants in biological processes. Generally speaking, proteins rarely act alone. They have many interactions with each other and form protein complexes to perform special biological functions or participate in the functional processes of living cells. For example, DNA replication, transcription, protein degradation and various other cellular functions are all performed by protein complexes rather than single proteins. Therefore, the identification of protein complexes in protein interaction networks has also become an important step in studying biological processes.

[0003] Traditional experimental methods are effective methods for detecting and identifying protein complexes, such as nuclear magnetic resonance (NMR) and mass spectrometry, but experimental methods are time-consuming and expensive. Therefore, in this case, algorithms based on machine learning and data mining are appropriate and alternative methods for detecting protein complexes. In particular, the efficient development of high-throughput technology has generated a large amount of protein-protein interaction (PPI) data, which has greatly promoted the detection of protein complexes from PPI networks through algorithms.

[0004] In the past few decades, people have proposed various algorithms to detect protein complexes. The existing protein complex recognition algorithms have been proven to be effective in applications such as identifying protein complexes and detecting gene co-expression modules, but there are still some problems. On the one hand, these recognition algorithms and their related variants all use low-order topological connectivity (i.e., first-order similarity) to detect protein complexes in PPI networks. Due to the complexity of biological networks, low-order topology alone is not enough to fully consider the overall structural information of complex networks. On the other hand, these algorithms only detect protein complexes in PPI networks from a topological perspective, and the recognition accuracy is low. Summary of the invention

[0005] In order to solve the problem that the existing methods do not fully consider the overall structural information of the complex network and only detect the protein complexes in the PPI network from a topological perspective, resulting in low recognition accuracy, the present invention provides a protein complex recognition method, system, device and storage medium.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] A method for identifying a protein complex comprises the following steps:

[0008] Obtain the required protein interaction network data, as well as the GO information and gene expression data corresponding to the protein;

[0009] For the protein interaction network, a weighted protein interaction network G is constructed by combining protein GO information and gene expression data;

[0010] The weighted protein interaction network G is represented as an adjacency matrix A, and the adjacency matrix A is normalized to obtain the adjacency similarity matrix S A ; S A Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K ; The obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL;

[0011] The fitness function is used to iterate the proteins in the initial protein cluster set; the protein clusters obtained by the updated iteration are screened to obtain the final protein complex.

[0012] Preferably, before constructing the weighted protein interaction network G, the process also includes removing self-loops and duplications in the GO information and gene expression data corresponding to the protein, and removing unused information in the GO information; wherein the unused information includes the annotation source, the function or location of the gene product, and the species identifier.

[0013] Preferably, obtaining the weights of the weighted protein interaction network G comprises the following steps:

[0014] The protein GO information and gene expression data are integrated to jointly measure the weights between nodes in the PPI network, as shown below:

[0015]

[0016] Where v and u represent two protein nodes in the network, TCS(v,u) represents the sum of the semantic similarities between the two protein nodes v and u, TCSScc(v,u), TCSSbp(v,u) and TCSSmf(v,u) represent the semantic similarities between node v and node u in terms of cellular components, biological processes and molecular functions, respectively; the gene expression profiles of v and u are represented by v = {v1,v2,...,v n} and u={u1,u2,...,u n}, n represents the time value;

[0017] The Pearson correlation coefficient was used to calculate the co-expression value between proteins:

[0018]

[0019] in, and Respectively represent the average value of gene expression data of the node; v i and u i They represent the values ​​of nodes v and u at time i in the gene expression data respectively; PCS(v,u) represents the co-expression value between nodes v and u;

[0020] The sum of the semantic similarity between the two protein nodes v and u is combined with the co-expression value between nodes v and u as the weight between the nodes:

[0021] Weight(v,u)=TCS(v,u)+PCS(v,u).

