A community query analysis method for gene regulatory networks

Abstract

The present invention provides a community query analysis method for gene regulatory networks, and relates to the field of computer technology. The method first reads gene regulatory network data, and counts basic information of the gene regulatory network; then determines a distance norm evaluation model corresponding to the gene regulatory network; determines whether to perform conventional kernel decomposition or p-restricted kernel decomposition on the distance norm evaluation model corresponding to the gene regulatory network according to whether a threshold p is input; finally, community query is performed on the kernel-decomposed gene regulatory network according to the query nodes and conditions input by the user, and the query results are evaluated. The community query is performed by using a distance norm model on a defined directed graph to improve the quality of the returned community query results.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a community query analysis method for gene regulatory networks. Background Art

[0002] In the real world, many complex systems can be represented by networks, exist in the form of networks, or can be transformed into networks. With the rapid development of information technology, various complex networks are becoming more and more common in practical applications. For example, gene regulatory networks: in the network, the expression level of a gene is affected by other genes, and the expression level of this gene affects the expression level of other genes. This mutually restrictive regulatory relationship between genes constitutes a complex gene regulatory network. Studying and analyzing gene regulatory networks and understanding regulatory relationships and mechanisms have a profound impact on the development of biology.

[0003] Community query is a hot topic in complex network analysis and has been applied to most real complex networks. A community in a complex network is an induced subgraph consisting of a group of nodes, and the nodes in the subgraph are tightly connected. Communities generally have important practical significance. For example, in a gene regulatory network, a community may be a target gene group for combined drug administration or a module for gene clustering analysis.

[0004] The US patent "US20140354649A1 provides a method for applying the k-core model to query on a large dynamic network", which describes the steps of identifying k-core subgraphs and maintaining k-core subgraphs in parallel when inserting new edges or deleting existing edges. This makes it possible to identify and monitor large-scale graph data in a scalable and efficient manner.

[0005] Chinese patent "CN104462260A provides a community search algorithm based on k-core", which includes generating a maximum spanning tree MST for a graph; finding a subtree connecting all query nodes on the maximum spanning tree MST; searching for a subtree containing the query node and returning the maximum k-core.

[0006] There are many community models, but the community query effect is not particularly ideal for specific gene regulatory networks. The current undirected graph distance norm model has a higher priority for hub node (nodes with a large degree distribution) genes (hub nodes and their surrounding nodes have more advantages in passing the model screening).

[0007] The methods provided by patent US20140354649A1 and patent CN104462260A are both based on the original k-core model. The k-core model is mainly applied to undirected graphs, performs poorly on directed graphs, and does not have the characteristics of the distance norm model. In other words, it cannot meet the model's priority bias for hub nodes in gene regulatory network community analysis. Summary of the invention

[0008] The technical problem to be solved by the present invention is to provide a community query analysis method for gene regulatory networks in response to the above-mentioned deficiencies in the prior art, propose a model suitable for directed gene regulatory networks, and at the same time tilt the priority of hub node screening to meet some analysis needs of genes.

[0009] In order to solve the above technical problems, the technical solution adopted by the present invention is: a community query analysis method for gene regulatory networks, comprising the following steps:

[0010] Step 1: Read the gene regulatory network data and perform statistics on the basic information of the gene regulatory network; the basic information of the gene regulatory network includes the maximum reachable path length, the number of edges, the number of points, and the maximum in-and-out degree of nodes in the network; the gene regulatory network is a directed graph model, each vertex in the graph represents a gene, and the directed edges between vertices represent the regulatory relationship between genes;

[0011] Based on the gene regulatory network data that has been read, whether to perform p-restricted core decomposition is determined by whether a threshold value p is input. If p is input, step 3 is executed to perform p-restricted core decomposition according to the limited conditions k, l, h; otherwise, step 2 is executed to perform conventional core decomposition according to the limited conditions k, l, h. The difference between p-restricted core decomposition and conventional core decomposition is that both limit the in-and-out degree of each node in the network within the distance h to at least k, l based on the (k, l)-h-core distance norm evaluation model corresponding to the gene regulatory network, but p-restricted core decomposition further limits the direct neighbor ratio of the node in the network to at least k*p and l*p.

