Fig. 3

Tensor decomposition reveals the spatiotemporal patterns of gene expression during mouse embryo development. a The schematic of spatiotemporal dataset of mouse development is represented by a three-order tensor (4 time points * 8 sub-tissues * 1000 highly variable genes), and then it is decomposed into seven latent factors. b Spatial locations of the sub-tissues at four time points. Each spot in the sub-tissues represents an ROI. c Loading vectors of space and time for each factor obtained by tensor decomposition. Higher loading values indicate larger contribution of the sub-tissues or time points to the expression variation of this factor. d Spatial expression of example genes for each factor. The contours of heart, lung, and midgut are colored by red, blue, and green curves, respectively. ROIs of gene expression are presented by cyan points. The darker the cyan color, the higher the gene expression level