|
Abstract:
|
Recent advances in spatial transcriptomics technologies have enabled characterization of gene expression patterns in the context of tissue microenvironment. Although methods have been developed to detect spatially variable genes, these methods often ignore cell type variations in a tissue. To elucidate spatial gene expression variation in the presence of cellular heterogeneity, we developed SpaGCN, a graph convolutional network approach that integrates information from gene expression, spatial location and histology. The aggregated gene expression through this graph is then fed into a deep embedding clustering algorithm to cluster spots into spatial domains. After spatial domains are identified, genes enriched in each domain are detected by differential expression analysis. To understand cell type composition variation in tissue, we further developed SpaDecon, a transfer learning-based approach that incorporates spatial location and histology information for cell type deconvolution. I will illustrate the applications of SpaGCN and SpaDecon through the analysis of spatial transcriptomics data generated from mouse and human brains.
|
