LncRNA marker prediction method based on hierarchical graph neural network

Abstract

The invention particularly relates to an lncRNA marker prediction method based on a hierarchical graph neural network. The method comprises the steps of construction of an lncRNA similarity network, construction of a disease similarity network, feature learning of lncRNA and diseases, construction of an lncRNA and disease association graph, and dimension feature extraction of a neural structure based on the graph. According to the method, the prediction performance in the aspect of identifying the association between the lncRNA and the disease pair can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of lncRNA marker prediction, and in particular to a lncRNA marker prediction method based on a hierarchical graph neural network. Background Art

[0002] In the related technologies, the technologies for predicting the association between lncRNA (Long non-coding RNA) and diseases are mainly based on machine learning and deep learning. For example, the SIMCLDA model, the lncRNA-disease association prediction model GCRFLDA based on graph convolution matrix, the DMFLDA model, and so on. Existing methods are insufficient in dealing with the multifaceted nature of lncRNA interactions, which span different biological pathways and molecular mechanisms, making the prediction and verification of the association between lncRNA and disease a difficult task. In addition, the heterogeneity of the disease itself, as well as its different genetic and environmental basis, adds another layer of complexity to identifying the specific role of lncRNAs in the pathogenesis of the disease.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0004] The present invention provides a lncRNA marker prediction method based on a hierarchical graph neural network, a storage medium, a computer program product, and an electronic device, which improves the prediction performance in identifying the association between lncRNA and disease pairs, thereby overcoming the defects existing in the prior art to a certain extent.

[0005] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0006] According to a first aspect of the present invention, a method for predicting lncRNA markers based on a hierarchical graph neural network is provided, the method comprising:

[0007] Construct lncRNA similarity network and disease similarity network;

[0008] Extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network respectively; and constructing a correlation feature graph using the lncRNA feature data and the disease feature data; and

[0009] Based on the associated source data of lncRNA and disease, an interconnection graph is constructed;

[0010] Using a feature extraction model based on a hybrid graph neural network architecture, dimensional feature extraction is performed on the associated feature graph data and the interconnected graph data to obtain first dimensional feature data corresponding to the associated feature graph data and second dimensional feature data corresponding to the interconnected graph data;

[0011] The attention mechanism layer is used to perform feature fusion processing on the first dimension feature data and the second dimension feature data to obtain fused feature data;

[0012] The multi-layer perceptron layer is used to classify the fused feature data and obtain the correlation probability between lncRNA and disease.

[0013] In some exemplary embodiments, constructing a lncRNA similarity network comprises:

[0014] The lncRNA data source is used to calculate the first similarity data between lncRNAs based on the Gaussian kernel function, and the ncRNA similarity network is constructed based on the first similarity data.

[0015] In some exemplary embodiments, the method further comprises:

[0016] Using the lncRNA data source and the omics data, calculating the second similarity data based on the Jaccard similarity coefficient between the lncRNA data source and various omics data; wherein the omics data includes: at least one of an mRNA data set and a protein data set;

[0017] A ncRNA similarity network is constructed based on the first similarity data and the second similarity data.

[0018] In some exemplary embodiments, constructing a disease similarity network comprises:

[0019] The disease data source is used to calculate the third similarity data between the diseases based on the FNSemSim function; and a disease similarity network is constructed based on the third similarity data.

[0020] In some exemplary embodiments, the method further comprises:

[0021] Using the disease data source, the relevant metabolite data set, and the circular RNA data set, respectively, to calculate fourth similarity data based on the Jaccard similarity coefficient between the disease data source and the relevant metabolite data set, and fifth similarity data based on the Jaccard similarity coefficient between the disease data source and the circular RNA data set;

[0022] A disease similarity network is constructed based on the third similarity data, the fourth similarity data and the fifth similarity data.

[0023] In some exemplary embodiments, extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network, respectively, comprises:

[0024] The RWR algorithm is used to aggregate the network nodes in the lncRNA similarity network and the disease similarity network to generate the corresponding lncRNA feature data and disease feature data;

[0025] The lncRNA feature data and disease feature data are denoised by a denoising autoencoder to obtain the lncRNA feature data and disease feature data after noise removal.

