Single cell transcriptome data clustering method and system based on pellet calculation

Abstract

The invention belongs to the technical field of bioinformatics and data mining, and discloses a single cell transcriptome sequencing data clustering method and system.The method comprises the steps that single cell transcriptome sequencing data are preprocessed, and preprocessing comprises data dimension reduction; constructing a stable pellet space based on the dimensionality-reduced data, wherein the pellet space is composed of a plurality of pellets covering the single cell sample; performing maximum pellet splitting on pellets in the stable pellet space to further refine the pellet set; the correction distance between the pellets and the balls is calculated to construct a correction distance matrix, and the correction distance integrates the Euclidean distance between the centers of the pellets and the radius information of the pellets and the balls; based on the corrected distance matrix, clustering the pellets by adopting a density clustering algorithm to obtain clustering labels of the pellets; and mapping the clustering labels of the pellets to the single cell samples covered by the pellets to generate a final single cell clustering result. According to the method, efficient, accurate and robust cell grouping is realized.

Description

Technical Field

[0001] This invention belongs to the field of bioinformatics and data mining technology, and particularly relates to a clustering analysis method for single-cell transcriptome sequencing (scRNA-seq) data. Specifically, it is a clustering method and system for single-cell transcriptome sequencing data based on particle sphere computation. Background Technology

[0002] The development of single-cell RNA sequencing (scRNA-seq) technology has made it possible to study gene expression at single-cell resolution, greatly advancing the exploration of cellular heterogeneity, developmental trajectories, and rare cell types. With the increase in sequencing throughput, the size of scRNA-seq datasets has rapidly expanded from a few thousand cells to hundreds of thousands or even millions. Cluster analysis, as a core data analysis step, aims to group cells with similar gene expression patterns into the same population; its accuracy and efficiency directly affect the reliability of downstream cell type identification and biological conclusions.

[0003] However, the inherent high dimensionality, high sparsity, and technical noise of scRNA-seq data pose significant challenges to cluster analysis. The high dimensionality is reflected in the fact that the expression profile of each cell consists of tens of thousands of genes; the sparsity stems from the limited sequencing capture efficiency, resulting in a large number of "zero" values ​​in the expression matrix. This is not biological non-expression, but rather technical artifacts, which severely interfere with the effectiveness of traditional distance measurement methods.

[0004] Existing scRNA-seq clustering methods typically follow a "dimensionality reduction first, then clustering" paradigm. For example, methods such as principal component analysis (PCA) or t-distributed random neighborhood embedding (t-SNE) are used to project the data into a low-dimensional space, and then algorithms such as K-means, hierarchical clustering, or spectral clustering are applied for grouping. While these methods are intuitive, when dealing with data with complex nonlinear structures, uneven density, or ambiguous boundaries, the clustering results are often highly sensitive to parameter selection and lack stability.

[0005] To better adapt to complex data distributions, density-based clustering algorithms, such as DBSCAN, have been introduced into single-cell analysis. These algorithms can identify clusters of arbitrary shapes without requiring a pre-specified number of clusters. However, traditional density-based clustering methods like DBSCAN suffer from high computational complexity and low efficiency when dealing with large-scale datasets. More importantly, setting density thresholds globally (such as Eps and MinPts parameters) is very difficult for different cell subpopulations with vastly different densities, easily leading to over-segmentation of high-density clusters and incorrect merging or labeling of low-density clusters as noise.

[0006] To address these issues, researchers have proposed various improvement strategies, but limitations remain. For example, graph-based clustering methods (such as those used in Seurat and Scanpy) construct K-nearest neighbor graphs (KNN graphs) between cells and perform community detection on the graph, exhibiting some robustness to local density changes. However, these methods incur significant computational overhead in graph construction and community partitioning, making it difficult to meet the real-time requirements of ultra-large-scale data analysis.

[0007] Granular ball computing, as a coarse-grained computing model, represents data by encapsulating raw data points into a series of "spheres" (defined by a center and radius). This significantly reduces the granularity of computation while preserving the spatial distribution and local structure information of the data, providing a new approach for processing large-scale complex data.

