A method, system, device and medium for predicting transcription factor target gene relationships

Abstract

The application provides a transcription factor target gene relationship prediction method, system, device and medium, and the method comprises the following steps: obtaining DNA sequences of transcription factors and DNA sequences of target genes of a to-be-predicted species and encoding to obtain encoded sequences; obtaining protein and genetic interaction network data and transcription factor target gene interaction network data of the to-be-predicted species; constructing a graph neural network model based on the encoded sequences, the protein and genetic interaction network data and the transcription factor target gene interaction network data, and performing model training to obtain a trained graph neural network model; and predicting unknown transcription factor target gene interaction relationships of the to-be-predicted species by using the trained graph neural network model. The application can avoid the problems of a large number of missing values of gene expression data and inaccurate binding site prediction; meanwhile, large-scale transcription factor-target gene potential relationship prediction can be realized according to the topological structure of an existing gene regulation network.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bioinformatics, and in particular to a transcription factor target gene relationship prediction method, system, device and medium based on a graph neural network. BACKGROUND

[0002] In the life process, organisms transcribe and translate genetic information through gene expression to exhibit corresponding traits. The expression process of genes is regulated by transcription factors. Gene regulation is a complex biological process. Under different time and space conditions, a single transcription factor can activate or inhibit the transcription of a large number of downstream target genes, and a target gene can be regulated by different transcription factors. The study of transcription factor target genes is of great significance to understanding the molecular mechanisms of organisms, the development process of life systems and disease diagnosis and treatment. In the study of genomics, there are many experimental and computational methods to predict target genes regulated by transcription factors, but the current methods are limited by data availability, lack of large-scale data, and large number of genes, so new data and methods are needed to realize target gene prediction.

[0003] The existing computational methods for predicting the relationship between transcription factors and target genes mainly include methods based on gene expression levels and methods for indirectly predicting transcription factor binding sites. Transcription factors can regulate the up-regulation and down-regulation of target gene expression, causing an increase or decrease in the protein produced by transcription and translation. Therefore, the correlation between transcription factors and target gene expression levels can be used to predict the regulatory relationship. Gene expression data can be divided into steady-state gene expression data and time series gene expression data. Methods based on information theory, methods based on Bayesian modeling and methods based on Boolean networks are widely used to predict target genes regulated by transcription factors using steady-state gene expression data. Methods based on linear or nonlinear ordinary differential equations and machine learning can use time series gene expression data to predict target genes regulated by transcription factors. The prediction of transcription factor target gene relationships can be indirectly inferred from their physical binding sites on the genome. Existing methods for predicting transcription factor binding sites mainly include machine learning and deep learning methods.

[0004] The document Reconstruction of gene regulatory networks from gene expression data using decoupled recurrent neural network model discloses inferring gene regulatory networks from gene expression data based on recurrent neural network model, which uses differential evolution algorithm to infer the regulatory factors of each gene, and uses experimental validation to analyze real gene expression profiles. In the method based on expression level prediction, the biggest challenge is that there are a large number of missing values in the expression data, which will cause too much information loss, thereby greatly reducing the accuracy of predicting regulatory relationships. The document Using Bayesian networks to construct gene regulatory networks from microarray data proposes a Bayesian model for constructing gene regulatory relationship networks from Saccharomyces cerevisiae and Escherichia coli gene expression data sets based on the ability of Bayesian networks to handle missing values. Differential equations can be used to model time-series gene expression data to predict target genes that may be regulated by transcription factors, and stochastic differential equations can simulate dynamic diffusion processes caused by irregular Brownian motion. The document A stochastic differential equation model for quantifying transcriptional regulatory network in Saccharomyces cerevisiae infers the transcriptional regulatory network of Saccharomyces cerevisiae based on stochastic differential equations, which uses a set of assumed regulatory factors to fit a generalized linear model to estimate the binding model of a specific target gene and a regulatory factor.

