Abstract

Investigations of spatial cellular composition of tissue architectures revealed by multiplexed in situ RNA detection often rely on inaccurate cell segmentation or prior biological knowledge from complementary single-cell sequencing experiments. Here, we present spage2vec, an unsupervised segmentation-free approach for decrypting the spatial transcriptomic heterogeneity of complex tissues at subcellular resolution. Spage2vec represents the spatial transcriptomic landscape of tissue samples as a graph and leverages a powerful machine learning graph representation technique to create a lower dimensional representation of local spatial gene expression. We apply spage2vec to mouse brain data from three different in situ transcriptomic assays and to a spatial gene expression dataset consisting of hundreds of individual cells. We show that learned representations encode meaningful biological spatial information of re-occurring localized gene expression signatures involved in cellular and subcellular processes. DATABASE: Spatial gene expression data are available in Zenodo database at https://doi.org/10.5281/zenodo.3897401. Source code for reproducing analysis results and figures is available in Zenodo database at http://www.doi.org/10.5281/zenodo.4030404.

Keywords

MeSH terms