Fig. 1

Overview of DcjComm. A DcjComm takes a single-cell gene expression matrix as input and then processes it through a preprocessing step to obtain the preprocessed matrix. B The NMF-based joint learning model. DcjComm performs dimension reduction by projected matrix decomposition and cell clustering by non-negative matrix factorization. C Visualization of cell clustering results using the Umap method. D Selection and analysis of functional gene modules. It mainly contains the identification of functional gene modules and evaluation of their quality. E The biological model of CCCs. The prior knowledge includes links between ligands, receptors, signaling pathways, transcription factors, and target genes. F The inference statistical model of CCCs. DcjComm models the probability of communication between cells and identifies significant communications. G The visualization methods for the results of CCCs inference. It mainly includes circle plot, heatmap, and ridge plot