Automating Polymer Molecular Dynamic Simulations: Simulation-Ready Ensembles from G-BigSMILES 2.0

Machine learning and informatics efforts in polymer science continue to be challenged by limited and fragmented datasets. While advanced algorithms now increasingly exploit datasets that, in some cases, include thousands of polymer entries, most polymer property databases remain modest in size and suffer from heterogeneous metadata and structure. Molecular simulations offer a promising route to expand machine-readable datasets; however, the setup of polymer simulations still represents a bottleneck for the high-throughput generation of data. In this work, we present an automatic pipeline for performing molecular dynamics simulations of polymer systems and calculating relevant properties. Our pipeline employs the G-BigSMILES 2.0 notation to generate ensembles of chains that accurately reflect a given molecular weight distribution, as well as the connectivity and composition of repeat units, enabling the description of a wide variety of polymer architectures. By incorporating these essential features of polymer materials, we aim to expand the scope of simulated data available for computational materials discovery.