Gene regulatory network inference system and method based on single-cell homeostasis and temporal sequence

Abstract

The present invention discloses a gene regulatory network inference system and method based on single-cell steady-state and time series. The present invention extracts the topological constraints of each target gene from the single-cell steady-state gene expression matrix, and on the basis of the topological constraints of each target gene, combines the single-cell steady-state gene expression data and the single-cell time series gene expression data to infer the gene regulatory relationship to obtain the gene regulatory network. Since the topological constraint extraction module in the present invention extracts the topological constraints of each target gene, the subsequent gene regulatory network inference is performed under the topological constraints, which improves the accuracy of the gene regulatory network inference and reduces the operating cost of the system. The present invention uses the Borda voting method to combine the regulatory relationships obtained from the single-cell steady-state gene expression data and the single-cell time series gene expression data to further improve the accuracy of the gene regulatory network inference system.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics, and more specifically, to a system and method for inferring gene regulatory networks based on single-cell homeostasis and temporal sequences in the field of bioinformatics data processing. This invention can be used to infer gene regulatory networks from gene expression data and to identify key regulatory factors in biological processes using the inferred networks. Background Art

[0002] Gene expression data, obtained by measuring gene expression levels under different experimental conditions, contains rich information about gene activity and can represent the current state of a cell's activity. Gene expression data can be used to infer network models of gene interactions, a process also known as reverse engineering. Single-cell RNA sequencing is the most popular single-cell sequencing technology, used to sequence and quantify RNA expression levels in each cell. It can capture the stochasticity and dynamics of gene expression while revealing transcriptome-wide intercellular variability at high resolution. Single-cell gene expression data is particularly promising for inferring gene regulatory networks. In recent years, numerous methods for inferring gene regulatory networks based on single-cell gene expression data have been proposed.

[0003] Northwestern Polytechnical University, in its patent application, "A Gene Regulatory Network Inference Method Based on a Linear Mixed Model" (patent application number CN 202211571759.6, application publication number CN 115831228 A), provides a gene regulatory network inference method based on a linear mixed model. This method takes single-cell gene expression data as input and represents the data using a mean, random effects, and noise. Both random effects and noise are matrix random variables that follow a matrix normal distribution. Parameter estimation then generates a row covariance matrix to represent the correlations between genes. This method has two shortcomings: first, it only considers steady-state data and fails to account for the temporal information inherent in the expression data. This results in incomplete single-cell gene expression data used for inference and fails to provide dynamic regulatory information about genes, affecting the accuracy of the inferred gene regulatory network. Second, this method, based solely on the data, somewhat ignores the inherent topological characteristics of the gene regulatory network, resulting in a lack of clear biological interpretation of its structure.

[0004] Nantong University provides a method for constructing a gene regulatory network based on scRNA-seq and dynamic time warping in its patent application "Gene regulatory network construction method based on scRNA-seq and dynamic time warping" (patent application number CN 201911012024.8, application publication number CN 110808083 A). The implementation steps of this method include the following: Step 1, single-cell pseudo-time analysis based on single-cell data; Step 2, dual DTW distance analysis based on positive and negative regulatory modes; Step 3, cluster analysis based on DTW distance and GRN module screening; Step 4, construction of an inter-gene similarity matrix based on DTW distance; Step 5, calculation of the gene-to-gene network connectivity threshold C through random permutation test; Step 6, gene functional analysis of the GRN module; Step 7, visualization of the pGRN network to obtain a GRN network diagram. This method has two shortcomings. First, it only considers the information contained in time-series data, ignoring the steady-state gene regulatory relationships contained in steady-state data. The single-cell gene expression data used before inference is incomplete and susceptible to noise, making the network stability impossible to guarantee. Second, although this method clusters the corresponding modules before inferring the gene regulatory network, it only uses simple similarity as the basis for determining whether genes within the module are related, making the resulting gene regulatory network less accurate and reliable. Summary of the Invention

[0005] The present invention aims to address the shortcomings of the above-mentioned prior art and provide a system and method for inferring gene regulatory networks based on single-cell steady-state and temporal data. This system is intended to address the issues of incomplete gene expression data during gene regulatory network inference and the low accuracy of gene regulatory network inference.

