Figure 1.

Schematic representation of TransformerST. A, The VisionTransformer encoder amalgamates spatial gene expression, spatial location and histology image, facilitating the exploration of image-gene expression co-expression. B, The Adaptive Graph Transformer model is employed to harness spatial neighboring dependence, enabling the association of spatial gene expression patterns at the original resolution. C, The Cross-scale Internal Graph Network is utilized for the super-resolved reconstruction of gene expression, taking concatenated embedding and histology image as inputs to elevate gene expression from multicellular to single-cell resolution.