Figure 5. Registration of Visium data to Allen Common Coordinate Framework and statistical analysis of aligned transcriptomic spots.

A. Nonlinear registration of the tissue image from a single Visium sample (A₁) and its transcriptomic spot coordinates (A₂) – shown as example: the gene Camk2n1 – to the template image (A₃), slice 70 from the Allen P56 Mouse Common Coordinate Framework (CCF. Due to the nonlinear nature of the registration, we were able to precisely align the sample image (A₄) to landmarks in the template image and apply that transformation to the spot coordinates (A₅). To account for different numbers of spots in individual samples, digital spots spaced at 150μm in a honeycomb were created for the template slioce. Each digital spot is populated with the log base 2 normalized transcriptomic counts from the 7 nearest spots from each sample in a group (A₇). This approach allows the comparison of gene expression across entire brain slices in an unrestricted inference space. B-G. Samples were split into non-sleep deprived (NSD, n=6, 42 sample spots per digital spot) and sleep deprived (SD, n=7, 49 sample spots per digital spot). The range of the color bar for the mean calculations is set from 0 to the maximum normalized gene count for that gene for all samples, while the t-statistic color bar is bounded to [−4,4], which is approximately the equivalent to the Šidák corrected p-value of < 2.50e-05. We show a selected group of 6 genes from the 428 DEGs (Sup. Table X) (B-G). Panel 1 shows for each gene (B1-G1) the mean normalized gene count in NSD, panel 2 depicts the mean normalized gene count in SD (B2-G2) and panel 3 shows the t-statistics (B3-G3). The following DEGs are depicted: B. Per1, 4 significant spots. C. Nr4a1, 29 significant spots. D. Homer1, 306 significant spots. E. Arc, 168 significant spots. F. Rbm3, 31 significant spots. G. Cirbp, 9 significant spots.