A miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multichannel heterogeneity graph

Abstract

This application relates to the field of bioinformatics, and more particularly to a method for inferring miRNA-enhanced gene regulatory networks based on knowledge graph embedding and multi-channel heterogeneous graphs. A multi-channel heterogeneous graph is constructed using transcription factors, target genes, and miRNAs as nodes, and the regulatory relationships between nodes as edges. In the pre-training stage, sequence-to-vector techniques are used to deeply mine and extract semantic information from nucleic acid sequences. Then, the multi-channel heterogeneous graph is decomposed into multiple subgraphs according to different types of regulatory relationships, and multi-channel GraphSAGE is used to learn the node features specific to each regulatory type. Finally, the connection probability values ​​between the embedding features of each node are used to determine the predicted connection relationships between nodes, achieving high-throughput miRNA regulatory relationship prediction. This aims to solve the problem of how to efficiently and accurately predict the regulatory relationships of high-throughput miRNAs.

Description

Technical Field

[0001] This application relates to the field of bioinformatics, and in particular to a method for inferring miRNA-enhanced gene regulatory networks based on knowledge graph embedding and multichannel heterogeneous graphs. Background Technology

[0002] Gene regulatory networks (GRNs) are complex network structures composed of genes and their products interacting with each other. They play a crucial role in regulating gene expression and function. Studying GRNs not only helps to elucidate the molecular mechanisms of disease development but also provides new research directions and ideas for personalized treatment. In the process of expression regulation, miRNAs (also known as microRNAs), as important regulatory factors, act as "molecular fine-tuners" in GRNs, influencing cellular function through multi-target, dynamic, and hierarchical regulatory mechanisms, thereby exerting a negative effect on gene expression regulation.

[0003] Currently, research on gene regulatory networks, particularly regarding the synergistic multi-level regulatory effects between miRNAs and other regulatory factors, is relatively scarce. Existing literature (Luo J, Ouyang W, Shen C, et al. Multi-Relation Graph Embedding for Predicting miRNA-Target Gene Interactions by Integrating Gene Sequence Information[J]. IEEE Journal of Biomedical and Health Informatics, 2022, 26: 4345-4353.) proposes a model based on multi-relation graph embedding to predict the interaction between miRNAs and target genes. By combining multi-relation graph convolution and a bidirectional long short-term memory (Bi-LSTM) network, the prediction accuracy of miRNA-target gene interactions is improved.

[0004] However, despite the rise of knowledge graphs and deep learning algorithms opening new avenues for integrating diverse biological data, feature extraction from high-throughput miRNA gene sequences and deep learning of complex regulatory relationships remain pressing challenges. To address this, this application proposes a novel gene regulatory network inference method to achieve efficient and accurate prediction of regulatory relationships. Summary of the Invention

[0005] The main objective of this application is to provide a method for inferring miRNA-enhanced gene regulatory networks based on knowledge graph embedding and multichannel heterogeneous graphs, aiming to solve the problem of how to efficiently and accurately predict the regulatory relationships of high-throughput miRNAs.

[0006] To achieve the above objectives, this application provides a method for inferring miRNA-enhanced gene regulatory networks based on knowledge graph embedding and multichannel heterogeneity maps, the method comprising:

[0007] A multichannel heterogeneous graph is constructed using transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges.

[0008] The gene sequence information of transcription factors, target genes, and miRNAs is segmented into DNA fragments of length k. The DNA fragments are pre-trained using a preset learning model so that the preset learning model can learn the knowledge graph embedding associated with the DNA fragments.

[0009] Within each channel, GraphSAGE is used to sample the neighboring nodes of any selected target node in the multi-channel heterogeneous graph, thereby capturing the features of the target node under one or more control relationships;

[0010] After traversing all nodes, determine the aggregated embedding representation of the neighboring nodes of the target node under each of the control relationships;

[0011] Based on the embedded representation, the connection probability value between any two nodes is calculated, and two nodes whose connection probability value is greater than a preset threshold are identified as target node pairs with a connection relationship.