[0022] Preferably, the weighted protein interaction network G is represented as an adjacency matrix A, and the adjacency matrix A is normalized to obtain an adjacency similarity matrix S A , specifically including:

[0023] The protein interaction network is represented as G = (V, E, A), where V represents the node set, E represents the continuous edge set between two nodes, and A is the adjacency matrix corresponding to the network;

[0024] Normalize the adjacency matrix A to obtain the adjacency similarity matrix S A , the adjacency similarity matrix S A Elements in for:

[0025]

[0026] Among them, max A v Represents the maximum value in a column corresponding to node v; A vu Represents the relationship between nodes v and u.

[0027] Preferably, the obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL, which specifically includes:

[0028] The total similarity matrix S is specifically expressed as follows:

[0029] S=S A +S 1 +S 2 ...S k ,

[0030] Among them, S 1, S 2 ...S k Represents the similarity matrix formed by each layer of convolution;

[0031] The total similarity matrix S is used as the matrix representation of the network;

[0032] The total similarity matrix S is normalized to generate a probability matrix M, and Markov clustering is used to perform cluster analysis on the probability matrix M to obtain a set of initial protein clusters.

[0033] Preferably, the method for constructing the fitness function includes:

[0034] Given an initial protein cluster C = (V c ,E c ,W c ), where V c is the set of proteins in cluster C, E c is the set of interactions in cluster C, W c is the set of weights between protein pairs in cluster C, and the cohesion score is defined as:

[0035]

[0036] Among them, W in (C) represents the sum of the weights of all edges in cluster C, W out (C) is the sum of the weights of the edges connecting the internal proteins in C to other proteins in the rest of the PPI network;

[0037] The weighted density Density is defined as:

[0038]

[0039] Where V c is the number of proteins in cluster C;

[0040] Based on the cohesion score and weighted density, a fitness function is constructed, which is:

[0041]

[0042] Among them, Cohension(C) represents the cohesion score and Density(C) represents the weighted density.

[0043] Preferably, the protein clusters obtained by the iterative update are screened to obtain the final protein complex, specifically:

[0044] A threshold was set and clusters with node numbers less than 2 and less than the threshold were directly deleted to obtain the final set of protein complexes.

[0045] The present invention also provides a protein complex recognition system, comprising:

[0046] Data acquisition module, used to obtain the required protein interaction network data, as well as GO information and gene expression data corresponding to proteins;

[0047] The network construction module is used to construct a weighted protein interaction network G by combining protein GO information and gene expression data for protein interaction networks;

[0048] Protein cluster acquisition module, used to represent the weighted protein interaction network G as an adjacency matrix A, and normalize the adjacency matrix A to obtain the adjacency similarity matrix S A ; S A Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K ; The obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL;

[0049] The screening module is used to update and iterate the proteins in the initial protein cluster set using the fitness function; and to perform screening operations on the protein clusters obtained by the updated iteration to obtain the final protein complex.

[0050] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement any one of the steps in the protein complex identification method.

[0051] The present invention also provides a computer-readable storage medium, on which a computer program is stored, characterized in that when the computer program is loaded by a processor, it can execute any one of the steps in the protein complex identification method.

[0052] The protein complex identification method provided by the present invention has the following beneficial effects:

[0053] This method first constructs a weighted protein interaction network G for the protein interaction network, combining the protein GO information and gene expression data. It not only considers the high-order structural information of the protein interaction network, but also considers multiple biological properties of proteins to compensate for the network data, thereby overcoming the limitations of low-order topological connectivity and the negative impact of high noise. Then, the weighted protein interaction network G is transformed, converting the network from low-order to high-order, obtaining an initial protein cluster set, and updating, iterating and screening the proteins in the initial protein cluster set to obtain the final protein complex. This method integrates biological properties and high-order structural information, combines biological-specific information, thereby improving the accuracy of protein complexes, improving the efficiency of prediction, and expanding the application scope and practicality of this technology in the field of biological information. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiment of the present invention and its design scheme, the following briefly introduces the drawings required for this embodiment. The drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 A flow chart of the protein complex identification method provided by the present invention.

