Cell type identification methods, systems and equipment based on multi-omics data integration

Abstract

The present invention discloses a cell type identification method, system, and device based on multi-omics data integration, relating to the field of cell type identification. The method can effectively mine the specific information of each omics data set and the shared information between different omics data sets; align the shared information between different omics data sets by applying instance-level comparative learning, thereby enhancing their consistency; and apply orthogonality constraints to enforce the separation of specific and shared information, thereby improving information purity and reducing redundancy and mutual contamination. The resulting sample embedding representation is complete, which is crucial for multi-omics data integration and cell type identification, and can significantly improve the accuracy of cell type identification.

Description

Technical Field

[0001] The present invention relates to the technical field of cell type identification, and in particular to a cell type identification method, system and device based on multi-omics data integration. Background Art

[0002] With the rapid development of high-throughput biotechnology, it has become increasingly feasible to obtain stable, reliable, and large-scale single-cell multi-omics data such as genome, transcriptome, proteome, and metabolome. Each omics contains partial information about the cell, and integrating multiple omics can provide a systematic approach to single-cell research. Multi-omics data can provide a more macroscopic perspective for further understanding and recognition of multiple cell types, and help to unravel the specific causal networks in each cell type. Since obtaining true cell type labels is difficult and costly, multi-omics clustering methods are often used in the study of multi-omics data integration and cell type identification. Although many multi-omics clustering methods have been developed, existing methods often do not consider the simultaneous learning of specific information and shared information in single-cell multi-omics data, ignoring the complementarity and consistency of multi-omics data, resulting in low accuracy in cell type identification. Summary of the Invention

[0003] In response to the problems raised in the above background technology, the present invention provides a cell type identification method, system and device based on multi-omics data integration to improve the accuracy of cell type identification.

[0004] To achieve the above object, the present invention provides the following solutions:

[0005] In one aspect, the present invention provides a method for cell type identification based on multi-omics data integration, comprising:

[0006] Use cell sequencing technology to obtain multi-omics data sets from multiple single-cell samples;

[0007] Extracting shared information and specific information of each omics data in the multi-omics data set using an autoencoder;

[0008] The extracted shared information is projected into the subspace using a nonlinear projection function to obtain the projected shared information;

[0009] Applying instance-level contrastive learning to strengthen the consistency of the projected shared information to obtain strengthened shared information;

[0010] Applying orthogonality constraints to separate the enhanced shared information from the specific information;

[0011] Construct a complete embedding representation of single-cell samples based on the separated shared information and specific information;

[0012] A clustering algorithm is used to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and the cluster labels corresponding to the clusters are obtained as the cell type identification results.

[0013] Optionally, the method of obtaining a multi-omics data set of multiple single-cell samples using cell sequencing technology specifically includes:

[0014] Use cell sequencing technology to obtain a multi-omics data set of multiple single-cell samples X = {X 1 , X 2 ,...,X V}; The v-th omics data included in the multi-omics data set X 1≤v≤V, V represents the number of omics included in the multi-omics data set X; Represents the v-th omics data X v The i-th single-cell sample data in , 1≤i≤N, N is the number of single-cell samples; the omics data includes genomics data, transcriptomics data, and metabolomics data.

[0015] Optionally, the extracting shared information and specific information of each omics data in the multi-omics data set using an autoencoder specifically includes:

[0016] Each omics data X in the multi-omics data set is v Input into two autoencoders to extract each omics data X v Shared information and specific information

[0017] Optionally, projecting the extracted shared information onto a subspace using a nonlinear projection function to obtain the projected shared information specifically includes:

[0018] Using nonlinear projection function Shared information will be extracted Projecting to the subspace, obtaining the shared information after projection; where the nonlinear projection function f(·) is a two-layer neural network; Extracted shared information The initial embedding representation of the i-th single-cell sample in; is the embedded representation of shared information after projection.

[0019] Optionally, applying an orthogonality constraint to separate the enhanced shared information from the specific information specifically includes:

[0020] Applying orthogonality constraints Separate the enhanced shared information from the specific information; wherein L ort represents the loss function; represents the Frobenius norm squared; C v Represents the shared information of N single-cell samples separated from the v-th omics data The shared information matrix composed of v Represents the specific information of N single cell samples separated from the v-th omics data A specific information matrix.

[0021] Optionally, constructing a complete embedded representation of the single-cell sample based on the separated shared information and specific information specifically includes:

[0022] Based on the separated shared information Using the formula Calculate the average value z of the embedding representation of shared information c，i ;

[0023] Using the formula The average value z of the embedding representation of shared information c，i Embedded representation of specific information from all omics data separated Perform head-to-tail splicing to construct the complete embedding representation z of the i-th single cell sample i ; Among them, || represents the splicing symbol;

[0024] The embedding representation matrix Z of all single-cell samples is expressed as Z = [z1,z2,...,z N ].