[0006] The idea of ​​achieving the purpose of the present invention is that the topological constraint extraction module of the present invention uses the WGCNA clustering method to extract the topological constraints of each target gene from the single-cell steady-state gene expression matrix, and performs subsequent gene regulatory network inference under the topological constraints. By extracting the topological constraints, the problem of the existing technology ignoring the inherent topological structure of the gene regulatory network is solved, so that the system of the present invention improves the accuracy of gene regulatory network inference. The inference problem is decomposed under the topological constraints, which can reduce the operating cost of the system. The present invention uses steady-state gene expression data and time-series gene expression data to infer the regulatory relationship between genes, and then combines the results inferred from the two data using the Borda voting method to comprehensively obtain the possibility ranking of the regulatory relationship between genes, thereby solving the problem that the existing technology only uses one type of data for network inference and can only obtain information about the gene regulatory network from a limited perspective.

[0007] The system of the present invention includes a topological constraint extraction module, a steady-state regulatory relationship inference module, a temporal regulatory relationship inference module, a regulatory relationship combination module, and a gene regulatory network generation module; wherein:

[0008] The topology constraint extraction module uses the WGCNA clustering method to extract the topology constraints of each target gene from the single-cell steady-state gene expression matrix;

[0009] The steady-state regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene, and composes the regulatory relationship matrix A of the single-cell steady-state gene expression data from the regulatory relationship vectors of all target genes;

[0010] The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of the target gene; the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data;

[0011] The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships between all genes;

[0012] The gene regulatory network generation module generates a gene regulatory network through parameter factors.

[0013] The steps of the gene regulatory network inference method based on single cell steady state and time sequence of the present invention include the following:

[0014] Step 1: Topological constraint extraction module, using the WGCNA clustering method, extracts the topological constraints of each target gene from the single-cell steady-state gene expression matrix;

[0015] Step 2: Infer the gene regulatory relationship matrix based on single-cell steady-state gene expression data:

[0016] The steady-state regulatory relationship inference module generates the regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene, and composes the regulatory relationship matrix A of the single-cell steady-state gene expression data from the regulatory relationship vectors of all target genes;

[0017] Step 3: Infer the gene regulatory relationship matrix based on single-cell temporal gene expression data:

[0018] The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene; the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data;

[0019] Step 4: The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships among all genes;

[0020] Step 5: The gene regulatory network generation module generates a gene regulatory network through parameter factors.

[0021] Compared with the prior art, the present invention has the following advantages:

[0022] First, the system of the present invention adopts a topological constraint extraction module to extract the topological constraints of each target gene and perform subsequent gene regulatory network inference under the topological constraints, overcoming the defect of the existing technology that ignores the inherent topological structure of the gene regulatory network. The system of the present invention improves the accuracy of gene regulatory network inference and reduces the operating cost of the system.

[0023] Second, the method of the present invention infers the regulatory relationship between genes based on single-cell steady-state gene expression data and single-cell temporal gene expression data, respectively, and uses the Borda voting method to combine the two regulatory relationships, overcoming the deficiency of the existing technology that only information about the gene regulatory network can be obtained from a limited perspective when using one type of data for network inference, so that the method of the present invention further improves the accuracy of gene regulatory network inference. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the system of the present invention;

[0025] Figure 2 It is a flow chart of the method of the present invention

[0026] Figure 3 It is the hierarchical clustering tree result diagram of the present invention. DETAILED DESCRIPTION

[0027] The embodiments and effects of the present invention are further described in detail below with reference to the accompanying drawings.

[0028] Reference Figure 1 , the system of the present invention is further described in detail.

