Single-cell omics feature amplification method and device and storage medium
By performing meta-cellular processing and dimensionality reduction on single-cell transcriptome data, the problems of data noise and high computational resource consumption were solved, enabling efficient training and analysis of machine learning classifiers.
Patent Information
- Application Number
- CN202511060046.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2025-07-29
- Filing Date
- 2025-07-30
- Publication Date
- 2025-11-07
AI Technical Summary
Single-cell transcriptomics technology suffers from data noise and high computational resource requirements, which affect the application and analysis efficiency of machine learning classifiers.
Meta-cell processing of single-cell transcriptome data, including clustering and merging into dense meta-cell dimensionality reduction matrices, is performed. PCA dimensionality reduction and filtering of non-marker genes are then used to generate a marker gene expression matrix for subsequent machine learning classifier model training.
It significantly reduces computational load, improves analytical efficiency and the accuracy and stability of machine learning classifiers, reduces the number of features after dimensionality reduction, and eliminates collinearity between features.
Smart Images
Figure CN120913664A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of bioinformatics, in particular to a single-cell omics feature amplification method, device and storage medium. BACKGROUND
[0002] The emergence of single-cell transcriptome technology has brought revolutionary breakthroughs to life science research. This technology can accurately obtain gene expression difference information of single cells at single-cell resolution, enabling scientists to deeply explore key biological phenomena such as the origin, function and differentiation of cells, as well as the interaction between cells in tissues and organs, and providing a new perspective for understanding the function of tissues and organs.
[0003] Although single-cell transcriptome technology has made significant progress, it still faces many challenges. On the one hand, due to technical bottlenecks, data noise problems are more prominent. For example, the widely existing dropout phenomenon, due to the small number of RNA transcriptome starting in a single cell, the randomness of gene expression and fragment capture, and the amplification bias, etc., leads to a high probability of misjudging non-zero values as zero in single-cell data. When dropout occurs in key genes, it will seriously affect the cell type annotation based on marker genes and downstream data processing and model fitting, increase the complexity of analysis and reduce the reliability of the results. On the other hand, with the development of single-cell sequencing technology, the cell capture throughput is continuously improved, and the massive data generated puts a huge pressure on the demand for computing resources for data analysis and storage. For example, although some single-cell sequencing platforms based on microwells or droplets can obtain high-throughput sequencing data of a large number of cells, powerful computing devices and complex algorithms are needed to process these data in the data processing and analysis process, which limits the wide application and further development of the technology.
[0004] In the field of artificial intelligence and data processing, classification task is one of the core applications of machine learning, aiming to accurately divide samples into different categories according to their features. With the development of machine learning technology, various classifiers have emerged, such as support vector machine (SVM), random forest, gradient boosting tree, multi-layer perception (MLP), convolutional neural network (CNN), etc. No matter which machine learning classification method, the training data itself needs to have less noise and more obvious data rules. However, in single-cell omics, there is a general noise problem in the data, and the cell-gene expression matrix obtained is usually a sparse matrix, which is very unfavorable for the application of machine learning classifiers. Moreover, the sample size (i.e. cell quantity) of single-cell omics is huge, with tens of thousands or even hundreds of millions of cells, and the calculation amount is even larger when applying machine learning classifiers. SUMMARY
[0005] The application provides a single-cell omics feature amplification method, device and storage medium to at least solve the above technical problems in the prior art.
[0006] According to a first aspect of the application, a single-cell omics feature amplification method is provided, comprising the following steps: S101, providing a gene expression matrix of a single-cell transcriptome, classifying cells in the gene expression matrix and labeling a classification label; S102, splitting the gene expression matrix according to the classification label, clustering cells labeled with the same classification label and generating a gene expression matrix of a meta-cell; S103, performing normalization processing on the gene expression matrix of the meta-cell and calculating marker genes of each classification label in the meta-cell; filtering non-marker genes in the gene expression matrix of the meta-cell to obtain a marker gene expression matrix; S104, calculating a meta-cell dimension reduction matrix based on the marker gene expression matrix, for subsequent model training of a machine learning classifier.
