Fig. 1: Overview of inputs and outputs of DOT and its optimization framework.

a From left to right: DOT takes two inputs: (i) spatially resolved transcriptomics data, which contains spatial measurements of genes at either high or low-resolution spots and their spatial coordinates, and (ii) reference singe-cell RNA-seq data, which contains single cells with categorical (e.g., cell type) or continuous (e.g., expression of genes that are missing in the spatial data) annotations. DOT employs several alignment objectives to locate the cell populations and the annotations therein in the spatial data. The alignment objectives ensure a high-quality transfer from different perspectives: b the expression profile of each spot in the spatial data (left) must be similar to the expression profile transferred to that spot from the reference scRNA-seq data (right), c the expression profile of each cell population in the reference data (left) must be similar to the expression profile of that cell population inferred in the spatial data (right), d expression map of each gene in the spatial data (left) must be similar to expression map of that gene as transferred from the reference data (right), e spots that are both adjacent and have similar expression profiles are likely to have similar compositions, and (f) if prior knowledge about the expected relative abundance of cell populations is available, the transfer should retain the given abundances.