Fig. 5: Spatial multi-omics analyses on the Xenium pancreatic cancer dataset.

a Overview of spatial multi-omics alignment. For each single cell, the information of (i) transcriptomics, (ii) stainings, and (iii) H&E is combined after the alignment of all the different layers. b H&E clusters of patch-level embeddings based on a pre-trained computer vision model (denoted as H&E niches). The figure shows the cells obtained from spatial transcriptomics data and coloured by the H&E patch cluster inside which they are included. c Proportion of cell types inside each H&E niche. The cell types are the cell types annotated using both spatial transcriptomics and protein information as in Fig. 3. d Differential gene expression performed on the H&E niches using single-cell resolution. e Distribution of TROP2 intensities per cell (N = 175,022) inside each H&E niche, showcasing the usage of the staining layer coupled with the H&E information. Source data for e are provided as a Source Data file.