[0008] Therefore, there is an urgent need to develop a novel clustering method that can effectively address the high-dimensionality, sparseness, and complex structural characteristics of scRNA-seq data, while balancing computational efficiency and clustering accuracy, in order to achieve stable and reliable cell clustering for large-scale single-cell datasets. Summary of the Invention

[0009] This invention aims to overcome the problems of difficulty in accurately identifying complex cluster structures in existing single-cell clustering methods when processing high-dimensional sparse data, and provides a single-cell transcriptome data clustering method and system based on particle-sphere computation to achieve efficient, accurate and robust cell grouping.

[0010] In a first aspect, the present invention provides a single-cell transcriptome data clustering method based on granulocyte-sphere computation, comprising the following steps: S1: Preprocessing single-cell transcriptome sequencing data, including data dimensionality reduction; S2: Construct a stable sphere space based on the dimensionality-reduced data, wherein the sphere space consists of multiple spheres covering a single-cell sample; S3: Perform maximum granular splitting on the granules in the stable granular space to further refine the granular set; S4: Calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles; S5: Based on the modified distance matrix, the density clustering algorithm is used to cluster the particles to obtain the cluster labels of the particles; S6: Map the clustering labels of the granules to the single-cell samples they cover to generate the final single-cell clustering results.

[0011] As an optional implementation of the first aspect of this application, in step S1, the preprocessing specifically includes: performing gene filtering on the single-cell transcriptome sequencing data to filter out genes expressed in cells below a preset proportion; normalizing and standardizing the filtered gene expression matrix; extracting highly variable genes using a feature selection method; and performing principal component analysis on the highly variable genes to reduce dimensionality.

[0012] As an optional implementation of the first aspect of this application, step S2, constructing a stable sphere space specifically includes: selecting an initial center point in the dimensionality-reduced data using the K-means++ algorithm, and dividing the single-cell samples to the initial center point according to the shortest distance principle to form an initial sphere set, wherein each sphere consists of its center, radius, and the set of sample points it covers; iteratively splitting the spheres in the initial sphere set until the sphere space reaches stability; the condition for the iterative splitting is based on the judgment of the number of samples covered by the sphere and the internal compactness.

[0013] As an optional implementation of the first aspect of this application, the judgment condition for the iterative splitting is specifically as follows: for any particle, if the number of samples it covers exceeds a preset threshold, or its internal compactness does not meet the preset splitting condition, then the particle is split into two sub-particles using the K-means++ algorithm; wherein, the internal compactness is measured by the median of the Euclidean distances between all pairs of sample points within the particle; the preset splitting condition is: the compactness of the current particle is greater than the sum of the compactnesses of the two sub-particles generated after its split.

[0014] As an optional implementation of the first aspect of this application, step S3, the maximum particle splitting specifically includes: repeatedly selecting the particle with the largest radius in the current particle set; performing K-means splitting on the selected particle with the largest radius to generate two new sub-particles to replace the original particle; repeating the particle splitting process until the preset maximum number of iterations is reached.

[0015] As an optional implementation of the first aspect of this application, in step S4, the formula for calculating the corrected distance between the two particles is defined as follows: ,in, For granules and Corrected distance between and These are the center vectors of the two spheres, respectively. and These are the radii of the two spheres, respectively. The Euclidean distance between the centers. It is the maximum of the two sphere radii.

[0016] As an optional implementation of the first aspect of this application, in step S5, the density clustering algorithm used is the OPTICS algorithm, which specifically includes: setting a minimum neighbor number threshold MinPts; calculating the core distance and reachability distance for each particle based on the modified distance matrix; constructing a distance reachability map based on the reachability distances of all particles; extracting the cluster structure by identifying significant changes in the distance reachability map, thereby assigning a clustering label to each particle.

[0017] Secondly, embodiments of this application provide a single-cell transcriptome data clustering system based on granulocyte-sphere computation, comprising: The data preprocessing module is used to preprocess single-cell transcriptome sequencing data, including data dimensionality reduction. The sphere space construction module is used to construct a stable sphere space based on the dimensionality-reduced data. The sphere space consists of multiple spheres covering a single-cell sample. The sphere splitting module is used to perform maximum sphere splitting on the spheres in the stable sphere space to further refine the sphere set; The distance calculation module is used to calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles; The clustering execution module is used to cluster the particles using a density clustering algorithm based on the modified distance matrix to obtain the cluster labels of the particles; The result generation module is used to map the clustering labels of the granules to the single-cell samples they cover, so as to generate the final single-cell clustering results.

