Tissue gene expression map generation method and device

Abstract

The invention discloses a tissue gene expression map generation method and device, relates to the technical field of image processing, and mainly aims to solve the problem of poor generation effectiveness of an existing tissue gene expression map. Comprising the following steps: acquiring multi-modal graph structure data of a tissue gene; determining a weighted graph representation of the multi-modal graph structure data; on the basis of a neural network model after model training is completed, map generation is conducted on the weighted map representation, an expression map of the tissue gene is obtained, an attention mechanism is introduced into the neural network model, and the attention weight of each layer in the neural network model is used for updating topological structure information of each layer.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method and device for generating an expression map of tissue genes. Background Art

[0002] The study of complex tissues is not limited to the analysis of molecular characteristics of a large number of cells, but also requires an understanding of how the spatial environment affects the state or function of cells. The transcriptional state of cells can be regulated by the Gene Regulatory Network (GRN), which is a collection of regulatory relationships between transcription factors (TF) and their downstream target genes. In recent years, breakthroughs in spatially resolved transcriptomics (SRT) technology have made it possible to perform transcriptome feature analysis while retaining positional information, thus providing unprecedented opportunities to capture transcriptional states with spatial context.

[0003] At present, the existing SRT data spatial domain detection calculation methods can be roughly divided into two categories: probabilistic models and deep learning models. Among them, probabilistic methods include hidden Markov random fields (HMRF), BayesSpace and SpatialPCA, etc., which use probabilistic graph models to integrate spatial information, and deep learning methods include SpaGCN, STAGATE, DeepST and GraphST, etc., which use graph convolutional networks and their derivative methods to capture dependencies driven by spatial data. However, these two types of models deal with spatial domain modeling problems through different methods, focusing on statistical probability or the learning ability of neural networks to understand the spatial context embedded in the data, relying on predefined graph structures, and cannot fully capture the complex and nonlinear relationship between gene expression and spatial context, nor can they fully represent actual biological interactions, greatly reducing the integration effect of gene expression data and spatial information. Therefore, a method for generating an expression map of tissue genes is urgently needed to solve the above problems. Summary of the invention

[0004] In view of this, the present application provides a method and device for generating an expression map of a tissue gene, the main purpose of which is to solve the problem of poor effectiveness of the existing generation of expression maps of tissue genes.

[0005] According to one aspect of the present application, a method for generating an expression map of a tissue gene is provided, comprising:

[0006] Obtain multimodal graph structure data of tissue genes;

[0007] determining a weighted graph representation of the multimodal graph structure data;

[0008] Based on the neural network model that has completed model training, the weighted graph representation is generated to obtain the expression map of the tissue gene. The attention mechanism is introduced into the neural network model, and the attention weight of each layer in the neural network model is used to update the topological structure information of each layer.

[0009] Furthermore, before generating a graph for the weighted graph representation based on the neural network model that has completed model training to obtain the expression graph of the tissue gene, the method further includes:

[0010] Obtaining gene expression training samples, wherein the gene expression training samples include graph representation samples of labeled graph expression and edge expression;

[0011] Constructing a neural network model that introduces an attention mechanism, wherein the attention weights of each layer in the neural network model include a bias term, and the bias term is determined based on position encoding and edge samples;

[0012] The neural network model is trained based on the gene expression training samples to obtain a neural network model that has completed model training.

[0013] Furthermore, the method further comprises:

[0014] A gene singular matrix is ​​obtained, and an adjacency matrix of the gene singular matrix is ​​decomposed based on a singular value decomposition algorithm to generate a position code, wherein the adjacency matrix is ​​determined based on a gene site and a neighboring gene site.

[0015] Furthermore, before performing model training on the neural network model based on the gene expression training samples, the method further includes:

[0016] Obtaining the adjacency matrix;

[0017] The expression sample is enhanced by using the adjacency matrix to obtain an enhanced expression sample, so as to perform model training based on the enhanced expression sample.

[0018] Furthermore, after obtaining the expression profile of the tissue gene, the method further comprises:

[0019] Determining allocation probabilities for clustering based on preset distributions matching different clustering tasks, and determining a target allocation distribution of the allocation probabilities based on a normalization algorithm;

[0020] Determining a clustering loss function based on the allocation probability and the target allocation distribution, and constraining a clustering model based on the clustering loss function;

[0021] The expression map is clustered based on the constrained clustering model to obtain a map clustering result of the clustering task.

