Cell clustering is a crucial step in current single-cell RNA sequencing (scRNA-seq) methods, where marker genes are identified and used for cell type annotation. However, this process can be time-consuming and laborious. To address this, biclustering algo- rithms have been developed to simultaneously identify functional gene sets and cell clusters. However, most existing biclustering algorithms are designed for microarray and bulk RNA sequencing data, and only a few are suitable for scRNA-seq analysis. These algorithms often suffer from issues such as limited scalability and accuracy. In this study, we propose Correspondence Analysis based biclustering on Networks (CAb- iNet), a graph-based biclustering approach specifically designed for scRNA-seq data. CAbiNet integrates multiple analysis steps by efficiently co-clustering cells and their marker genes, and visualizing the biclustering results in a non-linear embedding. We introduce two visualization approaches that enable the joint display of genes and cells in a two-dimensional space. Additionally, a random forest regression model is trained to predict the quality of clustering results, facilitating the selection of optimal parame- ters. CAbiNet fills the gap for a high-performing biclustering algorithm in scRNA-seq and spatial transcriptomics data analysis. It streamlines existing workflows and offers an intuitive and interactive visual exploration of cells and their marker genes in a single plot for efficient cell type annotation. CAbiNet is available as an R package on GitHub at https://github.com/VingronLab/CAbiNet.