[0018] Thirdly, embodiments of this application provide an electronic device, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the method described in the first aspect.

[0019] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0020] Compared with the prior art, the present invention has the following significant advantages: 1. High computational efficiency: This invention elevates the clustering object from the cellular level to the granular level by encapsulating a large number of single-cell samples into granules that are far fewer in number than the number of samples. This greatly reduces the complexity of distance calculation and density estimation, and significantly shortens the analysis time of large-scale datasets.

[0021] 2. High Clustering Accuracy: The "corrected distance" proposed in this invention comprehensively considers the spatial proximity of the granule center and the coverage range (radius) of the granule itself, which can more realistically reflect the structural relationship between clusters and avoid erroneous merging caused by differences in cluster density or size. Combined with the OPTICS algorithm, it can adaptively identify cell clusters of arbitrary shapes and different densities, improving the clustering accuracy.

[0022] 3. High robustness and low parameter dependence: The construction process of the spheres and the subsequent OPTICS clustering are not sensitive to the initial parameters. The construction of the stable sphere space and the maximum sphere splitting process are adaptive, and the OPTICS algorithm can extract the cluster structure from the reachability graph, avoiding the problem of setting a global density threshold Eps like DBSCAN, making the method stable on different datasets.

[0023] 4. Good structure preservation: As a coarse-grained representation of data, the sphere not only represents the central location of the data, but its radius also contains information about the distribution range and compactness of the local data. This method preserves the key local topological structure while reducing dimensionality, providing a solid foundation for accurate clustering. Attached Figure Description

[0024] Figure 1 This is a technical roadmap for a single-cell transcriptome data clustering method based on granulocyte computation according to an embodiment of the present invention; Figure 2 Distance reachability graph for the single-cell transcriptome dataset Heart_and_Aorta from 20 mouse organs, calculated using a granulosphere-based single-cell transcriptome data clustering method; Figure 3 The image shows the granulosphere-based results of the Heart_and_Aorta single-cell transcriptome dataset from 20 mouse organs, obtained through a granulosphere-based single-cell transcriptome data clustering method. Figure 4 Clustering results of a single-cell transcriptome sequencing data clustering method based on granulocyte-sphere computation for the mouse single-cell transcriptome dataset Heart_and_Aorta, representing 20 organs. Figure 5 Clustering results of the single-cell transcriptome dataset Heart_and_Aorta from 20 mouse organs after clustering using the Seurat method; Figure 6 A bar chart comparing the clustering results obtained by the method of this invention and the Seurat method for the dataset Heart_and_Aorta; Figure 7 This is a schematic diagram of the structure of a single-cell transcriptome data clustering system based on particle-sphere computation, provided in an embodiment of the present invention. Detailed Implementation

[0025] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] Example 1 Please see Figure 1 This is a flowchart illustrating a single-cell transcriptome data clustering method based on granulocyte-sphere computation, provided by an embodiment of the present invention. The method may include the following steps: S1: Preprocess the single-cell transcriptome sequencing data, including dimensionality reduction.

[0028] In this step, after downloading the raw expression matrix from a publicly available single-cell gene expression database, preprocessing is performed, specifically including: Gene filtering: To reduce noise and computational burden, genes with low expression are filtered out. Specifically, the number of cells in which each gene is expressed (expression value greater than 0) is counted, and only genes expressed in at least 1% of cells are retained.

[0029] Data normalization: To eliminate differences in sequencing depth between cells, expression data for each cell is normalized. A common method is to divide the expression count of each gene in each cell by the total expression count of that cell, multiply by a scaling factor (e.g., 10,000), and finally perform a natural logarithmic transformation on the result.

[0030] Feature selection: Highly variable genes (HVGs) with the greatest expression variation among cells were identified because these genes are most likely to contain biological signals that distinguish different cell types. In this embodiment, the variance-stabilizing transformation (VST) method was used to select the top 2000 genes with the highest variability as the feature set for subsequent analysis.

[0031] Data standardization (Scaling): Standardize the selected subset of highly variable genes so that the mean expression of each gene is 0 and the variance is 1 across all cells.