[0012] Optionally, the regulatory relationship includes at least one of the following:

[0013] The regulatory relationship between transcription factors and target genes;

[0014] Regulatory relationships among transcription factors;

[0015] The regulatory relationship between transcription factors and miRNAs;

[0016] The regulatory relationship between miRNAs and transcription factors;

[0017] The regulatory relationship between miRNAs and target genes.

[0018] Optionally, the training objective of the preset learning model is to minimize the negative log-likelihood loss.

[0019] Optionally, the expression for the embedding representation is:

[0020]

[0021] In the formula, This represents the aggregated embedding representation of the neighboring nodes of node v under relation r; Let be the embedding representation associated with neighbor node u in the k-th layer; MaxPool is the max pooling operation. and These represent the pooling weights and bias terms associated with relation r, respectively. For neighboring nodes, Represents the set of neighboring nodes under each relation. It is the ReLU activation function.

[0022] Optionally, the expression for calculating the connection probability value is:

[0023]

[0024] In the formula, Represents a node and The connection probability value between them For nodes Transpose of the associated embedded representation. For nodes Embedded representation.

[0025] Furthermore, to achieve the above objectives, this application also provides a miRNA-enhanced gene regulatory network inference model based on knowledge graph embedding and multi-channel heterogeneous graphs, wherein the miRNA-enhanced gene regulatory network inference model includes:

[0026] The multichannel heterogeneous graph construction module is used to construct multichannel heterogeneous graphs with transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges.

[0027] The pre-trained node feature embedding module is used to segment the gene sequence information of transcription factors, target genes and miRNAs into DNA fragments of length k. The DNA fragments are pre-trained by a preset learning model so that the preset learning model can learn the knowledge graph embedding associated with the DNA fragments.

[0028] A multi-channel heterogeneous GraphSAGE network module is used to sample the neighboring nodes of any selected target node in each channel using GraphSAGE, thereby capturing the features of the target node under one or more regulatory relationships; after traversing all nodes, the aggregated embedding representation of the neighboring nodes of the target node under each regulatory relationship is determined; based on the embedding representation, the connection probability value between any two nodes is calculated; and two nodes with a connection probability value greater than a preset threshold are determined as target node pairs with a connection relationship.

[0029] Optionally, the number of GraphSAGE layers in the multi-channel heterogeneous GraphSAGE network module is 2.

[0030] Optionally, the embedding dimension is set to 256 dimensions.

[0031] In addition, to achieve the above objectives, this application also provides a computer system comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multi-channel heterogeneous graph as described in any of the preceding claims.

[0032] In addition, to achieve the above objectives, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multi-channel heterogeneous graph as described in any of the preceding claims.

[0033] This application has at least the following beneficial effects:

[0034] 1. Using transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges, construct a multi-channel heterogeneous graph to improve the gene regulation network with a gene regulation knowledge graph covering transcription factors, miRNAs, and target genes.

[0035] 2. In the pre-training stage, sequence-to-vector technology is used to deeply mine and extract semantic information in nucleic acid sequences. Then, according to different types of regulatory relationships, the multi-channel heterogeneous graph is decomposed into multiple subgraphs, and multi-channel GraphSAGE is used to learn the node features unique to each type of regulation.

[0036] 3. Determine the predicted connection relationships between nodes based on the connection probability values ​​between the embedding features of each node, thereby achieving high-throughput prediction of miRNA regulatory relationships. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating the first embodiment of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multichannel heterogeneous graphs in this application.

[0038] Figure 2 This is a multi-channel heterogeneous diagram involved in the embodiments of this application.

[0039] Figure 3 This is a schematic diagram of the architecture of the miRNA-enhanced gene regulatory network inference model involved in the embodiments of this application;

[0040] Figure 4 The AUROC plot of the KGE-HGRN model involved in the embodiments of this application during five-fold cross-validation;

[0041] Figure 5 This is a schematic diagram showing the graph structure analysis results of different types of edges involved in the embodiments of this application;

[0042] Figure 6 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0043] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0044] To better understand the above technical solutions, exemplary embodiments of this disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of this disclosure to those skilled in the art.

[0045] First Embodiment

[0046] In gene regulatory networks, genes are represented as nodes, and regulatory relationships are represented by directed edges. Inference of gene regulatory networks can be transformed into a link prediction task, which aims to identify possible regulatory relationships between gene nodes.