[0007] In some embodiments of the first aspect of the application, in the S101, low-expression genes in the gene expression matrix are counted and filtered; comprising the following steps: S1011, calculating the expression proportion of each gene in the gene expression matrix by the following formula: P gene_i =N gene_i / N total *100% Wherein: P gene_i is the gene expression proportion of gene gene_i in all cells, N gene_i is the number of cells expressing gene gene_i, and N total is the total number of cells; S1012, if the expression proportion of a gene is lower than a set proportion threshold, the gene is determined as a low-expression gene; S1013, filtering the low-expression genes.
[0008] In some embodiments of the first aspect of the application, the S102 is implemented by the following S1021 to S1024. S1021, selecting cells with the same classification label in the gene expression matrix, performing data preprocessing, and generating a cell-PCA dimension reduction matrix for each classification label; S1022, calculating the most similar k cells of each cell based on the cell-PCA dimension reduction matrix; S1023, generating a gene expression matrix of the meta-cell, setting an expected number of meta-cells N; randomly selecting N cells in the cell-PCA dimension reduction matrix; in the N cells, merging the original gene expression matrix of each cell with the original gene expression matrix of the k most similar cells of the cell one by one to obtain and calculate the average expression of each gene to obtain a gene average expression matrix; S1024, repeating S1021 to S1023 for the single-cell transcriptome data of each classification label, that is, obtaining the gene expression count matrix of N meta-cells for each classification label.
[0009] In some embodiments of the first aspect of the application, in S1022, the method for calculating the k most similar cells of each cell is as follows: Calculate the Euclidean distance between each cell in the cell-PCA dimension reduction matrix, and based on the Euclidean distance between the cells, use the k-means algorithm to cluster the cells into several categories; for each cell, use the k-nearest neighbor algorithm to calculate the k cells with the smallest Euclidean distance to the cell as the k most similar cells of the cell.
[0010] In some embodiments of the first aspect of the application, the number N of meta-cells is 1% of the number of cells of the classification label; but when the number of cells of the classification label is less than 10,000, N is set to 100; when the number of cells of the classification label is greater than 100,000, N is set to 1,000.
[0011] In some embodiments of the first aspect of the application, in S1021, the data preprocessing includes normalizing, standardizing, high-variable gene calculation, and PCA dimension reduction of cells of the same classification label.
[0012] In some embodiments of the first aspect of the application, in S103, the method for normalizing the gene expression matrix of the meta-cell is as follows: NormExpr=log(1+(Count raw / Count total *10000)) Where: NormExpr is the expression after gene normalization, Count raw is the expression count of a single gene in a single meta-cell, Count total is the total expression count of all genes in a single meta-cell.
[0013] In some embodiments of the first aspect of the application, in S104, the method for calculating the meta-cell dimension reduction matrix based on the marker gene expression matrix is as follows: First, use PCA to calculate the projection matrix V of the marker gene expression matrix X meta Based on the formula Y meta =Xmeta Y can be calculated meta .
[0014] According to a second aspect of the present application, a single cell omics feature amplification system is provided, comprising: a data providing module configured to provide a gene expression matrix of a single cell transcriptome, classify cells in the gene expression matrix and label a classification label; a clustering module configured to split the gene expression matrix according to the classification label, cluster cells labeled with the same classification label and generate a gene expression matrix of a meta cell; a filtering module configured to normalize the gene expression matrix of the meta cell and calculate marker genes of each classification label in the meta cell, filter non-marker genes in the gene expression matrix of the meta cell and obtain a marker gene expression matrix; a generating module configured to calculate a meta cell dimension reduction matrix based on the marker gene expression matrix, for subsequent model training of a machine learning classifier.
[0015] According to a third aspect of the present application, an electronic device is provided, comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in the present application.
[0016] According to a fourth aspect of the present application, a non-transitory computer readable storage medium storing computer instructions is provided, the computer instructions being used to enable the computer to perform the method described in the present application.
[0017] Compared with the prior art, the present application has the following beneficial effects: The present application clusters and merges sparse single cell transcriptome data into dense meta cell dimension reduction matrix through meta cell processing, greatly reduces the number of samples (number of cells), reduces the number of features and eliminates the collinearity between features by reducing the dimension of ten thousand gene features to a small number of PC features through PCA. Therefore, compared with the prior art, the technical solution provided by the present application greatly reduces the amount of calculation in the analysis process, improves the analysis efficiency, and is conducive to increasing the accuracy and stability of the subsequent machine learning classifier.