[0032] Principal Component Analysis (PCA) Dimensionality Reduction: PCA is performed in the standardized high-dimensional feature space to project the data into a low-dimensional principal component space. In this embodiment, the first 20 principal components are extracted as the dimensionality-reduced data representation for subsequent granular sphere construction and clustering. The dimensionality-reduced dataset is denoted as […]. ,in It represents the first i A 20-dimensional vector for each cell.

[0033] S2: Construct a stable sphere space based on the dimensionality-reduced data, wherein the sphere space consists of multiple spheres covering a single-cell sample.

[0034] In the dimensionality-reduced data space, the original single-cell data points are encapsulated into a series of spheres. This step S2 includes the following steps S21 and S22.

[0035] S21: Initial set of particles generated.

[0036] The K-means++ algorithm is used to initialize the spheres. The initial number of spheres is set. An empirical choice is , where n is the total number of cells. K-means++ selects the initial center points in a probabilistic manner, ensuring they are well distributed across the data space and avoiding local optima problems that may result from random initialization.

[0037] For the dimensionality-reduced single-cell dataset Using the K-means++ approach to select Selecting a point as the initial center of the particle After establishing initial centers, the initial set of particles is obtained by dividing them into different centers according to the shortest distance principle. , where m represents the number of granules generated. All cell sample points Divide into the nearest center, forming There are 10 initial clusters. Each cluster constitutes an initial sphere. A sphere Defined by three elements: center ,radius and the set of sample points it covers .

[0038] center It is the mean of the coordinates of all the sample points it covers in the reduced-dimensional space: radius It is all the sample points within it from the center. Maximum Euclidean distance: in, Indicates granules Covered sample points, Indicates the first i The radius of each sphere, For the first granule j One sample point, Indicates the center of the grain.

[0039] S22: Iterative splitting constructs a stable sphere space.

[0040] The initial set of particles is iteratively refined to form a more stable particle space that better reflects the local structure of the data.

[0041] For each ball in the current set of balls The determination of whether splitting is necessary is based on a combination of the number of samples covered by granules and their internal compactness.

[0042] Define a threshold for the number of samples, for example, N=n / 2.

[0043] Define the internal compactness DM(X) of a sphere as the median of the Euclidean distances between all pairs of sample points within the sample set X covered by the sphere: Where X is the sample set currently covered by the particle. , These are the sample points contained within the sphere.

[0044] For a sphere If the number of samples it covers is less than the threshold N, or its compactness does not meet the stability condition, then it is split. The stability condition is: in, and It uses the K-means algorithm (k=2, where k represents the input parameters for executing K-means) to... The sample points of the two subsets obtained after splitting. If this condition is not met (i.e., the sum of the compactness of the two sub-spheres after splitting is less than or equal to the compactness of the original sphere), it means that splitting can improve internal homogeneity and is effective.

[0045] If splitting is required, the original sphere will be... Remove it from the set of particles and add the two new sub-particles (with newly calculated centers and radii) generated from its split to the set.

[0046] Repeatedly traverse all spheres and perform the above split judgment until no sphere splits in a complete traversal. At this point, the sphere space is considered to be stable.

[0047] S3: Perform maximum granular splitting on the granules in the stable granular space to further refine the granular set.

[0048] After the stable sphere space is constructed, some spheres may still be large, covering regions with heterogeneous internal structures. To further improve resolution, a maximum sphere splitting operation is performed.

[0049] This step is an iterative process: In each iteration, the radius is selected from all the current spheres. The largest ball.

[0050] For the sphere with the largest radius, use K-means (k=2) to split the sample points it covers into two subsets.

[0051] Calculate the center and radius of the new spheres corresponding to these two subsets, and replace the original large spheres with these two new spheres.

[0052] Repeat this process until the preset maximum number of iterations is reached (20 times in this embodiment). This step aims to specifically refine sparse or scattered regions in the data space.

[0053] S4: Calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles.

[0054] Traditional clustering algorithms typically only consider the distance between data points (or centroids). This invention introduces a "corrected distance" that considers not only the location of the cluster center but also the size of the cluster itself.

[0055] First, construct the Euclidean distance matrix D between the centers of all spheres. ]: To mitigate the influence of the spherical boundary coverage, a maximum radius matrix R is constructed. ]: ,in The maximum value among the radii of the two spheres; Corrected distance It is the sum of the Euclidean distance between the centers of the particles and the maximum radius between the particles: For any two spheres and Their centers are respectively and The radii are respectively and .