[0026] In some exemplary embodiments, the feature extraction model based on the hybrid graph neural network architecture includes: a first graph convolution layer, a graph attention layer, and a second graph convolution layer arranged in sequence;

[0027] Among them, the graph convolutional network learns node representation by utilizing the topological structure of the graph; the graph attention layer calculates node features by combining the attention mechanism.

[0028] According to a second aspect of the present invention, a storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned lncRNA marker prediction method based on hierarchical graph neural network is implemented.

[0029] According to a third aspect of the present invention, there is provided an electronic device, comprising:

[0030] Processor; and

[0031] A memory, configured to store executable instructions of the processor;

[0032] Wherein, the processor is configured to implement the above-mentioned lncRNA marker prediction method based on hierarchical graph neural network by executing the executable instructions.

[0033] According to a fourth aspect of the present invention, a computer program product is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned lncRNA marker prediction method based on hierarchical graph neural network is implemented.

[0034] The lncRNA marker prediction method based on hierarchical graph neural network provided by the embodiment of the present invention constructs lncRNA similarity network and disease similarity network, and extracts corresponding lncRNA feature data and disease feature data to construct associated feature graph data; and uses the associated source data of lncRNA and disease to construct interconnected graph data, so that the feature extraction model based on the hybrid graph neural network architecture can be used to extract dimensional features of the associated feature graph data and the interconnected graph data respectively, and the dimensional feature data is fused using the attention mechanism layer, and then the fused feature data is classified, so as to obtain the correlation probability prediction result of lncRNA and disease. This method effectively improves the prediction performance in identifying the association between lncRNA and disease pairs. Improves the accuracy and efficiency of the prediction results.

[0035] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present invention, and together with the specification are used to explain the principles of the present invention. Obviously, the accompanying drawings described below are only some embodiments of the present invention, and for those of ordinary skill in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0037] Figure 1 A schematic diagram schematically illustrates a method for predicting lncRNA markers based on a hierarchical graph neural network according to an exemplary embodiment of the present invention;

[0038] Figure 2 A schematic diagram schematically illustrates a framework of a model of an exemplary embodiment of the present invention;

[0039] Figure 3 The figure schematically shows the composition of an electronic device in an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0040] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present invention will be more comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0041] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0042] In related technologies, existing methods have shown inadequacy in dealing with the multifaceted nature of lncRNA interactions, which span different biological pathways and molecular mechanisms, making the prediction and validation of lncRNA-disease associations a daunting task. In addition, the heterogeneity of diseases themselves, with their different genetic and environmental bases, adds another layer of complexity to identifying the specific roles of lncRNAs in disease pathogenesis.

[0043] In view of the shortcomings and deficiencies of the prior art, this example embodiment provides a lncRNA marker prediction method based on a hierarchical graph neural network. The method can be executed on the terminal device side, or by the terminal device and the server side in collaboration. The user can initiate a data analysis request on the terminal device and send it to the server side, which performs the calculation and feeds back the results of the correlation analysis between ncRNA and the disease to the terminal device. Among them, the terminal device can be a computer, a mobile phone or other smart device.

[0044] refer to Figure 1 As shown, the prediction method may specifically include the following steps:

[0045] Step S11, constructing lncRNA similarity network and disease similarity network;

[0046] Step S12, extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network respectively; and constructing association feature graph data using the lncRNA feature data and the disease feature data; and

[0047] Step S13, constructing interconnection graph data based on the associated source data of lncRNA and disease;

[0048] Step S14, using a feature extraction model based on a hybrid graph neural network architecture, respectively extracting dimensional features from the associated feature graph data and the interconnected graph data to obtain first dimensional feature data corresponding to the associated feature graph data and second dimensional feature data corresponding to the interconnected graph data;

[0049] Step S15, using the attention mechanism layer to perform feature fusion processing on the first dimension feature data and the second dimension feature data to obtain fused feature data;

[0050] Step S16, using a multi-layer perceptron layer to classify the fused feature data to obtain the correlation probability between lncRNA and the disease.

