The article introduces StaVia, a novel computational framework designed for single-cell atlases, which are large datasets detailing cellular states and differentiation. StaVia addresses major challenges in trajectory inference by integrating spatial and temporal metadata with a method utilizing higher-order random walks with memory to accurately trace complex cell lineage pathways. The framework also includes a cartographic Atlas View for intuitive visualization of these large-scale trajectories, outperforming existing methods like UMAP and CellRank in maintaining both global continuity and local separation of cell types. The authors demonstrate StaVia's utility using extensive datasets, including murine gastrulation and the Zebrahub developmental atlas, highlighting its capacity for high-definition, spatially-aware trajectory analysis on complex omics data. The discussion also details the mathematical and algorithmic components underpinning the approach, such as Kernel Density Estimation for edge bundling in visualizations.

References:

• Stassen S V, Kobashi M, Lam E Y, et al. StaVia: spatially and temporally aware cartography with higher-order random walks for cell atlases[J]. Genome Biology, 2024, 25(1): 224.