[0029] The system of the present invention includes a topological constraint extraction module, a steady-state regulatory relationship inference module, a temporal regulatory relationship inference module, a regulatory relationship combination module, and a gene regulatory network generation module; wherein:

[0030] The topology constraint extraction module uses the WGCNA clustering method to extract the topology constraints of each target gene from the single-cell steady-state gene expression matrix.

[0031] The steady-state regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene, and composes the regulatory relationship vectors of all target genes into a regulatory relationship matrix A of single-cell steady-state gene expression data.

[0032] The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of the target gene; and the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data.

[0033] The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships among all genes.

[0034] The gene regulatory network generation module generates a gene regulatory network through parameter factors.

[0035] Reference Figure 2 , further describe the steps of the method of the present invention.

[0036] Step 1, the topological constraint extraction module, uses the WGCNA clustering method to extract the topological constraints of each target gene from the single-cell steady-state gene expression matrix.

[0037] The steps of the WGCNA clustering method are as follows:

[0038] In the first step, the steady-state gene expression matrix containing 100 genes was processed, the absolute value of the Pearson correlation between each element in the steady-state gene expression matrix was calculated, and the absolute values ​​of all correlations were combined into a symmetrical correlation matrix.

[0039] The second step is to convert the correlation matrix into the corresponding adjacency matrix.

[0040] The third step is to calculate the element values ​​in the topological overlap matrix through the adjacency matrix according to the following formula:

[0041]

[0042] Among them, TOM ij Represents the element value of the i-th row and j-th column in the topological overlap matrix, l pq Indicates the size of the interaction between the pth element and the qth element in the adjacency matrix and other elements, a pq represents the element value in the p-th row and q-th column of the adjacency matrix. The values ​​of p and q are equal to the corresponding values ​​of i and j. min(·) represents the minimum value operation. k p represents the sum of the values ​​of the elements in the pth row of the adjacency matrix, k q Represents the sum of the values ​​of the elements in the qth row of the adjacency matrix.

[0043] In the fourth step, the topological overlap matrix is ​​processed using a hierarchical clustering method. The distances between all classes are calculated based on the Pearson correlation coefficient. The classes with the closest distances are merged pairwise to obtain a hierarchical clustering tree. The threshold is set according to the closeness of the relationship between the classes to obtain the clustering results.

[0044] Reference Figure 3 , further describes the process of processing the topological overlap matrix using the hierarchical clustering method in an embodiment of the present invention.

[0045] Figure 3 Here is the corresponding hierarchical clustering dendrogram drawn by using AgglomerativeClustering and the dendrogram method available in scipy. Figure 3 (a) By continuously Figure 3 The operation described in (b) shows the completed hierarchical clustering dendrogram. Figure 3 (b) The horizontal axis represents the number of each gene, and the vertical axis represents the distance between the classes. The distance between the classes is calculated using the Pearson correlation coefficient to obtain the distance between all genes, and the classes with the closest distances are merged two by two. In the embodiment of the present invention, the distance between gene 1 and gene 3 is 0.41 at the closest. Gene 1 and gene 3 are connected by a line to form a new class. The distance between the class formed by gene 1 and gene 3 and the class formed by gene 15 alone is 0.57 at the closest. The class formed by gene 1 and gene 3 and gene 15 are connected by a line to form a new class. By analogy, a class containing all genes is finally obtained, and a multi-color hierarchical clustering tree diagram representing the close relationship between genes is obtained.

[0046] Depend on Figure 3 As can be seen in (a), the threshold horizontal line is drawn at point s to obtain the clustering result. The number of classes is the number of intersections with the threshold horizontal line and the vertical line. 22 30 1 24 1 1 37 1 46 15 48 1 27 1 23 10 3 13 47 36 28 36 48 1 27 1 23 10 3 13 47 36 28 40 1 22 30 1 24 1 1 37 1 46 15 48 1 27 1 23 10 3 13 47 36 28 50 1 40 1 41 1 42 100 40 1 22 30 1 24 1 1 37 1 46 15 48 1 27 1 23 10 3 13 47 36 28 3129 1 33 45 35]. The value of the first element of the clustering result is 1, which means that the first gene belongs to the set with class name 1.