[0022] Further, the determining of the weighted graph representation of the multimodal graph structure data includes:

[0023] Parsing gene loci and neighbor gene loci in the multimodal graph structure data, and obtaining an adjacency matrix based on the spatial distance between the gene loci and the neighbor gene loci;

[0024] The weighted graph representation is determined based on the gene expression data in the multimodal graph structure data and the adjacency matrix.

[0025] Furthermore, the method further comprises:

[0026] constructing a neighbor network based on the transcriptional representation data in the expression map, and performing spatial recognition processing based on the neighbor network to identify the spatial domain of the tissue gene;

[0027] A denoising process is performed based on the topological structure data in the expression map, and specific identification is performed based on the denoised topological structure data to determine specific differentially expressed genes.

[0028] According to another aspect of the present application, a device for generating an expression map of a tissue gene is provided, comprising:

[0029] An acquisition module, used to acquire multimodal graph structure data of tissue genes;

[0030] A determination module, configured to determine a weighted graph representation of the multimodal graph structure data;

[0031] A generation module is used to generate a graph for the weighted graph representation based on a neural network model that has completed model training, so as to obtain an expression graph of the tissue gene. An attention mechanism is introduced into the neural network model, and the attention weights of each layer in the neural network model are used to update the topological structure information of each layer.

[0032] Furthermore, the device also includes: a construction module, a training module,

[0033] The acquisition module is further used to acquire gene expression training samples, wherein the gene expression training samples include graph representation samples of labeled graph expression and edge expression;

[0034] The construction module is used to construct a neural network model that introduces an attention mechanism, wherein the attention weights of each layer in the neural network model include a bias term, and the bias term is determined based on position encoding and edge samples;

[0035] The training module is used to perform model training on the neural network model based on the gene expression training samples to obtain a neural network model that has completed model training.

[0036] Furthermore, the generation module is also used to obtain a gene singular matrix, and decompose the adjacency matrix of the gene singular matrix based on a singular value decomposition algorithm to generate a position code, wherein the adjacency matrix is ​​determined based on gene sites and neighboring gene sites.

[0037] Further,

[0038] The acquisition module is further used to acquire the adjacency matrix; enhance the expression sample through the adjacency matrix to obtain the enhanced expression sample, so as to perform model training based on the enhanced expression sample.

[0039] Furthermore, the device also includes:

[0040] The determination module is also used to determine the allocation probability for clustering based on matching the preset distribution of different clustering tasks, and determine the target allocation distribution of the allocation probability based on a normalization algorithm; determine the clustering loss function based on the allocation probability and the target allocation distribution, and constrain the clustering model based on the clustering loss function; cluster the expression map based on the constrained clustering model to obtain the map clustering result of the clustering task.

[0041] Furthermore, the determination module is also used to parse the gene loci and neighbor gene loci in the multimodal graph structure data, and obtain an adjacency matrix based on the spatial distance between the gene loci and the neighbor gene loci; and determine the weighted graph representation based on the gene expression data in the multimodal graph structure data and the adjacency matrix.

[0042] Furthermore, the device also includes:

[0043] A processing module is used to construct a neighboring network based on the transcriptional characterization data in the expression map, and perform spatial recognition processing based on the neighboring network to identify the spatial domain of the tissue gene; perform denoising processing based on the topological structure data in the expression map, and perform specific recognition based on the denoised topological structure data to determine specific differential genes.

[0044] According to another aspect of the present application, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for generating an expression map of tissue genes.

[0045] According to another aspect of the present application, a terminal is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus;

[0046] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned method for generating an expression map of tissue genes.

[0047] By means of the above technical solution, the technical solution provided by the embodiment of the present application has at least the following advantages:

[0048] The present application provides a method and device for generating an expression map of a tissue gene. Compared with the prior art, the embodiment of the present application obtains multimodal graph structure data of the tissue gene; determines a weighted graph representation of the multimodal graph structure data; generates a graph for the weighted graph representation based on a neural network model that has completed model training, and obtains an expression map of the tissue gene. An attention mechanism is introduced into the neural network model. The attention weights of each layer in the neural network model are used to update the topological structure information of each layer, and a self-attention mechanism is combined with reinforcement to achieve the purpose of iteratively evolving the topological structure information represented by the graph and the transcription signal representation. By replacing graph convolution with global self-attention, global information and spatial local information can be integrated, thereby improving the detection capability of spatial functional domains and improving the integration effect of gene expression data and spatial information.