[0056] The final corrected distance matrix for: The corrected distance is defined as the Euclidean distance between the centers of two grains, plus the radius of the larger of the two. This definition means that even if two grains with large radii have close centers, their corrected distance will be large, reflecting that the areas they cover may overlap or be adjacent, but should not be easily considered as part of the same compact cluster. Conversely, for two small and compact grains, the corrected distance is mainly determined by the distance between their centers.

[0057] Based on this formula, the corrected distance between all particle-sphere pairs is calculated, and a corrected distance matrix is ​​constructed. .

[0058] S5: Based on the modified distance matrix, the density clustering algorithm is used to cluster the particles to obtain the cluster labels of the particles.

[0059] Using the corrected distance matrix As input, the OPTICS (Ordering Points To Identify the Clustering Structure) algorithm is applied to cluster the spheres. OPTICS is a density clustering algorithm that generates an "accessibility" ranking of the data, from which cluster structures of different densities can be identified, and it is parameter-insensitive.

[0060] Set the parameter MinPts of the OPTICS algorithm, which is the minimum number of neighbors (particles) required to form a density core region.

[0061] For each sphere Calculate its core distance, which is defined as the distance between the grains. The corrected distance between it and its MinPts-th neighbor: in Indicates granules The corrected distance between the center of its MinPts-th neighbor; For each pair of particles ( , ),calculate Compared to Reachability Distance, defined as... The core distance and their ( and The maximum corrected distance between: in For granules and In the correction matrix The corrected distance in the middle.

[0062] The OPTICS algorithm outputs a sort of all spheres and their corresponding reachability distances. Plotting these reachability distances in order yields a distance reachability graph.

[0063] like Figure 2 As shown, valleys (low reachability regions) correspond to dense clusters, while peaks (high reachability) represent boundaries or sparse regions between clusters. This figure illustrates a modified distance reachability map constructed based on sphere centers, with different colors representing different clustering categories. The density ranking reflected in the reachability map clearly reveals the clustered distribution structure of spheres in the representation space. This method uses the modified distance matrix as input, ensuring that the distance between spheres not only reflects the spatial relationship between their centers but also comprehensively considers the influence of their coverage area (i.e., radius), thus more realistically depicting the compactness between spheres.

[0064] By analyzing the slope changes of the reachability graph or setting a threshold ξ, cluster structures can be automatically extracted from the graph, thereby assigning a cluster label (or labeling it as noise) to each particle. Figure 3 The diagram illustrates the results after assigning cluster labels to the sphere centers. The horizontal and vertical axes represent the first principal component (UMAP_1) and the second principal component (UMAP_2) after UMAP dimensionality reduction, respectively. Each data point corresponds to a sphere center, and different colors represent different clusters. This diagram clearly presents the clustering structure at the sphere level, reflecting the tight distribution and local density differences formed by the spheres in the high-dimensional expression space.

[0065] S6: Map the clustering labels of the granules to the single-cell samples they cover to generate the final single-cell clustering results.

[0066] The final step is to map the clustering results at the granular level back to the original single-cell level.

[0067] For each cluster label assigned granules All single-cell sample points x∈Xi covered by it are assigned the same clustering label. .

[0068] After traversing all spheres and completing the mapping, the final clustering result of all n single cells is obtained. For example... Figure 4 As shown in the figure, the horizontal and vertical axes represent the first principal component (UMAP_1) and the second principal component (UMAP_2) of the data after UMAP dimensionality reduction, respectively. This figure illustrates the single-cell clustering diagram obtained by mapping the granulocyte-sphere clustering results back to single cells, with different colors representing different cell clusters. It can be observed from the figure that the various clusters exhibit good spatial separation, with clear boundaries and compact clusters, indicating that this method improves the clustering resolution while preserving the spatial density information.

[0069] Comparison of experimental results To verify the effectiveness of the method of this invention, experiments were conducted using a single-cell transcriptome dataset of mouse heart and aorta (Heart_and_Aorta). The results of the method of this invention were compared with the widely used Seurat process (a graph-based clustering method).

[0070] Figure 4 The clustering results of the method of the present invention are shown. It can be seen that the boundaries of each type of cluster are clear, the clusters are compact, and the separation is good.