[0025] Optionally, the method of using a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters and obtaining cluster labels corresponding to the clusters as cell type identification results specifically includes:

[0026] The clustering algorithm is used to embed the complete representation of N single cell samples Z=[z1,z2,...,z N ] is divided into K clusters, and the cluster label corresponding to each cluster is obtained as the cell type identification result of the single cell sample contained in the cluster; the cell types include T cells, B cells, NK cells, and hepatocytes.

[0027] In another aspect, the present invention provides a cell type identification system based on multi-omics data integration, comprising:

[0028] Multi-omics data acquisition module, used to obtain multi-omics data sets of multiple single-cell samples using cell sequencing technology;

[0029] a shared information and specific information extraction module, configured to extract shared information and specific information of each omics data in the multi-omics data set using an autoencoder;

[0030] A shared information nonlinear projection module is used to project the extracted shared information into a subspace using a nonlinear projection function to obtain the projected shared information;

[0031] An instance-level contrastive learning module, configured to apply instance-level contrastive learning to enhance the consistency of the projected shared information to obtain enhanced shared information;

[0032] an orthogonality constraint module, configured to separate the enhanced shared information from the specific information by applying an orthogonality constraint;

[0033] A shared information and specific information splicing module is used to construct a complete embedding representation of single-cell samples based on the separated shared information and specific information;

[0034] The clustering and cell type identification module is used to use a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and obtain the cluster labels corresponding to the clusters as the cell type identification results.

[0035] On the other hand, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the cell type identification method based on multi-omics data integration when executing the computer program.

[0036] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which implements the cell type identification method based on multi-omics data integration when the computer program is executed.

[0037] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0038] The cell type identification method, system and equipment based on multi-omics data integration provided by the present invention can effectively mine the specific information of each omics data and the shared information between different omics data; by applying instance-level comparative learning to align the shared information between different omics, its consistency is enhanced; the application of orthogonality constraints to force the separation of specific information and shared information improves the purity of information and reduces redundancy and mutual contamination; the sample embedding representation obtained in this way is complete, and a more complete sample embedding representation is very important for multi-omics data integration and cell type identification, which can greatly improve the accuracy of cell type identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0040] Figure 1 This is a flow chart of a cell type identification method based on multi-omics data integration provided by the present invention. DETAILED DESCRIPTION

[0041] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] The purpose of the present invention is to provide a cell type identification method, system and device based on multi-omics data integration to greatly improve the accuracy of cell type identification.

[0043] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Figure 1 A flow chart of a cell type identification method based on multi-omics data integration provided by the present invention is provided in FIG. Figure 1 , a cell type identification method based on the integration of multi-omics data, including:

[0045] Step 1: Use cell sequencing technology to obtain multi-omics data sets from multiple single-cell samples.

[0046] Currently, cell sequencing technology has enabled the acquisition of multi-omics data such as genomics, transcriptomics, and metabolomics from single cells. Assume that the multi-omics data set of N single-cell samples is represented by X = {X 1 , X 2 ,...,X V}, where V represents the number of omics. (1≤v≤V) represents a cell sample set from the vth omics, which contains N single-cell samples, each with d v dimensional features. The v-th omics data X v The i-th single cell sample data in is represented as 1≤i≤N.

[0047] Step 2: Use an autoencoder to extract shared information and specific information of each omics data in the multi-omics data set.

[0048] For each omics data in the multi-omics data set X, each omics data X v Input into two autoencoders and extract the middle layer embedding representation as the shared information of each group and specific information To obtain the initial embedding representation. Each omics data X v There are corresponding autoencoders to adapt to different input data dimensions d v . Extract specific information Represents the unique information of each omics. Shared information It refers to some common properties between different omics. The shared information learned from different omics should be consistent with each other.

[0049] Step 3: Use the nonlinear projection function to project the extracted shared information into the subspace to obtain the projected shared information.

[0050] The projection process is as follows:

[0051]

[0052] The nonlinear projection function f(·) is a two-layer neural network; Extracted shared information The initial embedding representation of the i-th single-cell sample in; for The corresponding embedded representation of shared information after projection.

[0053] Step 4: Apply instance-level contrastive learning to strengthen the consistency of the projected shared information to obtain the strengthened shared information.

[0054] This paper applies contrastive learning to align shared information extracted from different omics data To reduce information loss, the extracted shared information is first projected into a subspace using a nonlinear projection function (1). Instance-level contrastive learning is then applied to the projected shared information to enforce the consistency of shared information across different omics data. Instance-level contrastive learning maximizes the cosine similarity between different instances of the same sample across different omics. After contrastive learning, the shared information across different omics data has a high cosine similarity, thus achieving alignment.

