Unsupervised Data-Driven Reconstruction of Molecular Motifs in Simple to Complex Dynamic Micelles

The reshuffling mobility of molecular building blocks in self-assembled micelles is a determinant key of many interesting properties, from emerging morphologies and surface compartmentalizations to dynamic reconfigurability and stimuli-responsiveness of these supramolecular soft particles. However, such complex structural dynamics is typically non-trivial to be elucidated, especially for multi-component assemblies. Here we show a machine-learning approach that allows to reconstruct the structural and dynamic complexity of mono- and bi-component surfactant micelles from high-dimensional data extracted from equilibrium molecular dynamics simulations. Unsupervised clustering of smooth overlap of atomic position (SOAP) data enables to identify the main local molecular environments in a micelle, and to retrace their composition and dynamics, in terms of the exchange of surfactants among micelle clusters. Provided that there is sufficient difference between surfactants that are mixed in a multi-component micelle, this approach can effectively recognize diverse surfactant types even in an exquisitely agnostic, unsupervised way: solely based on their relative displacements and dynamic motions, and without prior information on the molecular species present in the system.