Polybetaines, having both cationic and anionic moieties within the same pendant groups, present interesting properties such as antifouling, biocompatibility, tunable mechanical properties, and ionic conductivity. Studies of these materials have reported meaningful qualitative design rules; however, few quantitative models have been proposed relating macroscopic properties to monomer chemistry and system design. This is primarily due to the multiscale nature of these polymers and the lack of large, uniform datasets of polybetaine properties. Here, we present a multiscale machine learning (ML) approach incorporating intermediate-scale data from molecular dynamics simulations to elucidate quantitative structure-property relationships (QSPRs) describing polybetaine hydration and ionic conductivity. Common betaines are simulated in aqueous conditions with and without salt at the molecular, single-polymer, and multi-polymer scales to examine hydration, ion association, and ionic conductivity properties at each scale. These data are used as target variables for evaluating linear and non-linear ML models, while input features consist of betaine functional groups and cheminformatic descriptors, along with system properties. This method produces a large, controlled dataset of betaine and polybetaine properties and can reveal QSPRs of polybetaines while demonstrating the relevance of scale-specific features to macroscopic hydration and ionic conductivity properties.

Multiscale machine learning approach to modeling structure-property relationships of polybetaines

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.