Single cell data multi-clustering method and system based on multi-omics data fusion

Abstract

The invention belongs to the field of bioinformatics, and provides a single cell data multi-clustering method and system based on multi-omics data fusion, and the technical scheme is as follows: obtaining single cell transcriptome sequencing data and single cell apparent group data; mapping single-cell transcriptome sequencing data and single-cell apparent group data to respective nonlinear embedding spaces to obtain low-dimensional representation of corresponding data; extracting specific features of different omics data based on the low-dimensional representations of different omics, and reconstructing the low-dimensional representations based on the specific features of different omics data to obtain reconstructed low-dimensional features; extracting common characteristics of low-dimensional representation of different omics data; aggregating the reconstructed low-dimensional features and common features of low-dimensional representations of different omics data to generate a co-embedding representation; and projecting the co-embedding representation to a plurality of parallel potential subspaces, and generating a clustering result in the subspaces. And a new insight is provided for understanding the diversity effect of the cells from different angles.

Description

Technical Field

[0001] The present invention belongs to the field of bioinformatics, and in particular relates to a single cell data multi-clustering method and system based on multi-omics data fusion. Background Art

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

[0003] Single-cell RNA sequencing (scRNA-seq) can quantify the abundance of mRNA transcripts of genes in each cell, while the single-cell transposase-based chromatin accessibility assay (scATAC) can characterize the openness and epigenetic information of cis-regulatory elements in neighboring genes. Joint analysis of single-cell transcriptome data and single-cell epigenomic data helps to decipher gene regulatory relationships associated with cellular heterogeneity. However, different omics data have strong heterogeneity, and the high noise and sparsity of single-cell data also affect the fusion effect of multi-omics data. In addition, existing single-cell clustering methods only focus on the clustering of cell types and cannot effectively explore other cell clustering patterns for comprehensive analysis from other perspectives.

[0004] The integration of single-cell multi-omics data can reveal the regulatory basis of diverse characteristics such as different cell types and states. However, current single-cell multi-omics data fusion methods have certain limitations: shallow methods based on non-negative matrix decomposition or principal component analysis mostly project multi-omics data into a shared latent space, ignoring the specific information of different omics and the nonlinear attribute characteristics within the data; manifold alignment methods characterize the intrinsic cell structure by capturing the nonlinear geometric structure within the data, but they have high time complexity; deep learning-based methods use different deep generative networks to model single-cell multi-omics data, most of which focus on a shared representation and ignore the individuality of omics, and cannot integrate different levels of biological features to learn more discriminative representations. More importantly, existing methods cannot fully integrate genetic information from different omics to reveal potential alternative clusters with diversity and high quality, and it is difficult to reveal the different roles and characteristics of cells from different perspectives. Summary of the invention

[0005] In order to solve at least one technical problem existing in the above-mentioned background technology, the present invention provides a single-cell data multi-clustering method and system based on multi-omics data fusion, which guides the fusion and clustering of single-cell multi-omics data by learning high-quality common features and specific features from highly sparse heterogeneous single-cell multi-omics data, and discovers potential non-redundant cell clustering patterns in multi-omics data.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a single cell data multi-clustering method based on multi-omics data fusion, comprising the following steps:

[0008] Obtain single-cell transcriptome sequencing data and single-cell epigenetic data;

[0009] Mapping single-cell transcriptome sequencing data and single-cell epigenetic data into their respective nonlinear embedding spaces to obtain low-dimensional representations of the corresponding data;

[0010] Extracting specific features of different omics data based on low-dimensional representations of different omics data, reconstructing the low-dimensional representations based on the specific features of different omics data, and obtaining reconstructed low-dimensional features;

[0011] Extract common features of low-dimensional representations of different omics data;

[0012] Aggregate the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data to generate a co-embedded representation;

[0013] The co-embedded representation is projected into multiple parallel latent subspaces and clustering results are generated in the subspaces.

[0014] Furthermore, independent deep autoencoders were used to map single-cell multi-omics data into their respective nonlinear embedding spaces.

[0015] Furthermore, the normalized attention scores are used to reconstruct the low-dimensional representation by considering the similarities between cells.

[0016] Furthermore, the Siamese network is used to maximize the consistency between the low-dimensional representations of different omics data by sharing network parameters, and to extract the common features of the low-dimensional representations of different omics data.