[0056] Figure 2 The present invention provides an application flow chart of the protein complex identification method provided by the present invention. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the technical solution of the present invention and implement it, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the scope of protection of the present invention.

[0058] Example 1

[0059] The present invention provides a protein complex identification method, specifically a protein complex identification method based on high-order structural information, specifically as follows Figure 1 and Figure 2 As shown, the following steps are included:

[0060] S1: Obtain the required protein interaction network data, as well as the GO information and gene expression data corresponding to the protein, remove the self-loops and duplications, and remove the unused information in the GO information. Among them, the unused information includes the annotation source, the function or location of the gene product, and the species identifier, etc.

[0061] S2: For the protein interaction network, a weighted protein interaction network G is constructed by combining protein GO information and gene expression data.

[0062] The weights between nodes in the PPI network are measured by effectively integrating protein GO information and gene expression data. GO information is a useful and popular controlled biological terminology classification method that can be used to evaluate the functional relationship between different gene products. It mainly contains information on molecular function (MF), biological process (BP), and cellular component (CC). The specific expression is as follows:

[0063]

[0064] Gene expression data is the data obtained by measuring the expression (activity) levels of a large number of genes at the same time. These data can reflect the response of cells to specific treatments or what they do at a specific time point, etc. The present invention uses the Pearson correlation coefficient calculation method to calculate the co-expression value between proteins:

[0065]

[0066] Where v and u represent two protein nodes in the network, and their gene expression profiles are represented by v = {v1, v2, ..., v n} and u={u1,u2,...,u n}, n represents the time value. and They represent the average value of the gene expression data of the node respectively; PCS(v,u) represents the co-expression value between nodes v and u.

[0067] Here we consider combining the biological information of two proteins, that is, fusing equation (1) and equation (2), and we can get the weight between the two protein nodes:

[0068] Weight(v,u)=TCS(v,u)+PCS(v,u) (3)

[0069] S3: Represent the weighted protein interaction network G as an adjacency matrix A, and normalize the adjacency matrix A to obtain the adjacency similarity matrix S A , and S A Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K .

[0070] The protein interaction network obtained in the previous step is represented as G = (V, E, A), where V represents the node set, E represents the continuous edge set between two nodes, and A is the adjacency matrix corresponding to the network. In view of the significant differences in the weights of different connected edges, the adjacency matrix A is normalized according to formula (4) and the adjacency similarity matrix S is obtained: A , the adjacency similarity matrix S A Elements in The specific formula is as follows:

[0071]

[0072] Among them, maxA v Represents the maximum value in a column corresponding to node v; A vu Represents the relationship between nodes v and u.

[0073] Since the first-order similarity cannot accurately reflect the overall structure of the PPI network, the present invention uses graph convolution to measure multi-order similarity between nodes. Among them, the features of all nodes between the convolution layers are updated in an iterative manner, which is specifically defined as follows:

[0074]

[0075] Among them, node H K The feature matrix represents the convolutional layer of the Kth layer, H0 is the identity matrix I, f represents the inter-layer state update function, represents an adjacency matrix with self-loops, i.e.

[0076] During the gradual update process, each node aggregates the features of its neighboring nodes and the edge information of the connection to enrich its feature information. In this step, the node feature information in each convolutional layer is considered to measure the multi-order similarity between nodes. Considering that the larger the weight of the connected edge, the more significant the impact on the target node, the present invention associates the connected edge weight with the features of the adjacent nodes in each convolutional layer, and takes the sum of the influence values ​​of all adjacent nodes on the target node as the new feature vector of the target node. Therefore, the specific description of the node feature formula between convolutional layers is as follows.

[0077]

[0078] Where g(a,H i )=a×H i , N(v) represents the set of adjacent nodes including node v itself. Represents the relationship between node v and its adjacent node i.