[0047] Reference Figure 1 This embodiment proposes a method for inferring miRNA-enhanced gene regulatory networks based on knowledge graph embedding and multi-channel heterogeneity graphs. The method includes the following steps:

[0048] Step S10: Construct a multichannel heterogeneous graph using transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges.

[0049] In this step, a large-scale gene regulation knowledge graph containing multiple channels, namely a multichannel heterogeneous graph, is constructed by integrating data from the TRRUSTv2, RegNetwork, and DoRothEA databases.

[0050] In an alternative implementation, see Figure 2 The multichannel heterogeneous graph shown is represented as a directed graph in this knowledge graph. In this model, V, E, T, and R represent the node set, edge set, and their corresponding types, respectively. Each node v corresponds to a specific type t, and each edge e is associated with a specific edge type r. The model categorizes nodes into transcription factors. target genes and miRNAs The edge type E is classified according to their relationship: TF-TG TF-TF TF-miRNA miRNA-TF and miRNA-TG These refer to the regulatory relationships between transcription factors and target genes, the regulatory relationships among transcription factors, the regulatory relationships between transcription factors and miRNAs, the regulatory relationships between miRNAs and transcription factors, and the regulatory relationships between miRNAs and target genes, respectively. Furthermore, each node in the diagram is represented as vi, where i ranges from 1 to N, and N = |V| represents the total number of nodes.

[0051] In one alternative implementation, the constructed knowledge graph was deployed to the Neo4j platform, where path queries were performed. Taking transcription factor E2F1 and its target gene TOP2A as an example, the path length was set to no more than three hops, and some possible regulatory pathways between them were extracted.

[0052] Step S20: The gene sequence information of transcription factors, target genes and miRNAs is segmented into DNA fragments of length k. The DNA fragments are pre-trained using a preset learning model so that the preset learning model can learn the knowledge graph embedding associated with the DNA fragments.

[0053] In this embodiment, a sequence-to-vector technique is employed during the pre-training phase to convert the sequence into a low-dimensional numerical vector. This method effectively captures potential information and associations within the sequence while ensuring a close connection between this information and biological characteristics. In this process, k-mer counting is used, which divides the gene sequence information into DNA fragments of fixed length k, thereby improving the efficiency of sequence embedding. These DNA fragments are treated as "words," while the entire gene sequence information is considered a "document."

[0054] Furthermore, in some optional implementations, to process gene sequence information more efficiently, the Doc2Vec method from NLP is used in the preset learning model. It should be noted that the core idea of ​​the Doc2Vec stage is to generate a vector representation for each k-mer fragment in the sequence set. In this process, a PV-DBOW model can be used as the preset training model to optimize the vector representation of the entire sequence for training, with the goal of predicting k-mer fragments in the sequence. This method is similar to the Skip-gram model in Word2Vec. In this way, Doc2Vec can not only effectively capture the local features of RNA sequences but also deeply mine their global features, thus providing more comprehensive information for subsequent analysis.

[0055] Specifically, suppose a sequence c contains a series of k-mer segments. (where i ranges from 1 to N), the entire sequence is represented as a vector. Each k-mer fragment It can also be represented as a vector. The preset learning model's prediction objective is to predict the probability of each segment based on the vector representation of the sequence. This probability is normalized using the softmax function, as shown below:

[0056]

[0057] In the formula, v represents the set of all possible k-mer segments.

[0058] In some alternative implementations, to maximize the alignment between document vectors and their corresponding target fragment vectors while reducing similarity with non-target fragment vectors, the preset training model aims to minimize the following negative log-likelihood loss:

[0059]

[0060] In the formula, D represents the target fragment, and D represents the document (gene sequence information).

[0061] It should be noted that the negative log-likelihood loss function is continuously optimized using gradient descent. During this process, the negative log-likelihood measures the target segment. The goal is to maximize the similarity between document vector v(D) and each target fragment vector v(D). The inner product between the document vector v(D) and the fragment vector v(D) is calculated. Simultaneously, the softmax function normalizes all possible k-mer segments, transforming the objective into maximizing the probability of the target segment relative to all candidate segments. The loss function uses a logarithmic form, which can penalize incorrect predictions more significantly, thus encouraging the model to learn the relationship between documents and their segments more effectively during training. This is achieved by applying the document vector v(D) and the segment vector v(D) to the document vector v(D). ), and perform iterative updates, where Representing all potential k-mer segments, the model is able to progressively narrow the gap between predicted and actual values. Ultimately, this process generates effective embeddings that can simultaneously capture both global information and local features in the sequence.

