Figure 1. Scheme of BSP.

a) Definition of the patch in BSP. For spot $i$, a patch is defined as the set of all its neighboring spots within distance $D$. Local Mean ${\widehat{X}}_i^{\left(j\right)}$ is defined as the average expression of gene $j$ for all spots within the patch. b) Identification of SVGs with BSP. Gene $g_1$ indicates an SVG (red), while $g_n$ is a representative non-SVG gene (orange). Besides the original pattern as the left, we define a small patch of spots with a smaller radius $D_1$ (blue) in the middle column, and a big batch of spots with a larger radius $D_2$ (green) in the right column. c) The ratio $r_{D_1{D}_2}^{\left(j\right)}$ of the variance of the local mean between big batch ${\sigma }_{D_2}^{\left(j\right)}$ and the paired small batch ${\sigma }_{D_1}^{\left(j\right)}$ is chosen as the statistical value $r$. Gene $g_1$ (red bar) is statistically significant compared with the background gene $g_n$ (orange bar) on the fitted beta distribution of $r$.