[0051] Below, each step of the lncRNA marker prediction method based on a hierarchical graph neural network in this example implementation will be described in more detail with reference to the accompanying drawings and examples.

[0052] In step S11, a lncRNA similarity network and a disease similarity network are constructed.

[0053] Exemplarily, the constructing of the lncRNA similarity network includes: using the lncRNA data source, calculating first similarity data between lncRNAs based on a Gaussian kernel function, and constructing the ncRNA similarity network based on the first similarity data.

[0054] Specifically, the lncRNA data source can be a public data set, wherein the similarity between lncRNAs can be calculated using a Gaussian kernel function, and the similarity calculation results between two lncRNAs are used to construct an ncRNA similarity network.

[0055] The formula for the Gaussian kernel function includes:

[0056]

[0057] In a further embodiment, the method further includes: using the lncRNA data source and omics data to calculate the second similarity data based on the Jaccard similarity coefficient between the lncRNA data source and various omics data; wherein the omics data includes: at least one of the mRNA data set and the protein data set; and constructing the ncRNA similarity network based on the first similarity data and the second similarity data.

[0058] Specifically, considering the correlation between lncRNA and various omics data other than diseases, the similarity matrix between lncRNA and various omics data can also be calculated, including one or more of the mRNA data set and the protein data set.

[0059] Taking the relationship between lncRNA and mRNA as an example, Jaccard similarity can be used to measure the similarity with the mRNA dataset. The Jaccard similarity calculation formula can include:

[0060]

[0061] Exemplarily, the constructing of the disease similarity network includes: using a disease data source to calculate third similarity data between diseases based on the FNSemSim function; and constructing the disease similarity network based on the third similarity data.

[0062] Specifically, in order to evaluate the similarity between various diseases associated with lncRNA, the FNSemSim method can be used to calculate the similarity network of disease data. The FNSemSim method can integrate the functional and semantic similarities of diseases to calculate a comprehensive similarity score. After calculating the pairwise disease similarity, the minimum and maximum normalization method can be applied to appropriately scale these values.

[0063] In a further embodiment, the method further comprises:

[0064] Using the disease data source, the relevant metabolite data set, and the circular RNA data set, respectively, to calculate fourth similarity data based on the Jaccard similarity coefficient between the disease data source and the relevant metabolite data set, and fifth similarity data based on the Jaccard similarity coefficient between the disease data source and the circular RNA data set;

[0065] A disease similarity network is constructed based on the third similarity data, the fourth similarity data and the fifth similarity data.

[0066] Specifically, the Jaccard similarity coefficient can also be used to calculate the disease similarity between disease and disease-related metabolites, and between disease and circular RNA levels. These data can be used to construct a disease similarity network, thereby enriching the disease similarity network and ensuring the comprehensiveness of the data.

[0067] By integrating multiple biological data sets to construct a comprehensive similarity network of lncRNA and diseases, the correlation between lncRNA, disease, and other data can be reflected in subsequent calculation steps, thereby effectively improving the accuracy of the prediction results output by the model.

[0068] In step S12, corresponding lncRNA feature data and disease feature data are extracted from the lncRNA similarity network and the disease similarity network respectively; and association feature graph data are constructed using the lncRNA feature data and the disease feature data.

[0069] Exemplarily, extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network respectively includes:

[0070] Step S121, using the RWR algorithm to aggregate the network nodes in the lncRNA similarity network and the disease similarity network respectively, to generate corresponding lncRNA feature data and disease feature data;

[0071] Step S122, performing denoising processing on the lncRNA feature data and the disease feature data respectively through a denoising autoencoder to obtain the lncRNA feature data and the disease feature data after the noise is removed.

[0072] Specifically, after obtaining similarity network data from heterogeneous networks, feature learning of lncRNA and disease can be performed. For example, the RWR (Restart Random Walk Algorithm) method can be used to generate node embeddings within each similarity network, and the formula includes:

[0073] r=cWr+(1-c)e

[0074] Among them, the range of c is (0,1); W represents the transition probability matrix, e is the starting vector; if i is the starting node, then e[i] = 1, otherwise e[i] = 0, and r is a column vector.