[0062] Step S30: In each channel, GraphSAGE is used to sample the neighboring nodes of any selected target node, thereby capturing the features of the target node under one or more control relationships.

[0063] Step S40: After traversing all nodes, determine the aggregated embedding representation of the neighboring nodes of the target node under each of the control relationships;

[0064] Step S50: Based on the embedded representation, calculate the connection probability value between any two nodes, and determine the two nodes with the connection probability value greater than a preset threshold as target node pairs with a connection relationship.

[0065] In this embodiment, when performing link prediction, based on the framework established in steps S10 and S20, the features of the entire heterogeneous graph and all nodes are taken as input. The GraphSAGE module is used to extract and summarize the embedding features of each node. This module takes the embedding representation of all nodes as input and calculates the probability that there is a connection between any two nodes.

[0066] Since the way information propagates differs between different edge types in the graph, but these propagation mechanisms are interconnected, this embodiment uses the GraphSAGE network based on a multi-channel heterogeneous graph framework to handle this complex multi-relationship graph structure.

[0067] Considering the complexity of multi-relation graphs, the GraphSAGE network decomposes the multi-channel heterogeneous graph into multiple subgraphs, each containing only one specific type of edge. Then, within each subgraph, the GraphSAGE network arbitrarily selects the neighboring nodes of a target node from its neighborhood for learning, thereby capturing the features of the target node under one or more regulatory relationships. After traversing all nodes, the sampled neighbor node features are aggregated using an aggregation function to update the feature embedding representation of each node.

[0068] In some alternative implementations, the aggregation function selected is the max pooling function MaxPool.

[0069] For example, let the feature representation of the current node v be: Where d represents the dimension of the embedding space for each neighbor node. Its embedding representation is When multiple relations r affect the same node, This represents the set of neighboring nodes under each relation. Max pooling is performed on the neighboring nodes of each relation to obtain the aggregated embedding representation of the target node's neighboring nodes under each of the aforementioned regulatory relations:

[0070]

[0071] In the formula, This represents the aggregated embedding representation of the neighboring nodes of node v under relation r; Let be the embedding representation associated with neighbor node u in the k-th layer; MaxPool is the max pooling operation. and These represent the pooling weights and bias terms associated with relation r, respectively. For neighboring nodes, Represents the set of neighboring nodes under each relation. It is the ReLU activation function.

[0072] Furthermore, and optionally, after determining the embedding representation of each layer, the embedding representations of all relations are finally aggregated through a weighted summation to obtain a weighted feature representation. :

[0073]

[0074] It is important to emphasize that the model is able to capture unique neighborhood information for each type of relationship and generate a global representation of each node by integrating this information.

[0075] In some alternative implementations, the expression for calculating the connection probability value is:

[0076]

[0077] In the formula, Represents a node and The connection probability value between them For nodes Transpose of the associated embedded representation. For nodes Embedded representation.

[0078] Furthermore, in some optional implementations, a preset threshold is set to 0.5, that is, if the probability If the value exceeds 0.5, the node is considered... and nodes There is a connection between them. Conversely, if If the value is less than or equal to 0.5, it is considered that there is no connection between the two.

[0079] Furthermore, in some alternative implementations, in the probability prediction stage, the binary cross-entropy function is used to calculate the loss, which is used to measure the computational loss during model training, as shown in the following formula:

[0080]

[0081] In the formula, Loss represents the cross-entropy loss of the entire model, and V represents the set of all nodes in the graph. This refers to the label of the edge connecting node i and node j, while This indicates the probability that the model predicts whether there is a connection between node i and node j.