[0017] Furthermore, weighted fusion is used to aggregate the specific features and common features of multi-omics data to generate a co-embedded representation:

[0018] Z I =Z c +λ x Z gX +λ y Z gY ,

[0019] Among them, Z I is the co-embedding representation, Z c is the common feature representation, Z gX is the reconstructed low-dimensional feature of single-cell transcriptome data, Z gY is the reconstructed low-dimensional feature of single-cell epigenomic data, λ x is the weight coefficient of reconstructing low-dimensional features of single-cell transcriptome data, λ yWeight coefficients for reconstructing low-dimensional features from single-cell epigenomic data.

[0020] Furthermore, the co-embedded representation is projected into multiple parallel latent subspaces, and the process of generating clustering results in the subspaces includes:

[0021] The co-embedded representation matrix is ​​linearly transformed to obtain the Query matrix, Key matrix and Value matrix;

[0022] Combine the Query matrix, Key matrix, and Value matrix to obtain the features of each attention head in each subspace dimension;

[0023] Combine the features of each attention head in each subspace dimension to construct clustering loss and reconstruction loss;

[0024] The total loss of multiple clusters is obtained by summing the clustering loss and the reconstruction loss.

[0025] Optimize the total loss of multiple clusters to get the clustering results.

[0026] Furthermore, the total loss function of multi-clustering is:

[0027]

[0028] in, is the total loss, is the clustering loss, is the reconstruction loss, μ l,k (1≤k≤K l ) represents the kth cluster center of the lth cluster, λ is the regularization parameter, K l is the total number of cluster centers of the lth cluster, O L is the feature representation matrix of the Lth attention head, L is the number of attention heads, Z I is the co-embedding representation matrix.

[0029] The second aspect of the present invention provides a single cell data multi-clustering system based on multi-omics data fusion, comprising:

[0030] A data acquisition module, used to acquire single-cell transcriptome sequencing data and single-cell epigenetic data;

[0031] A low-dimensional feature extraction module is used to map single-cell transcriptome sequencing data and single-cell epigenetic data into their respective nonlinear embedding spaces to obtain low-dimensional representations of the corresponding data;

[0032] A feature reconstruction module is used to extract specific features of different omics data based on low-dimensional representations of different omics data, and reconstruct the low-dimensional representations based on the specific features of different omics data to obtain reconstructed low-dimensional features;

[0033] The embedding representation generation module is used to extract the common features of low-dimensional representations of different omics data; aggregate the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data to generate a co-embedded representation;

[0034] A multi-cluster mining module is used to project the co-embedded representation into multiple parallel latent subspaces and generate clustering results in the subspaces.

[0035] A third aspect of the present invention provides a computer-readable storage medium.

[0036] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the single-cell data multi-clustering method based on multi-omics data fusion as described above.

[0037] A fourth aspect of the present invention provides a computer device.

[0038] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the single-cell data multi-clustering method based on multi-omics data fusion as described above are implemented.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. The present invention guides the fusion and clustering of single-cell multi-omics data by learning high-quality common and specific features from highly sparse heterogeneous single-cell multi-omics data, and discovers potential non-redundant cell clustering patterns in multi-omics data.

[0041] 2. The present invention maps different omics data to their respective nonlinear embedding spaces, extracts specific features of different omics data, and solves the problem of ignoring the specific information of different omics data and the nonlinear attribute characteristics within the data.

[0042] 3. The present invention extracts the common features of low-dimensional representations of different omics data, aggregates the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data, and generates a co-embedded representation; the co-embedded representation is projected into multiple parallel latent subspaces, which solves the problem of ignoring the individuality of omics and failing to integrate biological features at different levels to learn more discriminative representations.

[0043] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0045] Figure 1 It is a flow chart of a single cell data multi-clustering method based on multi-omics data fusion provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0047] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0048] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0049] Regarding the limitations of the current single-cell multi-omics data fusion methods mentioned in the background technology: shallow methods based on non-negative matrix decomposition or principal component analysis mostly project multi-omics data into a shared latent space, ignoring the specific information of different omics and the nonlinear attribute characteristics within the data; manifold alignment methods characterize the intrinsic cell structure by capturing the nonlinear geometric structure within the data, but they have high time complexity; deep learning-based methods use different deep generative networks to model single-cell multi-omics data, most of which focus on a shared representation and ignore the individuality of omics, and cannot integrate biological features at different levels to learn more discriminative representations. More importantly, existing methods cannot fully integrate genetic information from different omics to reveal potential alternative clusters with diversity and high quality, and it is difficult to reveal the different roles and characteristics of cells from different perspectives.

