Speaker
Description
Active shape changes are essential for cell function, enabling motility, differentiation, and division. The active processes governing cell shape are tightly regulated by plasma membrane–bound proteins, which enable mechano- and chemosensitive responses as well as adhesion and force transmission to the environment.
We develop a trainable coarse-grained model of a cell that continuously adapts its shape by adapting mechanical properties and activity. Local adaptivity is controlled by artificial neural network policies that capture the dual role of the membrane as both a sensory interface and a coordinator of cell shape.
Using a neuroevolution algorithm, the model learns local adaptation rules that optimize the performance in complex tasks through emergent cooperative behavior of the cells constituent parts. This framework enables agent-based modeling of entire cells and tissues based on decentralized decision-making at the membrane level.
We apply this approach to problems in cell morphology, motility, and tissue formation, demonstrating its ability to inverse-design mechanical parameters for target shapes and to investigate how complex, hierarchical behaviors arise from local interactions and distributed information processing.