[0005] In addition to gene expression data, the correlation between histone modification and chromatin accessibility data across cell types can also be used to detect regulatory relationships between binding sites and target genes. In the method of indirectly predicting transcription factor binding sites, the document predicting transcription factor regulators and gene regulatory networks in Arabidopsis using publicly available binding site information predicts potential regulatory factors for a set of co-expressed or functionally related genes using publicly available transcription factor binding site information, which defines three types of gene regions related to gene translation start and end sites, calculates possible binding sites in these regions using probability mass function and hypergeometric distribution, and reconstructs the gene regulatory network of Arabidopsis thaliana according to the prediction results.

[0006] One of the main shortcomings of the gene expression-based prediction method is that the expression data is relatively sparse and has many missing values, which requires the use of some missing value imputation methods to improve the prediction accuracy. The method based on transcription factor binding sites indirectly uses the information of the binding sites. Since transcription factor binding sites are widely distributed on the genome, different transcription factors generally have different binding positions. Some transcription factors tend to bind to the promoter region close to the transcription start site of the target gene, and some transcription factors may bind to the enhancer region far from the transcription start site. Therefore, using indirect prediction of transcription start sites to infer the regulatory relationship of transcription factors may produce many false positives, and accurate prediction cannot be made. For the existing mathematical models and machine learning models, as the number of transcription factor-target gene relationships increases, the complexity of the model will increase exponentially, making it difficult to complete the whole genome regulatory relationship prediction task. SUMMARY

[0007] To solve the above problems of the prior art, the present application provides a transcription factor-target gene relationship prediction method, system, device and medium. The present application uses gene sequence data and known genetic and protein interaction network data, which can avoid the problems of many missing values in gene expression data and inaccurate binding site prediction. The graph neural network method is used to solve the link prediction problem in biological networks, which can well realize large-scale transcription factor-target gene relationship prediction.

[0008] The application is implemented by the following technical solutions:

[0009] A transcription factor-target gene relationship prediction method, comprising:

[0010] S1, obtaining the DNA sequence of the coding transcription factor and the DNA sequence of the target gene of the species to be predicted, and encoding the DNA sequence of the coding transcription factor and the DNA sequence of the target gene to obtain the encoded sequence; wherein the interaction relationship between the transcription factor and the target gene is known;

[0011] S2, obtaining the protein and genetic interaction network data of the species to be predicted; wherein the protein and genetic interaction network data correspond to the transcription factor and target gene in S1;

[0012] S3, obtaining the transcription factor-target gene interaction network data of the species to be predicted, wherein the transcription factor-target gene interaction network data is the interaction network data of the transcription factor and target gene in S1;

[0013] S4, constructing a graph neural network model based on the coding sequence obtained in S1, the protein and genetic interaction network data obtained in S2, and the transcription factor target gene interaction network data obtained in S3, and performing model training to obtain a trained graph neural network model;

[0014] S5, predicting the transcription factor target gene interaction relationship of the to-be-predicted species using the trained graph neural network model.

[0015] Preferably, in S1, the DNA sequences of the coding transcription factors and the target genes of the to-be-predicted species are obtained from the NCBI database.

[0016] Preferably, in S1, the DNA sequences of the coding transcription factors and the target genes are encoded using a k-mers encoding method.

[0017] Preferably, in S2, the protein and genetic interaction network data of the to-be-predicted species are obtained from the BioGRID database.

[0018] Preferably, S4 is specifically:

[0019] constructing a regulatory network graph of the graph neural network model based on the transcription factor target gene interaction network data; splicing the protein and genetic interaction network data and the coding sequence, extracting the features of the transcription factors and the target genes using a fully connected layer, performing feature fusion, using the obtained fusion features as the features of the gene nodes and the transcription factor nodes in the regulatory network graph, and constructing the graph neural network model; and performing model training on the graph neural network model.

[0020] Preferably, in S4, the graph neural network model is a graph neural network model based on personalized PageRank.