[0049] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present application. Also, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0051] Figure 1 A flow chart of a method for generating an expression profile of a tissue gene provided in an embodiment of the present application is shown;

[0052] Figure 2 A schematic diagram of a neural network model processing flow provided in an embodiment of the present application is shown;

[0053] Figure 3 A block diagram of a tissue gene expression profile generating device provided in an embodiment of the present application is shown;

[0054] Figure 4 A schematic diagram of the structure of a terminal provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0055] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0056] The present application embodiment provides a method for generating an expression map of a tissue gene, such as Figure 1 As shown, the method includes:

[0057] 101. Obtain multimodal graph structure data of tissue genes.

[0058] In the embodiment of the present application, the current execution subject, as the processing end for generating the expression map, can be a terminal device, or a cloud server, etc., in order to obtain the multimodal graph structure data of the tissue gene. Among them, the tissue gene can be the gene image of the tissue cell obtained by different species in the large-scale spatiotemporal omics data set through different dimensions, or it can be the gene image of the tissue cell collected in real time, and the embodiment of the present application is not specifically limited. The multimodal graph structure data includes the graph data of the characterization gene mechanism of different modes such as vector data, image data and spatial coordinate data, and the embodiment of the present application is not specifically limited.

[0059] 102. Determine a weighted graph representation of the multimodal graph structure data.

[0060] In the embodiment of the present application, after obtaining the multimodal graph structure data, in order to perform graph generation on it, the current execution end determines the weighted graph representation of the multimodal graph structure data, that is, the weighted graph representation includes the expression data of the tissue gene at the gene site or cell with the multimodal graph structure data and the adjacency matrix derived in space from different modalities, which can be expressed as G(X1,A), where X1∈R M×N , represented by the expression of M component genes in N spots / cells, A∈R N×N It is an adjacency matrix derived from spatial multimodal data, that is, the relationship between each spot and its spatial neighbors can be used to capture the inherent local structure of gene space in SRT data, which is not specifically limited in the embodiments of the present application.

[0061] 103. Generate a graph for the weighted graph representation based on the neural network model that has completed model training, and obtain an expression graph of the tissue gene.

[0062] In the embodiment of the present application, after obtaining the weighted graph representation, the weighted graph representation is input as a model input into the neural network model for graph generation. At this time, the attention mechanism is introduced into the neural network model, and the attention weights of each layer in the neural network model are used to update the topological structure information of each layer, wherein the topological structure information is used to characterize the structural relationship in the gene graph, and the embodiment of the present application does not make specific limitations. In addition, the attention weights in the self-attention mechanism are used to iteratively update the topological structure information of the graph representation between each layer, which can strengthen the self-attention module and realize efficient learning and updating of the expression graph of the entire model.

[0063] It should be noted that the neural network model in the embodiment of the present application is used as a spatial perception graph transformer to transform the weighted graph representation. After the i+1th layer of transformation, the expression map of the tissue gene is obtained. The expression map is represented by G(H L ,E L ), H L To characterize gene transcription signals, E L It is gene topological structure information, which is not specifically limited in the embodiments of the present application.

[0064] In another embodiment of the present application, for further definition and explanation, the step generates a graph based on the neural network model that has completed model training to generate the weighted graph representation, and before obtaining the expression graph of the tissue gene, the method further includes:

[0065] Obtain gene expression training samples;

[0066] Build a neural network model that introduces an attention mechanism;

[0067] The neural network model is trained based on the gene expression training samples to obtain a neural network model that has completed model training.

[0068] In order to realize the use of neural networks with attention mechanisms for graph generation, so as to accurately construct gene representations with clearer structures, the current execution end pre-acquires gene expression training samples. At this time, the gene expression training samples include graph representation samples of labeled graph expression and edge expression, which can be obtained from the SRT database. The SRT database includes data such as STOmics, SOAR, SpatialDB, CROST, 10x Genomics website (available through 10x Genomics) and Census (available through Cellxgene). For example, it contains 96,700,729 cells / points from 7,367 tissue sections, covering 365 tissue cells, including lungs, skin, brain, liver, kidneys, spinal cord and embryos, and covers normal tissues and various diseases, such as pancreatic ductal adenocarcinoma, amyotrophic lateral sclerosis, non-small cell lung cancer and hepatocellular carcinoma, which are not specifically limited in the embodiments of this application. The embodiments of this application are not specifically limited. Among them, the graph representation sample includes a weighted graph sample as a model input, represented as G(X0,A).

