Dataset --- "LYNX: a deep generative model for linking spatial dynamics and cell interactions in multimodal spatial data"

Dataset overview Pre-aligned multi-modal liver dataset for sample NIH_F5 with 2D (single tissue section) and a 3D (serial sections section_01…08) variants. Each variant pairs two spatially co-registered modalities — Xenium (spatial transcriptomics / RNA) and DESI (mass-spectrometry imaging / metabolomics), with cross-modal aligned spatial coordinates stored in the SpatialData .zarr stores (obsm/xenium_map, obsm/desi_map). data/LYNX_liver_dataset/ ├── NIH_F5_2D/ # single 2D section │ ├── xenium/ # Xenium (RNA) │ │ └── cell_feature_matrix.h5 # cell × gene counts (.h5ad) │ └── DESI/ # DESI (metabolomics) │ ├── NIH_F5.h5 # procesed pixel × ion intensity matrix │ └── NIH_F5.ome.tif # raw ion-image file │ └── NIH_F5_3D/ # serial 3D stack ├── xenium/ │ └── section_{}/ # one Xenium bundle per section │ └── sdata.zarr └── DESI/ └── section_{}_sdata.zarr # aligned DESI SpatialData store per section Formats: .zarr — SpatialData/OME-Zarr stores (images + AnnData tables with aligned spatial / xenium_map / desi_map coordinates); .h5 — cell×gene (Xenium) or pixel×ion (DESI) matrices; .ome.tif — morphology / DESI ion images.