[0079] Through the above formula, we can get the feature vector of each node in each convolution layer. As the number of convolution layers increases, the aggregation range of the nodes expands. The continuously updated feature vector of the node represents the relationship between the node and all the nodes in its aggregation range. The higher the edge weight or the more connected paths, the stronger the relationship between the nodes. These feature vectors are used to provide the relationship between the target node and other nodes in the K-order neighborhood, and the cosine similarity is used to calculate the K-order similarity of the nodes. The specific formula is as follows:

[0080]

[0081] Indicates that cosine similarity is used to calculate the similarity between the target node and other nodes in the k-order neighborhood.

[0082] S4: The high-order similarity matrix S of each layer obtained by the above operation K Combined to form the total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL

[0083] Neighborhood information of different scales (K-order neighborhood similarity, S K ) is the low-order neighborhood similarity information of the PPI network (neighborhood similarity, S A ). They measure node similarity from different perspectives and help reveal the global structural information of the network. In order to make full use of this multi-order similarity information, the low-order similarity information is integrated with the high-order similarity information to form the total similarity matrix S, and the total high-order similarity matrix S is formed by formula (8):

[0084] S=S A +S 1 +S 2 ...S k (8)

[0085] Among them, S 1 , S 2 ...S k Represents the similarity matrix formed by each convolution layer.

[0086] The total similarity matrix S is used as the matrix representation of the network. The higher the value in the matrix, the stronger the connection between the two nodes. Secondly, the high-order similarity matrix S is normalized using formula (9) to generate the probability matrix M. The Markov clustering MCL is used to perform cluster analysis on the probability matrix M to obtain the set of initial protein clusters.

[0087]

[0088] Among them, S ij Represents the elements in the total similarity matrix S.

[0089] S5: Use the fitness function to iteratively update the proteins in the initial protein cluster set.

[0090] In step S5, the fitness function is defined by combining the cohesion score with the weighted density for judgment. Given an initial protein cluster C = (V c ,E c ,W c ) where V c is the set of proteins in cluster C, E c is the set of interactions in the cluster, W c is the weight set between protein pairs in the cluster. The cohesion score can intuitively show the overall situation of the protein complex, but it cannot reflect the relationship between its internal nodes, which leads to the evaluation result being not comprehensive and accurate. The weighted density can better show the local situation of the protein complex. Therefore, the fitness function is defined by combining the cohesion score with the weighted density. The specific calculation is shown in formulas (10) and (11), and the final fitness function is shown in formula (12).

[0091]

[0092] Where W in (C) represents the sum of the weights of all edges in cluster C, W out (C) is the sum of the weights of the edges connecting the internal proteins in C to other proteins in the rest of the PPI network.

[0093] The weighted density is defined as follows:

[0094]

[0095] Where V c is the number of proteins in cluster C. The higher the density of a cluster, the more likely it is that the cluster represents a protein complex.

[0096] Based on the cohesion score and weighted density, the fitness function definition given by the present invention is:

[0097]

[0098] Among them, Cohension(C) represents the cohesion score and Density(C) represents the weighted density.

[0099] In general, a high-quality protein complex is a group that is densely interconnected but sparsely interconnected with the rest of the PPI network. According to Density (C), protein complexes with dense internal connections can be found. According to Cohension (C), protein complexes with densely interconnected nodes can be identified, which are sparsely interconnected with the rest of the PPI network. And combining in this way can better avoid biased evaluation results. Therefore, this fitness function can detect various topological properties of protein complexes in the PPI network.

[0100] S6: Screening operation is performed on the protein clusters obtained in step S5 to obtain the final protein complex.

[0101] Specifically, step S6 filters the obtained protein complex set. Considering that the protein complex is composed of a group of proteins with similar functions and structures, although the protein complex obtained above strengthens the connection between nodes, there are still clusters with low fitness. Therefore, a single protein cannot become a protein complex. Secondly, the fitness function includes global and local considerations of the protein complex, so a threshold is set and the obtained protein complex is calculated. If it is less than the threshold, it means that the internal structure is not similar, and it will be deleted.