[0069] It should be noted that if Figure 2 As shown in the figure, after the current execution end builds a neural network model with an attention mechanism, the neural network model is trained based on the gene expression training samples. During the training process, the attention weights of each layer in the neural network model contain bias terms. When each layer is trained, the output of the previous layer is expressed as a graph H. i and edge expression E i , as the input of this layer to train this layer. Among them, through the structural reinforcement self-attention module, H i and E i To update, H i It is expressed as:

[0070]

[0071] E i It is expressed as:

[0072] Among them, L is the number of layers of the neural network, e represents the update of the point pair topology structure, O is the output of each layer of the neural network, h represents the update of the transcriptome signal, Γ is the number of multi-head attention, V is the value matrix of the self-attention module, S is the attention weight of the self-attention module, and concact(*) is the row-wise splicing. At this time, the low-dimensional X1 of gene expression and the relationship matrix A between each point and its spatial neighbors are represented as H 0 and E 0 Then, the application of feed-forward sublayer (FFN) and layer normalization (LN) in the neural network model obtains the final graph representation as follows:

[0073]

[0074] In addition, in the embodiment of the present application, in order to facilitate the integration of global information and spatial local information, an attention machine architecture is introduced into the neural network model, that is, the attention process in the i-th layer and the t-th attention head is expressed as:

[0075]

[0076] in, are the query, key, and value matrices of the t-th head in layer i, respectively. The dimensions of query and key are d i , operator Θ is element-wise multiplication. It is a bias term added to the attention weight, which is determined based on the position encoding and edge samples so that the spatial information affects the global aggregation process. The sigmoid function σ(·) is used to gate the spatial information before aggregation, controlling the flow of information between sites. At the same time, the scaled dot product is constrained by the clip(·) operation to improve numerical stability. In addition, P S,E and P S,G is the learned projection matrix.

[0077] In another embodiment of the present application, for further definition and explanation, the steps further include:

[0078] A gene singular matrix is ​​obtained, and an adjacency matrix of the gene singular matrix is ​​decomposed based on a singular value decomposition algorithm to generate a position code.

[0079] In order to implement the position code P as the input of the attention mechanism and update the attention weights in each attention, the current execution end obtains the gene singular matrix, represented as U, V, U, V ∈ R M×r , respectively containing the first r left and right singular matrices, and then, based on the singular value decomposition algorithm (Singular Value Decomposition), the adjacency matrix A of the gene singular matrix is ​​decomposed to generate the position code, which is expressed as follows:

[0080]

[0081] Among them, Σ∈R γ×γ is a diagonal matrix, || represents column concatenation, W PE ∈R 2γ×N is the learned projection matrix, P∈R M×N Encode the position.

[0082] In another embodiment of the present application, for further definition and explanation, before the step of training the neural network model based on the gene expression training sample, the method further includes:

[0083] Obtaining the adjacency matrix;

[0084] The expression sample is enhanced by using the adjacency matrix to obtain an enhanced expression sample, so as to perform model training based on the enhanced expression sample.

[0085] In order to enhance the gene expression of each gene site and thus improve the accuracy of expression map generation, the current execution end generates an enhanced spatial expression. Specifically, the adjacency matrix A is first obtained, and the expression X1=X0+αX0A is enhanced, where α is an adjustable parameter for controlling the influence of spatial neighborhood similarity on spatial domain recognition. The embodiment of the present application is configured based on demand without specific limitation.

[0086] In another embodiment of the present application, for further definition and explanation, after obtaining the expression profile of the tissue gene, the method further comprises:

[0087] Determining allocation probabilities for clustering based on preset distributions matching different clustering tasks, and determining a target allocation distribution of the allocation probabilities based on a normalization algorithm;

[0088] Determining a clustering loss function based on the allocation probability and the target allocation distribution, and constraining a clustering model based on the clustering loss function;

[0089] The expression map is clustered based on the constrained clustering model to obtain a map clustering result of the clustering task.

