Idle data feature extraction method, spatial domain identification method and system

Abstract

The present application belongs to the field of spatial transcriptome spatial domain identification, and provides a spatial domain identification method and system to solve the problem of poor accuracy of spatial transcriptome spatial domain identification. The spatial domain identification method includes obtaining spatial transcriptome data of a pathological section, which includes an initial gene expression matrix and spatial position information; performing denoising processing on the initial gene expression matrix to obtain a denoised gene expression matrix; calculating the position code of each site according to the spatial position information, and splicing the position code with the denoised gene expression matrix to obtain an enhanced gene expression matrix; and using a mask autoencoder to encode the enhanced gene expression matrix to obtain a low-order latent representation as the extracted spatial transcriptome data feature, which can realize accurate identification of the spatial domain and provide a good upstream analysis basis for downstream tasks.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of spatial transcriptome spatial domain recognition, and particularly relates to a spatial transcriptome data feature extraction method, a spatial domain recognition method and system. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] In spatial transcriptome data analysis, one of the core tasks is the recognition of spatial domains. Spatial domains refer to regions with similar gene expression patterns in spatial organization, and the recognition results are of great significance to the study of tissue function and disease mechanism. However, due to the fact that spatial transcriptome data usually contains the expression of thousands of genes in thousands of sites (e.g., cells or tissues), it has high dimensionality and high sparsity, combined with the uncertainty of sequencing technology and the significant noise introduced by experimental conditions, which affects the accuracy of analysis; in addition, the gene expression relationship between cells or tissues is often nonlinear and high-order, and it is difficult to capture complex long-range interactions by relying only on simple Euclidean distance or local neighborhood relationship, which ultimately affects the accuracy of spatial transcriptome spatial domain recognition. In spatial transcriptome data analysis, one of the core tasks is the recognition of spatial domains. Spatial domains refer to regions with similar gene expression patterns in spatial organization, and the recognition results are of great significance to the study of tissue function and disease mechanism. However, due to the fact that spatial transcriptome data usually contains the expression of thousands of genes in thousands of sites (e.g., cells or tissues), it has high dimensionality and high sparsity, combined with the uncertainty of sequencing technology and the significant noise introduced by experimental conditions, which affects the accuracy of analysis; in addition, the gene expression relationship between cells or tissues is often nonlinear and high-order, and it is difficult to capture complex long-range interactions by relying only on simple Euclidean distance or local neighborhood relationship, which ultimately affects the accuracy of spatial transcriptome spatial domain recognition. SUMMARY

[0004] In order to solve the technical problems existing in the background art, the present application provides a spatial transcriptome data feature extraction method, a spatial domain recognition method and system, which can realize accurate recognition of spatial domains and provide a good upstream analysis foundation for downstream tasks.

[0005] In order to achieve the above purpose, the present application adopts the following technical solutions:

[0006] The first aspect of the present application provides a spatial transcriptome data feature extraction method.

[0007] A spatial transcriptome data feature extraction method, comprising:

[0008] obtaining spatial transcriptome data of a pathological section, which includes an initial gene expression matrix and spatial position information;

[0009] performing denoising processing on the initial gene expression matrix to obtain a denoised gene expression matrix;

[0010] calculating a position code for each site according to the spatial position information, and splicing the position code with the denoised gene expression matrix to obtain an enhanced gene expression matrix;

[0011] encoding the enhanced gene expression matrix using a mask autoencoder to obtain a low-order latent representation and use it as the extracted spatial transcriptome data feature.

[0012] ​As an implementation method, in the process of training the masked autoencoder, the adjacency matrix reconstructed by the low-order latent representation of the gene expression matrix is ​​used to construct an alignment loss function; the high-order latent representation and the low-order latent representation of the gene expression matrix are used to construct a spatial consistency loss function; the reconstructed enhanced gene expression matrix and the enhanced gene expression matrix before reconstruction are used to construct a second reconstruction loss function, and then the total loss function of the masked autoencoder is obtained from the alignment loss function, the spatial consistency loss function and the second reconstruction loss function, and then the masked autoencoder is trained.

[0013] As an implementation method, the reconstructed enhanced gene expression matrix is ​​obtained by decoding the low-order latent representation output by the masked autoencoder; the high-order latent representation of the gene expression matrix is ​​obtained by encoding the denoised gene expression matrix by a multi-scale hypergraph autoencoder.