[0018] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS
[0019] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description read in conjunction with the accompanying drawings, in which: In the drawings, identical or corresponding numbers refer to identical or corresponding parts.
[0020] Figure 1 A whole method flow chart of the embodiment one of the present application is shown.
[0021] Figure 2 A system principle diagram of the embodiment two of the present application is shown.
[0022] Figure 3 A composition structure schematic diagram of an electronic device of the embodiment three of the present application is shown. DETAILED DESCRIPTION
[0023] In order to make the purposes, features and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.
[0024] Embodiment one: Referring to Figure 1 The embodiment one provides a single cell omics feature amplification method, comprising the following steps: S101, providing a gene expression matrix of a single cell transcriptome, classifying and labeling the cells in the gene expression matrix to obtain a gene expression matrix of a certain single cell transcriptome as shown in Table 1: Table 1: Gene expression matrix of a certain single cell transcriptome
[0025] In the gene expression matrix, there are a certain amount of low expression genes, which will increase the data dimension (such as too many genes), cause the calculation amount to increase dramatically (such as clustering analysis, machine learning model training), and reduce the analysis efficiency. Therefore, preferably, in the step S101, the low expression genes in the gene expression matrix are counted and filtered; comprising the following steps: S1011, the expression proportion of each gene in the gene expression matrix is calculated by the following formula: P gene_i =N gene_i / N total *100% P = N gene_i is the proportion of gene expression of gene gene_i in all cells, N gene_i is the number of cells expressing gene gene_i, N total is the total number of cells; S1012, if the expression proportion of the gene is lower than the set proportion threshold, the gene is determined as a low expression gene; the proportion threshold can be 0.5%, 1% or 2%, preferably 1%.
[0026] S1013, filtering the low expression gene, that is, in the gene expression matrix shown in Table 1, deleting the row data corresponding to the low expression gene, obtaining the gene expression matrix after filtering the low expression gene as shown in Table 2.
[0027] Table 2: Gene expression matrix after filtering low expression genes
[0028] S102, splitting the gene expression matrix according to the classification label, clustering the cells labeled with the same classification label and generating the gene expression matrix of the meta-cell.
[0029] Specifically, the step S102 is realized by the following steps S1021 to S1024: S1021, selecting cells with the same classification label in the gene expression matrix, performing data preprocessing, including using the NormalizeData, FindVariableFeatures, ScaleData, RunPCA functions of the seurat software to perform normalization, standardization, high variable gene calculation, PCA dimension reduction and the like, obtaining the cell-PCA dimension reduction matrix of each classification label as shown in Table 3.
[0030] The core of feature dimension reduction is to reduce the number of features while retaining the key information of the data to the greatest extent, so as to improve the efficiency and generalization ability of the model. PCA (Principal Component Analysis) is an unsupervised feature dimension reduction method, which projects high-dimensional data to low-dimensional orthogonal space (principal component, i.e. PC) through linear transformation, and retains the maximum variance information. The core is the eigenvalue decomposition of the covariance matrix, and the data compression is realized by selecting the top several principal components (with the largest contribution rate). PCA algorithm itself has the advantages of removing noise, reducing redundant features, relieving dimension disaster, eliminating correlation between features, etc.
[0031] Table 3: Cell-PCA dimension reduction matrix of a single classification label obtained after data preprocessing
[0032] S1022, based on the cell-PCA dimensionality reduction matrix, the Euclidean distance between each cell in the cell-PCA dimensionality reduction matrix is calculated. Based on the Euclidean distance between cells, the cells are clustered into several categories using the k-means algorithm, and the number of clusters can be set to 10 or 15. And use k-nearest neighbor algorithm (KNN, k-nearest neighbor) to calculate the k cells (k is a hyperparameter, which can be adjusted according to the number of cells, set to any integer greater than 1, usually set to 25) most similar (i.e. minimum Euclidean distance) to each cell.
[0033] S1023, generate the gene expression matrix of the metacell. Set the expected number of metacells N (N takes the value of 100~1000, takes 1% of the number of cells of this classification label; but when the number of cells of this classification label is less than 10000, N is set to 100; when the number of cells of this cell type is greater than 100000, N is set to 1000).