[0050] The present invention uses deep autoencoders independent of omics data to learn low-dimensional representations of each omics; uses an attention mechanism to capture the individual characteristics of each omics, uses a Siamese network to capture the common characteristics between omics, and fuses individual characteristics and common characteristics to obtain information-rich co-embedded representations; uses an autoencoder network based on a multi-head attention mechanism to project the co-embedded representations into multiple significant latent subspaces; optimizes the clustering loss in each subspace, and finally generates diversified, high-quality cell clusters in an end-to-end manner.

[0051] Embodiment 1

[0052] This embodiment provides a single cell data multi-clustering method based on multi-omics data fusion, comprising the following steps:

[0053] Step 1: Obtain single-cell transcriptome sequencing data and single-cell epigenetic data;

[0054] In this example, single cell Assay for Transposase-Accessible Chromatin with high-throughput sequencing (scATAC-seq) was used to obtain single cell epigenetic data; single cell RNA sequencing data (single cell RNA-sequencing, scRNA-seq) was used to obtain single cell transcriptome data.

[0055] Step 2: Preprocess the acquired single-cell transcriptome sequencing data and single-cell epigenetic data to obtain the corresponding data matrices X and Y;

[0056] Among them, the preprocessing of single-cell transcriptome sequencing data includes: evaluating transcriptional activity through alignment and quantitative expression to eliminate technical variations; filtering low-quality cells and low-expressed genes, identifying highly variable genes, performing data normalization and scaling, and eliminating the effects of sequencing depth and cell size.

[0057] Preprocessing of single-cell epigenomic data includes:

[0058] Based on the acquired single-cell epigenetic data, low-quality cells are filtered out by calculating the number of reads for each cell and quality control indicators such as the proportion of mitochondrial genes; the reads are aligned to the reference genome using alignment software BWA or STAR, and reads with low scores are deleted based on the alignment score; duplicate reads are removed using tools such as samtools; open chromatin regions are identified using the MACS2 tool, and the detected open regions are associated with the nearest gene promoter regions to infer the activity status of the genes; a gene activity matrix is ​​created to represent the openness (i.e., activity) of genes in specific cells.

[0059] Step 3: Map the single-cell multi-omics data into their respective nonlinear embedding spaces to obtain a low-dimensional representation of the corresponding data;

[0060] In this embodiment, independent deep autoencoders are used to map single-cell multi-omics data into their own nonlinear embedding spaces to resist noise and outliers. The calculation process can be expressed as:

[0061] ZX =f EX (X); Z Y =f EY (Y) (1),

[0062] Among them, Z X is a low-dimensional representation of single-cell transcriptome data, Z Y is a low-dimensional representation of the single-cell epigenomic data, f Ex (·) and f EY (·) Two omics-independent deep autoencoders.

[0063] Independent deep autoencoders are initialized for different omics data, and high-dimensional single-cell multi-omics data are embedded into different low-dimensional representations to achieve preliminary information extraction and denoising.

[0064] Step 4: Extract the specific features of different omics data based on the low-dimensional representations of different omics, and extract the common features in the low-dimensional representations of different omics; embed the specific features and common features of multi-omics data into the same representation space through weighted fusion to generate a co-embedded representation with consistent information.

[0065] Specifically include:

[0066] Step 401: The low-dimensional representations of different omics are passed to independent attention layers respectively, and the attention weight representation A of the specific features of the single-cell transcriptome data and the single-cell epigenetic data is encoded. X and A Y , the calculation process is:

[0067]

[0068] Among them, A X and A Y The similarities between pairs of cells in different omics low-dimensional representations are quantified respectively; softmax(·) normalizes the weights to [0,1] to avoid the local optimal problem caused by excessive weights of some cells; d represents the dimension of the low-dimensional representation.

[0069] Step 402: reconstruct the low-dimensional features to obtain reconstructed low-dimensional features;

[0070] In this example, the normalized attention weights are used to reconstruct the low-dimensional representation by considering the similarities between cells:

[0071] Z gX =A X Z X ; Z gY =A Y Z Y (3)

[0072] Among them, Z gX is the reconstructed low-dimensional feature of single-cell transcriptome data, Z gY Reconstructing low-dimensional features for single-cell epigenomic data.

