A0183
Title: Reference-informed spatial domain detection for spatial transcriptomics
Authors: Xiang Zhou - University of Michigan (United States) [presenting]
Abstract: Spatial transcriptomics studies are becoming increasingly common and large, providing unprecedented opportunities for characterizing complex tissues' spatial and functional organization. A statistical method, IRIS, is presented that leverages single-cell RNA-seq (scRNA-seq) data to accurately and efficiently detect spatial domains on complex tissues in spatial transcriptomics. IRIS is capable of modelling multiple tissue slices jointly, explicitly accounts for the correlation both within and across slices, and takes advantage of numerous algorithmic innovations to achieve highly scalable computation. The advantages of IRIS through in-depth analysis of five spatial transcriptomics datasets from different technologies across distinct tissues and species are demonstrated. In real data applications, IRIS achieves 51\% - 94\% accuracy gain over existing methods in datasets with known ground truth. In addition, IRIS is faster than existing methods in moderate-sized datasets and is the only method applicable to large-scale spatial transcriptomics data collected today. As a result, IRIS captures the fine-scale structures of brain regions, reveals the spatial heterogeneity of tumour micro-environments, and characterizes the structural changes of the seminiferous tubes in the underlying testis diabetes, all at a speed and accuracy unattainable by existing approaches.