ONTraC characterizes spatially continuous variations of tissue microenvironment through niche trajectory analysis

Abstract

Recent technological advances enable mapping of tissue spatial organization at single-cell resolution, but methods for analyzing spatially continuous microenvironments are still lacking. We introduce ONTraC, a graph neural network-based framework for constructing spatial trajectories at niche-level. Through benchmarking analyses using multiple simulated and real datasets, we show that ONTraC outperforms existing methods. ONTraC captures both normal anatomical structures and disease-associated tissue microenvironment changes. In addition, it identifies tissue microenvironment-dependent shifts in gene expression, regulatory network, and cell–cell interaction patterns. Taken together, ONTraC provides a useful framework for characterizing the structural and functional organization of tissue microenvironments.