[0021] Preferably, S5 is specifically:

[0022] S51, obtaining the DNA sequences of the coding transcription factors and the DNA sequences of the candidate target genes of the to-be-predicted species, and encoding the DNA sequences of the coding transcription factors and the DNA sequences of the candidate target genes to obtain coding sequences;

[0023] S52, obtaining the protein and genetic interaction network data of the to-be-predicted species; wherein the protein and genetic interaction network data correspond to the transcription factors and the candidate target genes in S51;

[0024] S53, splicing the protein and genetic interaction network data in S52 and the coding sequences in S51, extracting the features of the transcription factors and the candidate target genes, performing feature fusion, and inputting the obtained fusion features into a graph neural network model to predict the transcription factor target gene interaction relationship.

[0025] A transcription factor target gene relationship prediction system, comprising a data acquisition module, a model construction and training module, and a prediction module;

[0026] The data acquisition module is configured to acquire DNA sequences of transcription factors, DNA sequences of target genes, protein and genetic interaction network data, and transcription factor target gene interaction network data of a to-be-predicted species;

[0027] The data processing module is configured to encode the DNA sequences of the transcription factors and the DNA sequences of the target genes to obtain encoded sequences;

[0028] The model construction and training module is configured to construct a graph neural network model based on the encoded sequences, the protein and genetic interaction network data, and the transcription factor target gene interaction network data, and to train the model to obtain a trained graph neural network model;

[0029] The prediction module is configured to predict the transcription factor target gene interaction relationship of the to-be-predicted species by using the trained graph neural network model.

[0030] Preferably, the model construction and training module comprises a model construction module and a training module.

[0031] The model construction module is configured to construct a regulatory network graph of the graph neural network model based on the transcription factor target gene interaction network data, splice the protein and genetic interaction network data and the encoded sequences, extract features of the transcription factors and the target genes by using a fully connected layer, perform feature fusion, and use the obtained fused features as features of gene nodes and transcription factor nodes in the regulatory network graph to construct the graph neural network model.

[0032] The training module is configured to train the constructed graph neural network model.

[0033] A computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the transcription factor target gene relationship prediction method when executing the computer program.

[0034] A computer-readable storage medium stores a computer program, and the computer program implements the steps of the transcription factor target gene relationship prediction method when executed by a processor.

[0035] Compared with the prior art, the present application has the following advantages:

[0036] The data adopted by the present application is gene sequence data and known genetic and protein interaction network data, and the data can be directly downloaded from a public database without biological experiments, so that the expression data and expression profile information of the gene can be obtained without a large time cost and experimental cost, and the problems of a large number of missing values of gene expression data and inaccurate prediction of binding sites can be avoided. Meanwhile, the present application uses prior transcription factor-target gene regulation relationship to try to find the binding mode of the transcription factor when regulating the target gene, and the graph neural network method in deep learning can realize large-scale transcription factor-target gene potential relationship prediction according to the topological structure of the existing gene regulation network, and has a high accuracy.

[0037] Further, the present application applies the graph neural network model based on personalized propagation to solve the biological regulation network prediction problem, and the method can solve the problem that too many layers of the graph neural network tend to be smooth. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 A schematic diagram for coding 3-mer of DNA sequence;

[0039] Figure 2 A schematic diagram for a transcription factor-target gene interaction prediction method of the present application;

[0040] Figure 3 A flowchart for a transcription factor-target gene interaction prediction method of the present application;

[0041] Figure 4 A comparison result of prediction effects of different graph neural network models;

[0042] Figure 5 A comparison result of sizes of AUROC of model prediction when different k values are used;

[0043] Figure 6 A comparison result of sizes of AUPR of model prediction when different k values are used. DETAILED DESCRIPTION

[0044] In order to further understand the present application, the present application is described below in combination with examples, and these descriptions are only used to further explain the features and advantages of the present application, and are not used to limit the claims of the present application.