[0090] In order to make the obtained expression map have stronger clustering performance and thus improve the applicability of different clustering tasks, after the current execution end obtains the expression map, it first determines the allocation probability for clustering based on the preset distribution matching different clustering tasks, and determines the target allocation distribution of the allocation probability based on the normalization algorithm. Among them, the preset distribution is preferably the Student's t distribution, so as to measure the probability of a point i being assigned to clustering task j through the Student's t distribution, and then based on the combined expression map H L Calculate the allocation probability, the calculation formula is:

[0091]

[0092] Among them, μ is the cluster center determined by the K-means clustering algorithm. H L The i-th column of ij] represents the distribution probability of all tissue genes, and ρ is the degree of freedom of Student's t distribution. Furthermore, the current execution end determines the clustering loss function based on the distribution probability and the target distribution distribution. At this time, the target distribution distribution is expressed as wheresj ＝ ∑ i q ij ; where each q ij The square of is normalized to generate the target distribution P, j' is the center point of each cluster. In addition, in the embodiment of the present application, based on minimizing the Kullback-Leibler (KL) divergence between the Q and P distributions, the confidence of the distribution is increased and the clustering loss function is determined, which is expressed as At this time, the loss function enhances the accuracy and stability of clustering by minimizing the difference between distributions Q and P, which helps to improve the performance of the clustering task. Therefore, the clustering model is constrained based on the clustering loss function, and the expression map is clustered based on the constrained clustering model to obtain the map clustering result of the clustering task.

[0093] In another embodiment of the present application, for further definition and explanation, the step of determining the weighted graph representation of the multimodal graph structure data includes:

[0094] Parsing gene loci and neighbor gene loci in the multimodal graph structure data, and obtaining an adjacency matrix based on the spatial distance between the gene loci and the neighbor gene loci;

[0095] The weighted graph representation is determined based on the gene expression data in the multimodal graph structure data and the adjacency matrix.

[0096] In order to better learn the gene structure, the weighted graph is represented as G(X0,A). At this time, the current execution end can calculate the Euclidean distance between each pair of gene spots based on the spatial coordinates and the corresponding histological image information, and select the k nearest spatial neighbors for each point. Then, based on the principal component analysis kernel PCA of gene expression, the cosine distance between the gene spot and the neighboring gene spot is calculated, and its exponential is converted into the similarity matrix A. The specific formula is expressed as: Among them, the matrix U∈R 15×NIt is a low-dimensional matrix composed of 15 principal components PCs. The matrix D represents the cosine distance of these adjacent spots as the spatial distance. In addition, the 15 principal components PCs are the first 15 principal components selected according to the corresponding eigenvalues ​​after X1 performs PCA dimensionality reduction. At this time, principal component analysis (PCA for short) is a commonly used data dimensionality reduction technology, which is used to convert high-dimensional data sets into low-dimensional space. After principal component analysis, multiple principal component PCs can be obtained.

[0097] In another embodiment of the present application, for further definition and explanation, the steps further include:

[0098] constructing a neighbor network based on the transcriptional representation data in the expression map, and performing spatial recognition processing based on the neighbor network to identify the spatial domain of the tissue gene;

[0099] De-noising is performed based on the topological structure data in the expression map, and specific identification is performed based on the de-noised topological structure data to determine specific differentially expressed genes.

[0100] In order to improve the effectiveness and accuracy of generating expression maps, the current execution end can also denoise the expression map. Specifically, first, based on the transcriptional representation data H in the expression map L Construct a neighbor network, that is, an adjacency matrix consisting of each spot and its 15 closest spots in expression representation, and perform spatial recognition processing based on this neighbor network. The Leiden algorithm can be used for calculation to identify the spatial domain of tissue genes. At the same time, based on the topological structure data E in the expression map L Perform denoising to obtain the denoised expression spectrum To enhance the spatial expression pattern and domain specificity, and to perform specific identification based on the denoised topological structure data to determine specific differentially expressed genes, which is not specifically limited in the embodiments of the present application.

[0101] The embodiment of the present application provides a method for generating an expression map of a tissue gene. Compared with the prior art, the embodiment of the present application obtains multimodal graph structure data of the tissue gene; determines a weighted graph representation of the multimodal graph structure data; generates a graph for the weighted graph representation based on a neural network model that has completed model training, and obtains an expression map of the tissue gene. An attention mechanism is introduced into the neural network model, and the attention weights of each layer in the neural network model are used to update the topological structure information of each layer, thereby realizing a combined and enhanced self-attention mechanism, achieving the purpose of iteratively evolving the topological structure information represented by the graph and the transcriptional signal representation, and by replacing graph convolution with global self-attention, global information and spatial local information can be integrated, thereby improving the detection capability of the spatial functional domain and improving the integration effect of gene expression data and spatial information.

