Spatial Transcriptomics Methods

Spatial Transcriptomics Methods#

Technology overview#

Spatial transcriptomics (ST) measures gene expression while preserving the spatial position of cells or spots within the tissue. Key platforms include:

Platform

Resolution

Typical use case

Visium (10x)

Multi-cell spots (~55 µm)

Whole-tissue profiling

Xenium / MERSCOPE

Single-cell resolved

High-plex FISH

Slide-seq

Near single-cell

Broad coverage

Key challenges#

  • Spot deconvolution (Visium): multiple cell types per spot → DestVI.

  • Segmentation noise (resolved ST): transcript assignment errors → ResolVI.

  • Niche modelling: capturing cellular microenvironment effects → scVIVA.

Neighbour graphs#

Spatial neighbour graphs encode tissue topology. spatialvi-tools computes these via model.compute_neighbors() using squidpy (CPU) or RAPIDS (GPU). The resulting index_neighbor and distance_neighbor arrays in adata.obsm are consumed by ResolVI and scVIVA during training.

SpatialData integration#

All models support SpatialData objects via setup_spatialdata() and from_spatialdata().