Fig. 2: Benchmarking mapping performance in synthetic spatial data – circle regime. | Nature Communications

Fig. 2: Benchmarking mapping performance in synthetic spatial data – circle regime.

From: Multi-scale and multi-context interpretable mapping of cell states across heterogeneous spatial samples

Fig. 2

We provide a Legend Key on the figure describing the nomenclature of Vesalius cost matrices as well as the meaning of acronyms. a Performance comparison in recovering cell types across samples using an ARI in synthetic spatial data. We compared Vesalius using 14 different cost matrices to SLAT, CytoSpace, CytoSpace with no cell type labels (noLab), Tangram, Scanorama, PASTE, and GPSA. The number of samples (n) used in each box plot is included in the axis labels. The box shows the interquartile range (IQR = Q3 − Q1), with a line at the median, and whiskers extending to the most extreme values within 1.5× IQR. CytoSpace obtains a perfect score since it only maps cells from the same cell type. b JI of cell interactions. To account for the spatial context of cells, we computed the JI between cell type labels in the neighborhood of mapped cells (k = 6). c Heat map representing p-values after performing a two-sided Wilcoxon rank sum test between tools on ARI scores. d Heat map representing p-values after performing a two-sided Wilcoxon rank sum test between tools on JI scores. e Example mapping event in the circle regime. In the circle regime, there are 5 randomly sized and randomly placed territories in a background territory. Each circle can contain 2 different cell types. We show mapping events for 3 Vesalius cost matrices (fncty, fnct, fnt). Source data are provided as a Source Data file.

Back to article page