Abstract
Single-cell transcriptome measurements are being applied at rapidly increasing scales to study cellular repertoires underpinning functions of complex tissues and organs, including mammalian brains. The transcriptional state of each cell, however, reflects a variety of biological factors, including persistent cell-type specific regulatory configurations, transient processes such as cell cycle, local metabolic demands, and extracellular signals. Depending on the biological setting, all such aspects of transcriptional heterogeneity can be of potential interest, but detecting complex heterogeneity structure from inherently uncertain single-cell data presents analytical challenges. We developed PAGODA to resolve multiple, potentially overlapping aspects of transcriptional heterogeneity by identifying known pathways or novel gene sets that show significant excess of coordinated variability among the measured cells. We demonstrate that PAGODA effectively recovers the subpopulations and their corresponding functional characteristics in a variety of single-cell samples, and use it to characterize transcriptional diversity of neuronal progenitors in the developing mouse cortex.