[0014] As an implementation method, a multi-scale hypergraph autoencoder is used to learn the high-order latent representation of the denoised gene expression matrix at different scales, and reconstruct the denoised gene expression matrix, and then a first reconstruction loss function is constructed to train the multi-scale hypergraph autoencoder.

[0015] As an implementation method, the spatial consistency loss function is:

[0016] ;

[0017] ;

[0018] ;

[0019] in, is the spatial consistency loss function; is the smoothing term; is the low-order latent representation of the gene expression matrix; is the high-order latent representation of the gene expression matrix; and All are intermediate parameters; 、 Respectively loci The corresponding vectors in the low-order latent representation and high-order latent representation of the gene expression matrix respectively; is the total number of sites; is the normalized exponential function.

[0020] As an implementation method, the alignment loss function is: ;

[0021] in, is the alignment loss function; 、 respectively , Laplacian matrix of the gene expression matrix; is a reconstructed adjacency matrix of the low-rank latent representation of the gene expression matrix; is an adjacency matrix of the gene expression matrix; denotes the F-norm.

[0022] The second aspect of the present application provides a idle data feature extraction system.

[0023] An idle data feature extraction system, comprising:

[0024] An idle data acquisition module for acquiring idle data of a pathological section, comprising an initial gene expression matrix and spatial position information;

[0025] A data denoising module for denoising the initial gene expression matrix to obtain a denoised gene expression matrix;

[0026] A position encoding module for calculating a position encoding for each site according to the spatial position information and splicing it with the denoised gene expression matrix to obtain an enhanced gene expression matrix;

[0027] A low-rank latent representation module for encoding the enhanced gene expression matrix using a mask autoencoder to obtain a low-rank latent representation as an extracted idle data feature.

[0028] The third aspect of the present application provides a spatial transcriptome spatial domain identification method.

[0029] A spatial transcriptome spatial domain identification method, comprising:

[0030] The low-rank latent representation of the idle data of the pathological section is extracted by using the steps in the idle data feature extraction method described above, and is taken as an idle data feature;

[0031] The sites of the idle data feature are clustered to obtain a spatial domain identification result.

[0032] The fourth aspect of the present application provides a spatial transcriptome spatial domain identification system.

[0033] A spatial transcriptome spatial domain identification system, comprising:

[0034] A feature extraction module for extracting the low-rank latent representation of the idle data of the pathological section by using the steps in the idle data feature extraction method described above, and taking it as an idle data feature;

[0035] A clustering module for clustering the sites of the idle data feature to obtain a spatial domain identification result.

[0036] A fifth aspect of the present application provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the spatial transcriptome spatial domain identification method based on a hypergraph autoencoder as described above when executing the program.

[0037] Compared with the prior art, the present application has the following advantages:

[0038] (1) The present application obtains a denoised gene expression matrix by denoising the initial gene expression matrix, calculates a position code for each site according to spatial position information, and splices the position code with the denoised gene expression matrix to obtain an enhanced gene expression matrix, then encodes the enhanced gene expression matrix using a mask autoencoder to obtain a low-order latent representation as extracted empty data features, and combines the position code with the denoised gene expression matrix, thereby improving the accuracy of empty data feature extraction.

[0039] (2) The present application combines multi-scale hypergraph graph construction with a multi-scale hypergraph autoencoder to solve the problem of difficult multi-level modeling of high-order relationships between sites, and achieves the effect of capturing high-order biological information; and uses spatial consistency loss and alignment loss to explicitly fuse the high-order latent representation and the low-order latent representation, thereby solving the problem of low interpretability of implicit fusion of low-order and high-order latent representations, and achieving the effect of capturing multi-level biological information.

[0040] The advantages of the additional aspects of the present application will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0041] The accompanying drawings, which form a part of the present application, are used to provide further understanding of the present application, and the illustrative embodiments of the present application and their description serve the purpose of explaining the present application, and do not constitute an improper limitation of the present application.

