Abstract
While single-cell measurement technologies provide an unprecedented opportunity to dissect developmental processes, revealing the mechanisms of cell fate decisions from single-cell RNA-seq data is challenging due to both cellular heterogeneity and transcriptional noise in the data. Here we developed Topographer, a bioinformatic pipeline, to construct an intuitive (i.e., every cell is equipped with both potential and pseudotime) developmental landscape, reveal stochastic dynamics of cell types, and infer both dynamic connections of marker gene networks and dynamic characteristics of transcriptional bursting kinetics across development. Applying this method to primary human myoblasts, we not only identified three known cell types but also estimated both their fate probabilities and transition probabilities among them. We found that the percent of the genes expressed in a bursty manner is significantly higher at the branch point than before or after branch, and there are apparent changes in both gene-gene and cell-cell correlations before and after branch. In general, single-cell transcriptome data with Topographer can well reveal the stochastic mechanisms of cell fate decisions from three different levels: cell lineage (macroscopic), gene network (mesoscopic) and gene expression (microscopic).