Unifying granular and foam-like representations of multicellular systems

Develop a single unified framework that integrates the granular perspective (cells represented as a point cloud with cell-level features) and the foam-like perspective (cell junction networks represented by edges and vertices) for mesoscale multicellular living systems, so that both cellular interactions and junction-level network information are coherently represented within one data structure.

Background

Historical approaches to modeling multicellular systems have arisen from two distinct physical pictures: a granular view that treats tissues as collections of cells with associated cell-level features, and a foam-like view that treats tissues as networks of cell edges and vertices reflecting junctions and boundary tensions. Each representation captures complementary aspects of tissue mechanics and morphogenesis, but they are typically developed and analyzed separately.

A unified framework would enable integrated geometric learning over both cell-level and junction-level information, supporting predictive modeling of developmental processes such as embryogenesis and local cell rearrangements. The paper motivates this need and proposes a dual-graph data structure and learning workflow to address it.