[0075] The topological relationship between each disease or lncRNA and other nodes in the similarity network is used to derive the probability distribution through multiple iterations and convergence of RWR. By aggregating the probability distribution of all nodes, the representative information of a specific node is formed. Finally, the characteristic representation of the disease and lncRNA is obtained.

[0076] In addition, DAE (denoising autoencoder) can be used to denoise disease feature data and lncRNA feature data respectively for further feature extraction.

[0077] Exemplarily, the extracted feature data can be used to construct a lncRNA-disease association feature graph RFG. Specifically, the features of lncRNA and disease can be connected as the features of the association node. For example, the cosine similarity between the features can be calculated and applied to construct the lncRNA-disease association feature graph. For any lncRNA-disease relationship node, the cosine similarity between its feature vector and the feature vectors of other nodes is calculated. Node PN i and PN j The similarity between is defined as:

[0078]

[0079] Among them, F(PN i ) and F(PN j ) represent the node PN i and node PN j The associated features.

[0080] In step S13, based on the associated source data of lncRNA and disease, interconnection graph data is constructed.

[0081] Exemplarily, the source data of association between lncRNA and disease may include data pairs formed by multiple lncRNA and disease, which are used to describe any lncRNA and its related disease. For example, the source data of association between lncRNA and disease includes data pairs consisting of 1821 lncRNA and 185 diseases.

[0082] Specifically, the lncRNA-disease association can be defined as a node, and the bipartite graph of the lncRNA-disease association can be converted into an interconnected graph containing lncRNA-disease relationship nodes. Among them, the association node between lncRNAL and disease D can be defined as PN L,D .

[0083] When constructing the interconnected graph, if two nodes share a lncRNA or a disease, there is an edge between them. L,D and PN L*,D* , their connectivity can be defined as:

[0084]

[0085] In step S14, a feature extraction model based on a hybrid graph neural network architecture is used to perform dimensional feature extraction on the associated feature graph data and the interconnected graph data, respectively, to obtain first dimensional feature data corresponding to the associated feature graph data and second dimensional feature data corresponding to the interconnected graph data.

[0086] Exemplarily, the feature extraction model based on the hybrid graph neural network architecture includes: a first graph convolution layer, a graph attention layer, and a second graph convolution layer arranged in sequence; wherein the graph convolution network learns node representation by utilizing the topological structure of the graph; and the graph attention layer calculates node features by combining the attention mechanism.

[0087] Specifically, the feature extraction process can use a feature extraction model of a three-layer hybrid graph neural network architecture, which consists of two graph convolutional layers and one graph attention layer.

[0088] Among them, the graph convolutional network GCN aims to learn node representation by leveraging the topological structure of the graph. The GCN layer can be expressed as:

[0089]

[0090] Among them, H (l) is the node feature matrix of the lth layer, is the adjacency matrix of the graph with self-connection added, yes The degree matrix, W (l) is the weight matrix of the lth layer, and σ is the nonlinear activation function.

[0091] The formula of the graph attention layer GAT includes:

[0092]

[0093] in, and is the feature vector of nodes i and j in layer l, is the attention coefficient between nodes i and j, which determines the importance of the feature of node j to node i, W (l) is the weight matrix of the lth layer.

[0094] In order to effectively combine the advantages of GCN and GAT, the GCN-GAT-GCN structure is used for feature extraction. The formula is:

[0095]

[0096] Where A represents the adjacency matrix of the interconnection graph (CG) or the relevance feature graph (RFG), and H l Represents the output of layer l. The graph attention layer GAT calculates node features by incorporating an attention mechanism, which enables the model to focus on more important nodes in the neighborhood.

[0097] In step S15, the attention mechanism layer is used to perform feature fusion processing on the first dimension feature data and the second dimension feature data to obtain fused feature data.

[0098] Exemplarily, in order to further improve the interpretability of the model, an attention mechanism layer is used to fuse all representations of the interconnected graph CG and the associated feature graph RFG.