[0082] In the technical solution provided in this embodiment, addressing the challenges of feature extraction from high-throughput gene sequences and the difficulty in learning complex regulatory relationships, a multi-channel heterogeneous graph is constructed using transcription factors, target genes, and miRNAs as nodes, and the regulatory relationships between nodes as edges. During the pre-training phase, sequence-to-vector techniques are employed to deeply mine and extract semantic information from nucleic acid sequences. Then, based on different types of regulatory relationships, the multi-channel heterogeneous graph is decomposed into multiple subgraphs, and multi-channel GraphSAGE is used to learn the node features specific to each regulatory type. Finally, the connection probability values ​​between the embedding features of each node are used to determine the predicted connection relationships between nodes, thus achieving high-throughput miRNA regulatory relationship prediction.

[0083] Second Embodiment

[0084] Based on the method in the first embodiment, as an implementation scheme, refer to... Figure 3 The diagram shown illustrates the architecture of a miRNA-enhanced gene regulatory network inference model. This embodiment provides a miRNA-enhanced gene regulatory network inference model based on knowledge graph embedding and multi-channel heterogeneous graphs, referred to as the KGE-HGRN model. This model includes:

[0085] The multichannel heterogeneous graph construction module is used to construct multichannel heterogeneous graphs with transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges.

[0086] The pre-trained node feature embedding module is used to segment the gene sequence information of transcription factors, target genes and miRNAs into DNA fragments of length k. The DNA fragments are pre-trained by a preset learning model so that the preset learning model can learn the knowledge graph embedding associated with the DNA fragments.

[0087] A multi-channel heterogeneous GraphSAGE network module is used to sample the neighboring nodes of any selected target node in each channel using GraphSAGE, thereby capturing the features of the target node under one or more regulatory relationships; after traversing all nodes, the aggregated embedding representation of the neighboring nodes of the target node under each regulatory relationship is determined; based on the embedding representation, the connection probability value between any two nodes is calculated; and two nodes with a connection probability value greater than a preset threshold are determined as target node pairs with a connection relationship.

[0088] Comparative Example 1

[0089] Based on the miRNA-enhanced gene regulatory network inference model proposed in the second embodiment, and considering that the effectiveness of the KGE-HGRN model is affected by several key factors, this embodiment conducted experiments on different neural network layer depths, embedding vector dimensions, and selected node feature aggregation strategies. To evaluate the impact of different layer numbers on the performance of the KGE-HGRN model, the embedding dimension was set to 256, and 5-fold cross-validation was performed using GraphSAGE models with 1, 2, and 3 layers. The results are shown in Table 1.

[0090] Table 1. Performance of the KGE-HGRN model at different GraphSAGE layers

[0091]

[0092] The results show that the model performs best with two layers on most evaluation metrics. However, when the number of layers is increased to three, the model performance drops significantly, which may be due to overfitting and thus affects the overall results.

[0093] Comparative Example 2

[0094] Based on the miRNA-enhanced gene regulatory network inference model proposed in the second embodiment, further, in order to evaluate the performance of the KGE-HGRN model, the embedding dimension was set to 64, 128, 256, 512 and 1024 respectively, and the results are shown in Table 2:

[0095] Table 2. Impact of different embedding dimensions on the performance of the KGE-HGRN model

[0096]

[0097] The results show that the model's performance gradually improves with increasing embedding dimension, reaching its best performance at 256 dimensions.

[0098] Comparative Example 3

[0099] Based on the miRNA-enhanced gene regulatory network inference model proposed in the second embodiment, the model further employs GraphSAGE for node feature learning. Each node collects and aggregates features from its neighboring nodes in each iteration. As the number of iterations increases, nodes gradually acquire information from more distant neighbors, thereby accessing a wider range of graph structure and feature information. This embodiment uses four aggregation strategies—MAX Pool, LSTM, mean, and GCN—to evaluate the framework's performance. The results are shown in Table 3.

[0100] Table 3. Impact of different aggregation strategies on model performance in GraphSAGE

[0101]

[0102] The results show that the model performs best when the aggregation strategy of GraphSAGE is set to MAX Pool.

[0103] Third Embodiment

[0104] As a verification embodiment, this embodiment compares the KGE-HGRN model constructed based on the method proposed in the first embodiment or the second embodiment with traditional common models.

