Extended Data Fig. 2: Analysis of spatial transcriptomics and villus zonation assembly.

a-d, UMAP of 15,145 spots (n = 8) coloured by metadata as indicated in the figure (Methods). e, Heat map showing relative gene expression per cluster. Columns represent 100 down-sampled spots from each cluster indicated by colour and title above columns. Row represents genes. f-h, Stacked bar plots showing the relative abundances of spots from each cluster by patient (f), crypt-villus axis zone (g) and stromal-epithelial axis zone (h). EEC.K\M – Enteroendocrine K \ M cells; ILF – Isolated lymphoid follicle; LP – Lamina propria; RBC – Red blood cells; V-Villus. SM – Submucosa. i, H&E image (up) and a colour-coded pixel classification results for the same image (down) with blow-ups (n = 8) (Methods). Scale bars 100 µm. j, Representative (n = 12) field of views showing H&E (left), certified pathologist annotations serving as ground truth (middle) and the pixel classifier prediction (right). Prediction shown as maximal probability given by the classifier, while probabilities were used to compute Visium spot composition (Methods). Scale bars 5 µm. k, Visium spots for white dashed area in i. spots are shown as pie charts coloured by the composition of pixels for each spot. Colour-code as in i. l, Visium spots on k, colour-coded by their inferred stromal-epithelial axis relative coordinate. m, Zonation profiles for genes peaking at different zones along the crypt-villus axis. Top panels show Visium spots coloured by log10(UMI fraction); bottom panels show zonation profile for P3 (Methods). Lines are normalized to mean, patches show SEM. n, Genes that are zonated along the epithelial stromal axis. Left panels show Visium spots coloured by log10(UMI fraction); right panels show expression levels stratified to a zone for P3 (n = 3105 spots, Methods). Error-bars show SEM. Apical – Apical epithelium; Basal – Basal epithelium; LP – Lamina propria; MMuc – Muscularis mucosa; SMuc – Submucosa. C – Crypt.