Pancreatic cancer histopathology/spatial transcriptomics data from Zhou et al.

This dataset contains the h5ad files for the 13 samples with histopathology and spatial transcriptomics features.

Each h5ad contains the following information necessary for running with the SpatialDIVA model:

• Raw ST counts per spot in the .X attribute of the anndata object
• Histopathology features extracted via the UNI foundation model (https://www.nature.com/articles/s41591-024-02857-3), stored in .obs columns beginning with "UNI". These features are also on a per-spot level.
• Spatial coordinates for each spot stored in .obsm["spatial"]
• Pathologist annotations on a per-spot level stored in .obs["is_tumor"]
• The most prevalent cell-type per spot stored in .obs["ST_celltype"]
• Raw de-convolution results for ST cell-types in various .obs columns ending with the cell-type names

The original data with the ST transcriptomic counts and histopathology - https://www.nature.com/articles/s41588-022-01157-1. We reannotated the dataset for both the ST-celltype annotations and pathologist annotations - details are outlined in the preprint https://www.biorxiv.org/content/10.1101/2025.02.19.638201v1.