Abstract: Single-cell RNA sequencing (scRNA-seq) can be used to study the gene expression of single cell and generate a large amount of single-cell gene expression data. This type of data generally has high-dimensional and complex structures, requiring dimension reduction and clustering analysis to reveal differences between cell types and states. A new scRNA-seq data analysis method (scLRRAGR) is proposed based on adaptive graph regularization low-rank representation. This method can fully utilize the global and local information of scRNA-seq data for graph learning, and capture the similarity and interaction between cells by adaptive graph regularization and the introduction of rank constraint. Its outcome can better reflect the clustering structure between cells and help to reveal differences between different cell types and states. When applying this method, scRNA-seq data can be transformed into a graph structure with each node representing a single-cell sample and edges representing similarities or interactions between cells. Then this method is used to learn and optimize this graph to obtain the optimal graph representation. Finally, typical clustering algorithms can use the optimal graph representation to recognize cell types and states. The experiment results show that the proposed method can significantly improve clustering performance on scRNA-seq datasets.