[0045] Part of the interaction relationship between transcription factors and target genes is known, and the purpose of the present application is to predict potential interaction relationships based on these known interactions. The present application aims to use data from public databases such as gene sequences to explore regulatory patterns in expression processes, thereby predicting interactions between transcription factors and genes. The present application uses a graph neural network method to solve the link prediction problem in biological networks, which can well achieve large-scale transcription factor target gene relationship prediction.

[0046] The present application uses a deep learning method to predict transcription factor target gene relationships, and the main research contents include:

[0047] (1) DNA sequences cannot be directly used as input to graph neural network models, and the present application uses k-mers encoding method to encode gene sequences composed of "A", "C", "G", "T", extract context information and implicit semantic information of gene sequences, so that the graph neural network model can capture the potential transcription factor and target gene binding mode information in the gene sequence.

[0048] (2) Introduce genetic and protein interaction network data, fuse it with the encoded gene sequence together as input features to build a graph neural network model. Genetic and protein interaction networks represent the functional correlation between genes and transcription factors, combining the functional features and sequence features of the gene to determine whether the gene is a target gene or a non-target gene of the transcription factor.

[0049] (3) The present application uses a graph neural network model based on personalized propagation to train known transcription factor target gene interaction network data, to explore the correlation between graph nodes, and to accurately predict potential target genes regulated by transcription factors according to known transcription factor target gene interaction network data.

[0050] As shown in Figure 2 and Figure 3 , the specific implementation process of the present application is as follows:

[0051] S1, first obtain the DNA sequence encoding the transcription factor and the DNA sequence of the target gene on the chromosome of the species to be predicted from the NCBI (National Center for Biotechnology Information) database, and use k-mers encoding method to embed the DNA sequence encoding the transcription factor and the DNA sequence of the target gene. Represented, get the encoding sequence. Among them, the interaction relationship between the transcription factor and the target gene is known.

[0052] In recent years, the use of DNA sequences as inputs to deep learning models has increased, indicating that DNA sequences contain information about the mechanisms of biological molecular interactions. One of the common representations of long DNA sequences is to decompose them into shorter k-mer components, and in many applications, short sequences k-mer are considered to be relevant. DNA2vec (Consistent vector representations of variable-length k-mers) is a method based on the word2vec word vector model (Efficient estimation of word representations in vector space) proposed in 2017, which uses the human genome sequence as a learning corpus to embed each k-mer into a 100-dimensional continuous vector. Therefore, the pre-trained k-mer vector in DNA2vec can be used to encode gene sequences. According to the k-mer assembly DNA sequence, a 3000bp sequence can be embedded in a (3000-k+1, 100) dimensional vector space, for example, with 3-mer, DNA2vec encodes gene sequences as shown in Figure 1

[0053] Figure 1 is a schematic diagram of encoding DNA sequences using k-mers (k is equal to 3 in the figure), first, an entire gene sequence is divided into subsequences composed of 3 bases using a sliding window, and then each subsequence is mapped to the corresponding vector according to the DNA2vec pre-trained vector.

[0054] S2, then, download the protein and genetic interaction network data of the predicted species from the BioGRID (Biological General Repository for Interaction Datasets) database. The protein and genetic interaction network data correspond to the transcription factors and target genes in S1.

[0055] Transcription factors with similar regulatory relationships and target genes co-regulated by the same transcription factor often have functional similarities. The protein and genetic interaction network data between genes contain relevant information about gene and protein functions, so the present application believes that protein and genetic interactions have a relatively important influence on the prediction of potential target genes regulated by transcription factors.

[0056] ​S3, obtaining transcription factor-target gene interaction network data of the to-be-predicted species from the TFlink and Yeastract databases, wherein the transcription factor-target gene interaction network data is the interaction network data of the transcription factor and the target gene in S1. Part of the transcription factor-target gene interaction network data is used as a training set to perform the step of S4, and part of the transcription factor-target gene interaction network data is used as a test set to perform the step of S5.