[0055] Step 5: Apply orthogonality constraints to separate the enhanced shared information from the specific information.

[0056] In order to reduce the redundancy and contamination between shared information and specific information, the present invention applies orthogonality constraints to separate these two types of information to improve the purity of the two types of information. The orthogonality constraint forces the shared information and specific information obtained from the autoencoder to meet the orthogonality, as shown in the following formula:

[0057]

[0058]

[0059]

[0060] Among them C v Represents the shared information of N samples extracted from the v-th omics The shared information matrix composed of v Represents the specific information of N samples extracted from the v-th omics The specific information matrix composed of L ort represents the loss function; stands for the Frobenius norm squared.

[0061] Step 6: Construct a complete embedding representation of the single-cell sample based on the separated shared information and specific information.

[0062] Shared information of each omics group after comparative learning are already highly similar, so their average is used to represent the shared information of the samples.

[0063]

[0064] The average value z of the embedding representation of shared information c，i and an embedded representation of all omics-specific information Perform head-to-tail splicing to represent the complete embedding representation of the single-cell sample. The formula is as follows:

[0065]

[0066] Therefore, the embedding representation matrix Z of all N samples can be expressed as:

[0067] Z=[z1,z2,...,z N ] T (7)

[0068] where z i represents the complete embedding representation of the i-th single-cell sample; || represents the splicing symbol.

[0069] Step 7: Use a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and obtain the cluster labels corresponding to the clusters as the cell type identification results.

[0070] The k-means clustering algorithm is an iterative clustering analysis algorithm. It divides the data into K groups, randomly selects K objects as initial cluster centers, calculates the distance between each object and each seed cluster center, and assigns each object to the cluster center closest to it. The cluster centers and the objects assigned to them represent a cluster.

[0071] The computer applies the K-means clustering algorithm on the embedding representation matrix Z to embed the representation z of all N single cell samples. i The cells are divided into K clusters to obtain clustering results. Each cell cluster is assigned a corresponding cluster label, i.e., a cell type label, thereby integrating multi-omics data and identifying cell types, thereby providing more accurate labels for cell annotation. Cell types (i.e., labels) can include T cells, B cells, NK cells, hepatocytes, etc. These cells can also be distinguished into more specific subtypes, such as CD45 and CD3 in T cells.

[0072] Compared to existing multi-omics data integration and cell type identification methods, the proposed method demonstrates higher accuracy and detection capabilities in cell type identification tasks. Cell type identification can reveal the distribution of cell types, providing more accurate labels for cell annotation and discovering novel cells with unknown functions, which is of great significance for clinical testing and information analysis.

[0073] Based on the method provided by the present invention, the present invention also provides a cell type identification system based on multi-omics data integration, comprising:

[0074] Multi-omics data acquisition module, used to obtain multi-omics data sets of multiple single-cell samples using cell sequencing technology;

[0075] a shared information and specific information extraction module, configured to extract shared information and specific information of each omics data in the multi-omics data set using an autoencoder;

[0076] A shared information nonlinear projection module is used to project the extracted shared information into a subspace using a nonlinear projection function to obtain the projected shared information;

[0077] An instance-level contrastive learning module, configured to apply instance-level contrastive learning to enhance the consistency of the projected shared information to obtain enhanced shared information;

[0078] an orthogonality constraint module, configured to separate the enhanced shared information from the specific information by applying an orthogonality constraint;

[0079] A shared information and specific information splicing module is used to construct a complete embedding representation of single-cell samples based on the separated shared information and specific information;

[0080] The clustering and cell type identification module is used to use a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and obtain the cluster labels corresponding to the clusters as the cell type identification results.

[0081] Furthermore, the present invention provides an electronic device, which may include: a processor, a communication interface, a memory, and a communication bus. The processor, the communication interface, and the memory communicate with each other via the communication bus. The processor may invoke a computer program in the memory to execute the aforementioned method for cell type identification based on multi-omics data integration.

[0082] In addition, when the computer program in the above-mentioned memory is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

[0083] Furthermore, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed, can implement the cell type identification method based on multi-omics data integration.

[0084] To fully explore and utilize the shared and specific information in different omics data, this paper proposes an unsupervised method based on shared and specific representation learning to achieve multi-omics data integration and cell type identification. Compared with existing multi-omics data integration and cell type identification methods, the method proposed in this paper can effectively mine the shared information between omics and the specific information within omics in multi-omics data, and reduce the redundancy and mutual interference between different types of information. It strengthens the consistency of shared information through instance-level comparative learning, and uses orthogonality constraints to enforce the separation of shared information and specific information, thereby improving information purity and greatly improving the accuracy of cell type identification.

