The Spatial Logic of Cell Organization in Complex Tissues

In biology we do not understand space as much as we understand gene expression. Live imaging, single-cell omics and deep learning are converging to provide unprecedented data and analysis to understand the spatial organization of complex tissues, but we still lack a general and effect paradigm to uncover patterns, let alone mechanistic insights from the rich multiscale and multimodal data. We hypothesize that informative patterns are more evident in some integrated high dimensional space, which in turn can be constructed through deep learning. Along these lines, I discuss our recent efforts on cell-based spatiotemporal learning for general representation of cell positions and to uncover collective cell behaviors.