Dpath reveals hierarchical lineages from Etv2 single cell transcriptome. Mus musculus

The molecular definition of cellular differentiation is of intense interest for developmental, stem cell and cancer biologists. While single cell RNA sequencing brought a transformational advance for global gene analyses, novel obstacles have emerged, including the computational management of dropout events, the reconstruction of biological pathways, and the isolation of target cell populations. Here, we provide solutions to these computational challenges. We propose a novel concept of metagene entropy that enables one to rank cells based on their differentiation potential using an algorithm named dpath. We also developed self-organizing map (SOM) and random walk with restart (RWR) algorithms to separate the progenitors from the differentiated cells and reconstruct the lineage hierarchies in an unbiased fashion. We tested these algorithms using single cells from Etv2-EYFP transgenic embryos and revealed specific molecular and signaling pathways that directed differentiation programs of the hematopoietic and endothelial lineages. This is the first software program that quantitatively assesses the progenitor and committed states in single cell RNA-seq datasets in a non-biased manner, and will be a powerful biological tool for the research community.