[0057] S4, constructing a regulatory network graph of a graph neural network model based on the transcription factor-target gene interaction network data; splicing the protein and genetic interaction network data and the coding sequence, using a fully connected layer to extract main features of the transcription factor and the target gene, performing feature fusion, taking the obtained fused features as features of gene nodes and transcription factor nodes in the regulatory network graph, and constructing the graph neural network model. The graph neural network model is trained.

[0058] The transcription factor-target gene interaction network can be represented as a graph structure, and the prediction of whether the transcription factor and the target gene have a regulatory relationship is defined as a link prediction problem. The present application preferably uses a graph neural network algorithm based on personalized propagation to perform network propagation between gene nodes and transcription factor nodes, update the features of each node based on all neighbor nodes, and obtain abstract features containing key information.

[0059] The graph neural network model based on personalized propagation combines PageRank and graph neural network (GCN), refers to the node propagation mode of PageRank, and forms an improved neighbor feature propagation mode based on personalized PageRank. The model generates a prediction according to the features of each node, then propagates them through a completely personalized PageRank scheme, and iteratively updates the features of the target node according to the features of the neighbor nodes and the features of the target node to generate a final prediction.

[0060] The final obtained features can be used to predict the interaction relationship between a pair of transcription factor-target genes using a bilinear decoder.

[0061] S5, evaluating the prediction effect of the trained graph neural network model on the test set.

[0062] S6, using the trained graph neural network model to identify potential transcription factor-target gene interaction relationships of the to-be-predicted species.

[0063] DNA sequences of the coding transcription factors and DNA sequences of the candidate target genes of the to-be-predicted species are obtained from the NCBI database, and the DNA sequences of the coding transcription factors and the DNA sequences of the candidate target genes are embedded and represented in a k-mers coding manner. The interaction relationship between the transcription factor and the candidate target gene is unknown.

[0064] The protein and genetic interaction network data of the predicted species are downloaded from the BioGRID database, wherein the protein and genetic interaction network data correspond to the transcription factor and the candidate target gene in the step.

[0065] The protein and genetic interaction network data are spliced with the embedding representation of the DNA sequence in the step, the main features of the transcription factor and the candidate target gene are extracted, feature fusion is performed, and then the graph neural network model is input to predict the transcription factor-target gene interaction relationship.

[0066] Figure 2 The complete flowchart of the present application is shown in the figure, module A represents the conversion of the DNA sequence of the transcription factor and the target gene and the genetic and protein interaction network data into numerical information and splicing; module B is feature extraction and feature fusion of the spliced results using a fully connected layer. Module C represents constructing a regulatory network graph based on known transcription factor-target gene interaction relationships; module D represents using the extracted features as the features of the nodes in the regulatory network graph, and using a personalized propagation-based graph neural network to perform information transmission and update node information. Module E is to use a bilinear decoder to predict the regulatory score between each pair of transcription factor-target gene, and use a cross-entropy loss function to train the graph neural network model.

[0067] In the present application, the predicted species can be Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens, Mus musculus, Rattus norvegicus and Saccharomyces cerevisiae.

[0068] The method of the present application has been verified by the following experiments:

[0069] The present application obtains the transcription factor-target gene regulatory relationship of Caenorhabditis elegans, Drosophila melanogaster, Homo sapiens, Mus musculus, Rattus norvegicus and Saccharomyces cerevisiae from the TFlink and Yeastract databases, and experiments are performed on the data of the six different species.

[0070] In the graph neural network model constructed in the present application, a personalized propagation-based graph neural network algorithm is used, which can solve the problem that too many layers of graph neural networks tend to be smooth. In order to demonstrate the effectiveness of the model, the model is compared with other three classical graph neural network models, and the experimental results are shown in Figure 4 Figure 4 ​AUROC and AUPR predicted by different methods on 6 species datasets are compared, and it can be seen from the figure that, compared with other models, the graph neural network model based on personalized PageRank of the application achieves the best performance in AUC and AUPR prediction in 6 datasets.

