Annotation transferring

RIF Transfer ensure coherence in annotation of SDI results on target slices with the annotated reference slices. we adopted a transfer learning strategy to address that. To account for batch effects, We deconstruct the gene expression on each slice into two components: gene-specific expression captured by MLP, and spot-specific expression captured by GNN. The sum of these two components constitutes the reconstructed gene expression.

• 10X Visium DLPFC
• 10X Visium HBRC
• Slide-seq V2 MOB