[0012] Step 2: Perform conventional core decomposition on the distance norm evaluation model (k,l)-h-core corresponding to the gene regulatory network;

[0013] Step 2.1: Define the distance norm evaluation model (k, l)-h-core for directed graphs; modify the original distance norm model applied to undirected graphs to apply to the distance norm model (k, l)-h-core for directed graphs, while considering the out-degree and in-degree of the nodes in the model;

[0014] First determine the in-degree indeg of node v within distance h h(v) and the out-degree outdeg of node v within distance h h (v) Definition; where index h (v) is the number of in-degree nodes of node v within distance h, outdeg h (v) is the number of out-degree nodes of node v within distance h;

[0015] The definition of the distance norm model (k, l)-h-core applied to directed graphs is given as follows: given a distance threshold h∈N+, an integer k≥0, and an integer l≥0, the (k, l)-core within the distance range h is the largest subgraph of the gene regulatory network that satisfies the conditions, such that any node v in the subgraph satisfies index h (v)≥k,outdeg h (v) ≥ l;

[0016] Step 2.2: According to the definition of the distance norm evaluation model (k, l)-h-core applied to the directed graph, and the query node input by the user and the set limiting conditions k, l, h values, the nodes that do not meet the limiting conditions are added to the queue to be deleted;

[0017] Traverse each node in the gene regulatory network and calculate the outdeg of the node within the distance h h (v) and indegree h (v), those in-degree index h (v) Nodes that do not meet the restriction condition k and outdeg h (v) Points that do not meet the restriction condition l are added to the queue to be deleted;

[0018] Step 2.3: If the queue to be deleted is not empty, remove the nodes to be deleted one by one, and then add the nodes to be deleted to the deleted node set; according to the newly added deleted nodes in the deleted node set, recalculate the out-degree outdeg of the nodes affected by the deleted nodes within the distance h. h (v) and indegree h (v) Then repeat step 2.2;

[0019] The pruning strategy is used to recalculate the out-degree of the nodes affected by the deleted node within the distance h by using the lower bound formula (1). h (v) and indegree h (v);

[0020] cur_outdeg h (v)>=bf_outdeg h (v)-outdeg h-1(u) (1)

[0021] Step 3: Perform p-restricted core decomposition on the distance norm evaluation model (k, l)-h-core corresponding to the gene regulatory network;

[0022] Step 3.1: Define the (k, l)-h-core of p-restricted core decomposition; Given a distance threshold h ∈ N+, a proportion threshold p ∈ [0, 1], an integer k ≥ 0, and an integer l ≥ 0, the p-restricted (k, l)-core within the h-distance range is the largest subgraph of the gene regulatory network that satisfies the condition, such that for any node v within the subgraph, indeg h (v) ≥ k, outdeg h (v) ≥ l, indeg 1 (v) ≥ k * p, outdeg 1 (v) ≥ l * p, where indeg 1 (v) and outdeg 1 (v) are the in-degree and out-degree of node v within distance 1 respectively;

[0023] Step 3.2: According to the definition of the (k, l)-h-core of p-restricted core decomposition, and by reading the query node input by the user and the set values of the limiting conditions k, l, h, add the nodes that do not meet the limiting conditions to the queue to be deleted;

[0024] Traverse each node in the gene regulatory network, calculate the out-degree outdeg h (v) and in-degree indeg h (v) of the node within distance h, and add the nodes whose in-degree indeg h (v) does not meet the limiting condition k, and the nodes whose out-degree outdeg h (v) does not meet the limiting condition l to the queue to be deleted; Add the nodes that satisfy indeg 1 (v) < k * p or outdeg 1 (v) < l * p to the queue to be deleted;