[0071] In the k-mer encoding mode, different k-mer encoding modes have different short sequences and different semantic information learned by the model, and the application compares the effects of the graph neural network model when different k-mer encoding DNA sequences are used in the data of 6 species, and the values of AUROC and AUPR predicted by the model are as shown in Figure 5 、 Figure 6 As can be seen from the figure, generally, the larger the k, that is, the longer the length of the subsequence, the better the meaning of the short sequence can be represented, and the higher the accuracy of the graph neural network model, and the better the prediction effect.

[0072] In another embodiment of the application, a transcription factor target gene relationship prediction system is provided, which can be used to implement the transcription factor target gene relationship prediction method described above, and specifically, the transcription factor target gene relationship prediction system comprises: a data acquisition module, a model construction and training module, and a prediction module.

[0073] The data acquisition module is used to acquire DNA sequences of transcription factors to be predicted, DNA sequences of target genes, protein and genetic interaction network data, and transcription factor target gene interaction network data of a to-be-predicted species.

[0074] The data processing module is used to encode the DNA sequences of the transcription factors and the DNA sequences of the target genes to obtain encoded sequences.

[0075] The model construction and training module is used to construct a graph neural network model based on the encoded sequences, the protein and genetic interaction network data, and the transcription factor target gene interaction network data, and to train the model to obtain a trained graph neural network model.

[0076] The prediction module is used to predict the transcription factor target gene interaction relationship of the to-be-predicted species by using the trained graph neural network model.

[0077] The model construction and training module comprises a model construction module and a training module.

[0078] The model construction module is configured to construct a regulatory network graph of a graph neural network model based on transcription factor target gene interaction network data; splice protein and genetic interaction network data with coding sequences, and use a fully connected layer to extract features of transcription factors and target genes, perform feature fusion, and use the obtained fused features as features of gene nodes and transcription factor nodes in the regulatory network graph to construct the graph neural network model.

[0079] The training module is configured to train the constructed graph neural network model.

[0080] In another embodiment of the present application, a computer device is provided, which includes a processor and a memory, the memory is configured to store a computer program, the computer program includes program instructions, and the processor is configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method process or a corresponding function. The processor in the embodiments of the present application can be used for the operation of the transcription factor target gene relationship prediction method.

[0081] In still another embodiment, the present application provides a storage medium, specifically a computer readable storage medium (Memory), which is a memory device in a computer system, for storing programs and data. It should be understood that the computer readable storage medium here can include both built-in storage medium in the computer system, and also can include the extended storage medium supported by the computer system. The computer readable storage medium provides a storage space, which stores an operating system of the terminal. In addition, one or more instructions adapted to be loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium here can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. The one or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the transcription factor target gene relationship prediction method in the above embodiments.

[0082] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can 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-ROMs, optical storage media, etc.) containing computer-usable program code.

[0083] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams 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 apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus for performing the functions specified in one or more flows and / or blocks.

[0084] These computer program instructions can also be stored in a computer readable memory capable of directing the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocksFigure 1 the function specified in the one or more blocks.

[0085] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide a process for implementing the flow Figure 1 the flow or flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.

[0086] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, but not to limit it, although the present application has been described in detail with reference to the above examples, those skilled in the art should understand: the specific embodiments of the present application can still be modified or replaced by the equivalent, without departing from the spirit and scope of the present application, any modification or equivalent replacement, which should be covered within the protection scope of the claims of the present application.

