Supplementary Figure 9: Assessment of statistical calibration of SpatialDE through data randomization.

(A) QQ-plot of expected P-values (Chi2 distribution with 1 degree of freedom) versus observed P-values from the SpatialDE tests on the breast cancer data. (B) To simulate data from an empirical null, without spatial structure, expression values were shuffled among the sampled coordinates. Shown is the resulting expression pattern for COL3A1 expression, as representative example. (C) SpatialDE negative log P-values for genes on shuffled data, with the number of detected SV genes (FD<0.05) being consistent with the selected false discovery rate. (D) Analogous QQ-plot as in A on shuffled expression values. SpatialDE P-values follow the null distribution, indicating that the model is calibrated.