Abstract
During developmental processes such as embryogenesis, how a group of cellsfold into specific structures, is a central question in biology that defineshow living organisms form. Establishing tissue-level morphology criticallyrelies on how every single cell decides to position itself relative to itsneighboring cells. Despite its importance, it remains a major challenge tounderstand and predict the behavior of every cell within the living tissue overtime during such intricate processes. To tackle this question, we propose ageometric deep learning model that can predict multicellular folding andembryogenesis, accurately capturing the highly convoluted spatial interactionsamong cells. We demonstrate that multicellular data can be represented withboth granular and foam-like physical pictures through a unified graph datastructure, considering both cellular interactions and cell junction networks.We successfully use our model to achieve two important tasks, interpretable 4-Dmorphological sequence alignment, and predicting local cell rearrangementsbefore they occur at single-cell resolution. Furthermore, using an activationmap and ablation studies, we demonstrate that cell geometries and cell junctionnetworks together regulate local cell rearrangement which is critical forembryo morphogenesis. This approach provides a novel paradigm to studymorphogenesis, highlighting a unified data structure and harnessing the powerof geometric deep learning to accurately model the mechanisms and behaviors ofcells during development. It offers a pathway toward creating a unified dynamicmorphological atlas for a variety of developmental processes such asembryogenesis.