Single-cell genomics technology allows multimodal profiling of millions of cells in temporal and spatial dimensions. Experimental limitations prevent the measurement of all-encompassing cellular conditions in their temporal or spatial tissue niche dynamics. The optimal transport theory has emerged as a powerful tool to overcome these constraints, allowing the discovery of the real cellular context. However, most of the currently available algorithmic implementations have not kept up with the increasing speed of the dataset, so that current methods cannot integrate multimodal or scale information into single cell atlases. Here, we present multi-omics single-cell optimal transport (moscot), a general and scalable framework for optimal transport applications in single-cell genomics, supporting multimodality in all applications. We demonstrate the ability of moscot to efficiently reconstruct the developmental pathways of 1.7 million cells in mouse embryos over 20 periods and identify the genes responsible for the formation of the first heart field. The construction of moscot can be used to transport cells on a spatial scale as well: To demonstrate this, we enrich datasets of spatial transcriptomics data by mapping multimodal information from single cell profiles in a mouse liver sample, and align multiple coronal sections of the mouse brain. We then present moscot.spatiotemporal, a new method that maximizes gene expression across all spatial and temporal dimensions to reveal the spatiotemporal dynamics of mouse embryogenesis. Finally, we dissect lineage relationships in a novel resolved pancreas development dataset using paired measurements of gene expression and chromatin accessibility, finding evidence of a shared lineage between delta and epsilon cells. Moscot is available as an easy-to-use, open-source python package with extensive documentation
† Foreign Donor Cooperation
ETH Zurich: Marius Lange
Google Research: Laetitia Meng-Papaxanthos
Hebrew University of Jerusalem: Zoe Piran, Mor Nitzan
Helmholtz Munich: Dominik Klein, Giovanni Palla, Marius Lange, Manuel Gander, Michael Sterr, Aimée Bastidas-Ponce, Marta Tarquis-Medina, Heiko Lickert, Mostafa Bakhti, Fabian J. Theis
Technical University of Munich (TUM): Dominik Klein, Giovanni Palla, Marius Lange, Heiko Lickert, Fabian J. Theis