[0102] Furthermore, as a response to the above Figure 1 The present invention provides a method for generating a tissue gene expression map. Figure 3 As shown, the device comprises:

[0103] An acquisition module 21 is used to acquire multimodal graph structure data of tissue genes;

[0104] A determination module 22, configured to determine a weighted graph representation of the multimodal graph structure data;

[0105] The generation module 23 is used to generate a graph for the weighted graph representation based on the neural network model that has completed model training, so as to obtain the expression graph of the tissue gene. The attention mechanism is introduced into the neural network model, and the attention weights of each layer in the neural network model are used to update the topological structure information of each layer.

[0106] Furthermore, the device also includes: a construction module, a training module,

[0107] The acquisition module is further used to acquire gene expression training samples, wherein the gene expression training samples include graph representation samples of labeled graph expression and edge expression;

[0108] The construction module is used to construct a neural network model that introduces an attention mechanism, wherein the attention weights of each layer in the neural network model include a bias term, and the bias term is determined based on position encoding and edge samples;

[0109] The training module is used to perform model training on the neural network model based on the gene expression training samples to obtain a neural network model that has completed model training.

[0110] Furthermore, the generation module is also used to obtain a gene singular matrix, and decompose the adjacency matrix of the gene singular matrix based on a singular value decomposition algorithm to generate a position code, wherein the adjacency matrix is ​​determined based on gene sites and neighboring gene sites.

[0111] Further,

[0112] The acquisition module is further used to acquire the adjacency matrix; enhance the expression sample through the adjacency matrix to obtain the enhanced expression sample, so as to perform model training based on the enhanced expression sample.

[0113] Furthermore, the device also includes:

[0114] The determination module is also used to determine the allocation probability for clustering based on matching the preset distribution of different clustering tasks, and determine the target allocation distribution of the allocation probability based on a normalization algorithm; determine the clustering loss function based on the allocation probability and the target allocation distribution, and constrain the clustering model based on the clustering loss function; cluster the expression map based on the constrained clustering model to obtain the map clustering result of the clustering task.

[0115] Furthermore, the determination module is also used to parse the gene loci and neighbor gene loci in the multimodal graph structure data, and obtain an adjacency matrix based on the spatial distance between the gene loci and the neighbor gene loci; and determine the weighted graph representation based on the gene expression data in the multimodal graph structure data and the adjacency matrix.

[0116] Furthermore, the device also includes:

[0117] A processing module is used to construct a neighboring network based on the transcriptional characterization data in the expression map, and perform spatial recognition processing based on the neighboring network to identify the spatial domain of the tissue gene; perform denoising processing based on the topological structure data in the expression map, and perform specific recognition based on the denoised topological structure data to determine specific differential genes.

[0118] The embodiment of the present application provides a device for generating an expression map of tissue genes. Compared with the prior art, the embodiment of the present application obtains multimodal graph structure data of tissue genes; determines a weighted graph representation of the multimodal graph structure data; generates a graph for the weighted graph representation based on a neural network model that has completed model training, and obtains an expression map of the tissue genes. An attention mechanism is introduced into the neural network model. The attention weights of each layer in the neural network model are used to update the topological structure information of each layer, thereby realizing a combined and enhanced self-attention mechanism to achieve the purpose of iteratively evolving the topological structure information represented by the graph and the transcriptional signal representation. By replacing graph convolution with global self-attention, global information and spatial local information can be integrated, thereby improving the detection capability of spatial functional domains and improving the integration effect of gene expression data and spatial information.

[0119] According to one embodiment of the present application, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for generating an expression map of a tissue gene in any of the above method embodiments.

[0120] Figure 4 A schematic diagram of the structure of a terminal provided according to an embodiment of the present application is shown, and the specific embodiment of the present application does not limit the specific implementation of the terminal.

[0121] like Figure 4 As shown, the terminal may include: a processor (processor) 302 , a communication interface (Communications Interface) 304 , a memory (memory) 306 , and a communication bus 308 .

