Fig. 1.

Analysis pipeline for scRNA-seq data for elucidating transcriptional heterogeneity. Without knowing cell identities, one may obtain gene expression profiles of single cells. After quality control, dimension reduction and unsupervised clustering are routinely applied to estimate cellular subpopulations that are used as cell identities in downstream analyses. The proposed methods enable statistically rigorous evaluation of cell identities improving unsupervised classification and feature selection