Accurate reconstruction of spatial cell-type maps and characterization of domain-specific functions based on a gene-aware heterogeneous network.

STGnet achieves the most accurate spatial cell type deconvolution and functional annotation, outperforming existing methods on simulated and real datasets.

• Why it matters: Understanding the spatial organization of cell types and functions is crucial for uncovering disease mechanisms and tissue pathology, but current methods lack integration of spatial and gene-level information.
• What they did: The authors developed STGnet, a gene-aware heterogeneous graph attention network that uses a hybrid pseudo-spot strategy to incorporate spatial adjacency, transcriptional similarity, and gene-spot associations for deconvolution and domain-specific function analysis.
• The result: STGnet provides highly accurate cell type maps and interpretable domain-specific genes, enabling the characterization of spatially ordered functional programs linked to disease progression and tissue organization.