SpatialEx 0.1.7

computational frameworks that leverage histology as a universal anchor to integrate spatial molecular data across tissue sections

The foundational SpatialEx model combines a pre-trained H&E foundation model with hypergraph learning and contrastive learning to predict single-cell omics profiles from histology, encoding multi-neighborhood spatial dependencies and global tissue context. Building upon SpatialEx, SpatialEx+ introduces an omics cycle module that encourages cross-omics consistency across adjacent sections via slice-invariant mapping functions, achieving seamless diagonal integration without requiring co-measured multi-omics data for training.