[0025] Step 3.3: If the queue to be deleted is not empty, dequeue the nodes to be deleted one by one, and then add the nodes to be deleted to the set of deleted nodes; According to the newly added deleted nodes in the set of deleted nodes, recalculate the out-degree outdeg h (v) and in-degree indeg h (v) of those nodes that are affected by the deleted nodes within distance h, and calculate indeg 1 (v) and outdeg 1 (v) according to the neighbors directly affected by the deleted nodes, and then repeat Step 3.2;

[0026] The out-degree of the nodes affected by the deleted node within the distance h h (v) and indegree h (v) Still using the pruning strategy to recalculate using formula (1);

[0027] Step 4: Perform community query on the gene regulatory network after nuclear decomposition according to the query nodes and conditions input by the user. If there is no community that meets the conditions, prompt the user, otherwise output the community that meets the conditions;

[0028] Step 4.1: First determine whether the query node is in the deleted node set. If it is, there is no community that meets the conditions and the query ends. Otherwise, add the query node to the query queue and continue to step 4.2.

[0029] Step 4.2: Determine whether the query queue is empty. If the query queue is empty, the query ends and step 4.4 is executed. If the query queue is not empty, step 4.3 is executed.

[0030] Step 4.3: Dequeue the nodes in the query queue one by one, and then add the dequeued nodes to the set of visited nodes, traverse the direct in-degree and out-degree neighbor node sets of the dequeued node, if the neighbor node of the dequeued node is not in the deleted node, then determine whether the neighbor node is in the set of visited nodes, if not, add it to the query queue, and determine whether the edge formed by the dequeued node and its neighbor node is in the set of visited edges, if not, add it to the set of visited edges, and then re-execute step 4.2;

[0031] Step 4.4: Output the set of edges that have been visited, which is the community query result represented; finally, calculate the basic information of the community query result to assist in the analysis and statistics of the community query result;

[0032] Step 5: Use the proportion of unreachable paths in the community, the proportion of nodes with zero out-degree and in-degree of community members, and the similarity of community members as evaluation indicators to evaluate the community query results;

[0033] The proportion of unreachable paths in the community is shown in the following formula:

[0034]

[0035] Among them, URP is the proportion of unreachable paths in the community, C represents the community,

[0036] The proportion of nodes with out-degree 0 and in-degree 0 in the community is the proportion of nodes with out-degree 0 to the total number of nodes in the community, and the proportion of nodes with in-degree 0 to the total number of nodes;

[0037] The similarity of community members includes the in-degree similarity CMS in and the out-degree similarity CMS out of community members. The CMS in and CMS out of a community C are calculated as shown in the following formula:

[0038]

[0039]

[0040] Among them, CMS_in(C) and CMS_out(C) represent the in-degree similarity and out-degree similarity of the members of community C, in(u) and in(v) represent the in-degree node sets of nodes u and v, respectively, and Out(u) and Out(v) represent the out-degree node sets of nodes u and v, respectively.

[0041] Finally, the community query results, as well as basic information and statistical indicators related to the community query results are output.

[0042] The beneficial effect of adopting the above technical solution is that: the community query analysis method for gene regulatory networks provided by the present invention modifies the original distance norm model (k,h)-core (only restricting the out-degree to at least k) based on the undirected graph of the gene regulatory network into a distance norm model (k,l)-h-core model (restricting the out-degree to at least k and the in-degree to at least l) applied to the directed gene regulatory network, so as to solve the problem that "the community query effect for specific gene regulatory networks in many current community models is not particularly ideal". The existing undirected graph distance norm model has a higher priority tilt for hub node (node ​​with large degree distribution) genes (hub nodes and their surrounding nodes have more advantages in passing the model screening). Such characteristics make it easier to obtain community query results and the community pays more attention to hub nodes. However, the model is applied to directed graphs and simply ignores the direction of the edge. The effect is not good. The method of the present invention performs community query by defining the distance norm model on the directed graph, improves the quality of the returned community query results, and proposes some pruning strategies and extended queries based on this model. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of a community query analysis method for gene regulatory networks provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of a distance norm model applied to a directed graph provided by an embodiment of the present invention;