[0034] Randomly select N cells in the cell-PCA dimensionality reduction matrix (N cells do not overlap, and must be distributed in the 10 k-means clusters obtained in step S1022).
[0035] In the N cells, as shown in Table 4, the original gene expression matrix of each cell and the k cells most similar to it is merged to obtain the average expression of each gene, and the average gene expression matrix as shown in Table 5 is obtained.
[0036] Table 4: Randomly selected cells and their most similar k cells' original expression profile
[0037] Table 5: Average gene expression matrix
[0038] Therefore, N cells generate N gene average expression matrices, each of which is a gene expression matrix of a metacell, that is, N gene expression matrices of metacells as shown in Table 6 are generated.
[0039] Table 6: Gene expression count matrix of N metacells
[0040] S1024, repeat steps S1021 to S1023 for each classification label of single cell transcriptome data, that is, N gene expression count matrices of metacells can be obtained for each classification label, and after merging, the complete metacell gene expression matrix corresponding to the single cell transcriptome as shown in Table 7 is obtained.
[0041] Table 7: Complete metacell gene expression matrix
[0042] S103, normalizing the gene expression matrix of the meta cell and calculating the marker genes of each classification label in the meta cell. Filtering non-marker genes in the gene expression matrix of the meta cell to obtain a marker gene expression matrix Xmeta.
[0043] Specifically comprising the following steps: S1031, normalizing the gene expression matrix of the meta cell, and normalizing the expression amount of each gene of each meta cell by the following formula: NormExpr=log(1+(Count raw / Count total *10000)) Wherein: NormExpr is the expression amount of the gene after normalization, Count raw is the expression count of a single gene in a single meta cell, and Count total is the total expression count of all genes in a single meta cell. Obtain the normalized gene expression matrix of the meta cell as shown in Table 8; Table 8: Normalized gene expression matrix of the meta cell
[0044] S1032, calculating the marker genes of each classification label based on the normalized gene expression matrix of the meta cell.
[0045] Using the FindAllMarkers function of the seurat software and the wilcox difference test method to calculate the avg_log2FC (log2 processed difference multiple) and p value (i.e. significance level) of each gene under each classification label in the normalized gene expression matrix of the meta cell - essentially wilcox rank sum test is performed one by one between the meta cell of a single classification label and the genes of the remaining other meta cells to obtain the genes specifically highly expressed in each classification label. In each classification label, the genes are sorted in ascending order of p value, and for genes with the same p value, they are sorted in descending order of avg_log2FC. As shown in Table 9, only the top 50 marker genes (the number of marker genes retained can be set by oneself) are retained for each classification label, and the marker genes of all classification labels are combined into a marker gene set.
[0046] Table 9: TOP50 marker genes under each classification label
[0047] Removing non-marker genes from the normalized gene expression matrix of the meta cell, obtaining the normalized marker gene expression matrix X meta.
[0048] Table 10: Normalized marker gene expression matrix
[0049] S104, using PCA and marker gene expression matrix X meta The meta-cell reduced dimension matrix Yscis calculated, which can be used for model training of subsequent machine learning classifiers.
[0050] First, the projection matrix V of the marker gene expression matrix X is calculated using PCA meta PCA uses a general calculation method, and the process includes calculating the covariance matrix, eigenvalue decomposition, selecting principal components, etc. In this step, only the projection matrix V of the first 50 PCs is retained. Based on the formula Y meta = X meta *V, the meta-cell reduced dimension matrix Y meta shown in Table 11 can be calculated, and the matrix Y meta can be used for model training of subsequent machine learning classifiers.
[0051] Table 11: Meta-cell reduced dimension matrix
[0052] Example II: Please refer to Figure 2 , this embodiment II provides a single cell omics feature amplification system, comprising the following functional modules: data providing module, clustering module, filtering module, generating module.
[0053] The data providing module provides a gene expression matrix of a single cell transcriptome, classifies and labels the cells in the gene expression matrix; Specifically, the data providing module further statistically analyzes and filters low-expression genes in the gene expression matrix: The expression proportion of each gene in the gene expression matrix is calculated by the following formula: P gene_i =N gene_i / N total *100% Wherein: P gene_i is the gene expression proportion of gene_i in all cells, N gene_i is the number of cells expressing gene_i, and N total is the total number of cells; If the expression proportion of the gene is lower than the set proportion threshold, the gene is determined as a low-expression gene; The low-expression genes are filtered.
