CALISTA: Clustering And Lineage Inference in Single-Cell Transcriptional Analysis

Abstract Recent advances in single-cell transcriptional profiling technologies provide a means to shed new light on the dynamic regulation of physiological processes at the single-cell level. Single-cell gene expression data taken during cell differentiation have led to novel insights on the functional role of cell-to-cell variability. Here, the clustering of cells, the reconstruction of cell lineage progression and developmental trajectory, and the identification of key genes involved in the cell fate decision making, are among the most common-yet-challenging data analytical tasks. In this work, we developed CALISTA (Clustering And Lineage Inference in Single-Cell Transcriptional Analysis) for clustering and lineage inference analysis using single-cell transcriptional profiles. CALISTA adopts a likelihood-based approach using the two-state model of stochastic gene transcriptional bursts to capture the heterogeneity of single-cell gene expression. We demonstrated the efficacy of CALISTA by applying the method to single-cell gene expression datasets (RT-qPCR and RNA-sequencing) from four cell differentiation systems without any lineage bifurcation and with one or multiple lineage bifurcations. CALISTA is freely available on https://github.com/CABSEL/CALISTA.