[0045] Figure 3 A comparison diagram of CMS results of community queries on three gene regulatory networks using two different methods provided in an embodiment of the present invention, wherein (a) is a comparison diagram of the in-degree similarity CMS in of community members; and (b) is a comparison diagram of the out-degree similarity CMS out of community members. DETAILED DESCRIPTION

[0046] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0047] In this embodiment, a community query analysis method for gene regulatory networks is provided. Figure 1 As shown, the following steps are included:

[0048] Step 1: Read the gene regulatory network data and collect statistics on the basic information of the gene regulatory network; the basic information of the gene regulatory network includes the maximum reachable path length, number of edges, number of points, and maximum in-and-out degree of nodes in the network;

[0049] The gene regulatory network is a directed graph model. Each vertex in the graph represents a gene. The directed edges between the vertices represent the regulatory relationship between genes. There may be some other attribute values ​​on the edges, such as the probability value representing the probability of regulation of the regulatory relationship between genes, and the positive and negative values ​​of the regulatory relationship representing the promotion or inhibition of the regulatory relationship. The model definition of this algorithm is a directed graph model, which will consider the edges of the in- and out-degrees separately. Therefore, compared with the original algorithm, the weights on the edges can be retained when reading. Although these attribute values ​​will not be processed, they will be retained and displayed to the user for analysis when querying;

[0050] Based on the gene regulatory network that has been read, whether to perform p-constrained core decomposition (p-constrained core decomposition) is determined by whether the threshold p is input. If p is input, step 3 is executed to perform p-constrained core decomposition according to the constraints k, l, h; otherwise, step 2 is executed to perform conventional core decomposition according to the constraints k, l, h; the difference between p-constrained core decomposition and conventional core decomposition is that both limit the in-and-out degree of each node in the network within the distance h to at least k, l according to the (k, l)-h-core distance norm evaluation model corresponding to the gene regulatory network, but p-constrained core decomposition further limits the direct neighbor ratio of the node in the network to at least k*p and l*p;

[0051] This example uses three gene regulatory networks of different scales: Ecoli, Breast_cancer and Ecoli_mutifactorial. The basic information of these three networks is shown in Table 1:

[0052] Table 1 Basic information of gene regulatory networks of three different scales

[0053] Network Name Points Number of edges Average reachable path Maximum in-degree / out-degree Ecoli 64 224 3.7 6 / 13 Breast_cancer 574 3502 4.5 9 / 108 Ecoli_mutifactorial 1564 3648 2.4 11 / 430

[0054] Step 2: Perform conventional core decomposition on the distance norm evaluation model (k, l)-h-core corresponding to the gene regulatory network;

[0055] Step 2.1: Define the distance norm evaluation model (k, l)-h-core for directed graphs; modify the original distance norm model applied to undirected graphs to apply to the distance norm model (k, l)-h-core for directed graphs, while considering the out-degree and in-degree of the nodes in the model;

[0056] First determine the in-degree indeg of node v within distance h h (v) and the out-degree outdeg of node v within distance h h (v) Definition; where index h (v) is the number of in-degree nodes of node v within distance h, outdeg h (v) is the number of out-degree nodes of node v within distance h;

[0057] like Figure 2 As shown, index h (x) is the number of in-degree nodes of point x within distance h (i.e., the in-degree in h-neighborhood), for example, index 1 (v) is the number of in-degree nodes of node v within distance 1. Taking point x in the figure as an example, the in-degree of point x under the condition of h = 1 is 1 (only point F). At this time, indeg 1 (x) is the in-degree of point x in the traditional sense. When h = 2, the in-degree node of x in the first-level neighbor is only F, and in the 2-neighborhood, there are two cases: 1-neighborhood F in-degree and out-degree points A and B. At this time, considering the directionality of regulatory information in the gene regulatory network, it is natural to ignore the edge F->B with inconsistent directions (because at this time, for x, B is the 2-neighborhood point of x, and it is difficult to affect x through F). At this time, indeg 2 (x) = 2 (point F and point A). Similarly, outdeg h (x) is calculated in the same way, outdeg 1 (x) = 1, outdeg 2 (x) = 2 (points E and D).

