Inter-cellular interaction reconstruction method and system based on graph convolution and long short-term memory attention network

Abstract

The present application relates to a kind of intercellular interaction reconstruction method and system based on graph convolution and long short-term memory attention network.The method comprises the following steps: collecting spatial transcriptome data under single cell or subcellular resolution, obtaining the gene expression matrix corresponding to each cell;Determine the spatial proximity between each cell, determine whether there is interaction between cells, thereby constructing cell graph;Using graph convolution and long short-term memory attention network combination, form feature extraction network to extract features from cell graph, obtain latent feature representation;Using inner product to decode latent feature representation, generate new adjacency matrix, obtain the intercellular interaction after reconstruction.By introducing the combination of graph convolution and long short-term memory attention network, the spatial correlation in cell graph structure can be captured while learning intercellular interaction, and the intercellular interaction is reconstructed, so that the intercellular interaction and the organization homeostasis of organism can be better understood.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cell biological recognition, in particular to a cell-cell interaction reconstruction method and system based on graph convolution and long short-term memory attention network. BACKGROUND

[0002] In biological systems, cells directly or indirectly transmit information to each other by secreting signal molecules, and cell-cell interactions can be short-distance contact-dependent interactions or can be achieved through secreted signal molecules. Multicellular life depends on the coordination of cell activities, and the coordination of cell activities depends on the cell-cell interactions between different cell types and tissues of an organism, which can coordinate the development, balance of the organism and to some extent affect the physiological functions between cell tissues. In addition, intercellular regulation can maintain the normal function of tissues and organs, and their disorder can lead to a variety of diseases, including cancer, autoimmune diseases and infections. Therefore, identifying cell-cell interactions is crucial to understanding the role of individual cells in maintaining tissue homeostasis and responding to the microenvironment. The study of cell-cell interactions is highly dependent on spatial location information of cells, and it also requires high gene expression in individual cells. With the continuous development of spatially resolved transcriptome analysis technology, current studies can measure gene expression profiles at single-cell or subcellular resolution while retaining spatial location information of cells.

[0003] Traditional cell-cell interaction prediction methods are often based on network, machine learning method, and spatial information method. Among them, the network-based cell-cell interaction prediction method mainly studies the potential cell-cell interaction by calculating the interaction score of the ligand-receptor pair of two cell types; the machine learning method mainly identifies cell types by clustering, calculates the ligand-receptor pair score of two cell types, and infers the cell-cell interaction according to the score; the spatial information method mainly characterizes the spatial location information and spatially adjacent cell clusters, and discovers the interaction between different cell types according to single-cell transcriptome data, spatial transcriptome data, etc. However, the network-based and machine learning network do not consider the overall location distribution of cells in the tissue, which is prone to false interaction phenomenon; the spatial information method does not consider the time dependence of cell-cell interaction, and the effect of inferring cell-cell interaction needs to be improved.

[0004] Therefore, in the traditional method of obtaining cell-cell interaction, the time dependence of cell-cell interaction and the overall location distribution of cells in the tissue are not considered, and the effect of inferring cell-cell interaction is not good. SUMMARY

[0005] Based on this, in order to solve the above technical problems, a cell interaction reconstruction method and system based on graph convolution and long short-term memory attention network are provided, which can better capture spatial correlation and reconstruct cell interaction, better understand cell-cell interaction and study the organization homeostasis of organisms.

[0006] A cell interaction reconstruction method based on graph convolution and long short-term memory attention network, the method comprising:

[0007] Collecting spatial transcriptome data at single cell or subcellular resolution, and converting the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell;

[0008] Determining the spatial proximity between each cell according to the spatial transcriptome data, and judging whether there is interaction between cells based on the spatial proximity to obtain a judgment result;

[0009] Taking each cell as a vertex of a graph structure, if the judgment result is that there is interaction between cells, constructing an edge of the graph structure for the two cells with interaction according to the spatial connection network, taking a matrix describing the cells and the edge as an adjacency matrix, taking the gene expression matrix of each cell as a graph node feature of the graph structure to obtain a node feature matrix, and constructing a cell graph;

[0010] Combining graph convolution and long short-term memory attention network to form a feature extraction network, and extracting features of the cell graph through the feature extraction network to obtain a latent feature representation;

[0011] Decoding the latent feature representation using inner product to generate a new adjacency matrix, and obtaining the reconstructed cell interaction.

