Abstract
AbstractSpatial omics analyze gene expression and interaction dynamics in relation to tissue structure and function. However, existing methods cannot model the intrinsic and spatial-induced variation in spatial omics data, thus failing to identify true spatial interaction effects. Here, we present Spatial Interaction Modeling using Variational Inference (SIMVI), an annotation-free framework that disentangles cell intrinsic and spatial-induced latent variables for modeling gene expression in spatial omics data. SIMVI enables novel downstream analyses, such as clustering and differential expression analysis based on disentangled representations, spatial effect (SE) identification, SE interpretation, and transfer learning on new measurements / modalities. We benchmarked SIMVI on both simulated and real datasets and show that SIMVI uniquely generates highly accurate SE inferences in synthetic datasets and unveils intrinsic variation in complex real datasets. We applied SIMVI to spatial omics data from diverse platforms and tissues (MERFISH human cortex / mouse liver, Slide-seqv2 mouse hippocampus, Spatial-ATAC-RNA-seq) and revealed various region-specific and cell-type-specific spatial interactions. In addition, our experiments on MERFISH human cortex and spatial-ATAC-RNA-seq showcased SIMVI’s power in identifying SEs for new samples / modalities. Finally, we applied SIMVI on a newly collected CosMx melanoma dataset. Using SIMVI, we identified immune cells associated with spatial-dependent interactions and revealed the underlying spatial variations associated with patient outcomes.
Publisher
Cold Spring Harbor Laboratory
Cited by
1 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献