Presenter(s): Jake Olsen—Chemistry and Mathematics
Faculty Mentor(s): Marina Guenza, Jake Searcy
Session 1: It’s a Science Thing
Polymeric systems, things like proteins, DNA, and synthetic plastics, are of great interest for their applications in material design and the biomedical industry . Therefore, having time-efficient and inexpensive approaches to investigate these systems on multiple scales, from the microscopic to the macroscopic level, is of great importance and necessity . Molecular dynamics (MD) simulations are one such tool for investigating these systems; however, MD simulations that simulate polymer systems in their atomistic (AT) representation are unable to reach the time scale necessary so that the system exhibits the correct chain characteristics . Thus, coarse-graining (CG) methods, a process by which the local degrees of freedom are averaged out, are applied to improve upon computational time . Unfortunately, the computational gain is coupled with the loss of statistical information from the CG process . Therefore, to regain the lost AT information the CG trajectories need to be transformed back to an AT representation . This process is known as backmapping . Utilizing state- of-the-art machine learning techniques coupled with AT data, we have developed a backmapping procedure for CG polymeric systems . The model, centered around a recurrent neural network (RNN), shows strong agreement with the AT data across many statistical quantities prompting further investigation and development of the model .