[0012] In one embodiment, converting the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell comprises:

[0013] Finding a fuzzy-defined cell in the spatial transcriptome data, and removing the fuzzy-defined cell from the spatial transcriptome data to obtain updated spatial transcriptome data;

[0014] Converting the updated spatial transcriptome data to obtain a gene expression matrix corresponding to each cell.

[0015] In one embodiment, after converting the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell, the method further comprises:

[0016] Performing a logarithmic change normalization operation on each gene expression matrix.

[0017] In one of the embodiments, a cell graph is constructed, including:

[0018] A vertex cell as a vertex of a graph structure is determined, and a target edge of the cell graph structure is found which has an interaction with the vertex cell;

[0019] A matrix describing the vertex cell and the target edge is taken as a target adjacency matrix, and a target node feature matrix of the graph structure is determined according to a gene expression matrix corresponding to the vertex cell;

[0020] The vertex, the target edge, the target adjacency matrix, and the target node feature matrix are combined to construct a cell graph.

[0021] In one of the embodiments, a graph convolution and a long short-term memory attention network are combined to form a feature extraction network, including:

[0022] The graph convolution and the long short-term memory attention network are combined in series to form the feature extraction network.

[0023] In one of the embodiments, the long short-term memory attention network includes a long short-term memory network, an activation function, an attention mechanism, and a skip connection.

[0024] The long short-term memory network is used to learn the time dependence of the interaction between cells; the attention mechanism is used to enhance the feature representation of the network; and the skip connection is used to add the output feature of the long short-term memory network to the output feature of the attention mechanism.

[0025] The graph convolution is used to capture the spatial correlation of the graph structure and learn the relationship between cells.

[0026] In one of the embodiments, the feature extraction network is used to extract features of the cell graph to obtain a latent feature representation, including:

[0027] The cell graph is input into the feature extraction network, the feature matrix in the cell graph is optimized by the long short-term memory attention network in the feature extraction network, and an optimized feature matrix is obtained.

[0028] The optimized feature matrix and the adjacency matrix are input into the graph convolution network, the spatial correlation of the graph structure is captured by the graph convolution network, and a latent feature representation is obtained according to the spatial correlation of the graph structure.

[0029] In one of the embodiments, the latent feature representation is obtained according to the spatial correlation of the graph structure, including:

[0030] The spatial correlation of the graph structure is processed by the activation function, added to the input feature matrix using a skip connection, and captures different levels of features;

[0031] The different levels of features are input again into the long short-term memory attention network and the graph convolution network for encoding optimization, and are processed by regularization to obtain latent feature representation.

[0032] A cell-cell interaction reconstruction system based on graph convolution and long short-term memory attention network, the system comprises:

[0033] A data collection module is configured to collect spatial transcriptome data at single-cell or sub-cellular resolution, and convert the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell;

[0034] An interaction judgment module is configured to determine the spatial proximity between cells according to the spatial transcriptome data, and determine whether there is interaction between cells based on the spatial proximity to obtain a judgment result;

[0035] A cell graph construction module is configured to take each cell as a vertex of a graph structure, and if the judgment result is that there is interaction between cells, construct a graph structure edge for the two cells with interaction according to a spatial connection network, take a matrix describing the cells and edges as an adjacency matrix, take the gene expression matrix of each cell as a graph node feature of the graph structure to obtain a node feature matrix, and construct a cell graph;

[0036] A feature extraction module is configured to combine graph convolution and long short-term memory attention network to form a feature extraction network, and extract features of the cell graph through the feature extraction network to obtain latent feature representation;

[0037] A cell-cell interaction reconstruction module is configured to decode the latent feature representation using inner product to generate a new adjacency matrix, and obtain reconstructed cell-cell interaction.

