Gene interaction relation prediction method and device based on graph neural network

Abstract

The invention discloses a gene interaction relationship prediction method and device based on a graph neural network, and the method comprises the steps: carrying out the preprocessing of the transcription expression data of each experimental group, including the standardization of a gene ID, and representing the gene transcription expression data of different experimental groups through a unified gene ID; obtaining original data for subsequent analysis; performing gene similarity calculation according to each sample group to obtain similarity data between genes in the sample groups; performing statistics and screening on data of all sample groups, and selecting gene pairs meeting set conditions as a basis for constructing a graph data structure; further processing the original data to ensure that all sample groups only contain genes defined in the graph structure; then, carrying out standardization processing on the data of the sample groups, so that the data dimensions of all the sample groups are consistent; and carrying out model training on the generated graph data set to finally obtain edge weight data in a graph structure, and accurately reflecting an interaction relationship between genes. The method can be used for analyzing gene expression data, predicting the interaction relationship between genes, and carrying out unsupervised learning under the condition of no label data.

Description

Technical Field

[0001] The present invention relates to the field of graph neural networks (GNNs), and in particular to a method and device for predicting gene interaction relationships based on a graph attention network (GAT). Background Art

[0002] In bioinformatics research, the analysis of gene expression data is an important way to explore gene functions and their interactions. Traditional gene similarity calculation methods are usually based on statistics and distance metrics, but these methods have limitations when dealing with complex gene relationships. In recent years, graph neural network technology has shown significant advantages in processing graph structured data, especially in gene network analysis, which can effectively mine potential gene interactions.

[0003] Graph Attention Network (GAT) is a type of graph neural network. It adaptively assigns weights to adjacent nodes, making the network more flexible in processing node features. However, most existing GAT models rely on supervised learning, and in many genetic data analysis scenarios, there is often a lack of sufficient labeled data. Therefore, unsupervised GAT models have become an important development direction.

[0004] The GAT model in the prior art usually faces problems such as high computational complexity and difficulty in determining the direction of gradient descent when processing gene expression data. Therefore, a model architecture that can simplify the calculation process and effectively process unlabeled data is needed. The present invention proposes an unsupervised gene relationship prediction method based on a graph attention network, which simplifies the calculation through ResNet and pooling operations, and combines the addition and deletion of edges to improve the accuracy and efficiency of gene correlation prediction. Summary of the invention

[0005] The present invention aims to overcome the technical problems of low accuracy and efficiency in gene correlation prediction in the prior art, and to provide a method and device for predicting gene interaction relationships based on graph neural networks that is simple, easy to operate and capable of processing complex gene expression data.

[0006] The method of the present invention combines unsupervised learning technology, calculates the similarity between gene expression data, and predicts edges (edge ​​addition and edge deletion) on the graph structure, thereby achieving effective prediction of gene correlation. This technology is suitable for gene function analysis and gene expression data mining in the field of bioinformatics.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is that the first aspect of the present invention provides a gene interaction relationship prediction method based on a graph neural network, comprising the following steps:

[0008] a. Preprocess the transcriptional expression data of each experimental group, including standardizing the gene ID, and representing the gene transcriptional expression data of different experimental groups with a unified gene ID to ensure consistent data format. Then, group the cell samples in the experimental group, and one sample group contains multiple samples, and finally obtain the raw data for subsequent analysis;

[0009] b. For the above raw data, calculate the gene similarity for each sample group and obtain the similarity data between genes in the sample group. By counting and screening the data of all sample groups, select gene pairs that meet the set conditions as the basis for constructing the graph data structure to ensure that the selected gene pairs can reflect the actual biological significance;

[0010] c. Based on the graph structure obtained by the above screening, the original data is further processed to ensure that all sample groups contain only the genes defined in the graph structure. Subsequently, the data of these sample groups are standardized to make the data dimensions of all sample groups consistent. Through this standardization process, multiple graph data groups with the same structure but different node characteristics (i.e., gene expression characteristics) are formed for subsequent graph neural network model training;

[0011] d. Perform model training on the graph data set generated above. The model used is the graph attention network (GAT) and a simple edge prediction algorithm. Through multiple iterative training of the model, the loss function is gradually optimized, and finally the edge weight data in the graph structure is obtained, which can accurately reflect the interaction relationship between genes.