[0042] Figure 1 is an empty data feature extraction method flowchart of an embodiment of the present application;

[0043] Figure 2 is a training process of a mask autoencoder and a multi-scale hypergraph autoencoder of an embodiment of the present application;

[0044] Figure 3 is a schematic structural diagram of an empty data feature extraction system of an embodiment of the present application;

[0045] Figure 4 is a multi-scale hypergraph autoencoder architecture diagram of an embodiment of the present application;

[0046] Figure 5This is a flow chart of a method for identifying spatial domains of a spatial transcriptome according to an embodiment of the present invention;

[0047] Figure 6 This is a schematic diagram of the structure of a spatial transcriptome spatial domain identification system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0049] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0050] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0051] The rapid development of high-throughput sequencing technology has greatly advanced research in genomics and transcriptomics. The emergence of spatial transcriptomics, in particular, has enabled researchers in medicine and bioinformatics to explore gene expression within the spatial context of tissues. This technology not only reveals the spatial distribution of gene expression within tissues but also provides a new perspective for understanding tissue structure, gene function, and pathogenic mechanisms, becoming an increasingly important component in clinical disease prevention and treatment.

[0052] Explanation of terms:

[0053] (1) Spatial domain identification: identifying regions with similar gene expression patterns in spatial organization.

[0054] (2) Evaluation criteria for spatial domain recognition:

[0055] Assumptions are sites that belong to the same spatial domain and the same true label class quantity, belongs to the same spatial domain and does not belong to the same real label class quantity, belong to different spatial domains but belong to the same real label class quantity, does not belong to a spatial domain and does not belong to a real label class quantity.

[0056] (3) Adjusted Rand Index :

[0057] Definition: where the adjusted Rand index is , is the expected Rand index.

[0058] Purpose: Evaluate the consistency between the spatial domain recognition result and the real label class.

[0059] (4) Normalized Mutual Information :

[0060] Definition: where are the cross-entropy of the spatial domain recognition distribution and the real label class distribution, respectively. .

[0061] Purpose: Evaluate the similarity between the spatial domain recognition result and the real label class.

[0062] (5) (Fowlkes-Mallows index):

[0063] Definition: ;

[0064] Purpose: Evaluate the consistency between the clustering result and the real label.

[0065] In one or more embodiments, as shown in the following, a method for extracting idle data features according to an embodiment of the present application includes: Figure 1

[0066] S101: Obtain idle data of a pathological section, including a gene expression matrix and spatial position information.

[0067] The complete spatial transcriptome sequencing data corresponding to one pathological section includes a gene expression matrix and spatial position information .

[0068] S102: Perform data denoising processing on the gene expression matrix to obtain a denoised gene expression matrix;

[0069] Before performing data denoising processing on the gene expression matrix, it further includes:

[0070] Perform normalization and logarithmic conversion on the gene expression matrix to obtain a preprocessed gene expression matrix;

[0071] ​Before normalization and log-conversion of the gene expression matrix, it also includes:

[0072] According to the set screening conditions, the gene expression matrix is screened to realize the quality control of the gene expression matrix.

[0073] From the initial gene expression matrix Screening the gene expression matrix that meets the conditions, and output the gene expression matrix that meets the conditions , wherein the screening conditions are as follows:

[0074] Gene At least 1% of have expression;

[0075] The total expression count of gene is not less than a set value, such as 200.

[0076] The gene expression matrix that meets the conditions is normalized and log-converted, and the normalized gene expression matrix is output.

[0077] The normalized gene expression matrix is further processed by data denoising using graph Fourier transform.

[0078] Distance metric: based on , the Euclidean distance between any two points is calculated;

[0079] Construct graph : is a set (denoted as ) of , and there is an edge between and if and only if is the nearest neighbor of or is the nearest neighbor of , , is composed of the set , and is obtained;

[0080] Construct the adjacency matrix , the diagonal matrix :

[0081] ;

[0082] Diagonal matrix , wherein​​​ For the degree of

[0083] Fourier transform and eigen decomposition: computing Laplacian matrix The implementation of Fourier transform in graph is based on the eigen decomposition of the Laplacian matrix , which gives eigenvalues and eigenvectors , where the eigenvalues represent frequencies;

[0084] Graph Fourier transform: project the original data into frequency space using , the graph Fourier transform of the normalized gene expression matrix is ;

[0085] Low-pass filter: the low-pass filter weights each frequency component, with expression , where , , is the smoothing parameter ;

[0086] Inverse transform of filtered signal: inverse transform the filtered signal from frequency space back to original space ;

[0087] , the output is the denoised gene expression matrix and the adjacency matrix .

[0088] S103: Calculate the position encoding for each site according to the spatial position information, and concatenate it with the denoised gene expression matrix to obtain the enhanced gene expression matrix.

[0089] The goal of position encoding is to provide a unique representation for each position in , which is based on sine and cosine functions:

[0090] ;

[0091] where is the position index, is the index of the position encoding dimension (sine for even dimensions and cosine for odd dimensions), i.e., the th site; is the total dimension of the position encoding.

