Fig. 2: Superior performance of SpaTalk over existing methods.

a Schematic diagram for generating simulated spot data. Cells were split according to the fixed spatial distance and then merged for the single-cell ST data with known cell types. b Performance comparison of SpaTalk with other existing cell-type deconvolution methods (RCTD, Seurat, SPOTlight, deconvSeq, Stereoscope, Cell2location). The asterisk represents the top-ranked method for each dataset. NA, not available. c Schematic illustration of the procedure and rationale for single-cell ST data to evaluate predicted LRIs that mediate spatially resolved cell–cell communications. d and e Performance comparison of SpaTalk with existing cell–cell communication inference methods (Giotto, SpaOTsc, NicheNet, CytoTalk, CellCall, CellPhoneDB, and CellChat) on the STARmap and seqFISH+ datasets. The P-value represents the difference of spatial distances between sender-receiver and all cell–cell pairs assessed with the Wilcoxon test. For the boxplots (minima, 25th percentile, median, 75th percentile, and maxima) from SpaTalk to CellChat, the numbers of data points for the STARmap dataset are 15, 5, 676, 537, 75, 31, 7, and 166, respectively; the numbers of data points for the seqFISH+ OB dataset are 1,559, 38, 3,972, 1,223, 375, 675, 10, and 404, respectively; the numbers of data points for the seqFISH+ SVZ dataset are 9817, 50, 3337, 1424, 1553, 7798, 258, and 291, respectively. f Schematic illustration of the procedure and rationale for single-cell ST data to evaluate predicted downstream target and pathways underlying LRIs. g Performance comparison of SpaTalk on the inferred downstream targets with other methods (NicheNet, CytoTalk, and CellCall) over the STARmap and seqFISH+ datasets. The P-value represents the significance of enriched pathways or biological processes from the KEGG and Reactome databases using inferred downstream targets with the Fisher-exact test. For the boxplots (minima, 25th percentile, median, 75th percentile, and maxima) from SpaTalk to CellCall, the numbers of data points for the STARmap dataset are 25, 1203, 66, and 68, respectively; the numbers of data points for the seqFISH+ OB dataset are 23,132, 19,209, 5168, and 27,668, respectively; the numbers of data points for the seqFISH+ SVZ dataset are 121,693, 25,216, 15,454, and 225,231, respectively.