A Unified Modular Framework to Incorporate Structural Dependency in Spatial Omics Data

Abstract

AbstractSpatial omics technologies, such as spatial transcriptomics, allow the identification of spatially organized biological processes, while presenting computational challenges for existing analysis approaches that ignore spatial dependencies. Here we introduce Smoother, a unified and modular framework that integrates positional information into non-spatial models via spatial priors and losses. In simulated and real datasets, we show that Smoother enables spatially aware data imputation, cell-type deconvolution, and dimensionality reduction with high accuracy.