[0092] Concatenate and to obtain the enhanced gene expression matrix:

[0093] ; Represents the connection function.

[0094] S104: Encode the enhanced gene expression matrix using a masked autoencoder to obtain a low-order latent representation as the extracted idle data feature.

[0095] Combine Figure 2 When training the masked autoencoder, the low-order latent representation output by the masked autoencoder is decoded to obtain a reconstructed enhanced gene expression matrix; the adjacency matrix reconstructed by the low-order latent representation of the gene expression matrix is ​​used to construct an alignment loss function, the high-order latent representation and the low-order latent representation are used to construct a spatial consistency loss function, and the reconstructed enhanced gene expression matrix and the enhanced gene expression matrix before reconstruction are used to construct a second reconstruction loss function, and finally the total loss function of the masked autoencoder is obtained.

[0096] The reconstructed enhanced gene expression matrix is ​​obtained by decoding the low-order latent representation output by the masked autoencoder; the high-order latent representation of the gene expression matrix is ​​obtained by encoding the denoised gene expression matrix by a multi-scale hypergraph autoencoder.

[0097] In the specific implementation process, a multi-scale hypergraph autoencoder is used to learn the high-order latent representation of the denoised gene expression matrix at different scales, and reconstruct the denoised gene expression matrix, and then construct the first reconstruction loss function to train the multi-scale hypergraph autoencoder.

[0098] Distance metric: based on , calculate any two The Euclidean distance between

[0099] The process of constructing a multi-scale hypergraph is:

[0100] First, we use a multi-scale hypergraph As an example, let’s explain how to construct a hypergraph: , select the one closest to it indivual Constitute a hyperedge (this indivual and Together they form a hyperedge). Multiscale hypergraph incidence matrix express:

[0101] ;

[0102] is the multi-scale hypergraph incidence matrix The elements in ; the hyperedge set can be represented ,in, , multi-scale hypergraph .

[0103] from data adaptively determine the number , generate a hypergraph adjacency matrix, and concatenate it by column to obtain a multi-scale hypergraph adjacency matrix , multi-scale hypergraph . , ; , ; the diagonal matrix of , ; is an element in the multi-scale hypergraph adjacency matrix ; is a preset weight function; is an edge between sites.

[0104] Hyperedge convolution layer: the hyperedge convolution layer can be represented as:

[0105] ;

[0106] where is the representation of the hypergraph at the layer, is the feature dimension, , is the diagonal matrix of the hyperedge weight, is a learnable parameter during the training process, is a bias vector; is the diagonal matrix of the site set .

[0107] In combination Figure 4 , the structure of the multi-scale hypergraph autoencoder includes an encoder and a decoder;

[0108] The encoder has 3 layers in total, which are a hyperedge convolution layer, an activation layer and a hyperedge convolution layer in sequence, and can be expressed as:

[0109] ;

[0110] where , are learnable weight matrices; represents the hyperedge convolution layer.

[0111] The decoder has the same structure as the encoder, and can be expressed as:

[0112] ;

[0113] where , is the learnable weight matrix. The module output is the high-order latent representation and the reconstructed .

[0114] The loss function of the multi-scale hypergraph autoencoder is the first reconstruction loss function.

[0115] The first reconstruction loss function is constructed as follows: The corresponding vectors in , are , respectively. The loss of the multi-scale hypergraph autoencoder module is:

[0116] .

[0117] The mask autoencoder (MAE) is composed of a mask mechanism, an encoder and a decoder.

[0118] Mask mechanism: mask according to the set proportion, and get .

[0119] Encoder: encode the data input to the encoder to get the low-order latent representation , which can be represented as:

[0120] ;

[0121] , is the weight matrix, , is the bias vector.

[0122] Decoder: the structure of the decoder is consistent with that of the mask encoder, which can be represented as:

[0123] ;

[0124] , is the learnable weight, , is the bias vector. The module output is , .

[0125] The total loss function of the mask autoencoder (MAE) is ​​The second reconstruction loss function, the alignment loss function and the spatial consistency loss function.