[0047] In the fifth step, based on the clustering results, the homologous gene set and heterologous gene set for each target gene are obtained, and these homologous gene sets and heterologous gene sets are used as topological constraints for the target gene. In this embodiment, gene 1 belongs to the gene set named "1," and all genes in this set except gene 1 are homologous to gene 1. Excluding genes in the gene set named "1," the remaining genes are heterologous to gene 1.

[0048] Step 2: Infer the gene regulatory relationship matrix based on single-cell steady-state gene expression data.

[0049] The steady-state regulatory relationship inference module generates the regulatory relationship vector between each target gene and other genes under the topological constraints of the target gene, and composes the regulatory relationship vectors of all target genes into the regulatory relationship matrix A of the single-cell steady-state gene expression data.

[0050] The steps of generating the regulatory relationship vector between each target gene and other genes under the topological constraints of the target gene are as follows:

[0051] In the first step, under the topological constraints of the similar genes of each target gene, the ridge regression method is used to solve the dependence of the expression level of each target gene on the expression level of the similar genes of the target gene, and obtain the possibility vector of the similar genes regulating the target gene.

[0052] In the second step, under the topological constraints of each target gene heterologous gene, the Lasso regression method is used to solve the dependence of the expression level of each target gene on the expression level of the target gene heterologous gene, and obtain the possibility vector of the heterologous gene regulating the target gene.

[0053] In the third step, the possibility vector of target genes regulated by similar genes is concatenated with the possibility vector of target genes regulated by heterogeneous genes to obtain the gene regulation relationship vector of each target gene under topological constraints.

[0054] Step 3: Infer the gene regulatory relationship matrix based on single-cell temporal gene expression data.

[0055] The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene; the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data.

[0056] The steps of generating the regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene are as follows:

[0057] In the first step, under the topological constraints of the similar genes of each target gene, the ridge regression method is used to solve the dependence of the expression level of each target gene at the current time point on the cumulative expression levels of the similar genes of the target gene at the three time points before the current time point, and obtain the possibility vector of the similar genes regulating the target gene.

[0058] The cumulative expression level is obtained by the following formula:

[0059]

[0060] Among them, f accum (·) represents the cumulative function, Z1 represents the first data value participating in the cumulative effect, and Z r represents the rth data value participating in the cumulative effect, δ represents the attenuation factor, ∑ represents the summation operation, Z w Represents the w-th data value, m-1 is the exponent of δ, and the values ​​of m and w are equal.

[0061] In the second step, under the topological constraints of each target gene heterologous gene, the Lasso regression method is used to solve the dependency of the expression level of each target gene at the current time point on the cumulative expression level of the target gene heterologous gene at the three time points before the current time point, and the possibility vector of the heterologous gene regulating the target gene is obtained.

[0062] In the third step, the possibility vectors of target genes regulated by similar genes and the possibility vectors of target genes regulated by heterogeneous genes are combined to obtain the gene regulation relationship vector of each target gene under topological constraints.

[0063] Step 4: The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships between all genes.

[0064] The steps of obtaining the probability ranking of regulatory relationships among all genes are as follows:

[0065] In the first step, the smallest element value in the control relationship matrix A and B is set to 1, the second smallest element value is set to 2 points, and so on, to obtain two score matrices.

[0066] In the second step, the corresponding element values ​​in the two score matrices are added together to obtain a regulatory relationship possibility matrix M.

[0067]

[0068] Among them, M ij is the element in the i-th row and j-th column of the regulation relationship possibility matrix M.