[0012] Preferably, in step a, the specific operation method of the preprocessing process of transcription expression data is:

[0013] a1. Since different experimental groups use different sequencing technologies, the formats of gene expression data are inconsistent. Therefore, the data of all experimental groups need to be uniformly converted into the same gene ID or other consistent standard data formats to ensure the comparability of subsequent analysis.

[0014] a2. According to different experimental condition variables (such as cell type, experimental environment, etc.), each set of experimental data is grouped and processed. Each experimental data can be divided into multiple sample groups, and the condition variable can be the cell type (such as normal cells, tumor cells, etc.);

[0015] a3. Arrange the grouped data to form a set of sample groups corresponding to each cell type, ensuring that the gene expression data in each sample group can reflect the biological characteristics of the type.

[0016] Preferably, in step b, the specific operation methods of similarity calculation and data screening are as follows:

[0017] b1. For each sample group under each cell category, calculate the similarity between the gene pairs in the group in turn. The data of each sample is regarded as the characteristics of the gene, and an appropriate similarity calculation method (such as Pearson correlation coefficient, Euclidean distance, etc.) is used. The selection of these methods should be determined according to the characteristics of the data (such as whether the data is normally distributed, whether the data type is an integer or a decimal, etc.), and the similarity data between genes in the sample group is calculated and obtained;

[0018] b2. For each sample group, perform a preliminary screening of the similarity results based on a pre-set threshold, and retain gene pairs with similarity greater than or equal to the threshold. The selection of the threshold is very critical and should be tested and adjusted to ensure that the number of gene pairs screened is moderate, without losing key information or being excessively redundant;

[0019] b3. Perform statistical screening operations on the sample groups of each cell category, count the gene pairs of all sample groups under the same category, and select the gene pairs whose occurrence frequency is greater than the set threshold. The choice of threshold will also have an important impact on the formation of the graph structure. The gene pairs that are finally retained will constitute the edges in the graph data, which is the core of the graph structure.

[0020] Preferably, in step c, the specific operation method of the standardization processing and graph data generation is as follows:

[0021] c1. When screening gene pairs, some sample groups may not contain the screened genes, so these genes need to be supplemented. You can use the trained model to complete the data, or directly use 0 values ​​to ensure that each sample group contains the required gene pairs. Through such screening and supplementation, you can finally obtain a gene data set containing all graph structure nodes;

[0022] c2. Standardize the gene expression data of each sample group to unify its dimensions. Standardization methods can include selecting statistical features such as median, mean, variance, or using principal component analysis (PCA) to extract principal components and calculate their contribution rates to determine the rationality of principal component selection. Ensure data consistency and comparability through standardization;

[0023] c3. Arrange the generated graph structure and standardized data to finally form an adjacency matrix represented by a sparse matrix and multiple feature matrix data groups. This data group is used for subsequent model training and ensures the consistency of data in each sample group during the training process.

[0024] Preferably, in step d, the specific operation methods of model training and edge prediction are as follows:

[0025] d1. Use the modified Graph Attention Network (GAT) model for training. Since the data is unlabeled, an unsupervised method is used for training. The key to training is to define the loss function of GAT. By calculating the graph data loss between each sample group, the total loss value is obtained to determine the direction of gradient descent. In order to simplify the amount of data calculation, ResNet and pooling operations are first performed on the graph data, and then the loss between the graph data is calculated. According to the complexity of the experimental data, the appropriate number of convolutional layers and feature outputs are selected, and the learning rate and gradient descent parameters are adjusted to ensure that the model can converge effectively in unsupervised learning.

[0026] d2. During the model training process, the attention mechanism in the GAT model is used to calculate the attention weight αij of each gene pair, and the sparse adjacency matrix is ​​screened to delete the edges whose attention weight is lower than the set threshold. The selection of the threshold is determined through multiple experiments, and the minimization of the loss value is used as the judgment criterion;

[0027] d3. For the deleted edges, based on the feature matrix output by the last convolution layer of the model, calculate the similarity of the gene pairs and select the edges that are greater than the set threshold and not included in the original adjacency matrix for addition. Retrain the model, adjust the threshold, and continue to optimize the loss value;

[0028] d4. The edge prediction process can be iterated multiple times until the loss value no longer decreases, and the final output is the gene relationship data for a certain cell category.