[0126] The second reconstruction loss function: In , the corresponding vectors are , respectively. The second reconstruction loss function is:

[0127] ;

[0128] The alignment loss function :

[0129] Take as the node, construct the reconstructed adjacency matrix by . When is the neighbor of , then , otherwise . The module alignment loss is:

[0130] ;

[0131] wherein, is the alignment loss function; , are the Laplacian matrices of , respectively; is the reconstructed adjacency matrix of the low-order latent representation of the gene expression matrix; is the adjacency matrix of the gene expression matrix; denotes the F-norm. The alignment loss encourages the modules to generate embeddings with similar graph structure as their input.

[0132] The spatial consistency loss function is:

[0133] ;

[0134] ;

[0135] ;

[0136] wherein, is the spatial consistency loss function; is the smoothing term; is the low-order latent representation of the gene expression matrix; is the high-order latent representation of the gene expression matrix; and are intermediate parameters; , are the first sites are the corresponding vectors in the low-order latent representation and the high-order latent representation of the gene expression matrix, respectively; is the total number of sites; is a normalized exponential function.

[0137] The spatial consistency loss function is used to integrate the high-order latent representation into the low-order latent representation, so that the trained mask autoencoder obtains the low-order latent representation after integrating the high-order latent representation.

[0138] The total loss function of the mask autoencoder is: ; wherein, and are constant coefficients.

[0139] Corresponding to the above method, as shown in Figure 3 , a system for extracting idle data features is also provided, comprising:

[0140] An idle data acquisition module 301 is configured to acquire idle data of a pathological section, which includes an initial gene expression matrix and spatial position information;

[0141] A data denoising module 302 is configured to perform denoising processing on the initial gene expression matrix to obtain a denoised gene expression matrix;

[0142] A position encoding module 303 is configured to calculate position encoding for each site according to the spatial position information, and splice the position encoding with the denoised gene expression matrix to obtain an enhanced gene expression matrix;

[0143] A low-order latent representation module 304 is configured to encode the enhanced gene expression matrix by using a mask autoencoder to obtain a low-order latent representation and use the low-order latent representation as extracted idle data features.

[0144] In the process of training the mask autoencoder, an adjacency matrix reconstructed by using the low-order latent representation of the gene expression matrix is used to construct an alignment loss function; the high-order latent representation and the low-order latent representation of the gene expression matrix are used to construct a spatial consistency loss function; the enhanced gene expression matrix after reconstruction and the enhanced gene expression matrix before reconstruction are used to construct a second reconstruction loss function, and then the total loss function of the mask autoencoder is obtained from the alignment loss function, the spatial consistency loss function and the second reconstruction loss function, and the mask autoencoder is trained.

[0145] Specifically, the reconstructed enhanced gene expression matrix is obtained by decoding the low-order latent representation output by the mask autoencoder; and the high-order latent representation of the gene expression matrix is obtained by encoding the denoised gene expression matrix by the multi-scale hypergraph autoencoder.

[0146] The multi-scale hypergraph autoencoder is trained by using the multi-scale hypergraph autoencoder to learn the high-order latent representation of the denoised gene expression matrix at different scales, reconstructing the denoised gene expression matrix, and constructing a first reconstruction loss function.

[0147] It should be noted that the specific implementation process of each module in the embodiment of the present application corresponds to the steps in the above-mentioned idle data feature extraction method one by one, and the specific implementation process is the same, which will not be described in detail here.

[0148] As shown in Figure 5 , the embodiment of the present application provides a spatial transcriptome spatial domain identification method, comprising:

[0149] S501: The low-order latent representation of the idle data of the pathological section is extracted by using the steps in the above-mentioned idle data feature extraction method, and is taken as the idle data feature;

[0150] S502: The sites of the idle data feature are clustered to obtain a spatial domain identification result.

[0151] As shown in Figure 6 , in one or more embodiments, a spatial transcriptome spatial domain identification system is also provided, comprising:

[0152] The feature extraction module 601 is used to extract the low-order latent representation of the idle data of the pathological section by using the steps in the above-mentioned idle data feature extraction method, and is taken as the idle data feature;

[0153] The clustering module 602 is used to cluster the sites of the idle data feature to obtain a spatial domain identification result.

[0154] The input of the clustering module 602 is , and (K-means clustering), a model-based clustering method (based on model clustering), and (Gaussian mixture model) are selected to cluster , and the clustering label is the spatial domain identification result.

[0155] Model evaluation: the consistency of the clustering result and the true label is evaluated by using , , For the convenience of understanding and better display of the model performance, the visualization of the spatial domain identification result and the true label is provided.

[0156] Downstream tasks: based on the spatial domain recognition result, biological function analysis such as gene enrichment analysis and signal pathway prediction is performed.

