Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: scTour: a deep learning architecture for robust inference and accurate prediction of cellular dynamics

Fig. 2

scTour robustly captures the cellular dynamics during dentate gyrus neurogenesis. a UMAP visualizations of the cell types from the granule cell lineage (4007 cells) [15], and the developmental pseudotime, transcriptomic vector field, and latent representations inferred by scTour. Leftmost panel shows the PCA space-based UMAP with the arrow indicating the differentiation from nIPCs to mature granule cells. b PCA space-based UMAP embedding showing the cell types (colors, 15,174 cells) [15] along the pyramidal and granule cell lineages (arrows). c As in b, but colored by sample batches. d As in b, but colored by the developmental pseudotime derived from the scTour model. e Developmental ordering of cells by the pseudotime inferred from scTour. Cells are colored from top to bottom by pseudotime, sample batches, and cell types. f UMAP visualizations of the latent representations learned from scTour, with colors denoting the cell types and sample batches (inset). g Streamline visualization of the transcriptomic vector field from scTour on the same embedding as in f, with cells color-coded by the inferred pseudotime. h Developmental ordering of cells by the pseudotime estimated from scTour models trained using a range of cell subsets (1 to 95% of total cells from top to bottom). Cells are colored by cell types. i UMAP visualizations of the latent representations, developmental pseudotime (colors), and transcriptomic vector field (streamlines) learned from the scTour model trained based on 20% of total cells. The inset shows the same plot but color-coded by cell types

Back to article page