A µ FE Simulation-based Surrogate Machine Learning Model to Predict Mechanical Functionality of Protein Networks from Live Confocal Imaging

Understanding sub-cellular mechanics is crucial for a better understanding of a variety of biological functions and dysfunctions. A structure-function analysis of the cytoskeletal protein networks provides not only ways to deduce from its structure insights into its mechanical behaviour but potentially also new insights into sub-cellular processes such as mechano-transduction, stiffness-induced cytoskeletal restructuring and stiffness changes, or mechanical aspects of cell-biomaterial interactions. Recently, fluorescence imaging has become a powerful tool to study protein network structures at high resolution. Yet, automated tools for quantitative functional analysis of these complex structures, are missing. These, however, are needed to relate structural characteristics to cellular functionality. Here, we present a machine learning framework that combines 3D imaging and mechanical modelling on the nano scale, enabling prediction of mechanical behaviour of protein networks and the subsequent automatic extraction of structural features of which one can deduce mechanical characteristics. This study demonstrates the method’s applicability to investigate the skeleton’s functionality of the Filamentous Temperature Sensitive Z (FtsZ) family inside organelles (here, chloroplasts) of the moss .