[0069] In the third step, all elements in the regulatory relationship possibility matrix M are divided into the following order: the gene corresponding to the row is the target gene, the gene corresponding to the column is the regulatory gene, and the element value at the corresponding position is the regulatory possibility of the regulatory gene regulating the target gene, as shown in Table 1:

[0070] Table 1 Regulation possibility between genes

[0071] target genes Regulatory genes Possibility of regulation Gene 1 Gene 1 <![CDATA[M 11 ]]> Gene 1 Gene 2 <![CDATA[M 12 ]]> …… …… …… Gene 100 Gene 100 <![CDATA[M 100 100 ]]>

[0072] The fourth step is to sort the table from large to small according to the possibility of regulation to obtain the possibility ranking of the relationship between genes.

[0073] Step 5: Generate gene regulatory network through parameter factors.

[0074] The gene regulatory network generation module sets the number of regulatory edges in the real network to the parameter factor value and selects the top k gene pairs ranked by gene regulation likelihood, where the value of k is equal to the parameter factor value. In this embodiment of the present invention, there are 651 real regulatory edges, and the parameter factor is set to 651. Edges are set between the target gene and the regulatory gene in each of the selected gene pairs, and all gene pairs after the edges are set form the gene regulatory network.

Claims

1. A gene regulatory network inference system based on single-cell steady-state and temporal sequence, comprising a topological constraint extraction module, a regulatory relationship combination module, and a gene regulatory network generation module, characterized in that: It also includes a steady-state regulatory relationship inference module and a temporal regulatory relationship inference module; wherein: The topology constraint extraction module uses the WGCNA clustering method to extract the topology constraints of each target gene from the single-cell steady-state gene expression matrix; The steady-state regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene, and composes the regulatory relationship matrix A of the single-cell steady-state gene expression data from the regulatory relationship vectors of all target genes; The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of the target gene; the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data; The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships between all genes; The gene regulatory network generation module generates a gene regulatory network through parameter factors.

2. A gene regulatory network inference method based on single-cell steady-state and temporal sequence according to the network inference system of claim 1, characterized in that: Based on the extraction of topological constraints for each target gene, the gene regulatory relationships are inferred by combining single-cell steady-state gene expression data and single-cell temporal gene expression data to obtain a gene regulatory network. The steps of this inference method include the following: Step 1: Topological constraint extraction module, using the WGCNA clustering method, extracts the topological constraints of each target gene from the single-cell steady-state gene expression matrix; Step 2: Infer the gene regulatory relationship matrix based on single-cell steady-state gene expression data: The steady-state regulatory relationship inference module generates the regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene, and composes the regulatory relationship matrix A of the single-cell steady-state gene expression data from the regulatory relationship vectors of all target genes; Step 3: Infer the gene regulatory relationship matrix based on single-cell temporal gene expression data: The temporal regulatory relationship inference module generates a regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene; the regulatory relationship vectors of all target genes are combined into a regulatory relationship matrix B of the single-cell temporal gene expression data; Step 4: The regulatory relationship combination module uses the Borda voting method to score the regulatory relationship matrices A and B respectively to obtain the probability ranking of the existence of regulatory relationships among all genes; Step 5: The gene regulatory network generation module generates a gene regulatory network through parameter factors.

3. The gene regulatory network inference method based on single-cell steady-state and temporal sequence according to claim 2, characterized in that: The steps of the WGCNA clustering method described in step 1 are as follows: In the first step, the absolute value of the Pearson correlation between each element in the steady-state gene expression matrix is ​​calculated, and the absolute values ​​of all correlations are combined into a symmetrical correlation matrix; The second step is to convert the correlation matrix into the corresponding adjacency matrix; The third step is to calculate the element values ​​in the topological overlap matrix through the adjacency matrix according to the following formula: Among them, TOM ij Represents the element value of the i-th row and j-th column in the topological overlap matrix, l pq Indicates the size of the interaction between the pth element and the qth element in the adjacency matrix and other elements, a pq represents the element value in the p-th row and q-th column of the adjacency matrix. The values ​​of p and q are equal to the corresponding values ​​of i and j. min(·) represents the minimum value operation. k p represents the sum of the values ​​of the elements in the pth row of the adjacency matrix, k q represents the sum of the values ​​of the elements in the qth row of the adjacency matrix; The fourth step is to use the hierarchical clustering method to process the topological overlap matrix, calculate the distance between all classes based on the Pearson correlation coefficient, merge the classes with the closest distances into two pairs, and obtain a hierarchical clustering tree. The threshold is set according to the closeness of the relationship between the classes to obtain the clustering result. In the fifth step, based on the clustering results, the homologous gene set and heterologous gene set of each target gene are obtained respectively, and the homologous gene set and heterologous gene set are used as the topological constraints of the target gene.