[0029] The second aspect of the present invention relates to a method and device for predicting gene interaction relationships based on a graph neural network, including a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the code, it is used to implement the method for predicting gene interaction relationships based on a graph neural network of the present invention.

[0030] The third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, can implement the gene interaction relationship prediction method based on graph neural network of the present invention.

[0031] The present invention generates multiple graph data groups with the same structure but different node features by preprocessing the gene transcription expression data, calculating similarity, constructing the graph structure, and standardizing the gene transcription expression data. The model is trained using a graph attention network (GAT), and the loss function is defined in combination with an unsupervised learning method. Specifically, the loss value is obtained by calculating the graph data loss between different sample groups, thereby determining the direction of gradient descent. During the training process, in order to simplify the calculation, the graph data is first subjected to ResNet and pooling operations, and then the loss between the graph data is calculated. During training, the model adjusts the number of convolutional layers, feature output, learning rate, and gradient descent parameters according to the complexity of the data to ensure effective convergence. This method can be used for the analysis of gene expression data, predicting the interaction between genes, and can perform unsupervised learning in the absence of labeled data.

[0032] Compared with the prior art, the advantages and positive effects of the present invention are:

[0033] The present invention predicts gene correlation through an unsupervised graph attention network (GAT) model, which overcomes the reliance of the prior art on a large amount of labeled data and can effectively predict in unlabeled gene data. In addition, the present invention proposes an edge prediction mechanism, which adaptively adds or deletes edges in the graph according to the relationship between gene data, thereby improving the flexibility of the model and the accuracy of prediction. In order to simplify the calculation process, the present invention introduces ResNet and pooling operations into the model, which reduces the amount of data calculation while ensuring the effective convergence of the model, thereby improving the calculation efficiency. It is particularly suitable for large-scale high-dimensional gene expression data processing, and significantly improves the performance and efficiency of gene interaction relationship mining. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the method of the present invention. Specific implementation methods

[0036] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described below in conjunction with the accompanying drawings and embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways than those described herein. Therefore, the present invention is not limited to the specific embodiments of the following disclosure.

[0038] Example 1

[0039] like Figure 1This embodiment aims to achieve the possibility of more accurate detection of gene correlation. To this end, this embodiment provides a gene interaction relationship prediction method based on graph neural network, including the following steps:

[0040] First, experimental data is obtained from a public database. In this embodiment, all data related to renal cancer crawled from the GEO database are selected. These data are usually messy, so standardized preprocessing is required. The downloaded data may contain gene expression files from multiple different platforms. First, all gene expressions need to be converted into a unified identification form, such as Entrez Gene, Symbol, etc. If the platform file lacks information on the corresponding gene, it can be converted through other databases, but some data may be lost in the process. For example, when converting a Symbol to a unified ID, some Symbols may not be able to find corresponding items in the ID database, and this loss is difficult to avoid in complex data sets.

[0041] Next, review the experimental description file of the data to ensure that you understand the experimental design of the data, including the experimental subjects, experimental variables, experimental groups, and control groups. Group the data according to the experimental variables. For example, the data can be divided into five categories: cancer, metastasis, normal, primary, and tumor. Each category contains multiple sample group files. After further processing, each file will form a graph data.

[0042] For each sample group file, the similarity between gene expression data is calculated, and gene expression is used as the node feature. To this end, two similarity calculation methods are selected: Pearson correlation coefficient is used for normal distribution data, and Spearman correlation coefficient is used for non-normal distribution data. The similarity threshold is set according to the calculation results to filter out significantly correlated gene pairs. For example, the threshold is set to 0.3 to filter out gene pairs with a similarity greater than the threshold.

