A1024
Title: Spatial deconvolution and cell type-specific spatially variable gene detection in spatial transcriptomics
Authors: Haohao Su - Michigan State University (United States)
Yuehua Cui - Michigan State University (United States)
Yuehua Cui - Michigan State University (United States) [presenting]
Abstract: Spatial transcriptomics (ST) provides crucial insights into tissue-specific gene expression patterns in various cancer studies. Most ST data, such as that obtained from the 10x Visium platform, is captured at a spot resolution that measures gene expression across multiple cells, often originating from various cell types. Deconvolution of such multi-cellular data to infer cell type compositions is crucial for further downstream analysis. Recent methodological developments have greatly advanced the detection of spatially variable genes (SVGs), whose expression patterns are non-random across tissue locations. Given that many SVGs correlate with cell type compositions, a unified approach is introduced to identify both SVGs and cell type-specific SVGs (ctSVGs), integrated with ST deconvolution, under a linear mixed-effect model framework. The method, termed STANCE, ensures tissue rotation-invariant results, with a two-stage testing strategy: Initial SVG/ctSVG detection followed by ctSVG-specific testing. Its performance is demonstrated through extensive simulations and analyses of public datasets. Downstream analyses reveal STANCE's potential in spatial transcriptomics analysis.
