Single Cell RNA-seq Clustering Method Based On Self-renewal Of Cell Relationship Matrix

Single-cell transcriptome sequencing technology is a sequencing technology that can detect gene expression at the single-cell level.This technology is widely used in many biological branches such as developmental biology,tumor biology,immunology,and neurobiology.With the development of technology,the number of cells detected and sequencing coverage of single cell transcriptome sequencing technology are increasing,providing important technical support for the human cell atlas project.When using single-cell transcriptome sequencing technology to identify cell types,because there is no relevant cell type label when separating cells,we must first use unsupervised clustering methods to divide cells into several cell groups before subsequent analysis.Since the unsupervised clustering algorithm has no training sample set,the analysis results of most algorithms are easily changed by the initial parameter values.We need to develop a more stable clustering algorithm.This topic is based on the analysis of the cell-cell similarity matrix.The cell-cell similarity matrix is updated by redefining the similarity between cells by finding the angle cosine of the similarity matrix row vector.We use the law of large numbers of statistics to prove that this operation will establish stable and orderly statistics to describe the similarity relationship between cells,and based on the order of this relationship and whether the corresponding two cells are in A new clustering algorithm has been developed for the relationship of the same category,Chebyshev's large number clustering algorithm.Next,do simulation test experiments in the simulation data set to verify that each step of the clustering algorithm can achieve the desired ideal effect,and finally complete the clustering analysis.In the next step,we put the developed algorithm in the real data set for testing.The test results show that the clustering algorithm can identify the cell status of most cells in the mouse bladder tissue in the mouse cell map.After completing the clustering,we slightly modify the parameters of the Chebyshev large number clustering algorithm,and we will find some cell types related to a variety of cell states based on the clustering results,which is determined by cell differentiation and fate May provide new research ideas.In summary,the Chebyshev large number clustering algorithm is a clustering algorithm with good accuracy,strong stability,and a wide range of applications.It canperform well on discrete or continuous cell-cell relationship matrices.Clustering ability.The algorithm has a strong ability to mine the feature distribution of the relationship between cells.It can start from the general cell-cell similarity relationship matrix and continuously dig out its internal cell relationship laws.Not only is it possible to classify a large number of cells into different modes,but it also provides solutions to analyze the correlation between these modes.The algorithm may provide a reliable basis for future development of a dynamic model of embryo or disease occurrence.