Characterization of cell-fate decision landscapes by estimating transcription factor dynamics

Abstract: Modulation of gene expression during differentiation by transcription factors promotes cell diversity. Despite their role in cell fate decisions, no experimental assays estimate their regulatory activity in a high-throughput manner and at the single-cell resolution. We present FateCompass for identifying lineage-specific transcription factors across differentiation. It uses single-cell transcriptomics data to infer differentiation trajectories and transcription factor activities. We combined a probabilistic framework with RNA velocities or a differentiation potential to estimate transition probabilities and perform stochastic simulations. Also, we learned transcription factor activities using a linear model of gene regulation. Considering dynamic changes and correlations, we identified lineage-specific regulators. We applied FateCompass to an islet cell formation dataset from the mouse embryo, and we found known and novel potential cell-type drivers. Also, when applied to a differentiation protocol dataset towards beta-like cells, we pinpointed undescribed regulators of an off-target population, which were experimentally validated. Thus, as a framework for identifying lineage-specific transcription factors, FateCompass could have implications on hypothesis generation to increase the understanding of the gene regulatory networks driving cell fate choices. Highlights We developed FateCompass, a flexible pipeline to estimate transcription factor activities during cell-fate decision using single-cell RNA seq data. FateCompass outlines gene expression stochastic trajectories by infusing the direction of differentiation using RNA velocity or a differentiation potential when RNA velocity fails. Transcription factor dynamics allow the identification of time-specific regulatory interactions. FateCompass predictions revealed known and novel cell-subtype-specific regulators of mouse pancreatic islet cell development. Differential motif analysis predicts lineage-specific regulators of stem cell-derived human β- cells and sheds light on the cellular heterogeneity of β-cell differentiation protocols. Experimental validation supports the proposed GRN controlling SC-EC differentiation predicted by FateCompass.