[0073] The attention mechanism in the encoding module not only measures the importance of biological signals in the intrinsic feature space of different omics and extracts omics-specific features, but also explores the representation relationship between cells and features from a global perspective.

[0074] Step 403, extracting common features of low-dimensional representations of different omics data;

[0075] In this embodiment, the Siamese network is used to maximize the consistency between low-dimensional representations of different omics data by sharing network parameters. The calculation process is:

[0076]

[0077] Among them, f sia (·) is a shared multi-layer perceptron used to map the low-dimensional representation of heterogeneous omics into the same semantic space, Z c It is the common feature representation of the integration.

[0078] The Siamese network is used to extract common features in the low-dimensional representations of different omics, ensuring that similar cells are close to each other and dissimilar cells are far away from each other in the genetic signals of different omics.

[0079] The present invention utilizes deep neural networks and contrastive learning mechanisms to extract specific features and shared features of single-cell multi-omics data, thereby improving the quality of feature extraction, automatically constructing features, and saving human resources.

[0080] Step 404: Use weighted fusion to aggregate the specific features and common features of multi-omics data to generate a more discriminative co-embedded representation:

[0081] Z I =Z c +λ x Z gX +λ y Z gY (5)

[0082] Among them, λ x is the weight coefficient of reconstructing low-dimensional features of single-cell transcriptome data, λ y Weight coefficients for reconstructing low-dimensional features from single-cell epigenomic data.

[0083] The specific and common features of different omics are integrated through weighted fusion to obtain information-rich co-embedded representation.

[0084] Step 5: Project the co-embedded representation into multiple parallel latent subspaces and generate high-quality and diverse cell clusters by optimizing the clustering loss;

[0085] The specific steps include:

[0086] Step 501: Obtain the Query matrix by linear transformation of the co-embedded representation matrix Key Matrix and Value matrix

[0087] The transformation formula for the lth representation is:

[0088]

[0089] in, and is the transformation weight matrix.

[0090] Step 502: Through the linear transformation matrix obtained in step 501, the feature space representation matrix generated by each attention head can be obtained:

[0091]

[0092] Where m represents the dimension of each subspace. By changing the constraints of the weight matrix, different attention heads will focus on different aspects of the co-embedded representation, which helps to generate diverse clustering patterns.

[0093] Using a multi-head attention mechanism to project the co-embedded representation into different salient subspaces to mine different and meaningful alternative clusters is beneficial for comprehensively mining the complex clustering patterns within single-cell multi-omics data.

[0094] Step 503: Generate a high-quality and diversified feature subspace by jointly optimizing the clustering loss and the reconstruction loss. The specific calculation process is as follows:

[0095]

[0096] in, is the total loss, is the clustering loss, is the reconstruction loss, μ l,k (1≤k≤K l ) represents the kth cluster center of the lth cluster, λ is the regularization parameter, K l is the total number of cluster centers of the lth cluster, O L is the representation matrix of the Lth attention head, L is the number of attention heads, Z I is the co-embedding representation matrix.

[0097] Step 504: Joint Optimization Using co-embedding representation Z I Multiple salient subspaces can be obtained And generate multiple high-quality clusters in it in an end-to-end manner;

[0098] Specifically, update the lth cluster C l When , the label assigned to the i-th sample can be recorded as

[0099] Clustering loss is used to generate high-quality clustering in an end-to-end manner without introducing an additional clustering process; a variety of clustering structures with diversity and quality are discovered, providing new insights into the diverse roles of cells from different perspectives.

[0100] Embodiment 2

[0101] This embodiment provides a single cell data multi-clustering system based on multi-omics data fusion, including:

[0102] A data acquisition module, used to acquire single-cell transcriptome sequencing data and single-cell epigenetic data;

[0103] A low-dimensional feature extraction module is used to map single-cell transcriptome sequencing data and single-cell epigenetic data into their respective nonlinear embedding spaces to obtain low-dimensional representations of the corresponding data;

[0104] A feature reconstruction module is used to extract specific features of different omics data based on low-dimensional representations of different omics data, and reconstruct the low-dimensional representations based on the specific features of different omics data to obtain reconstructed low-dimensional features;

[0105] The embedding representation generation module is used to extract the common features of low-dimensional representations of different omics data; aggregate the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data to generate a co-embedded representation;

[0106] A multi-cluster mining module is used to project the co-embedded representation into multiple parallel latent subspaces and generate clustering results in the subspaces.