[0157] In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by a communication section, and / or installed from a detachable medium. When the computer program is executed by a central processing unit, various functions defined in the apparatus of the present application are performed.

[0158] The present application is described with reference to the flowcharts and / or block diagrams of the methods, apparatus (system) and computer program products of embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, a special purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions, which are executed via the processor of the computer or other programmable data processing apparatus, generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in a flow or multiple flows and / or blocks Figure 1 The functions specified in a flow or multiple flows and / or blocks

[0159] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of idle data feature extraction, the method comprising: The method comprises the following steps: acquiring empty run data of pathological sections, which comprises an initial gene expression matrix and spatial position information; performing denoising processing on the initial gene expression matrix to obtain a denoised gene expression matrix; calculating a position code for each site according to the spatial position information, and splicing the position code with the denoised gene expression matrix to obtain an enhanced gene expression matrix; encoding the enhanced gene expression matrix by using a mask autoencoder to obtain a low-order latent representation and taking the low-order latent representation as an extracted empty run data feature; in the process of training the mask autoencoder, constructing an alignment loss function by using an adjacency matrix reconstructed by the low-order latent representation of the gene expression matrix, constructing a spatial consistency loss function by using a high-order latent representation and the low-order latent representation of the gene expression matrix, and constructing a second reconstruction loss function by using a reconstructed enhanced gene expression matrix and the enhanced gene expression matrix before reconstruction, and then obtaining a total loss function of the mask autoencoder from the alignment loss function, the spatial consistency loss function and the second reconstruction loss function, and training the mask autoencoder; the reconstructed enhanced gene expression matrix is obtained by decoding the low-order latent representation output by the mask autoencoder; and the high-order latent representation of the gene expression matrix is obtained by encoding the denoised gene expression matrix by using a multi-scale hypergraph autoencoder.

2. The idle data feature extraction method of claim 1, wherein, The multi-scale hypergraph autoencoder is used to learn the high-order latent representation of the denoised gene expression matrix at different scales, reconstruct the denoised gene expression matrix, and then construct a first reconstruction loss function to train the multi-scale hypergraph autoencoder.

3. The idle data feature extraction method of claim 1, wherein, The spatial consistency loss function is: ； ； ； wherein, is a spatial consistency loss function; is a smoothing term; and are intermediate parameters; , are the first sites are the corresponding vectors in the low-order and high-order latent representations of the gene expression matrix, respectively; is the total number of sites; is a normalized exponential function.

4. The idle data feature extraction method of claim 1, wherein, The alignment loss function is: ; wherein, is an alignment loss function; , are respectively , is a Laplacian matrix of is a reconstructed adjacency matrix from a low-rank latent representation of the gene expression matrix; is an adjacency matrix of the gene expression matrix; denotes the F-norm.

5. An idle data feature extraction system based on the idle data feature extraction method according to any one of claims 1 to 4, characterized by The method comprises the following steps: an empty run data acquisition module is configured to acquire empty run data of pathological sections, which comprises an initial gene expression matrix and spatial position information; a data denoising module is configured to perform denoising processing on the initial gene expression matrix to obtain a denoised gene expression matrix; a position code module is configured to calculate a position code for each site according to the spatial position information, and splice the position code with the denoised gene expression matrix to obtain an enhanced gene expression matrix; a low-order latent representation module is configured to encode the enhanced gene expression matrix by using a mask autoencoder to obtain a low-order latent representation and take the low-order latent representation as an extracted empty run data feature.

6. A spatial transcriptome spatial domain identification method, characterized in that, The method comprises the following steps: a low-order latent representation of empty run data of pathological sections is extracted by using the steps in the empty run data feature extraction method according to any one of claims 1-4, and taken as an empty run data feature; sites of the empty run data feature are clustered to obtain a spatial domain recognition result.

7. A spatial transcriptome spatial domain identification system, comprising: The method comprises the following steps: a feature extraction module is configured to extract a low-order latent representation of empty run data of pathological sections by using the steps in the empty run data feature extraction method according to any one of claims 1-4, and take the low-order latent representation as an empty run data feature; a clustering module is configured to cluster sites of the empty run data feature to obtain a spatial domain recognition result.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the steps in the empty run data feature extraction method according to any one of claims 1-4; or the processor executes the program to implement the steps in the spatial transcriptome spatial domain recognition method according to claim 6.

Images