[0038] The above cell-cell interaction reconstruction method and system based on graph convolution and long short-term memory attention network can capture spatial correlation in the cell graph structure while learning cell-cell interaction by introducing graph convolution and long short-term memory attention network combination, thereby training the ability to extract and optimize features, improving overall computing efficiency and computing accuracy, better capturing spatial correlation, reconstructing cell-cell interaction, and better understanding cell-cell interaction and studying the homeostasis of biological organisms. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1An application environment diagram of an intercellular interaction reconstruction method based on a graph convolution and a long short-term memory attention network in an embodiment;

[0040] Figure 2 A flowchart of an intercellular interaction reconstruction method based on a graph convolution and a long short-term memory attention network in an embodiment;

[0041] Figure 3 A schematic diagram of a feature extraction enhancement network framework in an embodiment;

[0042] Figure 4 A structural block diagram of an intercellular interaction reconstruction system based on a graph convolution and a long short-term memory attention network in an embodiment;

[0043] Figure 5 An internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0044] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0045] The intercellular interaction reconstruction method based on a graph convolution and a long short-term memory attention network provided by the embodiments of the present application can be applied in an application environment as shown in Figure 1 . As shown in Figure 1 , the application environment includes a computer device 110. The computer device 110 can collect spatial transcriptome data at single-cell or subcellular resolution, convert the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell, determine the spatial proximity between each cell according to the spatial transcriptome data, and determine whether there is an interaction between the cells based on the spatial proximity to obtain a determination result. The computer device 110 can take each cell as a vertex of a graph structure, if the determination result is that there is an interaction between the cells, construct an edge of the graph structure for the two cells with the interaction according to a spatial connection network, take a matrix describing the cells and the edges as an adjacency matrix, take the gene expression matrix of each cell as a graph node feature of the graph structure, obtain a node feature matrix, and construct a cell graph. The computer device 110 can combine a graph convolution and a long short-term memory attention network to form a feature extraction network, and perform feature extraction on the cell graph through the feature extraction network to obtain a latent feature representation. The computer device 110 can use an inner product to decode the latent feature representation to generate a new adjacency matrix, and obtain a reconstructed intercellular interaction. The computer device 110 can be, but is not limited to, various personal computers, notebook computers, smart phones, robots, unmanned aerial vehicle devices, and the like.

[0046] In one embodiment, as shown in Figure 2 A method for intercellular interaction reconstruction based on graph convolution and long short-term memory attention network is provided, comprising the following steps:

[0047] In step 202, spatial transcriptome data at single cell or subcellular resolution is collected, and gene expression matrices corresponding to respective cells are obtained by conversion based on the spatial transcriptome data.

[0048] The computer device can collect and obtain data. Specifically, in the present embodiment, the computer device can collect spatial transcriptome data at single cell or subcellular resolution.

[0049] In one embodiment, the method for intercellular interaction reconstruction based on graph convolution and long short-term memory attention network can further comprise a data processing process, and the specific process comprises: finding ambiguously defined cells in the spatial transcriptome data, and removing the ambiguously defined cells from the spatial transcriptome data to obtain updated spatial transcriptome data; and converting the updated spatial transcriptome data to obtain gene expression matrices corresponding to respective cells.

[0050] In one embodiment, the computer device can perform logarithmic change normalization operation on each gene expression matrix.

[0051] Specifically, in the present embodiment, the computer device can collect a spatial transcriptome dataset at single cell or subcellular resolution, remove ambiguously defined cells, and perform logarithmic change normalization operation on the gene expression matrices.

