Self-Guided Molecular Simulation Methods

This work reviews self-guided (SG) molecular simulation methods and illustrates the characteristics and applications of these methods through several example simulations. The main characteristic of SG methods is that past motion in simulations is used to guide future motion. Two forms of these methods are self-guided molecular dynamics (SGMD) and self-guided Langevin dynamics (SGLD). SG methods achieve an enhanced conformational search through promoting low-frequency motion. A simple local averaging scheme is used to extract low-frequency properties from past simulation trajectories to promote low-frequency motion, which significantly enhances conformational search efficiency with little overhead in computing cost. Based on a generalized Langevin equation (GLE), an SGLD-GLE simulation method is developed, which has enhanced conformational searching ability and at the same time can vigorously sample the canonical ensemble. A reformulation of the SG methods leads to a quantitative relation between the guiding parameters and the conformational distribution, which allows the SG methods to be combined with the replica exchange scheme to perform replica-exchanging self-guided simulations (RXSGMD/RXSGLD). RXSGMD/RXSGLD are much more efficient than temperature-based replica exchange methods, especially for large systems.