[0054] The clustering module splits the gene expression matrix according to the classification labels, clusters cells with the same classification label, and generates a gene expression matrix of metacells; The clustering module is specifically implemented by the following method: Select cells with the same classification label in the gene expression matrix, and perform data preprocessing, including normalization, standardization, high-variation gene calculation, and PCA dimensionality reduction of cells with the same classification label, a cell-PCA dimensionality reduction matrix for each classification label; Based on the cell-PCA dimensionality reduction matrix, the most similar k cells of each cell are calculated; The method for calculating the most similar k cells of each cell is as follows: Calculate the Euclidean distance between each cell in the cell-PCA dimensionality reduction matrix, and use the k-means algorithm to cluster the cells into several categories based on the Euclidean distance between the cells; for each cell, use the k-nearest neighbor algorithm to calculate the k cells with the smallest Euclidean distance from the cell as the most similar k cells of the cell.
[0055] Generate a gene expression matrix of metacells, set the expected number of metacells N; the number of metacells N is 1% of the number of cells of the classification label; but when the number of cells of the classification label is less than 10000, N is set to 100; when the number of cells of the classification label is greater than 100000, N is set to 1000. Randomly select N cells in the cell-PCA dimensionality reduction matrix; in the N cells, merge the original gene expression matrix of each cell with the most similar k cells of each cell to obtain the average expression of each gene, and calculate the average expression of each gene to obtain a gene average expression matrix; Repeat the above method for each classification label of single-cell transcriptome data, i.e., each classification label can obtain a gene expression count matrix of N metacells.
[0056] The filtering module normalizes the gene expression matrix of the metacells and calculates the marker genes of each classification label in the metacells; filters non-marker genes in the gene expression matrix of the metacells to obtain a marker gene expression matrix; The method for normalizing the gene expression matrix of the metacells is as follows: NormExpr=log(1+(Count raw / Count total *10000)) Where: NormExpr is the expression of the gene after normalization, Count raw is the expression count of a single gene in a single metacell, and Count total is the total expression count of all genes in a single metacell.
[0057] The generation module calculates a meta-cell dimensionality reduction matrix based on the marker gene expression matrix, which is used for subsequent model training of the machine learning classifier.
[0058] First, PCA is used to calculate the marker gene expression matrix X. meta The projection matrix V; based on formula Y meta =X meta *V can be used to calculate the dimensionality reduction matrix Y of the metacell. meta .
[0059] The implementation principles and effects of each functional module in this embodiment two are described in steps S101 to S104 of embodiment one, and will not be repeated here.
[0060] Example 3: According to an embodiment of this application, this third embodiment also provides an electronic device and a readable storage medium.
[0061] Figure 3 A schematic block diagram of an example electronic device that can be used to implement embodiments of this application is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.
[0062] like Figure 3 As shown, the device includes a computing unit that can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) or a computer program loaded from a storage unit into random access memory (RAM). The RAM can also store various programs and data required for device operation. The computing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.
[0063] Multiple components in the device are connected to the I / O interface, including: input units such as keyboards and mice; output units such as various types of displays and speakers; storage units such as disks and optical discs; and communication units such as network interface cards (NICs), modems, and wireless transceivers. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.
[0064] The computing unit can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, and the like. The computing unit performs various methods and processes described above, such as the single-cell omics feature amplification method described in Embodiment I. For example, in some embodiments, the single-cell omics feature amplification method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device via a ROM and / or the communication unit. When the computer program is loaded onto the RAM and executed by the computing unit, one or more steps of the single-cell omics feature amplification method described above can be performed. Alternatively, in other embodiments, the computing unit can be configured to perform the single-cell omics feature amplification method by any other suitable means, such as by means of firmware.
[0065] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.
[0066] Program code for carrying out methods of the present application can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, partially on a machine and partially on a remote machine or entirely on a remote machine or server.
[0067] In the context of this application, a machine-readable medium can be a tangible medium that can contain or store program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable storage medium can include but are not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0068] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.
[0069] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.
[0070] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.