[0052] In step 204, spatial proximity between respective cells is determined according to the spatial transcriptome data, and whether there is interaction between the cells is judged based on the spatial proximity, to obtain a judgment result.

[0053] The computer device can determine whether there is interaction between the cells according to the spatial proximity between the cells.

[0054] In step 206, each cell is taken as a vertex of a graph structure, and if the judgment result is that there is interaction between the cells, an edge of the graph structure is constructed for the two cells with interaction according to a spatial connection network, a matrix describing the cells and the edge is taken as an adjacency matrix, gene expression matrices of the respective cells are taken as graph node features of the graph structure, a node feature matrix is obtained, and a cell graph is constructed.

[0055] The computer device can regard each cell as a vertex of a graph structure, and there is an edge between the two cells if there is an interaction, and there is no edge between the two cells if there is no interaction. The edges of the graph structure are constructed according to the spatial connection network. The matrix describing the cells and the edges is called an adjacency matrix. The gene expression matrix of each cell in the graph is regarded as a graph node feature. The adjacency matrix and the graph node feature matrix together construct a cell-level graph structure, that is, a cell graph.

[0056] Specifically, in one embodiment, the cell interaction reconstruction method based on the graph convolution and long short-term memory attention network can further include a process of constructing a cell graph, and the specific process includes: determining a vertex cell as a vertex of a graph structure, and finding a cell having an interaction with the vertex cell to construct a target edge of the graph structure; regarding a matrix describing the vertex cell and the target edge as a target adjacency matrix, and determining a target node feature matrix of the graph structure according to a gene expression matrix corresponding to the vertex cell; and combining the vertex, the target edge, the target adjacency matrix, and the target node feature matrix to construct a cell graph.

[0057] In step 208, the graph convolution and long short-term memory attention network are combined to form a feature extraction network, and the cell graph is feature-extracted through the feature extraction network to obtain a latent feature representation.

[0058] In one embodiment, the computer device can combine the graph convolution and long short-term memory attention network in series to form a feature extraction network. The feature extraction network can perform feature extraction and feature enhancement.

[0059] In this embodiment, the computer device can construct a graph convolution and long short-term memory attention network to perform feature extraction on the constructed cell graph. The feature extraction network used is a combination of a long short-term memory attention network (module) and a graph convolution network in series.

[0060] In one embodiment, the long short-term memory attention network includes a long short-term memory network, an activation function, an attention mechanism, and a skip connection. The long short-term memory network is used to learn the time dependence of the interaction between cells. The attention mechanism is used to enhance the feature representation of the network. The skip connection is used to add the output features of the long short-term memory network and the output features of the attention mechanism. The graph convolution is used to capture the spatial correlation of the graph structure and learn the relationship between the cells.

[0061] The long short-term memory attention module and the graph convolution network are combined in series, and the network framework diagram of this part is as follows: Figure 3The feature matrix X and the adjacency matrix A are input into a feature extraction network formed by a combination of a graph convolution and a long short-term memory attention network, first processed in a linear layer, then passed through a long short-term memory attention module, and then processed by a graph convolution and an activation function, and iterated; finally, the latent features are obtained by passing through the long short-term memory attention module and the graph convolution and regularizing processing. The long short-term memory attention module is composed of a long short-term memory network, an activation function, an attention mechanism, and a skip connection. The long short-term memory network can learn the time dependence of cell-cell interactions. In order to make the network pay more attention to useful features, an attention mechanism can be introduced into the long short-term memory network in this embodiment to make the network strengthen the feature weights related to cell-cell interactions and enhance the feature representation of the network.

[0062] In one embodiment, the method for reconstructing cell-cell interactions based on a graph convolution and a long short-term memory attention network can further include a feature extraction process by a feature extraction network, and the specific process includes: inputting a cell graph into the feature extraction network, optimizing the feature matrix in the cell graph by the long short-term memory attention network in the feature extraction network to obtain an optimized feature matrix; inputting the optimized feature matrix and the adjacency matrix into the graph convolution network, capturing the spatial correlation of the graph structure by the graph convolution network, and obtaining the latent feature representation according to the spatial correlation of the graph structure.