[0105] (1) Dataset construction and experimental design

[0106] In this embodiment, the pre-constructed knowledge graph contains 1,647 transcription factors, 18,393 target genes, and 807 miRNAs, along with 519,464 regulatory relationships among them. Several commonly used evaluation metrics in gene regulatory network inference tasks are employed to evaluate the KGE-HGRN model, including F1 score, accuracy, precision, recall, AUROC, and AUPRC. First, the model's performance is benchmarked and compared with existing GRN inference methods. Next, ablation experiments are performed, and model parameters for the knowledge graph and framework are compared to rigorously evaluate its effectiveness. Finally, this section focuses on the E2F1 case study to further validate the effectiveness of the KGE-HGRN model.

[0107] In bioinformatics, positive samples typically refer to gene pairs known to have a regulatory relationship, while negative samples refer to gene pairs that do not. To construct a negative sample set, the model randomly selects gene pairs from unlabeled data. However, to avoid mistakenly selecting positive samples during the selection process, the negative samples are resampled before each model calculation to ensure the accuracy and diversity of the samples.

[0108] The model employs K-fold cross-validation to evaluate its performance. The dataset is divided into several subsets. In each epoch, the model is trained on one subset and tested on another. This process is repeated multiple times, using a different subset as the test set each time. The commonly used 5-fold cross-validation method is selected. Each training epoch lasts for 300 epochs to ensure that the model's variance is reduced to below 0.01.

[0109] (2) Experimental results and analysis

[0110] To test the impact of different values ​​of K on the experimental results in K-fold cross-validation, specific K values ​​were set to 3, 5, 7, and 10, respectively. The relevant results are shown in Table 4.

[0111] Table 4. Comparison of different cross-validation methods used in the KGE-HGRN model

[0112]

[0113] The results show that the best results were achieved when K=5, which is consistent with the experimental setup of most current studies.

[0114] Furthermore, the AUROC plot of the KGE-HGRN model under five-fold cross-validation is shown below. Figure 4 As shown, KGE-HGRN was compared with six other existing methods, including KGE-TGI, MGDHGS, EdgeConv, Graph Isomorphic Network (GIN), GAT, and GATv2.

[0115] The results are shown in Table 5 below:

[0116] Table 5. Performance comparison of the KGE-HGRN model with other methods

[0117]

[0118] The results show that the model outperforms other models in terms of AUROC, AUPRC, accuracy, precision, and F1 score. Although EdgeConv achieved a recall of 0.95, its precision was low, indicating that the framework may have generated more false positives. In terms of AUROC and AUPRC, the KGE-HGRN model outperformed MGDHGS and KGE-TGI by 12%, 3%, 8%, and 2%, respectively, further demonstrating the model's advantages and effectiveness.

[0119] In the technical solution provided in this embodiment, the advantage of the KGE-HGRN model lies in its ability to effectively learn unique features under different regulatory types by utilizing multi-channel GraphSAGE. Thanks to this strategy, the model can more accurately capture the multi-level structural features of biological networks. By utilizing multi-relationship heterogeneous knowledge graphs, the model can learn complementary features from additional regulatory information, thereby playing a key role in improving prediction accuracy.

[0120] Fourth embodiment

[0121] As a verification embodiment, this application considers miRNAs to play a crucial role in gene regulatory networks; therefore, the regulatory relationships between miRNAs, transcription factors (TFs), and target genes (TGs) are incorporated into this model. Transcription factors have dual regulatory roles: they not only regulate mRNA transcription but also affect the transcriptional level of miRNAs. On the other hand, some miRNAs can target mRNAs encoding transcription factors, thereby regulating the expression of these transcription factors. Furthermore, transcription factors and miRNAs can synergistically regulate the same target gene, where transcription factors control gene expression at the transcriptional level, while miRNAs finely regulate gene expression at the post-transcriptional level.

[0122] To evaluate the impact of incorporating miRNA regulatory information on model performance, graph structure analysis of different edge types was performed in this embodiment:

[0123] I. The regulatory relationship between transcription factors and target genes, i.e., TF-TG ;

[0124] II. Regulatory relationships between transcription factors, i.e., TF-TF ;

[0125] III. The regulatory relationship between transcription factors and miRNAs, i.e., TF-miRNA ;

[0126] IV. The regulatory relationship between miRNAs and transcription factors, i.e., miRNA-TF ;

[0127] V. The regulatory relationship between miRNA and target genes, i.e., miRNA-TG .