[0043] After obtaining the gene-gene similarity data for each sample group, statistical analysis was started. In the cancer category, there are 30 sample group files. The number of gene pairs in each sample group was counted, and the gene pairs with more than 70% of the occurrences were retained. Finally, 2,772,471 gene pairs were obtained, forming a preliminary graph structure. This method can effectively deal with noise and incompleteness in the data. For sample groups that do not contain certain gene pairs, they may not be detected due to noise in the data or similarity not reaching the threshold. To solve this problem, zero padding is performed on the sample groups that do not contain these genes.

[0044] After obtaining consistent gene pairs for all sample groups in each category (such as cancer), the genes were uniformly screened and arranged in a fixed order to ensure that each sample group contained the same 16,841 genes. Then, the IDs in the gene pairs were replaced with the gene numbers in the sample group, ultimately forming gene expression data and corresponding graph structure data in the same order.

[0045] Since the number of samples in each sample group is inconsistent, it is necessary to unify the dimensions of the gene features. Principal component analysis (PCA) was used for feature extraction, the expression data was compressed into three principal components, and the principal component contribution of all sample groups was verified to be greater than 95%. In this way, 30 graph data with consistent dimensions were obtained, the feature matrix size of which was 16841×3, and the adjacency matrix was represented by a sparse matrix.

[0046] In terms of model selection, the Graph Attention Network (GAT) model was selected. Since the data is unlabeled, a suitable unsupervised loss function is selected to train the model. The similarity between graph data is used as the loss function, but due to the large number of nodes in the graph, directly calculating the similarity will result in too high a computational complexity. For this reason, ResNet and pooling operations are introduced in the loss function to simplify the calculation and reduce the dimension, so as to perform similarity calculation.

[0047] After the model training reaches loss convergence, the edge prediction operation is further performed. This process is divided into two parts: edge deletion and edge addition. The purpose is to make the distance of graph data closer by optimizing the edge structure. Specifically, the maximum threshold is first screened according to the threshold, the edge deletion operation is performed, and then the edge addition operation is performed. It is worth noting that the threshold here is in percentage form, because the weight a ij The absolute value of is difficult to set, and the percentage method is easier to operate. As independent modules, the deletion and addition can be used repeatedly according to the convergence of the model until the loss cannot be further reduced. Finally, the optimized gene interaction relationship diagram is obtained.

[0048] Example 2

[0049] This embodiment relates to a gene interaction relationship prediction device based on a graph neural network, including a memory and one or more processors. The memory stores executable code. When the processor executes the code, it is used to implement the gene interaction relationship prediction method based on a graph neural network of Example 1.

[0050] Example 3

[0051] This embodiment relates to a computer-readable storage medium on which a program is stored. When a processor executes the program, the method for predicting gene interaction relationships based on graph neural networks of Example 1 can be implemented.

[0052] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A gene interaction relationship prediction method based on graph neural network, characterized in that: The method comprises the following steps: a. Preprocess the transcriptional expression data of each experimental group, including standardizing the gene ID, and representing the gene transcriptional expression data of different experimental groups with a unified gene ID to ensure the consistency of data format; then, group the cell samples in the experimental group, and one sample group contains multiple samples, and finally obtain the raw data for subsequent analysis; b. For the above raw data, calculate the gene similarity for each sample group to obtain the similarity data between genes in the sample group; by counting and screening the data of all sample groups, select gene pairs that meet the set conditions as the basis for constructing the graph data structure to ensure that the screened gene pairs can reflect the actual biological significance; c. Based on the graph structure obtained by the above screening, the original data is further processed to ensure that all sample groups contain only the genes defined in the graph structure; then, the data of these sample groups are standardized to make the data dimensions of all sample groups consistent; through this standardization process, multiple graph data groups with the same structure but different node characteristics (i.e., gene expression characteristics) are formed for subsequent graph neural network model training; d. Perform model training on the graph data set generated above. The model used is the graph attention network GAT and a simple edge prediction algorithm. Through multiple iterative training of the model, the loss function is gradually optimized, and finally the edge weight data in the graph structure is obtained, which can accurately reflect the interaction relationship between genes.

