Journal: Nature Communications
Article Title: Reconstructing single-cell resolution from spatial transcriptomics with CellRefiner
doi: 10.1038/s41467-026-70090-2
Figure Lengend Snippet: a MERFISH slice from the mouse hypothalamic preoptic region used as ground truth for benchmarking CellRefiner using a known cell-to-spot mapping. Zoomed-in view shows a region of the tissue structure around the ependymal cells (yellow) for creation of simulated data. Cells are aggregated to form pseudo-Visium spot data, which is used to initialize the physical model, followed by the spatially refined output. b Spatially perturbed MERFISH data before and after CellRefiner processing, with ependymal cells in red, and density estimation of the first-time step and last time step of the refinement process. c Refinement error calculated as KL-divergence between density estimations of CellRefiner output and ground truth over simulation iterations for ependymal cells. d Performance comparison of spatial mapping methods using multiple metrics. Comparison of CellRefiner, CellTrek, CytoSPACE, and Tangram on four single-cell resolution spatial datasets. Bar heights represent mean values of metrics with error bars representing standard error of the mean. For the three metrics across cell types, each dot represents one cell type and n indicates the number of cell types. For Euclidean distance across cells, n indicates the number of cells. Source data are provided in the Source Data file. e Reconstructed single-cell resolution spatial data by CellRefiner of MERFISH, seqFISH, Slide-seqV2, and STARmap data.
Article Snippet: The mouse hippocampus Slide-seqV2 data is available at the Broad Institute Single Cell Portal ( https://singlecell.broadinstitute.org/single_cell/study/SCP815/sensitive-spatial-genome-wide-expression-profiling-at-cellular-resolution#study-summary ) with preprocessed data available via the Squidpy package .
Techniques: Comparison, Single Cell
