Fig. 3: Comparing tissue cohorts on the basis of spatial structure.

a,b, Spatial power analysis provides insight into experimental design. Schematic of design to distinguish cohorts (or individual samples) (a) or select cohort size to detect a spatial feature of interest (b). c,e,g, Larger FOVs distinguish tissues with different AESs. Distribution of AESs for FOVs of 5% (c), 7.5% (e), or 10% (g) of the absolute tissue size, drawn from the original spleen tissue (cyan) or a modified data set in which the overall number of CD4⁺–CD8⁺ T cell adjacencies was reduced by 37% but the relative frequency of cell types and absolute structure were preserved (magenta). Dashed lines, probability density functions for the original (red) and modified (black) tissues. d,f,h, Power analysis for comparison of distributions of AES between original and modified tissues. P value (y axis) in a t-test comparing the AES distribution of original and modified tissues (as in c,e,g), for different numbers of FOVs (x axis) that were 5% (d), 7.5% (f), or 10% (h) of the absolute tissue size. i,j, Power analysis for number of tissues required to detect a significant cell–cell adjacency. i, Distributions of number of unique cell–cell adjacencies (y axis) detected as significant (P < 0.01, permutation test, one-sided) in a set of 20 ISTs (x axis), for 729 possible cell–cell adjacencies (light gray) and for the 50 cell–cell adjacencies that were significant in all three real spleen tissues (dark gray). j, Probability of observing a significant adjacency (y axis) for a different number of tissues sampled (x axis) for adjacencies recovered in different numbers of ISTs (line color).