[0102] In this embodiment, clusters with node numbers less than 2 and less than a threshold are directly deleted to obtain the final protein complex set.

[0103] The protein complex set obtained by the recognition method based on high-order structural information makes the protein functions and structures in the complex more similar, and the accuracy of the protein complex is higher. Not only can potential drug targets be discovered through key protein complexes, but it also helps in the study of the molecular mechanism of diseases, promoting a deeper understanding and application of biological systems.

[0104] Based on the same inventive concept, the present invention also provides a protein complex identification system, including a data acquisition module, a network construction module, a protein cluster acquisition module and a screening module.

[0105] Specifically, the data acquisition module is used to obtain the required protein interaction network data, as well as GO information and gene expression data corresponding to the protein.

[0106] The network construction module is used to construct a weighted protein interaction network G based on the protein GO information and gene expression data.

[0107] The protein cluster acquisition module is used to represent the weighted protein interaction network G as an adjacency matrix A and normalize the adjacency matrix A to obtain the adjacency similarity matrix S A ; SA Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K ; The obtained high-order similarity matrix S of each layer K They are combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL.

[0108] The screening module is used to update and iterate the proteins in the initial protein cluster set using the fitness function; the protein clusters obtained by the updated iteration are screened to obtain the final protein complex.

[0109] Each module in the above protein complex identification system can be implemented in whole or in part by software, hardware and a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0110] The present invention also provides a computer device, including a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps in the protein complex identification method embodiment. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0111] Furthermore, the present invention also provides a non-temporary computer-readable storage medium containing instructions, and a computer program is stored on the storage medium. For example, a memory containing instructions, the above instructions can be executed by a processor of a computer device to complete the above method. For example, a non-temporary computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a tape, a floppy disk, and an optical data storage device. When the computer program is executed by the processor, the steps in the protein complex identification method embodiment can be implemented. The specific implementation method can be found in the method embodiment, which will not be repeated here.

[0112] It will be appreciated by those skilled in the art that embodiments of the present invention may provide methods, systems or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0113] The present invention is described with reference to flowcharts and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as a combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0114] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0115] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0116] It should be pointed out that the specific implementation methods described above can enable those skilled in the art to understand the invention more comprehensively, but do not limit the invention in any way. Therefore, although the invention has been described in detail in this specification and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and scope of the invention are included in the protection scope of the patent for the invention. Any figure mark in the claims should not be regarded as limiting the claims involved. Any simple change or equivalent replacement of the technical solution that can be obviously obtained by any technician familiar with the field within the technical scope disclosed in the present invention belongs to the protection scope of the present invention.

Claims

1. A method for identifying a protein complex, characterized in that: The following steps are involved: Obtain the required protein interaction network data, as well as the GO information and gene expression data corresponding to the protein; For the protein interaction network, a weighted protein interaction network G is constructed by combining protein GO information and gene expression data; The weighted protein interaction network G is represented as an adjacency matrix A, and the adjacency matrix A is normalized to obtain the adjacency similarity matrix S A ; S A Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K ; The obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL; The proteins in the initial protein cluster set are updated and iterated using the fitness function; The protein clusters obtained by the updated iteration are screened to obtain the final protein complex.

2. The protein complex identification method according to claim 1, characterized in that: Before constructing the weighted protein interaction network G, the self-loops and duplications in the GO information and gene expression data corresponding to the protein are removed, and the unused information in the GO information is removed; the unused information includes the annotation source, the function or location of the gene product, and the species identifier.