[0058] The definition of the distance norm model (k,l)-h-core applied to directed graphs is given as follows: given a distance threshold h∈N+, an integer k≥0, and an integer l≥0, the (k,l)-core (that is, (k,l)-h-core) within the distance range h is the largest subgraph of the gene regulatory network that satisfies the conditions, such that any node v in the subgraph satisfies index h (v)≥k,outdeg h (v) ≥ l;

[0059] Step 2.2: According to the definition of the distance norm evaluation model (k, l)-h-core applied to directed graphs, and reading the query node input by the user and the set limiting conditions k, l, h values, this embodiment sets h=2, k=4, l=4, and adds the nodes that do not meet the limiting conditions to the queue for deletion;

[0060] Traverse each node in the gene regulatory network and calculate the outdeg of the node within the distance h h (v) and indegree h (v), those in-degree index h (v) Nodes that do not meet the restriction condition k and outdeg h (v) Points that do not meet the restriction condition l are added to the queue to be deleted;

[0061] Step 2.3: If the queue to be deleted is not empty, remove the nodes to be deleted one by one, and then add the nodes to be deleted to the deleted node set; according to the newly added deleted nodes in the deleted node set, recalculate the out-degree outdeg of the nodes affected by the deleted nodes within the distance h. h (v) and indegree h (v) Then repeat step 2.2;

[0062] The pruning strategy is used to recalculate the out-degree of the nodes affected by the deleted node within the distance h by using the lower bound formula (1). h (v) and indegree h (v) According to the proposed lower bound formula (1), the number of repeated calculations is reduced and the decomposition process is accelerated;

[0063] This embodiment takes the calculation of the out-degree of a node within the distance h as an example. Suppose the deleted node is u, the affected node is v, and the out-degree value of v within the distance h before being affected is bf_outdeg h (v), the actual value of the out-degree of node v after the current update is cur_outdeg h(v), when the distance between node v and node u is 1, there are only two cases. The first is that all the child nodes of node u within the distance range of h-1 are connected to v through u. In this case, cur_outdeg h (v) = bf_outdeg h (v)-outdeg h-1 (u), the second is that not all of the child nodes of node u are connected to node v through u, that is, node v and the child nodes of u have other connections. In the second case, there is an over-deletion, so cur_outdeg h (v)>bf_outdeg h (v)-outdeg h-1 (u); When the distance between node v and node u is greater than 1, the nodes whose distance from node v to u exceeds h are all over-deleted nodes, and there is still cur_outdeg h (v)>=bf_outdeg h (v)-outdeg h-1 (u). So we have:

[0064] cur_outdeg h (v)>=bf_outdeg h (v)-outdeg h-1 (u) (1)

[0065] Therefore, according to formula (1), we can get the actual outdeg of point v after point u is deleted. h The lower bound of (v), that is, the lower bound is bf_outdeg h (v)-outdeg h-1 (u), when deleting node u, if the lower bound of point v is bf_outdeg h (v)-outdeg h-1 (u) is still not less than the limit, then we can directly replace the lower bound without calculating outdeg at this time h (v) Because the actual outdeg h (v) must also be greater than the limiting condition, and when the lower bound is less than the limiting condition, the true value outdeg h (v) is not necessarily less than the limit and needs to be recalculated.

[0066] Step 3: Perform p-restricted core decomposition on the distance norm evaluation model (k,l)-h-core corresponding to the gene regulatory network;

[0067] Step 3.1: Define the (k, l)-h-core of the p-restricted core decomposition; Given a distance threshold h ∈ N+, a proportion threshold p ∈ [0, 1], an integer k ≥ 0, and an integer l ≥ 0, the p-restricted (k, l)-core within the h-distance range is the largest subgraph of the gene regulatory network that satisfies the condition such that for any node v in the subgraph, indeg h (v) ≥ k, outdeg h (v) ≥ l, indeg 1 (v) ≥ k * p, outdeg 1 (v) ≥ l * p;