[0063] In this embodiment, the long short-term memory attention module uses a skip connection to add the output features of the attention mechanism to the output features of the long short-term memory network, avoiding the influence of gradient disappearance on feature representation and thus affecting the overall model performance. In summary, the long short-term memory network with an attention mechanism can better focus on effective information and improve the overall computational efficiency. The feature matrix optimized by the long short-term memory attention module is input into the graph convolution network coding module together with the adjacency matrix constructed in the second step to further encode and optimize the features.

[0064] In one embodiment, the method for reconstructing cell-cell interactions based on a graph convolution and a long short-term memory attention network can further include a process of obtaining a latent feature representation, and the specific process includes: adding the spatial correlation of the graph structure processed by the activation function to the input feature matrix using a skip connection to capture features at different levels; inputting the features at different levels into the long short-term memory attention network and the graph convolution network again for encoding and optimization, and performing regularization processing to obtain the latent feature representation.

[0065] In the feature extraction using the feature extraction network, the specific process can include: inputting the cell graph into the feature extraction network, and the input node feature matrix is first optimized by the long short-term memory attention module; the optimized feature matrix is input into the graph convolution coding module together with the adjacency matrix to capture the spatial correlation of the graph structure; then the activation function is processed to add the input feature matrix through the jump connection, capture the features at different levels, avoid the degradation of the network, and further promote the information transmission, so that the information can be better propagated and shared in the network; the feature matrix optimized by the first stage coding layer is input into the long short-term memory attention module and the graph convolution network for coding optimization, and the latent feature representation is obtained after regularization processing.

[0066] In the present embodiment, the long short-term memory network can learn the time dependence of cell-cell interaction, and the attention mechanism is introduced to better focus on effective information and improve the overall calculation efficiency. The feature matrix optimized by the long short-term memory attention module is input into the graph convolution coding module together with the adjacency matrix, and then the spatial correlation of the graph structure is captured to learn the relationship between cells, and the above operation is repeated again to further enhance the weight of the features related to cell-cell interaction, and finally the latent feature representation is generated.

[0067] In step 210, the latent feature representation is decoded using inner product to generate a new adjacency matrix, and the reconstructed cell-cell interaction is obtained.

[0068] The latent feature representation extracted and enhanced by the graph convolution and long short-term memory attention module integrates spatial information and time dependence. Then, the computer device can decode the latent feature representation using inner product to generate a new adjacency matrix, and obtain the reconstructed cell-cell interaction based on the cell-cell interaction reconstruction method of the graph convolution and long short-term memory attention network.

[0069] In one embodiment, a cell-cell interaction reconstruction model is created and trained according to a cell-cell interaction reconstruction method based on a graph convolution and long short-term memory attention network. The attention mechanism module is introduced into the long short-term memory network to focus more on the learning of effective information, and combined with the graph convolution network to capture the spatial correlation in the graph structure while learning the time pattern, improve the ability of the cell-cell interaction reconstruction model to extract and optimize features, improve the calculation efficiency and accuracy of the overall model, and improve the robustness and noise resistance of the model; can help researchers understand the cell communication network and reveal the cell signaling mechanism.

[0070] Specifically, in the present embodiment, when evaluating the cell-cell interaction reconstruction model, the area under the receiver operating characteristic curve (AUROC), average precision (AP), and accuracy (ACC) are mainly used to evaluate the performance of the model. AUROC is calculated based on the receiver operating characteristic curve (ROC), which is drawn according to the true positive rate (TPR) and the false positive rate (FPR), and the TPR is the recall. AP measures the performance of the model at all possible thresholds, and the precision-recall curve is the weighted average of the precision achieved at each threshold, and the average precision is the area under the precision-recall curve. The calculation formula is as follows: where TP represents true positive, TN represents true negative, FP represents false positive, and FN represents false negative. If there is an edge between cells, it is positive, and if there is no edge between cells, it is negative. N is the number of threshold values, and i represents the ith threshold value.

