Spatial Transcriptomics: Identifying Spatially Variable Genes

Presentation overview

Yu Chen will discuss two statistical approaches to identifying spatially variable genes in spatial transcriptomics. These data combine gene-expression measurements with tissue locations, making it possible to study how expression changes across space. The presentation will distinguish overall spatial variation from variation within a particular cell type.

Weber and colleagues introduce nnSVG, which uses nearest-neighbor Gaussian processes to identify spatially variable genes while reducing computational cost. Shang, Wu, and Zhou introduce Celina, which uses a spatially varying coefficient model to detect cell type-specific spatial expression patterns. Together, the papers motivate a discussion of scalability, cell-type composition, and the biological questions addressed by different definitions of spatial variation.