spatial-domain-stagate

Spatial domain detection identifies spatially coherent regions in single-slice spatial transcriptomics data by applying a graph-attention auto-encoder that jointly models gene expression and spatial neighborhood structure to generate per-cell or per-spot domain labels.

Yes, the STAGATE workflow works on both cell-level and spot-level spatial transcriptomics data. You need to provide x,y coordinates in your AnnData object to successfully generate the spatial domain assignments.

STAGATE clustering assigns domain labels by learning spatially informed embeddings through a graph-attention auto-encoder, then applying Gaussian Mixture or Leiden-based clustering to output per-cell/spot labels in adata.obs['spatial_domain'].

No, you do not need to strictly predefine the number of spatial domains. The STAGATE workflow allows flexible domain counting via the n_domains parameter or supports automatic domain detection directly from the data.

Modeling spatial neighbor relationships is best handled by jointly integrating gene expression and spatial coordinates using a graph-attention auto-encoder, which learns spatially informed embeddings for accurate spatial domain identification.

Running STAGATE for spatial domain detection requires the STAGATE package, with optional support from STAGATE_pyG, to process the AnnData object and output spatial domain assignments in adata.obs.