[0071] To verify the influence of each module in the feature extraction and enhancement network on the performance of the cell-cell interaction reconstruction model, an ablation experiment was performed on the seqFISH data set of the mouse visual cortex, and the training set and test set were divided in a ratio of 7:3. The ablation experiment results were evaluated using the AUROC, AP, and ACC indicators, and the results are shown in the following table:

[0072]

[0073]

[0074] As shown in the above table, in the performance evaluation of the cell-cell interaction reconstruction model, the existence of cell-cell interaction is positive, and the non-existence of cell-cell interaction is negative, which is converted into a binary classification task. The reconstructed cell-cell interaction is evaluated using the AP, AUROC, and ACC evaluation indicators. The adjacency matrix obtained by the inner product decoding of the decoder is visualized, and the reconstructed cell-cell interaction map is obtained.

[0075] It should be understood that, although the various steps in the above flow chart are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the above flow chart may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0076] In one embodiment, Figure 4 As shown, a cell-to-cell interaction reconstruction system based on graph convolution and long short-term memory attention network is provided, including: a data collection module 410, an interaction judgment module 420, a cell map construction module 430, a feature extraction module 440 and a cell-to-cell interaction reconstruction module 450, wherein:

[0077] A data collection module 410 is used to collect spatial transcriptome data at single-cell or subcellular resolution, and convert the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell;

[0078] An interaction determination module 420 is configured to determine the spatial proximity between cells based on the spatial transcriptome data, and determine whether there is interaction between the cells based on the spatial proximity to obtain a determination result;

[0079] The cell graph construction module 430 is configured to use each cell as a vertex of a graph structure. If the result of the determination is that there is an interaction between the cells, the module constructs edges of the graph structure for the two interacting cells based on the spatial connection network. The module uses the matrix describing the cells and edges as the adjacency matrix, and uses the gene expression matrix of each cell as the graph node feature of the graph structure to obtain a node feature matrix, thereby constructing a cell graph.

[0080] A feature extraction module 440 is configured to combine graph convolution and long short-term memory attention networks to form a feature extraction network, and extract features from the cell graph through the feature extraction network to obtain a latent feature representation;

[0081] The cell-to-cell interaction reconstruction module 450 is used to decode the latent feature representation using the inner product, generate a new adjacency matrix, and obtain the reconstructed cell-to-cell interaction.

[0082] In an embodiment, the data collection module 410 is further configured to find ambiguously defined cells in the spatial transcriptome data, and remove the ambiguously defined cells from the spatial transcriptome data to obtain updated spatial transcriptome data; and perform transformation based on the updated spatial transcriptome data to obtain gene expression matrices corresponding to the cells.

[0083] In an embodiment, the data collection module 410 is further configured to perform log change normalization operation on each gene expression matrix.

[0084] In an embodiment, the cell graph construction module 430 is further configured to determine a vertex cell as a vertex of a graph structure, and find cells that have interactions with the vertex cell to construct target edges of the graph structure; determine a target adjacency matrix describing the vertex cell and the target edges, and determine a target node feature matrix of the graph structure according to a gene expression matrix corresponding to the vertex cell; and combine the vertex, the target edges, the target adjacency matrix, and the target node feature matrix to construct the cell graph.

[0085] In an embodiment, the feature extraction module 440 is further configured to combine a graph convolution and a long short-term memory attention network in series to form a feature extraction network.