[0107] Embodiment 3

[0108] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the single-cell data multi-clustering method based on multi-omics data fusion as described above are implemented.

[0109] Embodiment 4

[0110] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the single-cell data multi-clustering method based on multi-omics data fusion as described above are implemented.

[0111] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, 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 device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0115] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A single-cell data multi-clustering method based on multi-omics data fusion, characterized in that: The steps include: Obtain single-cell transcriptome sequencing data and single-cell epigenetic data; Mapping single-cell transcriptome sequencing data and single-cell epigenetic data into their respective nonlinear embedding spaces to obtain low-dimensional representations of the corresponding data; Extracting specific features of different omics data based on low-dimensional representations of different omics data, reconstructing the low-dimensional representations based on the specific features of different omics data, and obtaining reconstructed low-dimensional features; Extract common features of low-dimensional representations of different omics data; Aggregate the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data to generate a co-embedded representation; The co-embedded representation is projected into multiple parallel latent subspaces and clustering results are generated in the subspaces.

2. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: Independent deep autoencoders are used to map single-cell multi-omics data into their own nonlinear embedding spaces.

3. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: Using the normalized attention scores, a low-dimensional representation is reconstructed by considering the similarities between cells.

4. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: The Siamese network is used to maximize the consistency between low-dimensional representations of different omics data by sharing network parameters, and to extract common features of low-dimensional representations of different omics data.

5. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: The specific and common features of multi-omics data are aggregated using weighted fusion to generate a co-embedded representation: WITH I =Z c +λ x WITH gX +λ y WITH gY , Among them, Z I is the co-embedding representation, Z c is the common feature representation, Z gx is the reconstructed low-dimensional feature of single-cell transcriptome data, Z gY is the reconstructed low-dimensional feature of single-cell epigenomic data, λ x is the weight coefficient of reconstructing low-dimensional features of single-cell transcriptome data, λ y Weight coefficients for reconstructing low-dimensional features from single-cell epigenomic data.

6. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: The co-embedded representation is projected into multiple parallel latent subspaces. The process of generating clustering results in the subspaces includes: The co-embedded representation matrix is ​​linearly transformed to obtain the Query matrix, Key matrix and Value matrix; Combine the Query matrix, Key matrix, and Value matrix to obtain the features of each attention head in each subspace dimension; Combine the features of each attention head in each subspace dimension to construct clustering loss and reconstruction loss; The total loss of multiple clusters is obtained by summing the clustering loss and the reconstruction loss. Optimize the total loss of multiple clusters to get the clustering results.

7. The single cell data multi-clustering method based on multi-omics data fusion according to claim 1, characterized in that: The total loss function for multi-clustering is: in, is the total loss, is the clustering loss, is the reconstruction loss, μ l,k (1≤k≤K l ) represents the kth cluster center of the lth cluster, λ is the regularization parameter, K l is the total number of cluster centers of the lth cluster, O L is the feature representation matrix of the Lth attention head, L is the number of attention heads, Z I is the co-embedding representation matrix.

8. Single cell data multi-clustering system based on multi-omics data fusion, characterized by: include: A data acquisition module, used to acquire single-cell transcriptome sequencing data and single-cell epigenetic data; A low-dimensional feature extraction module is used to map single-cell transcriptome sequencing data and single-cell epigenetic data into their respective nonlinear embedding spaces to obtain low-dimensional representations of the corresponding data; A feature reconstruction module is used to extract specific features of different omics data based on low-dimensional representations of different omics data, and reconstruct the low-dimensional representations based on the specific features of different omics data to obtain reconstructed low-dimensional features; The embedding representation generation module is used to extract the common features of low-dimensional representations of different omics data; aggregate the reconstructed low-dimensional features and the common features of low-dimensional representations of different omics data to generate a co-embedded representation; A multi-cluster mining module is used to project the co-embedded representation into multiple parallel latent subspaces and generate clustering results in the subspaces.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the multi-clustering of single-cell data based on multi-omics data fusion as described in any one of claims 1 to 7 are implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the multi-clustering of single-cell data based on multi-omics data fusion as described in any one of claims 1 to 7 are implemented.

Images