[0071] It should be understood that the various forms of flow shown above can be used to reorder, add, or delete steps. For example, the steps described in this application can be performed in parallel, in series, or in a different order, as long as the desired results of the technical solutions of the present application can be achieved, which are not limited herein.
[0072] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.
[0073] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.
Claims
1. A method of single cell omics feature amplification, characterized in that, The method comprises the following steps: S101, providing a gene expression matrix of single-cell transcriptome, classifying cells in the gene expression matrix and labeling a classification label; S102, splitting the gene expression matrix according to the classification label, clustering cells labeled with the same classification label and generating a gene expression matrix of meta-cells; S103, performing normalization processing on the gene expression matrix of the meta-cells and calculating marker genes of each classification label in the meta-cells; In Filtering non-marker genes in the gene expression matrix of the meta-cells to obtain a marker gene expression matrix; S104, calculating a meta-cell dimension reduction matrix based on the marker gene expression matrix for subsequent model training of a machine learning classifier.
2. The method of single-cell-omics feature amplification of claim 1, wherein, In the S101, statistics and filtering are performed on low-expression genes in the gene expression matrix; The method comprises the following steps: S1011, calculating the expression proportion of each gene in the gene expression matrix by the following formula: P gene_i =N gene_i / N total *100% wherein: P gene_i is the proportion of gene expression of gene_i in all cells, N gene_i is the number of cells expressing gene_i, N total is the total number of cells; S1012, if the expression proportion of a gene is lower than a set proportion threshold, the gene is determined as a low-expression gene; S1013, filtering the low-expression genes.
3. The method of single-cell-omics feature amplification of claim 1, wherein, The S102 is implemented through the following S1021 to S1024; S1021, selecting cells with the same classification label in the gene expression matrix, performing data preprocessing, and generating a cell-PCA dimension reduction matrix of each classification label; S1022, calculating the most similar k cells of each cell based on the cell-PCA dimension reduction matrix; S1023, generating a gene expression matrix of meta-cells, setting an expected number of meta-cells N, randomly selecting N cells in the cell-PCA dimension reduction matrix, and merging the original gene expression matrix of each cell with the most similar k cells to obtain the average expression of each gene to obtain a gene average expression matrix; S1024, repeating S1021 to S1023 for single-cell transcriptome data of each classification label, that is, N gene expression count matrices of meta-cells can be obtained for each classification label.
4. The method of single-cell-omics feature amplification of claim 3, wherein, In the S1022, the method for calculating the most similar k cells of each cell is as follows: Calculate the Euclidean distance between each cell in the cell-PCA dimension reduction matrix, and use the k-means algorithm to cluster the cells into several categories based on the Euclidean distance between the cells; for each cell, use the k-nearest neighbor algorithm to calculate the k cells with the smallest Euclidean distance to the cell as the most similar k cells of the cell.
5. The method of single-cell-omics feature amplification of claim 3, wherein, The number N of the meta-cells is 1% of the number of cells of the classification label; when the number of cells of the classification label is less than 10,000, N is set to 100; when the number of cells of the classification label is greater than 100,000, N is set to 1,000.
6. The method of single-cell-omics feature amplification of claim 3, wherein, In the S1021, the data preprocessing includes normalization, standardization, high-variable gene calculation and PCA dimension reduction of the cells with the same classification label.
7. The method of single-cell-omics feature amplification of claim 1, wherein, In the S103, the method for performing normalization processing on the gene expression matrix of the meta-cells is as follows: NormExpr = log(1 + (Count raw / Count total * 10000)) wherein: NormExpr is the normalized expression of the gene, Count raw is the expression count of a single gene in a single cell, Count total is the sum of expression counts of all genes in a single cell.
8. The method of single-cell-omics feature amplification of claim 1, wherein, In the S104, the method for calculating a meta-cell dimension reduction matrix based on the marker gene expression matrix is as follows: First, the projection matrix V of the marker gene expression matrix X is calculated using PCA meta ; based on the formula Y meta = X meta * V, the meta-cell dimension reduction matrix Y meta can be calculated.
9. An electronic device, comprising: comprise: at least one processor; and a memory connected in communication with the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.
10. A non-transitory computer-readable storage medium having stored thereon computer instructions, wherein, The computer instructions are for causing the computer to perform the method of any one of claims 1-8.