[0085] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0086] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A cell type identification method based on multi-omics data integration, characterized in that: include: Use cell sequencing technology to obtain multi-omics data sets from multiple single-cell samples; Extracting shared information and specific information of each omics data in the multi-omics data set using an autoencoder; The extracted shared information is projected into the subspace using a nonlinear projection function to obtain the projected shared information; The method of projecting the extracted shared information onto the subspace using a nonlinear projection function to obtain the projected shared information specifically includes: Using nonlinear projection function Shared information will be extracted Projecting to the subspace, obtaining the shared information after projection; where the nonlinear projection function f(·) is a two-layer neural network; Extracted shared information The initial embedding representation of the i-th single-cell sample in; is the embedded representation of shared information after projection; Applying instance-level contrastive learning to strengthen the consistency of the projected shared information to obtain strengthened shared information; Applying orthogonality constraints to separate the enhanced shared information from the specific information; The applying orthogonality constraint to separate the enhanced shared information from the specific information specifically includes: Apply orthogonality constraints Separate the enhanced shared information from the specific information; wherein L ort represents the loss function; represents the Frobenius norm squared; C v Represents the shared information of N single-cell samples separated from the v-th omics data The shared information matrix composed of v Represents the specific information of N single cell samples separated from the v-th omics data The specific information matrix composed of; V represents the number of omics included in the multi-omics data set X; Construct a complete embedding representation of single-cell samples based on the separated shared information and specific information; A clustering algorithm is used to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and the cluster labels corresponding to the clusters are obtained as the cell type identification results.

2. The cell type identification method based on multi-omics data integration according to claim 1, characterized in that The multi-omics data set of multiple single-cell samples obtained by using cell sequencing technology specifically includes: Use cell sequencing technology to obtain a multi-omics data set of multiple single-cell samples X = {X 1 , X 2 ,...,X V }; The v-th omics data included in the multi-omics data set X 1≤v≤V; Represents the v-th omics data X v The i-th single-cell sample data in , 1≤i≤N, N is the number of single-cell samples; the omics data includes genomics data, transcriptomics data, and metabolomics data.

3. The cell type identification method based on multi-omics data integration according to claim 2, characterized in that: The extracting shared information and specific information of each omics data in the multi-omics data set using an autoencoder specifically includes: Each omics data X in the multi-omics data set is v Input into two autoencoders to extract each omics data X v Shared information and specific information 4. The cell type identification method based on multi-omics data integration according to claim 3, characterized in that: The complete embedding representation of the single-cell sample is constructed based on the separated shared information and specific information, specifically including: Based on the separated shared information Using the formula Calculate the average value z of the embedding representation of shared information c，i ; Using the formula The average value z of the embedding representation of shared information c，i Embedded representation of specific information from all omics data separated Perform head-to-tail splicing to construct the complete embedding representation z of the i-th single cell sample i ; Among them, || represents the splicing symbol; The embedding representation matrix Z of all single-cell samples is expressed as Z = [z1,z2,...,z N ].

5. The cell type identification method based on multi-omics data integration according to claim 4, characterized in that: The method uses a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and obtains cluster labels corresponding to the clusters as cell type identification results, specifically including: The clustering algorithm is used to embed the complete representation of N single cell samples Z=[z1,z2,...,z N ] is divided into K clusters, and the cluster label corresponding to each cluster is obtained as the cell type identification result of the single cell sample contained in the cluster; the cell types include T cells, B cells, NK cells, and hepatocytes.

6. A cell type identification system based on multi-omics data integration, characterized in that: For implementing the cell type identification method based on multi-omics data integration according to any one of claims 1 to 5, the cell type identification system based on multi-omics data integration comprises: Multi-omics data acquisition module, used to obtain multi-omics data sets of multiple single-cell samples using cell sequencing technology; a shared information and specific information extraction module, configured to extract shared information and specific information of each omics data in the multi-omics data set using an autoencoder; A shared information nonlinear projection module is used to project the extracted shared information into a subspace using a nonlinear projection function to obtain the projected shared information; An instance-level contrastive learning module, configured to apply instance-level contrastive learning to enhance the consistency of the projected shared information to obtain enhanced shared information; an orthogonality constraint module, configured to separate the enhanced shared information from the specific information by applying an orthogonality constraint; A shared information and specific information splicing module is used to construct a complete embedding representation of single-cell samples based on the separated shared information and specific information; The clustering and cell type identification module is used to use a clustering algorithm to divide the complete embedded representations of multiple single-cell samples into multiple clusters, and obtain the cluster labels corresponding to the clusters as the cell type identification results.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the cell type identification method based on multi-omics data integration according to any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the cell type identification method based on multi-omics data integration according to any one of claims 1 to 5 is implemented.

Images