The paper describes TopoLa, a novel computational framework that enhances cell representations for single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics (ST) data by leveraging Topology-encoded Latent Hyperbolic Geometry. The framework introduces two main components: TopoLa distance (TLd), a metric that quantifies geometric distance and topological relationships between cells in a latent hyperbolic space, and TopoConv, a spatial convolution technique that uses TLd to refine cell representations. The authors demonstrate TopoLa's universal applicability and robustness by integrating it with existing state-of-the-art models (like SIMLR, scGNN, scGPT, GraphST, and IRIS) across seven critical biological tasks, including clustering, multi-batch/multi-omic integration, and rare cell identification, showing significant performance improvements across all metrics. This approach provides a new theoretical foundation for analyzing complex cellular network geometries, establishing TopoLa as a valuable tool for advancing cell science and bioinformatics.

References:

• Zheng K, Wang S, Xu Y, et al. TopoLa: A Universal Framework to Enhance Cell Representations for Single-cell and Spatial Omics through Topology-encoded Latent Hyperbolic Geometry[J]. bioRxiv, 2025: 2025.07. 23.666288.