2. The method for predicting gene interaction relationships based on graph neural network according to claim 1, characterized in that: In step a, the preprocessing process of transcription expression data specifically includes the following steps: a1. Since different experimental groups use different sequencing technologies, the formats of gene expression data are inconsistent. Therefore, the data of all experimental groups need to be uniformly converted into the same gene ID or other consistent standard data formats to ensure the comparability of subsequent analysis. a2. According to different experimental condition variables, each set of experimental data is divided into sample groups; each experimental data is divided into multiple sample groups, and the condition variable is the cell type; a3. Arrange the grouped data to form a set of sample groups corresponding to each cell type, ensuring that the gene expression data in each sample group can reflect the biological characteristics of the type.

3. The method for predicting gene interaction relationships based on graph neural network according to claim 1, characterized in that: The specific steps of similarity calculation and data screening in step b are as follows: b1. For each sample group under each cell category, the similarity between the gene pairs in the group is calculated in turn; the data of each sample is regarded as the characteristics of the gene, and the similarity calculation method is selected according to the characteristics of the data to calculate and obtain the similarity data between genes in the sample group; b2. For each sample group, perform a preliminary screening of the similarity results according to a pre-set threshold, and retain gene pairs with similarity greater than or equal to the threshold. The selection of the threshold is very critical and should be tested and adjusted to ensure that the number of gene pairs screened is moderate, without losing key information or being excessively redundant. b3. Perform statistical screening operations on the sample groups of each cell category, count the gene pairs of all sample groups under the same category, and select the gene pairs whose occurrence frequency is greater than the set threshold. The choice of threshold will also have an important impact on the formation of the graph structure. The gene pairs that are finally retained will constitute the edges in the graph data, which is the core of the graph structure.

4. The method for predicting gene interaction relationships based on graph neural network according to claim 1, characterized in that: The specific operations of standardization processing and graph data generation in step c are as follows: c1. When screening gene pairs, some sample groups may not contain the screened genes, so these genes need to be supplemented. Use the trained model to complete the data, or directly use 0 values ​​to fill in, to ensure that each sample group contains the required gene pairs. Through such screening and supplementation, a gene data set containing all graph structure nodes is finally obtained. c2. Standardize the gene expression data of each sample group to unify its dimensions; Standardization methods may include selecting statistical features such as median, mean, and variance, or using principal component analysis (PCA) to extract principal components and calculate their contribution rates to determine the rationality of principal component selection; standardization can ensure data consistency and comparability; c3. Arrange the generated graph structure and the standardized data, and finally form an adjacency matrix represented by a sparse matrix and multiple feature matrix data groups; This data set is used for subsequent model training and ensures the consistency of data for each sample group during the training process.

5. The method for predicting gene interaction relationships based on graph neural network according to claim 1, characterized in that: The specific steps of model training and edge prediction in step d are as follows: d1. Use the modified graph attention network GAT model for training; since the data has no labels, an unsupervised method is used for training; define the loss function of GAT, calculate the graph data loss between each sample group, and obtain the total loss value to determine the direction of gradient descent; in order to simplify the amount of data calculation, first perform ResNet and pooling operations on the graph data, and then calculate the loss between the graph data; according to the complexity of the experimental data, select the appropriate number of convolutional layers and feature outputs, adjust the learning rate and gradient descent parameters, and ensure that the model can effectively converge in unsupervised learning; d2. During the model training process, the attention mechanism in the GAT model is used to calculate the attention weight αij of each gene pair, and the sparse adjacency matrix is ​​screened to delete the edges whose attention weight is lower than the set threshold. The selection of the threshold is determined through multiple experiments, and the minimization of the loss value is used as the judgment criterion. d3. For the deleted edges, based on the feature matrix output by the last convolution layer of the model, calculate the similarity of the gene pairs, and select the edges that are greater than the set threshold and not included in the original adjacency matrix for addition; Retrain the model, adjust the threshold, and continue to optimize the loss value; d4. The edge prediction process can be iterated multiple times until the loss value no longer decreases, and the final output is the gene relationship data for a certain cell category.

6. A gene interaction relationship prediction device based on graph neural network, comprising a memory and one or more processors, wherein the memory stores executable code, and when the processor executes the code, it is used to implement the method described in any one of claims 1 to 5.

7. A computer-readable storage medium having a program stored thereon, which can implement the method according to any one of claims 1 to 5 when a processor executes the program.

Images