[0099] Node PN i Two representations of and Let the vector α i Represents their attention vector, which can be obtained by the attention function The formula is:

[0100]

[0101] in, and They are and The corresponding attention score.

[0102] For the definition of the attention function, first calculate Weight score The formula is:

[0103]

[0104] in, is a weight matrix, is a deviation vector, is the shared attention vector. l is the dimension of the attention layer. h is the PN learned from the convolutional graph learner i The dimension of the final representation of . Through the same operation, we can also get The value of node i

[0105] Then, the obtained weight value is normalized to the attention score through the softmax function. Therefore, the following equation can be used to calculate

[0106]

[0107] Similarly, we can calculate Therefore, PN i The final representation of and The weighted sum of is defined as:

[0108]

[0109] In step S16, the fused feature data is classified and processed using a multi-layer perceptron layer to obtain the correlation probability between lncRNA and the disease.

[0110] For example, a multi-layer perceptron layer can be used as a classification layer. Finally, a multi-layer perceptron MLP is used as a classification model to obtain the correlation probability between lncRNA and disease.

[0111] Exemplarily, in order to verify the predictive performance of this method in identifying the association between lncRNA and disease pairs, five-fold and ten-fold cross-validation techniques were used. In order to scientifically compare the performance of various models, this method used six evaluation indicators: AUC, AUPR, precision, recall, F1 score and Matthews correlation coefficient (MCC).

[0112] Reference Table 1 shows the evaluation results of the method of the present invention under five-fold cross validation and ten-fold cross validation, wherein the average AUC and AUPR are both around 0.97, and other evaluation scores also basically exceed 0.9, which proves the reliability of the method.

[0113] Table 1

[0114]

[0115]

[0116] Referring to Table 2, which is a comparison between the method of the present invention and other methods, it can be seen that except for the accuracy, other performance indicators of this method are better than those of other methods.

[0117] Table 2

[0118]

[0119] After comparing with other methods, this method also conducted ablation experiments and comparisons with different classification methods. In the ablation experiment, the different modules in the method were cross-validated with leave-one-out method to evaluate the contribution of different modules to the method. We also conducted experiments to see whether the method is best to use the multi-layer perceptron as a classifier, and compared it with the multi-layer perceptron by replacing other classifiers: decision tree, random forest, logistic regression, SVM, XGB and HGL.

[0120] Table 3

[0121]

[0122] Table 4 shows the verification of the classifier effect. It can be found that the multi-layer perceptron used in this method performs well in the final prediction task.

[0123] Table 4

[0124]

[0125] The method provided in the embodiment of the present invention is mainly divided into: construction of lncRNA similarity network, construction of disease similarity network, feature learning of lncRNA and disease, construction of lncRNA and disease association graph, and dimensional feature extraction of graph-based neural structure.

[0126] refer to Figure 2 As shown in the figure, firstly, the similarity network of lncRNA and disease is constructed, and the features of lncRNA and disease are obtained by using RWR and DAE. Then, the interconnected graph (CG) and the associated feature graph (RFG) of lncRNA and disease are constructed. Then, the feature extraction is performed using a three-layer hybrid graph neural network architecture, and the interpretability is enhanced by an attention mechanism. Finally, the results are predicted using a multi-layer perceptron. This method integrates multiple biological datasets to construct a comprehensive similarity network of lncRNA and disease, provides a powerful, scalable and interpretable framework to predict the association between lncRNA and disease, and uses a feature extraction model based on a hierarchical graph neural network to accurately extract features, which solves the limitations of existing technologies.

[0127] It should be noted that the above figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0128] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0129] Figure 3 A schematic diagram of an electronic device suitable for implementing an embodiment of the present invention is shown.

[0130] It should be noted that Figure 3 The electronic device 1000 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0131] like Figure 3 As shown, the electronic device 1000 includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage part 1008 to the random access memory (RAM) 1003. In the RAM 1003, various programs and data required for system operation are also stored. The CPU 1001, ROM 1002 and RAM 1003 are connected to each other through a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0132] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage section 1008 as needed.

