A0847
Title: TopSpace: Bayesian spatial topic modeling for unsupervised discovery of spatial tissue structures in multiplex images
Authors: Junsouk Choi - University of Michigan (United States) [presenting]
Abstract: The recent development of multiplex imaging technologies allows for measuring the expression of tens of protein markers at single-cell resolution while preserving spatial information of cells, enabling direct observation of cellular phenotypes, spatial distributions, and interactions in intact tissues. A key research question in analyzing such data is to identify higher-order patterns of tissue organization, which holds systematic implications for disease pathology and clinical outcomes. To address this, TopSpace, a novel Bayesian topic model, is proposed to identify the higher-order architecture of tissues and recover signatures of characteristic cellular microenvironments that are potential determinants of patient outcomes. The proposed approach infers the local distribution of cellular phenotypes to represent the cellular microenvironment and incorporates spatial information via Gaussian processes to ensure spatial coherence among neighboring microenvironments. By applying the proposed TopSpace to publicly available multiplexed imaging data, higher-order architectures are uncovered within lung cancer tissues and identified tertiary lymphoid structures, which are strongly associated with patient survival.