Claims

1. A method for predicting a relationship between a transcription factor and a target gene, characterized by, The method comprises the following steps: S1, obtaining the DNA sequences of the transcription factor and the target gene of the predicted species, and encoding the DNA sequences of the transcription factor and the target gene to obtain an encoded sequence; wherein the interaction relationship between the transcription factor and the target gene is known; S2, obtaining the protein and genetic interaction network data of the predicted species; wherein the protein and genetic interaction network data correspond to the transcription factor and the target gene in S1; S3, obtaining the transcription factor-target gene interaction network data of the predicted species, wherein the transcription factor-target gene interaction network data is the interaction network data of the transcription factor and the target gene in S1; S4, constructing a regulatory network graph of a graph neural network model based on the transcription factor-target gene interaction network data; splicing the protein and genetic interaction network data and the encoded sequence, and using a fully connected layer to extract the features of the transcription factor and the target gene, and performing feature fusion, wherein the obtained fusion features are used as the features of the gene nodes and the transcription factor nodes in the regulatory network graph, and a graph neural network model is constructed; the graph neural network model is trained to obtain a trained graph neural network model; the graph neural network model is a graph neural network model based on personalized PageRank; S5, predicting the unknown transcription factor-target gene interaction relationship of the predicted species by using the trained graph neural network model.

2. The method according to claim 1, wherein In S1, the DNA sequences of the transcription factor and the target gene of the predicted species are obtained from the NCBI database; and in S2, the protein and genetic interaction network data of the predicted species are obtained from the BioGRID database.

3. The method according to claim 1, wherein In S1, the DNA sequences of the transcription factor and the target gene are encoded by using a k-mers encoding method.

4. The method according to claim 1, wherein S5 specifically comprises: S51, obtaining the DNA sequences of the transcription factor and the candidate target gene of the predicted species, and encoding the DNA sequences of the transcription factor and the candidate target gene to obtain an encoded sequence; S52, obtaining the protein and genetic interaction network data of the predicted species; wherein the protein and genetic interaction network data correspond to the transcription factor and the candidate target gene in S51; S53, splicing the protein and genetic interaction network data in S52 and the encoded sequence in S51, extracting the features of the transcription factor and the candidate target gene, performing feature fusion, and inputting the obtained fusion features into a graph neural network model to predict the transcription factor-target gene interaction relationship.

5. A transcription factor target gene relationship prediction system, comprising: The method comprises the following steps: A data acquisition module, a model construction and training module, and a prediction module; The data acquisition module is used to obtain the DNA sequences of the transcription factor, the DNA sequences of the target gene, the protein and genetic interaction network data, and the transcription factor-target gene interaction network data of the predicted species; The data processing module is used to encode the DNA sequences of the transcription factor and the target gene to obtain an encoded sequence; The model construction and training module constructs a regulatory network graph of the graph neural network model based on the transcription factor target gene interaction network data; splices the protein and genetic interaction network data and the coding sequence, extracts the features of the transcription factor and the target gene using a fully connected layer, performs feature fusion, and uses the obtained fusion features as the features of the gene nodes and the transcription factor nodes in the regulatory network graph to construct the graph neural network model; the graph neural network model is trained to obtain the trained graph neural network model; the graph neural network model is a graph neural network model based on personalized PageRank; The prediction module is configured to predict the transcription factor target gene interaction relationship of an unknown species to be predicted using the trained graph neural network model.

6. The transcription factor target gene relationship prediction system according to claim 5, wherein The model construction and training module includes a model construction module and a training module. The model construction module is configured to construct a regulatory network graph of the graph neural network model based on the transcription factor target gene interaction network data; splice the protein and genetic interaction network data and the coding sequence, extract the features of the transcription factor and the target gene using a fully connected layer, perform feature fusion, and use the obtained fusion features as the features of the gene nodes and the transcription factor nodes in the regulatory network graph to construct the graph neural network model. The training module is configured to train the constructed graph neural network model.

7. A computer device, comprising: The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the transcription factor target gene relationship prediction method according to any one of claims 1 to 4.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the transcription factor target gene relationship prediction method according to any one of claims 1 to 4.

Images