[0071] Figure 5 The results of single-cell clustering obtained using the Seurat standard process are shown, where the horizontal and vertical axes represent the first principal component (UMAP_1) and the second principal component (UMAP_2) after UMAP dimensionality reduction, respectively. The comparison reveals that the Seurat method exhibits blurred boundaries and inter-cluster overlap in some clusters. In contrast, the method of this invention, based on granulocyte construction, fully utilizes local cell density and structural information, resulting in more stable and accurate clustering results. This allows for more effective identification of potential cell subpopulation structures, enhancing the interpretability and biological significance of the clustering.

[0072] To quantitatively evaluate clustering performance, clustering accuracy (ACC) and adjusted RAND index (ARI) were used. Clustering accuracy is defined as the maximum matching accuracy between the true labels and the predicted labels, and is calculated as follows: in For predicting labels, For real labels, This is an indicator function (1 if they match, 0 otherwise). This is the optimal matching function for Hungarian languages.

[0073] in Represents real clusters p With prediction clusters q The number of intersections, and Respectively p The first real cluster and the second q Number of samples for each predicted cluster.

[0074] Figure 6 The results comparing the method of this invention with Seurat's method on three key evaluation metrics are presented: clustering accuracy (ACC), adjusted RAND index (ARI), and clustering task runtime. It can be seen that the method of this invention significantly outperforms the Seurat method in both ACC and ARI, two metrics that measure clustering accuracy and consistency, indicating that the invention can more accurately and consistently identify cell subpopulation structures.

[0075] In terms of operational efficiency, the method of this invention effectively reduces the scale of data involved in clustering calculations by constructing a granular-spherical structure for coarse-grained modeling. Compared with methods that directly perform density estimation and clustering on all single-cell samples, this invention significantly reduces the overall computational overhead, and the time required to complete clustering is much shorter than the traditional graph clustering process used by Seurat, demonstrating good practicality and scalability.

[0076] In summary, Figure 6 This indicates that the method of the present invention is superior to existing methods not only in terms of the accuracy and consistency of clustering results, but also has significant advantages in computational efficiency, making it suitable for large-scale, highly complex single-cell transcriptome data analysis scenarios.

[0077] Example 2 Please see Figure 7 The diagram shown is a schematic representation of a single-cell transcriptome data clustering system based on granulocyte-sphere computation, as proposed in the second embodiment of this application. This system includes the following key modules: Data preprocessing module 100 is used to preprocess single-cell transcriptome sequencing data, the preprocessing including data dimensionality reduction; The sphere space construction module 200 is used to construct a stable sphere space based on the dimensionality-reduced data, wherein the sphere space is composed of multiple spheres covering a single-cell sample; The sphere splitting module 300 is used to perform maximum sphere splitting on the spheres in the stable sphere space to further refine the sphere set; The distance calculation module 400 is used to calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles; The clustering execution module 500 is used to cluster the particles using a density clustering algorithm based on the modified distance matrix to obtain the clustering labels of the particles; The result generation module 600 is used to map the clustering labels of the granules to the single-cell samples they cover, so as to generate the final single-cell clustering results.

[0078] The single-cell transcriptome data clustering system based on particle sphere computing in this application embodiment can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. For example, mobile electronic devices can be mobile phones, tablets, laptops, PDAs, in-vehicle electronic devices, wearable devices, ultra-mobile personal computers (UMPCs), netbooks, or personal digital assistants (PDAs), etc., while non-mobile electronic devices can be servers, network-attached storage (NAS), personal computers (PCs), etc. This application embodiment does not impose specific limitations.

[0079] The single-cell transcriptome data clustering system based on granulocyte-sphere computation in this application embodiment can be a device with an operating system. This operating system can be Android, iOS, or other possible operating systems; this application embodiment does not specifically limit it.

[0080] This application provides a single-cell transcriptome data clustering system based on granulocyte computation, which can achieve... Figure 1 The various processes implemented by the single-cell transcriptome data clustering method based on particle sphere computation in the method embodiment are not described in detail here to avoid repetition.

