Graph Attention Network generates Super-resolution Spatial Transcriptomic data

Spatial transcriptomics allow uncovering the spatial origin of RNA molecules within a tissue slide. However, acquiring informative signals remains a challenge, with a trade-off between sequencing depth and spatial resolution. Sequencing-based technologies provide unbiased transcriptional profiles, but their minimal workable spatial unit compromises more than one cell. While deconvolution methods have been proposed to estimate the cell-type composition of each spot, we believe that the resolution can be further improved by generating enhanced subpots in place of the original spots, such that cells in the original spot can be allocated into the newly generated subspots. We propose a computational method based on graph and attention learning, named Square, that leverages message passing for information sharing between neighboring spots. Even though this rearrangement of cells can be solely spatially approximated, a resolution enhancement is achieved. We show that the proposed approach is capable of deciphering the composition of spots in ST and Visium samples, whilst imputing sparse profiles and amplifying the signal in them. Newly generated subspots have been empirically and biologically validated. The gap between both spatial transcriptomic modalities is then closed, generating high-throughput cellular-scale outputs.