3. The protein complex identification method according to claim 2, characterized in that: Obtaining the weights of the weighted protein interaction network G comprises the following steps: The protein GO information and gene expression data are integrated to jointly measure the weights between nodes in the PPI network, as shown below: Where v and u represent two protein nodes in the network, TCS(v,u) represents the sum of the semantic similarities between the two protein nodes v and u, TCSScc(v,u), TCSSbp(v,u) and TCSSmf(v,u) represent the semantic similarities between node v and node u in terms of cellular components, biological processes and molecular functions, respectively; the gene expression profiles of v and u are represented by v = {v1,v2,…,v n } and u={u1,u2,...,u n }, n represents the time value; The Pearson correlation coefficient was used to calculate the co-expression value between proteins: in, and Respectively represent the average value of gene expression data of the node; v i and u i They represent the values ​​of nodes v and u at time i in the gene expression data respectively; PCS(v,u) represents the co-expression value between nodes v and u; The sum of the semantic similarity between the two protein nodes v and u is combined with the co-expression value between nodes v and u as the weight between the nodes: Weight(v,u)=TCS(v,u)+PCS(v,u).

4. The protein complex identification method according to claim 1, characterized in that: The weighted protein interaction network G is represented as an adjacency matrix A, and the adjacency matrix A is normalized to obtain an adjacency similarity matrix S A , specifically including: The protein interaction network is represented as G = (V, E, A), where V represents the node set, E represents the continuous edge set between two nodes, and A is the adjacency matrix corresponding to the network; Normalize the adjacency matrix A to obtain the adjacency similarity matrix S A , the adjacency similarity matrix S A Elements in for: Among them, maxA v Represents the maximum value in a column corresponding to node v; A vu Represents the relationship between nodes v and u.

5. The protein complex identification method according to claim 4, characterized in that: The obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL, which specifically includes: The total similarity matrix S is specifically expressed as follows: S=S A +S 1 +S 2 ...S k , Among them, S 1 , S 2 ...S k Represents the similarity matrix formed by each layer of convolution; The total similarity matrix S is used as the matrix representation of the network; The total similarity matrix S is normalized to generate a probability matrix M, and Markov clustering is used to perform cluster analysis on the probability matrix M to obtain a set of initial protein clusters.

6. The protein complex identification method according to claim 1, characterized in that: The method for constructing the fitness function includes: Given an initial protein cluster C = (V c ,E c ,W c ), where V c is the set of proteins in cluster C, E c is the set of interactions in cluster C, W c is the set of weights between protein pairs in cluster C, and the cohesion score is defined as: Among them, W in (C) represents the sum of the weights of all edges in cluster C, W out (C) is the sum of the weights of the edges connecting the internal proteins in C to other proteins in the rest of the PPI network; The weighted density Density is defined as: Where V c is the number of proteins in cluster C; Based on the cohesion score and weighted density, a fitness function is constructed, which is: Among them, Cohension(C) represents the cohesion score and Density(C) represents the weighted density.

7. The protein complex identification method according to claim 1, characterized in that: The protein clusters obtained by the iterative update are screened to obtain the final protein complex, specifically: A threshold was set and clusters with node numbers less than 2 and less than the threshold were directly deleted to obtain the final set of protein complexes.

8. A protein complex recognition system, characterized in that: include: Data acquisition module, used to obtain the required protein interaction network data, as well as GO information and gene expression data corresponding to proteins; The network construction module is used to construct a weighted protein interaction network G by combining protein GO information and gene expression data for protein interaction networks; Protein cluster acquisition module, used to represent the weighted protein interaction network G as an adjacency matrix A, and normalize the adjacency matrix A to obtain the adjacency similarity matrix S A ; S A Perform graph convolution as a feature matrix and calculate the high-order similarity matrix S of each layer of graph convolution K ; The obtained high-order similarity matrix S of each layer K Combined to form a total similarity matrix S, and the initial protein cluster set is obtained from the total similarity matrix S through Markov clustering MCL; A screening module is used to iteratively update the proteins in the initial protein cluster set using a fitness function; The protein clusters obtained by the updated iteration are screened to obtain the final protein complex.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is loaded into a processor, the computer program can execute the steps of the method according to any one of claims 1 to 7.