Coarse-grained Molecular Dynamics Modeling of Actin Self-assembly

Actin filaments, microtubules, and intermediate filaments are major components of the cytoskeleton. The self-assembly process of actin filaments, where actin proteins transition from their free monomer form (G-actin) to be part of a long, double-stranded, helical polymer microfilament (F-actin), is of fundamental importance to biology and also serves as a platform to inspire the design of new biomimetic materials. A coarse-grained model is developed to capture the geometric features of G-actin and the assembly characteristics of microfilaments. The model monomer has the shape of a slightly bent, twisted rod with binding sites on its lateral and top/bottom surfaces. Such monomers are found to form a variety of structures, including double-stranded actin filaments. A machine learning approach is developed to classify the monomer states in a self-assembling system by computing the Smooth Overlap of Atomic Positions (SOAP) descriptor for each monomer. Probabilistic Analysis of Molecular Motifs (PAMM), based on a principal component analysis along with an unsupervised clustering algorithm, is adopted to identify free monomers, monomers in the middle and at the end of a filament, and defects. We further define a metric on the basis of such classification information to capture the type and extent of the self-assembly ordering of a given system. This metric enables us to quantify the relation between monomer-monomer binding strengths, self-assembly structures, and defects. Overall, we obtain an understanding of the self-assembly process of actin filaments.