Identifying different cell types is a prerequisite step in the analysis of single-cell RNA sequencing (scRNA-seq) data, with clustering being a common technique utilized for this purpose. However, high dropout rates inherent in scRNA-seq data and complex intercellular relationships become main challenges in scRNA-seq data analysis. To address these issues, we proposed a novel model based on zero-inflated negative binomial (ZINB) distribution and graph attention network for scRNA-seq data clustering (scZGA). scZGA consists of three key modules. The first module captures the global probabilistic structure using a ZINB model. The second module constructs the graph with Pearson’s correlation coefficient, and employs a graph autoencoder with residual connection to learn important neighbor relationships while preserving topological structure information simultaneously. The final module conducts deep clustering through a self-optimizing embedding algorithm. With these improvements, clustering results show that scZGA consistently achieves higher scores across six scRNA-seq datasets by using evaluation metrics such as normalized mutual information and adjusted rand index.
Kan et al. (Fri,) studied this question.