scPEGEnhanced Graph Convolutional Sparse Subspace Clustering Method for scRNA-Seq Data

The identification of cell types by clustering single-cell RNA sequencing (scRNA-seq) data is a fundamental step in the downstream analysis of single-cell data. However, great challenges remain owing to the inherent characteristics of scRNA-seq data, including high dimensionality, high noise, and high sparsity. In this study, we propose a proximity enhanced graph convolutional sparse subspace clustering method scPEGSSC for scRNA-seq data. Method scPEGSSC generates the similarity matrix with the self-expression matrix (SEM) learned from a graph autoencoder, and enhances it further through its square. Experiments were performed on thirteen real biological datasets. The experimental results indicate compared with eleven state-of-the-art single-cell clustering methods, method scPEGSSC have attained superior performance across most datasets.