[0068] Step 3.2: According to the definition of the (k, l)-h-core of the p-restricted core decomposition, and by reading the query node input by the user and the set k, l, h values of the limiting conditions, add the nodes that do not meet the limiting conditions to the deletion queue;

[0069] Traverse each node in the gene regulatory network, calculate the out-degree outdeg h (v) and in-degree indeg h (v) of the node within the distance h, and add the nodes whose in-degree indeg h (v) does not meet the limiting condition k, and the nodes whose out-degree outdeg h (v) does not meet the limiting condition l to the deletion queue; Add the nodes that satisfy indeg 1 (v) < k * p or outdeg 1 (v) < l * p to the deletion queue;

[0070] Step 3.3: If the deletion queue is not empty, dequeue the nodes to be deleted one by one, and then add the nodes to be deleted to the set of deleted nodes; According to the newly added deleted nodes in the set of deleted nodes, recalculate the out-degree outdeg h (v) and in-degree indeg h (v) of those nodes that are affected by the deleted nodes within the distance h, and calculate indeg 1 (v) and outdeg 1 (v) according to the neighbors directly affected by the deleted nodes, and then repeat Step 3.2;

[0071] The out-degree outdeg h (v) and in-degree indeg h (v) of the nodes affected by the deleted nodes within the distance h are still recalculated using the pruning strategy through the lower bound formula (1);

[0072] Step 4: Perform community query on the gene regulatory network after nuclear decomposition according to the query nodes and conditions input by the user. If there is no community that meets the conditions, prompt the user, otherwise output the community that meets the conditions;

[0073] Step 4.1: First determine whether the query node is in the deleted node set. If it is, there is no community that meets the conditions and the query ends. Otherwise, add the query node to the query queue and continue to step 4.2.

[0074] Step 4.2: Determine whether the query queue is empty. If the query queue is empty, the query ends and step 4.4 is executed. If the query queue is not empty, step 4.3 is executed.

[0075] Step 4.3: Dequeue the nodes in the query queue one by one, and then add the dequeued nodes to the set of visited nodes, traverse the direct in-degree and out-degree neighbor node sets of the dequeued node, if the neighbor node of the dequeued node is not in the deleted node, then determine whether the neighbor node is in the set of visited nodes, if not, add it to the query queue, and determine whether the edge formed by the dequeued node and its neighbor node is in the set of visited edges, if not, add it to the set of visited edges, and then re-execute step 4.2;

[0076] Step 4.4: Output the set of edges that have been visited, which is the community query result represented; finally, calculate the basic information of the community query result to assist in the analysis and statistics of the community query result;

[0077] Step 5: Use the proportion of unreachable paths in the community, the proportion of nodes with zero out-degree and in-degree of community members, and the similarity of community members as evaluation indicators to evaluate the community query results;

[0078] The calculation formula for the proportion of unreachable paths in a community is as follows:

[0079]

[0080] Among them, URP is the proportion of unreachable paths in the community, C represents the community,

[0081] The proportion of nodes with out-degree 0 and in-degree 0 in the community is the proportion of nodes with out-degree 0 to the total number of nodes in the community, and the proportion of nodes with in-degree 0 to the total number of nodes;

[0082] The similarity of community members includes the in-degree similarity CMS (community members similarity) in and the out-degree similarity CMS out of community members. The CMS in and CMS out of a community C are calculated as shown in the following formula:

[0083]

[0084]

[0085] Among them, CMS_in(C) and CMS_out(C) represent the in-degree similarity and out-degree similarity of the members of community C, in(u) and in(v) represent the in-degree node sets of nodes u and v, respectively, and Out(u) and Out(v) represent the out-degree node sets of nodes u and v, respectively.