[0086] In an embodiment, the long short-term memory attention network includes a long short-term memory network, an activation function, an attention mechanism, and a skip connection; the long short-term memory network is configured to learn time dependence of interactions between cells; the attention mechanism is configured to enhance feature representation of the network; the skip connection is configured to add output features of the long short-term memory network to output features of the attention mechanism; and the graph convolution is configured to capture spatial correlation of the graph structure and learn relationships between cells.

[0087] In an embodiment, the feature extraction module 440 is further configured to input the cell graph into the feature extraction network, optimize a feature matrix in the cell graph by the long short-term memory attention network in the feature extraction network to obtain an optimized feature matrix, and input the optimized feature matrix and an adjacency matrix into a graph convolution network to capture spatial correlation of the graph structure and obtain latent feature representation according to the spatial correlation of the graph structure.

[0088] In an embodiment, the feature extraction module 440 is further configured to add the spatial correlation of the graph structure processed by an activation function to the input feature matrix by a skip connection to capture features at different levels; and input the features at different levels into the long short-term memory attention network and the graph convolution network again for encoding optimization, and perform regularization processing to obtain the latent feature representation.

[0089] In an embodiment, a computer device is provided, which can be a terminal, and an internal structure diagram of the computer device can be as shown in Figure 5As shown in the figure. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a cell interaction reconstruction method based on graph convolution and long short-term memory attention network. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0090] Those skilled in the art can understand that, Figure 5 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0091] In one embodiment, a computer device is provided, comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the cell interaction reconstruction method based on graph convolution and long short-term memory attention network.

[0092] In one embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by the processor to implement the steps of the cell interaction reconstruction method based on graph convolution and long short-term memory attention network.

[0093] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, databases, or other media in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0094] The technical features of the above embodiments can be combined in any way. In order to make the description simple, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist, they should be considered as the scope of the present application.

[0095] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of the patent of the present application should be subject to the appended claims.

Claims

1. A cell-cell interaction reconstruction method based on graph convolution and long short-term memory attention network, characterized by: The method comprises: Collect spatial transcriptome data at single-cell or subcellular resolution, and convert the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell; Determining the spatial proximity between the cells according to the spatial transcriptome data, and judging whether there is interaction between the cells based on the spatial proximity to obtain a judgment result; Taking each of the cells as a vertex of a graph structure, if the judgment result is that there is an interaction between the cells, constructing edges of a graph structure for the two interacting cells based on a spatial connection network, using a matrix describing cells and edges as an adjacency matrix, and using the gene expression matrix of each cell as a graph node feature of the graph structure to obtain a node feature matrix, and constructing a cell graph; A feature extraction network is formed by combining graph convolution and long short-term memory attention network, including: combining graph convolution and long short-term memory attention network in series to form a feature extraction network; the long short-term memory attention network includes a long short-term memory network, an activation function, an attention mechanism, and a jump connection; the long short-term memory network is used to learn the time dependency of the interaction between cells; the attention mechanism is used to enhance the feature representation of the network; the jump connection is used to add the output feature of the long short-term memory network to the output feature of the attention mechanism; the graph convolution is used to capture the spatial correlation of the graph structure and learn the relationship between cells; the cell graph is extracted by the feature extraction network to obtain a potential feature representation, including: the cell The graph is input into the feature extraction network, and the feature matrix in the cell graph is optimized by the long short-term memory attention network in the feature extraction network to obtain an optimized feature matrix; the optimized feature matrix and the adjacency matrix are input into the graph convolution network together, and the spatial correlation of the graph structure is captured by the graph convolution network, and a potential feature representation is obtained according to the spatial correlation of the graph structure, including: the spatial correlation of the graph structure is processed by the activation function and added to the input feature matrix using a jump connection to capture features at different levels; the features at different levels are input again into the long short-term memory attention network and the graph convolution network for encoding optimization, and are subjected to regularization processing to obtain a potential feature representation; The inner product is used to decode the latent feature representation to generate a new adjacency matrix and obtain the reconstructed cell-to-cell interactions.

