ABSTRACT
Molecular dynamics (MD) simulations have been widely applied in many biological applications in recent years due to the advancement of novel algorithms and high-performance computing. Both structural and functional roles of many biomolecules such as proteins have been revealed, thanks to the advances in molecular modeling techniques. In order to understand molecular functions, it is crucial to adequately sample their conformational space. MD simulation is a powerful tool to explore the conformational space, but when applied to complex systems such as proteins, they are often trapped in local minima of the rugged free energy landscape. Recently, extremely long MD simulations enabled by specic hardware such as Anton [1,2] and Blue Gene [3] or a large ensemble of individual simulations generated by the distributed computing platform [4-9] start to make it possible to study the folding of small proteins in the explicit solvent. However, it is still a dicult task for conventional MD simulations to satisfactorily sample the conguration space of proteins due to the trapping problem. Generalized ensemble (GE) sampling algorithms such as replica exchange method (REM) (or parallel tempering) [10-13] and simulated tempering (ST) [14,15] were developed to overcome the sampling problem by inducing a random walk in temperature space in the expanded ensemble. In these methods, conformations sampled at high temperatures may be exchanged to lower ones and thus help the system to escape the local energy minima, because energetic barriers can be easily crossed with larger thermal uctuations at high temperatures. ese temperature-based GE algorithms have been widely applied in simulating proteins and other biological macromolecules [16-23]. In this chapter, we rst review the basic theory of MD simulations. en we introduce various GE-enhanced sampling algorithms, followed by some examples. Next, we discuss some limitations of the GE algorithms and recent development to overcome them. We end with a review of a few Hamiltonian REMs. e development of GE-enhanced sampling algorithms is a fast-evolving eld, and hence, it is not possible to exhaustively review all the variants of GE algorithms due to the limit of space in this chapter.