[0086] Finally, the community query results are output, as well as basic information and statistical indicators related to the community query results, such as Figure 3 The CMS is shown as an example.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A community query analysis method for gene regulatory networks, characterized by: Read gene regulatory network data and collect basic information of gene regulatory network; Determine the distance norm evaluation model corresponding to the gene regulatory network; Perform kernel decomposition on the distance norm evaluation model corresponding to the gene regulatory network; Perform community query on the gene regulatory network after nuclear decomposition according to the query nodes and conditions input by the user; The following steps are involved: Step 1: Read the gene regulatory network data and collect statistics on the basic information of the gene regulatory network; Based on the gene regulatory network data that has been read, whether to perform p-restricted nuclear decomposition is determined based on whether a threshold p is input. If p is input, step 3 is executed to perform p-restricted nuclear decomposition according to the limited conditions k, l, h; otherwise, step 2 is executed to perform conventional nuclear decomposition according to the limited conditions k, l, h; The basic information of the gene regulatory network includes the maximum reachable path length, the number of edges, the number of points in the network, and the maximum in-and-out degree of nodes in the network; the gene regulatory network is a directed graph model, in which each vertex represents a gene, and the directed edges between vertices represent the regulatory relationship between genes; Step 2: Perform conventional core decomposition on the distance norm evaluation model (k,l)-h-core corresponding to the gene regulatory network; Step 3: Perform p-restricted core decomposition on the distance norm evaluation model (k,l)-h-core corresponding to the gene regulatory network; The difference between p-restricted core decomposition and conventional core decomposition is that both of them limit the in-and-out degree of each node in the network within the distance h to at least k, l according to the (k, l)-h-core distance norm evaluation model corresponding to the gene regulatory network, but p-restricted core decomposition further limits the direct neighbor proportion of the node in the network to at least k*p and l*p; Step 4: Perform community query on the gene regulatory network after nuclear decomposition according to the query nodes and conditions input by the user. If there is no community that meets the conditions, prompt the user, otherwise output the community that meets the conditions; Step 5: Output the community query results, as well as basic information and statistical indicators related to the community query results.

2. A community query analysis method for gene regulatory networks according to claim 1, characterized in that: The specific method of step 2 is: Step 2.1: Define the distance norm evaluation model (k, l)-h-core for directed graphs; modify the original distance norm model applied to undirected graphs to apply to the distance norm model (k, l)-h-core for directed graphs, while considering the out-degree and in-degree of the nodes in the model; First determine the in-degree indeg of node v within distance h h (v) and the out-degree outdeg of node v within distance h h (v) Definition; where index h (v) is the number of in-degree nodes of node v within distance h, outdeg h (v) is the number of out-degree nodes of node v within distance h; The definition of the distance norm model (k, l)-h-core applied to directed graphs is given as follows: given a distance threshold h∈N+, an integer k≥0, and an integer l≥0, the (k, l)-core within the distance range h is the largest subgraph of the gene regulatory network that satisfies the conditions, such that any node v in the subgraph satisfies index h (v)≥k,outdeg h (v) ≥ l; Step 2.2: According to the definition of the distance norm evaluation model (k, l)-h-core applied to the directed graph, and the query node input by the user and the set limiting conditions k, l, h values, the nodes that do not meet the limiting conditions are added to the queue to be deleted; Traverse each node in the gene regulatory network and calculate the outdeg of the node within the distance h h (v) and indegree h (v), those in-degree index h (v) Nodes that do not meet the restriction condition k and outdeg h (v) Points that do not meet the restriction condition l are added to the queue to be deleted; Step 2.3: If the queue to be deleted is not empty, remove the nodes to be deleted one by one, and then add the nodes to be deleted to the deleted node set; according to the newly added deleted nodes in the deleted node set, recalculate the out-degree outdeg of the nodes affected by the deleted nodes within the distance h. h (v) and indegree h (v) Then repeat step 2.2; The pruning strategy is used to recalculate the out-degree of the nodes affected by the deleted node within the distance h by using the lower bound formula (1). h (v) and indegree h (v); cur_outdeg h (v) >= bf_outdeg h (v)- outdeg h-1 (u) (1).

