Using co-localization priors and microenvironment statistics to reconstruct tissue organization from single-cell data
Abstract: Unveiling spatial expression patterns across tissues has been key for studying developmental processes, division of labor mechanisms, as well as variations in health and disease. Along the rapid development of improved experimental assays, computational methods have been shown to successfully recover spatial information from non-spatial single-cell data using reference atlases and/or assumptions about tissue organization such as relative smoothness of expression. However, spatial reconstruction can still be challenging for complex tissues, especially given a limited reference atlas. Here we show how information about tissue microenvironments statistics, such as cell type neighborhoods, or co-localization priors, can enhance tissue reconstruction in such cases. Specifically, we incorporate co-localization priors as a generalization of novoSpaRc, an optimal transport based framework for tissue reconstruction given single-cell data, which relies at its core on an interpolation between a structural correspondence assumption between expression and physical space and a potential reference atlas. We demonstrate that incorporating cell type co-localization priors can enhance the reconstruction of the mammalian organ of Corti and testicular spatial structure.