(376ac) The Use of Experiment Directed Simulations to Improve the Accuracy of Simulations

Experiment Directed Simulation (EDS) is a single replica biasing technique that is used to improve the agreement between simulations and experiments. Experiment Directed Simulation is a maximum entropy biasing technique that makes the average of simulated collective variable agree with a reference which usually comes from experiments. Here, we present a broad overview of EDS applications in peptide modeling in GROMACS simulator and EDS implementation in a modified simulation engine, HOOMD-TF. EDS had been used to improve the simulations of water molecules, lithium ions, and coarse-grained proteins in the past. We used EDS to improve all-atom simulations of GYG peptide and Aβ21−30 peptide by matching chemical shifts. Aβ21−30 peptide is 10 residue long fragment of amyloid peptide (Aβ42). We also used EDS in coarse-grained modeling of Aβ42 amyloid peptide self-assembly. The amyloid peptides are interesting because they are linked to Alzheimer’s Disease that which has many variations and does not have potent cure. By adding experimental results directly to simulation with EDS, we were able to better characterize and understand structure and dynamics of these peptides. We recently developed HOOMD-TF software package, that combines GPU-accelerated computational, statistical, and analytical capabilities of TensorFlow (TF) library with the simulation engine called HOOMD-blue. We implemented EDS in HOOMD-TF, for making positions of Lennard Jones particles to follow special geometry in HOOMD. We showed that EDS can be applied in two simulation engines to simulate properties and behaviors of molecules that agree with experiments.