3. A community query analysis method for gene regulatory networks according to claim 2, characterized in that: The specific method of step 3 is: Step 3.1: Define the (k, l)-h-core of the p-restricted core decomposition; given a distance threshold h∈N+, a ratio threshold p∈[0,1], an integer k≥0, an integer l≥0, the p-restricted (k, l)-core within the h distance range is the largest subgraph of the gene regulatory network that satisfies the conditions, so that any node v in the subgraph satisfies, index h (v)≥k,outdeg h (v)≥l,indeg 1 (v)≥k*p,outdeg 1 (v)≥l*p, where index 1 (v) and outdeg 1 (v) are the in-degree and out-degree of node v within distance 1 respectively; Step 3.2: According to the definition of (k, l)-h-core of p-restricted core decomposition, and reading the query node input by the user and the set limiting conditions k, l, h values, the nodes that do not meet the limiting conditions are added to the queue to be deleted; Traverse each node in the gene regulatory network and calculate the out-degree outdeg h (v) and in-degree indeg h (v). Add the nodes whose in-degree indeg h (v) does not meet the limit condition k, and the nodes whose out-degree outdeg h (v) does not meet the limit condition l to the deletion queue; Add the nodes that satisfy indeg 1 (v) < k * p or outdeg 1 (v) < l * p to the deletion queue; Step 3.3: If the queue to be deleted is not empty, remove the nodes to be deleted one by one, and then add the nodes to be deleted to the deleted node set; according to the newly added deleted nodes in the deleted node set, recalculate the out-degree outdeg of the nodes affected by the deleted nodes within the distance h. h (v) and indegree h (v) and calculate the index based on the neighbors directly affected by the deleted node 1 (v) and outdeg 1 (v) Then repeat step 3.2; The out-degree of the nodes affected by the deleted node within the distance h h (v) and indegree h (v) The pruning strategy is still used to recalculate using the lower bound formula (1).

4. A community query analysis method for gene regulatory networks according to claim 3, characterized in that: The specific method of step 4 is: Step 4.1: First determine whether the query node is in the deleted node set. If it is, there is no community that meets the conditions and the query ends. Otherwise, add the query node to the query queue and continue to step 4.

2. Step 4.2: Determine whether the query queue is empty. If the query queue is empty, the query ends and step 4.4 is executed. If the query queue is not empty, step 4.3 is executed. Step 4.3: Dequeue the nodes in the query queue one by one, and then add the dequeued nodes to the set of visited nodes, traverse the direct in-degree and out-degree neighbor node sets of the dequeued node, if the neighbor node of the dequeued node is not in the deleted node, then determine whether the neighbor node is in the set of visited nodes, if not, add it to the query queue, and determine whether the edge formed by the dequeued node and its neighbor node is in the set of visited edges, if not, add it to the set of visited edges, and then re-execute step 4.2; Step 4.4: Output the set of edges that have been visited, which is the community query result; Finally, the basic information of the community query results is calculated to assist in the analysis and statistics of the community query results.

5. A community query analysis method for gene regulatory networks according to any one of claims 1 to 4, characterized in that: The method also uses the proportion of unreachable paths in the community, the proportion of nodes with 0 out-degree and in-degree of community members, and the similarity of community members as evaluation indicators of community query results to evaluate the community query results.

6. A community query analysis method for gene regulatory networks according to claim 5, characterized in that: The proportion of unreachable paths in the community is shown in the following formula: Among them, URP is the proportion of unreachable paths in the community, C represents the community, The proportion of nodes with out-degree 0 and in-degree 0 in the community is the proportion of nodes with out-degree 0 to the total number of nodes in the community, and the proportion of nodes with in-degree 0 to the total number of nodes; The similarity of community members includes the in-degree similarity CMSin and the out-degree similarity CMSout of community members. The CMSin and CMSout of a community C are calculated as shown in the following formula: Among them, CMS_in(C) and CMS_out(C) represent the in-degree similarity and out-degree similarity of the members of community C respectively, in(u) and in(v) represent the in-degree node sets of nodes u and v respectively, and Out(u) and Out(v) represent the out-degree node sets of nodes u and v respectively.

Images