[0128] The analysis results are shown below Figure 5 In the figure, ACC represents accuracy, Pre represents precision, and F1 is a metric that comprehensively considers model precision and recall (F1 Score). It uses a harmonic mean to balance precision and recall, thus providing a more comprehensive evaluation.

[0129] ;

[0130] AUROC represents the area under the Receiver Operating Characteristic Curve (ROC Curve), which is obtained by plotting the true positive rate (TPR, also known as recall) and false positive rate (FPR, 1 - true negative rate) at different thresholds; while AUPRC is calculated by calculating the precision at different recall levels, plotting the curve in the precision-recall space, and then calculating the area under the curve.

[0131] In this embodiment, the research subject was the E2F1 gene, an important member of the E2F transcription factor family, involved in the regulation of DNA replication and the cellular response to DNA damage. Dysregulation of E2F1 activity is closely related to the progression of various cancers. In the experiment, all known regulatory relationships were used for training, while the remaining unknown relationships served as a candidate test set. The KGE-HGRN model was used to identify potential regulatory relationships, and the candidate relationships were ranked according to their predicted probability, with the top ten candidate relationships ultimately selected. (See Table 6 below.)

[0132] Table 6. Genes predicted to have regulatory relationships with E2F1

[0133]

[0134] The results showed that by reviewing relevant literature on PubMed to verify these predictions, it was found that 70% (i.e., 7 out of 10) of the predicted regulatory relationships were directly or indirectly confirmed in existing studies.

[0135] As one implementation scheme, Figure 6 This is a schematic diagram of the hardware operating environment of the computer system involved in the embodiments of this application.

[0136] like Figure 6As shown, the computer system may include: a processor 1001, such as a CPU; a memory 1005; a user interface 1003; a network interface 1004; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0137] Those skilled in the art will understand that Figure 6 The computer system architecture shown does not constitute a limitation on the computer system and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0138] like Figure 6 As shown, the memory 1005, as a storage medium, may include an operating system, a network communication module, a user interface module, and computer programs. The operating system is a program that manages and controls the hardware and software resources of the computer system, as well as the operation of the computer programs and other software or programs.

[0139] exist Figure 6 In the computer system shown, the user interface 1003 is mainly used to connect to the terminal and communicate with the terminal; the network interface 1004 is mainly used to communicate with the backend server; and the processor 1001 can be used to call the computer program stored in the memory 1005.

[0140] In this embodiment, the computer system includes: a memory 1005, a processor 1001, and a computer program stored in the memory and executable on the processor, wherein:

[0141] When processor 1001 calls a computer program stored in memory 1005, it performs the following operations:

[0142] A multichannel heterogeneous graph is constructed using transcription factors, target genes, and miRNAs as nodes and the regulatory relationships between nodes as edges.

[0143] The gene sequence information of transcription factors, target genes, and miRNAs is segmented into DNA fragments of length k. The DNA fragments are pre-trained using a preset learning model so that the preset learning model can learn the knowledge graph embedding associated with the DNA fragments.

[0144] Within each channel, GraphSAGE is used to sample the neighboring nodes of any selected target node in the multi-channel heterogeneous graph, thereby capturing the features of the target node under one or more control relationships;

[0145] After traversing all nodes, determine the aggregated embedding representation of the neighboring nodes of the target node under each of the control relationships;

[0146] Based on the embedded representation, the connection probability value between any two nodes is calculated, and two nodes whose connection probability value is greater than a preset threshold are identified as target node pairs with a connection relationship.

[0147] Furthermore, those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in a computer system to implement the process steps of the embodiments of the above methods.

[0148] Therefore, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the various steps of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multi-channel heterogeneous graphs as described in the above embodiments.

[0149] The computer-readable storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0150] It should be noted that, since the storage medium provided in the embodiments of this application is the storage medium used to implement the methods of the embodiments of this application, those skilled in the art can understand the specific structure and variations of the storage medium based on the methods described in the embodiments of this application, and therefore will not be repeated here. All storage media used in the methods of the embodiments of this application fall within the scope of protection of this application.