[0133] In particular, according to an embodiment of the present invention, the process described below with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a storage medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part 1009, and / or installed from a removable medium 1011. When the computer program is executed by a central processing unit (CPU) 1001, various functions defined in the system of the present application are executed.

[0134] It should be noted that the storage medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any storage medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the storage medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0135] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0136] The units involved in the embodiments of the present invention may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, limit the units themselves.

[0137] It should be noted that, as another aspect, the present application also provides a storage medium, which can be included in an electronic device; or it can exist independently without being installed in the electronic device. The above storage medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method described in the following embodiments. For example, the electronic device can implement the following Figure 1 The individual steps of the method are shown.

[0138] In one embodiment, the present application provides a computer program product, including a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0139] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0140] Other embodiments of the invention will readily occur to those skilled in the art after considering the specification and practicing the invention herein. This application is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the claims.

[0141] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A lncRNA marker prediction method based on hierarchical graph neural network, characterized in that: The method comprises: Construct lncRNA similarity network and disease similarity network; Extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network respectively; and constructing association feature graph data using the lncRNA feature data and the disease feature data; and Based on the associated source data of lncRNA and disease, the interconnection graph data is constructed; Using a feature extraction model based on a hybrid graph neural network architecture, dimensional feature extraction is performed on the associated feature graph data and the interconnected graph data to obtain first dimensional feature data corresponding to the associated feature graph data and second dimensional feature data corresponding to the interconnected graph data; The attention mechanism layer is used to perform feature fusion processing on the first dimension feature data and the second dimension feature data to obtain fused feature data; The multi-layer perceptron layer is used to classify the fused feature data and obtain the correlation probability between lncRNA and disease.

2. The method according to claim 1, characterized in that The constructing of the lncRNA similarity network comprises: The lncRNA data source is used to calculate the first similarity data between lncRNAs based on the Gaussian kernel function, and the ncRNA similarity network is constructed based on the first similarity data.

3. The method according to claim 2, characterized in that The method further comprises: Using the lncRNA data source and the omics data, calculating the second similarity data based on the Jaccard similarity coefficient between the lncRNA data source and various omics data; wherein the omics data includes: at least one of an mRNA data set and a protein data set; A ncRNA similarity network is constructed based on the first similarity data and the second similarity data.

4. The method according to claim 1, characterized in that: The constructing of a disease similarity network comprises: The disease data source is used to calculate the third similarity data between the diseases based on the FNSemSim function; and a disease similarity network is constructed based on the third similarity data.

5. The method according to claim 1, characterized in that The method further comprises: Using the disease data source, the relevant metabolite data set, and the circular RNA data set, respectively, to calculate fourth similarity data based on the Jaccard similarity coefficient between the disease data source and the relevant metabolite data set, and fifth similarity data based on the Jaccard similarity coefficient between the disease data source and the circular RNA data set; A disease similarity network is constructed based on the third similarity data, the fourth similarity data and the fifth similarity data.

6. The method according to claim 1, characterized in that The extracting corresponding lncRNA feature data and disease feature data from the lncRNA similarity network and the disease similarity network respectively includes: The RWR algorithm is used to aggregate the network nodes in the lncRNA similarity network and the disease similarity network to generate the corresponding lncRNA feature data and disease feature data; The lncRNA feature data and disease feature data are denoised by a denoising autoencoder to obtain the lncRNA feature data and disease feature data after noise removal.

7. The method according to claim 1, characterized in that The feature extraction model based on the hybrid graph neural network architecture includes: a first graph convolution layer, a graph attention layer, and a second graph convolution layer arranged in sequence; Among them, the graph convolutional network learns node representation by utilizing the topological structure of the graph; the graph attention layer calculates node features by combining the attention mechanism.

8. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting lncRNA markers based on a hierarchical graph neural network as described in any one of claims 1 to 7 is implemented.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the lncRNA marker prediction method based on a hierarchical graph neural network described in any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the lncRNA marker prediction method based on hierarchical graph neural network described in any one of claims 1 to 7 by executing the executable instructions.

Images