We applied CellOracle to a scRNA-seq atlas of mouse gastrulation and organogenesis by Pijuan-Sala et al.30 (link). This single-cell profiling of WT cells highlighted a continuous differentiation trajectory across the early development of various cell types (Extended Data Fig. 9a ). In addition, the developmental effects of Tal1 KO, a TF known to regulate early haematoendothelial development64 (link),65 (link), were investigated in this study. We validated the CellOracle simulation using these Tal1 KO ground-truth scRNA-seq data. The data were generated from seven chimeric E8.5 embryos of WT and Tal1 KO cells (25,307 cells and 26,311 cells, respectively). We used the R library, MouseGastrulationData (https://github.com/MarioniLab/MouseGastrulationData ), to download the mouse early gastrulation scRNA-seq dataset. This library provides the GEM and metadata. We used the Tal1 chimera GEM and cell-type annotation, “cell type.mapped”, provided by this library. Data were normalized with SCTransform66 (link). The GEM was converted to the AnnData format and processed in the same way as the Paul et al. dataset. For the dimensionality reduction, we used UMAP using the PAGA graph for the initialization (maxiter=500, min_dist=0.6). We removed the extraembryonic ectoderm (ExE), primordial germ cell (PGC) and stripped nuclei clusters which lie outside the main differentiation branch. After removing these clusters, we used the WT cell data for the simulations (24,964 cells). GRN calculations and simulations were performed as described above using the default parameters. We used the base GRN generated from the mouse sci-ATAC-seq atlas dataset. We constructed cluster-wise GRN models for 25 cell states. Then, we simulated Tal1 KO effects using the WT scRNA-seq dataset. For the late-stage-specific Tal1 conditional KO simulation, we set Tal1 expression to be zero in the blood progenitor and erythroid clusters to analyse the role of Tal1 in late erythroid differentiation stages (Extended Data Fig. 9i,j ).
Free full text: Click here