[0081] Optionally, embodiments of this application also provide an electronic device, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the various processes of the above-described embodiment of a single-cell transcriptome data clustering method based on particle sphere computation, and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0082] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described embodiment of a single-cell transcriptome data clustering method based on particle sphere computation, and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0083] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0084] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0085] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0086] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A single-cell transcriptome data clustering method based on granulocyte-sphere computation, characterized in that, Includes the following steps: S1: Preprocessing single-cell transcriptome sequencing data, including data dimensionality reduction; S2: Construct a stable sphere space based on the dimensionality-reduced data, wherein the sphere space consists of multiple spheres covering a single-cell sample; S3: Perform maximum granular splitting on the granules in the stable granular space to further refine the granular set; S4: Calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles; S5: Based on the modified distance matrix, the density clustering algorithm is used to cluster the particles to obtain the cluster labels of the particles; S6: Map the clustering labels of the granules to the single-cell samples they cover to generate the final single-cell clustering results.

2. The method according to claim 1, characterized in that, In step S1, the preprocessing specifically includes: Gene filtering is performed on the single-cell transcriptome sequencing data to remove genes expressed in cells below a preset proportion; The filtered gene expression matrix was normalized and standardized. High-variability genes were extracted using feature selection methods; Principal component analysis was performed on the highly variable genes to reduce dimensionality.

3. The method according to claim 1, characterized in that, In step S2, constructing a stable granular space specifically includes: The K-means++ algorithm is used to select initial centroids in the dimensionality-reduced data, and single-cell samples are divided into initial centroids according to the shortest distance principle to form an initial set of spheres. Each sphere consists of its center, radius, and the set of sample points it covers. The spheres in the initial sphere set are iteratively split until the sphere space reaches stability; the conditions for the iterative splitting are based on the judgment of the number of samples covered by the spheres and the internal compactness.

4. The method according to claim 3, characterized in that, The specific judgment condition for the iterative split is as follows: For any given sphere, if the number of samples it covers exceeds a preset threshold, or if its internal compactness does not meet a preset splitting condition, then the K-means++ algorithm is used to split the sphere into two sub-spheres. The internal compactness is measured by the median of the Euclidean distances between all pairs of sample points within the sphere; the preset splitting condition is that the compactness of the current sphere is greater than the sum of the compactnesses of the two sub-spheres generated after its split.

5. The method according to claim 1, characterized in that, In step S3, the splitting of the largest granule specifically includes: In the current set of particles, repeatedly select the particle with the largest radius; Perform K-means splitting on the selected sphere with the largest radius to generate two new sub-spheres to replace the original sphere; Repeat the granulation process until the preset maximum number of iterations is reached.

6. The method according to claim 1, characterized in that, In step S4, the formula for calculating the corrected distance between two particles is defined as follows: in, For granules and Corrected distance between and These are the center vectors of the two spheres, respectively. and These are the radii of the two spheres, respectively. The Euclidean distance between the centers. It is the maximum of the two sphere radii.

7. The method according to claim 1, characterized in that, In step S5, the density clustering algorithm used is the OPTICS algorithm, which specifically includes: Set the minimum number of neighbors threshold MinPts; Based on the corrected distance matrix, the core distance and reachability distance are calculated for each sphere; Construct a distance reachability graph based on the reachability of all particles; Cluster structures are extracted by identifying significant changes in the reachability graph, thereby assigning cluster labels to each sphere.

8. A single-cell transcriptome data clustering system based on granulocyte-sphere computation, characterized in that, include: The data preprocessing module is used to preprocess single-cell transcriptome sequencing data, including data dimensionality reduction. The sphere space construction module is used to construct a stable sphere space based on the dimensionality-reduced data. The sphere space consists of multiple spheres covering a single-cell sample. The sphere splitting module is used to perform maximum sphere splitting on the spheres in the stable sphere space to further refine the sphere set; The distance calculation module is used to calculate the corrected distance between particles to construct a corrected distance matrix, wherein the corrected distance integrates the Euclidean distance between the centers of particles and the radius information of the particles; The clustering execution module is used to cluster the particles using a density clustering algorithm based on the modified distance matrix to obtain the cluster labels of the particles; The result generation module is used to map the clustering labels of the granules to the single-cell samples they cover, so as to generate the final single-cell clustering results.

9. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the steps of a single-cell transcriptome data clustering method based on granulocyte computation as described in any one of claims 1-7.

10. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of a single-cell transcriptome data clustering method based on granulocyte computation as described in any one of claims 1-7.

Images