4. The gene regulatory network inference method based on single-cell steady-state and temporal sequence according to claim 2, characterized in that: The steps for generating the regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene described in step 2 are as follows: In the first step, under the topological constraints of the same genes of each target gene, the ridge regression method is used to solve the dependence of the expression level of each target gene on the expression levels of its same genes, and obtain the possibility vector of the same genes regulating the target gene; In the second step, under the topological constraints of each target gene heterologous gene, the Lasso regression method is used to solve the dependence of the expression level of each target gene on the expression level of the target gene heterologous gene, and obtain the possibility vector of the heterologous gene regulating the target gene; In the third step, the possibility vector of target genes regulated by similar genes is concatenated with the possibility vector of target genes regulated by heterogeneous genes to obtain the gene regulation relationship vector of each target gene under topological constraints.

5. The gene regulatory network inference method based on single-cell steady-state and temporal sequence according to claim 2, characterized in that: The steps for generating the regulatory relationship vector between each target gene and other genes under the topological constraints of each target gene described in step 3 are as follows: In the first step, under the topological constraints of the target gene's similar genes, the ridge regression method is used to solve the dependence of the expression level of each target gene at the current time point on the cumulative expression levels of the target gene's similar genes at the three time points before the current time point, and obtain the possibility vector of the similar genes regulating the target gene; The cumulative expression level is obtained by the following formula: Among them, f accum (·) represents the cumulative function, Z1 represents the first data value participating in the cumulative effect, and Z r represents the rth data value participating in the cumulative effect, δ represents the attenuation factor, ∑ represents the summation operation, Z w represents the wth data value, m-1 is the exponent of δ, and the values ​​of m and w are equal; In the second step, under the topological constraints of each target gene heterologous gene, the Lasso regression method is used to solve the dependency of the expression level of each target gene at the current time point on the cumulative expression level of the target gene heterologous gene at the three time points before the current time point, and the possibility vector of the heterologous gene regulating the target gene is obtained; In the third step, the possibility vectors of target genes regulated by similar genes and the possibility vectors of target genes regulated by heterogeneous genes are combined to obtain the gene regulation relationship vector of each target gene under topological constraints.

6. The gene regulatory network inference method based on single-cell steady-state and temporal sequence according to claim 2, characterized in that: The steps for obtaining the probability ranking of regulatory relationships among all genes described in step 4 are as follows: The first step is to set the smallest element value in the control relationship matrix A and B to 1, the second smallest element value to 2 points, and so on, to obtain two score matrices; The second step is to add the corresponding element values ​​in the two score matrices to obtain a regulatory relationship possibility matrix M; The third step is to transform all elements of the regulatory relationship possibility matrix M into a corresponding table, with the gene corresponding to the row as the target gene, the gene corresponding to the column as the regulatory gene, and the element value at the corresponding position as the regulatory possibility of the regulatory gene regulating the target gene; The fourth step is to sort the table from large to small according to the possibility of regulation to obtain the possibility ranking of the relationship between genes.

7. The gene regulatory network inference method based on single-cell steady-state and temporal sequence according to claim 2, characterized in that: The generation of a gene regulatory network by parameter factors described in step 5 refers to setting the number of regulatory edges in the real network to the parameter factor value, selecting the top k gene pairs ranked by the regulatory possibility between genes, where the value of k is equal to the parameter factor value, setting edges between the target gene and the regulatory gene of each pair in the selected gene pairs, and forming a gene regulatory network with all gene pairs after the edges are set.

Images