Internal KCN: generative algorithm for realistic astrocytes
TITLE:
Biophysically Constrained Generative Algorithm Synthesizing Realistic Astrocytes in silico.
ABSTRACT:
Astrocytes are intricately branched cells that infiltrate the neuropil, enveloping neuronal processes and modulating neural function through diverse molecular pathways governed by their intracellular signalling. The arrangement of these branches defines the astrocyte's modulatory domain and has been proposed to influence the propagation of astrocyte intracellular signals. Therefore, understanding the shape and structure of this anatomical domain is key to codifying the physiological relevance of neuron-astrocyte interactions. However, this knowledge is currently missing as we lack a detailed characterization of astrocyte branching morphology. To address this gap, we utilize a comprehensive library of morphometric measures to extract distinct anatomical features of astrocytes resolved by electron microscopy. These features are then integrated into machine learning frameworks to develop generative algorithms capable of synthesizing realistic astrocytic tree structures. Our simulations reveal that astrocyte branch arrangement is optimized to balance spacefilling properties with minimizing total cytosolic volume. At the same time, they provide a general prescription to perform automated classification of astrocyte types and distinguish astrocytes from neurons. Our preliminary data represents the first step towards a detailed characterization of astrocyte morphology and highlights the potential biophysical principles underpinning astrocyte branch organization.