Topography Aware Optimal Transport for Alignment of Spatial Omics Data

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown

Published: April 20, 2025

Abstract Understanding the spatial organization of tissues is essential for uncovering cellular communication, developmental processes, and disease mechanisms. Recent advances in omics technologies have provided unprecedented insight into tissue organization, but challenges remain aligning slices integrating complementary single-cell data. Here, we propose TOAST (Topography-aware Optimal Alignment Spatially-resolved Tissues), a novel OT-based framework that extends classical Fused Gromov-Wasserstein (FGW) objective to more comprehensively model heterogeneity local molecular interactions. By introducing coherence , quantified through entropy neighborhoods, neighborhood consistency which preserves expression profiles neighboring spots, TOAST’s function significantly improves alignment spatially resolved mapping between Through comprehensive evaluations on both simulated real-world datasets, including human brain cortex Visium data, Axolotl Stereo-seq mouse embryo seqFISH Imaging Mass Cytometry from multiple cancer types, demonstrate our method consistently outperforms traditional FGW other methods. Specifically, accuracy slice alignment, better cell type compositions, recovers lineage trajectories reconstructs relationships transcriptomics constraints OT, provides principled approach enhance biological interpretability data facilitate multimodal integration.

Language: Английский

Learning Latent Trajectories in Developmental Time Series with Hidden-Markov Optimal Transport

Lecture notes in computer science, Journal Year: 2025, Volume and Issue: unknown, P. 367 - 370

Published: Jan. 1, 2025

Language: Английский