Reconstructing regulatory differentiation networks from time series single cell expression data

We developed and tested a computational method for reconstructing dynamic regulatory networks from single cell time series data. Unlike prior methods for pseudo-temporal ordering of such data, our method uses static information about targets of TFs to improve both the learning of a branching model and the identification of TFs that regulate various stages in the process. Applying our method to single cell lung development data from multiple laboratories allowed us to reconstruct developmental pathways for a number of different types of lung epithelial cells. As we show, the reconstructed models both capture known biology (in terms of cell groupings and temporal assignment of events) and raise new hypothesis about the roles that certain TFs play in the development of specific cell types. We validated these predictions using both staining and expression experiments identifying new roles for several TFs in regulating lung development.