2. The cell-cell interaction reconstruction method based on graph convolution and long short-term memory attention network according to claim 1 is characterized in that The gene expression matrix corresponding to each cell is obtained by conversion based on the spatial transcriptome data, including: Searching for fuzzy-defined cells in the spatial transcriptome data, and removing the fuzzy-defined cells from the spatial transcriptome data to obtain updated spatial transcriptome data; The updated spatial transcriptome data is converted to obtain the gene expression matrix corresponding to each cell.

3. The cell-cell interaction reconstruction method based on graph convolution and long short-term memory attention network according to claim 1 is characterized in that After converting the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell, the method further includes: A logarithmic change normalization operation was performed on each of the gene expression matrices.

4. The cell-cell interaction reconstruction method based on graph convolution and long short-term memory attention network according to claim 1, characterized in that The cell map is constructed, including: Determine a vertex cell that is a vertex of a graph structure, and find a cell that interacts with the vertex cell to construct a target edge of the graph structure; Using the matrix describing the vertex cells and the target edges as the target adjacency matrix, and determining the target node feature matrix of the graph structure according to the gene expression matrix corresponding to the vertex cells; The vertices, target edges, target adjacency matrix, and target node feature matrix are combined and constructed to obtain a cell graph.

5. A cell-cell interaction reconstruction system based on graph convolution and long short-term memory attention network, characterized by: The system comprises: A data collection module is used to collect spatial transcriptome data at single-cell or subcellular resolution, and convert the spatial transcriptome data to obtain a gene expression matrix corresponding to each cell; an interaction determination module, configured to determine the spatial proximity between cells based on the spatial transcriptome data, and determine whether there is interaction between cells based on the spatial proximity to obtain a determination result; a cell graph construction module, configured to use each of the cells as a vertex of a graph structure; if the judgment result indicates that there is an interaction between the cells, construct edges of a graph structure for the two interacting cells based on a spatial connection network; use a matrix describing cells and edges as an adjacency matrix; use the gene expression matrix of each cell as a graph node feature of the graph structure; obtain a node feature matrix; and construct a cell graph; A feature extraction module is used to combine graph convolution and long short-term memory attention network to form a feature extraction network, including: combining graph convolution and long short-term memory attention network in series to form a feature extraction network; the long short-term memory attention network includes a long short-term memory network, an activation function, an attention mechanism, and a jump connection; the long short-term memory network is used to learn the temporal dependency of interactions between cells; the attention mechanism is used to enhance the feature representation of the network; the jump connection is used to add the output features of the long short-term memory network and the output features of the attention mechanism; the graph convolution is used to capture the spatial correlation of the graph structure and learn the relationship between cells; the cell graph is extracted by the feature extraction network to obtain a potential feature representation, including: Inputting the cell map into the feature extraction network, optimizing the feature matrix in the cell map through the long short-term memory attention network in the feature extraction network to obtain an optimized feature matrix; inputting the optimized feature matrix together with the adjacency matrix into the graph convolution network, capturing the spatial correlation of the graph structure through the graph convolution network, and obtaining a potential feature representation based on the spatial correlation of the graph structure, including: processing the spatial correlation of the graph structure through the activation function and adding it to the input feature matrix using a jump connection to capture features at different levels; inputting the features at different levels again into the long short-term memory attention network and the graph convolution network for encoding optimization, and obtaining a potential feature representation after regularization processing; The cell-cell interaction reconstruction module is used to decode the latent feature representation using the inner product, generate a new adjacency matrix, and obtain reconstructed cell-cell interactions.

6. The cell-cell interaction reconstruction system based on graph convolution and long short-term memory attention network according to claim 5, characterized in that The data collection module is further configured to search for fuzzy-defined cells in the spatial transcriptome data and remove the fuzzy-defined cells from the spatial transcriptome data to obtain updated spatial transcriptome data; The updated spatial transcriptome data is converted to obtain the gene expression matrix corresponding to each cell.

Images