[0122] The processor 302 , the communication interface 304 , and the memory 306 communicate with each other via the communication bus 308 .

[0123] The communication interface 304 is used to communicate with other devices such as clients or other servers.

[0124] The processor 302 is used to execute the program 310, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for generating the expression map of tissue genes.

[0125] Specifically, the program 310 may include program codes, which include computer operation instructions.

[0126] The processor 302 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0127] The memory 306 is used to store the program 310. The memory 306 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0128] The program 310 may be specifically configured to enable the processor 302 to perform the following operations:

[0129] Obtain multimodal graph structure data of tissue genes;

[0130] determining a weighted graph representation of the multimodal graph structure data;

[0131] Based on the neural network model that has completed model training, the weighted graph representation is generated to obtain the expression map of the tissue gene. The attention mechanism is introduced into the neural network model, and the attention weight of each layer in the neural network model is used to update the topological structure information of each layer.

[0132] Obviously, those skilled in the art should understand that the above modules or steps of the present application can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present application is not limited to any specific combination of hardware and software.

[0133] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for generating an expression profile of a tissue gene, characterized in that: include: Obtain multimodal graph structure data of tissue genes; determining a weighted graph representation of the multimodal graph structure data; Based on the neural network model that has completed z-model training, the weighted graph representation is generated to obtain the expression map of the tissue gene. The attention mechanism is introduced into the neural network model, and the attention weight of each layer in the neural network model is used to update the topological structure information of each layer.

2. The method according to claim 1, characterized in that Before generating a graph based on the weighted graph representation using the neural network model for which model training has been completed to obtain the expression graph of the tissue gene, the method further comprises: Obtaining gene expression training samples, wherein the gene expression training samples include graph representation samples of labeled graph expression and edge expression; Constructing a neural network model that introduces an attention mechanism, wherein the attention weights of each layer in the neural network model include a bias term, and the bias term is determined based on position encoding and edge samples; The neural network model is trained based on the gene expression training samples to obtain a neural network model that has completed model training.

3. The method according to claim 2, characterized in that The method further comprises: A gene singular matrix is ​​obtained, and an adjacency matrix of the gene singular matrix is ​​decomposed based on a singular value decomposition algorithm to generate a position code, wherein the adjacency matrix is ​​determined based on a gene site and a neighboring gene site.

4. The method according to claim 2, characterized in that: Before performing model training on the neural network model based on the gene expression training samples, the method further includes: Obtaining the adjacency matrix; The expression sample is enhanced by using the adjacency matrix to obtain an enhanced expression sample, so as to perform model training based on the enhanced expression sample.

5. The method according to claim 1, characterized in that After obtaining the expression profile of the tissue gene, the method further comprises: Determining allocation probabilities for clustering based on preset distributions matching different clustering tasks, and determining a target allocation distribution of the allocation probabilities based on a normalization algorithm; Determining a clustering loss function based on the allocation probability and the target allocation distribution, and constraining a clustering model based on the clustering loss function; The expression map is clustered based on the constrained clustering model to obtain a map clustering result of the clustering task.

6. The method according to claim 1, characterized in that Determining the weighted graph representation of the multimodal graph structure data includes: Parsing gene loci and neighbor gene loci in the multimodal graph structure data, and obtaining an adjacency matrix based on the spatial distance between the gene loci and the neighbor gene loci; The weighted graph representation is determined based on the gene expression data in the multimodal graph structure data and the adjacency matrix.

7. The method according to any one of claims 1 to 6, characterized in that: The method further comprises: constructing a neighbor network based on the transcriptional representation data in the expression map, and performing spatial recognition processing based on the neighbor network to identify the spatial domain of the tissue gene; De-noising is performed based on the topological structure data in the expression map, and specific identification is performed based on the de-noised topological structure data to determine specific differentially expressed genes.

8. A device for generating an expression profile of a tissue gene, characterized in that: include: An acquisition module, used to acquire multimodal graph structure data of tissue genes; A determination module, configured to determine a weighted graph representation of the multimodal graph structure data; A generation module is used to generate a graph for the weighted graph representation based on a neural network model that has completed model training, so as to obtain an expression graph of the tissue gene. An attention mechanism is introduced into the neural network model, and the attention weights of each layer in the neural network model are used to update the topological structure information of each layer.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

10. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method of claim 1.

Images