[0151] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations 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, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0153] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0154] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0155] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. This application can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0156] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0157] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for inferring miRNA-enhanced gene regulatory network based on knowledge graph embedding and multi-channel heterogeneous graph, characterized in that, The method comprises the following steps: A multi-channel heterogeneous graph is constructed by taking transcription factors, target genes and miRNAs as nodes and the regulatory relationships between the nodes as edges; The gene sequence information of the transcription factors, target genes and miRNAs is divided into DNA fragments with a length of k, and a preset learning model is used to pre-train the DNA fragments, so that the preset learning model learns the knowledge graph embedding associated with the DNA fragments; In each channel, GraphSAGE is used to sample the adjacent nodes of a target node selected at random in the multi-channel heterogeneous graph, so as to capture the features of the target node under one or more regulatory relationships; the number of layers of GraphSAGE is 2; After all nodes are traversed, the embedding representation of the neighbor nodes of the target node under each of the regulatory relationships after aggregation is determined; According to the embedding representation, the connection probability value between any two nodes is calculated, and two nodes with a connection probability value greater than a preset threshold are determined as a target node pair with a connection relationship; The regulatory relationships include at least one of the following: The regulatory relationship between a transcription factor and a target gene; The regulatory relationship between transcription factors; The regulatory relationship between a transcription factor and a miRNA; The regulatory relationship between a miRNA and a transcription factor; The regulatory relationship between a miRNA and a target gene; The training target of the preset learning model is to minimize the negative log-likelihood loss; The expression of the embedding representation is: ； wherein, denotes the aggregated embedding representation of the neighbor nodes of the node r under the relation v ; is the embedding representation associated with the neighbor node k in the u layer; MaxPool is a max-pooling operation, and represent the pooling weight and bias term associated with the relation r , is the neighbor node, denotes the set of neighbor nodes under each relation, is a ReLU activation function; the embedding dimension is set to 256 dimensions.

2. The method of claim 1, wherein, The calculation expression of the connection probability value is: ； wherein denotes a connection probability value between nodes and , is the transpose of the embedding representation associated with node , is the embedding representation of node .

3. A miRNA-enhanced gene regulation network inference model based on knowledge graph embedding and multi-channel heterogeneous graph, characterized in that, The miRNA-enhanced gene regulation network inference model comprises: A multi-channel heterogeneous graph construction module is configured to construct a multi-channel heterogeneous graph by taking transcription factors, target genes and miRNAs as nodes and the regulatory relationships between the nodes as edges; A node feature embedding module based on pre-training is configured to divide the gene sequence information of the transcription factors, target genes and miRNAs into DNA fragments with a length of k, and pre-train the DNA fragments by using a preset learning model, so that the preset learning model learns the knowledge graph embedding associated with the DNA fragments; A multi-channel heterogeneous GraphSAGE network module is configured to sample the adjacent nodes of a target node selected at random in each channel by using GraphSAGE, so as to capture the features of the target node under one or more regulatory relationships; after all nodes are traversed, the embedding representation of the neighbor nodes of the target node under each of the regulatory relationships after aggregation is determined; according to the embedding representation, the connection probability value between any two nodes is calculated; and two nodes with a connection probability value greater than a preset threshold are determined as a target node pair with a connection relationship; The number of layers of GraphSAGE in the multi-channel heterogeneous GraphSAGE network module is 2; The expression of the embedding representation is: ； wherein, denotes the aggregated embedding representation of the neighbor nodes of the node r under the relation v ; is the embedding representation associated with the neighbor node k in the u layer; MaxPool is a max-pooling operation, and represent the pooling weight and bias term associated with the relation r , is the neighbor node, denotes the set of neighbor nodes under each relation, is a ReLU activation function; the embedding dimension is set to 256 dimensions.

4. A computer system, characterized by The computer system comprises a memory, a processor, and a computer program stored on the memory and capable of running on the processor, and the computer program, when executed by the processor, implements the steps of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multi-channel heterogeneous graph according to any one of claims 1 to 2.

5. A computer readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program, when executed by the processor, implements the steps of the miRNA-enhanced gene regulatory network inference method based on knowledge graph embedding and multi-channel heterogeneous graph according to any one of claims 1 to 2.

Images