ABSTRACT
Atomistic molecular dynamics (MD) simulations [Allen and Tildesley 1987; Frenkel and Smit 2002] remain one of the most powerful tools for computationally modeling biomolecular and condensed phase systems. By providing at each timestep in the simulation the coordinates, velocities, and forces for each atom within a system, atomistic MD simulations allow the direct investigation of molecular structure and dynamics with Angstrom-level detail and femtosecond resolution. Although conventional MD methods cannot readily describe phenomena such as bond cleavage that require explicit treatment of quantum mechanical effects [although see, e.g., Hammes-Schiffer 2006; Voth 2006; Warshel 2002], classical atomistic force elds have been carefully parameterized to accurately reproduce many experimental observables [Gnanakaran et al. 2003; Jorgensen, Maxwell, and Tirado-Rives 1996; Lindahl and Edholm 2001; MacKerell et al. 1998]. With available computational resources [Mervis 2001; Reed 2003] and software [Brooks et al. 1983; Case et al. 2005; Phillips et al. 2005; Smith and Forester 1996; Van der Spoel et al. 2005], classical MD simulations routinely model the equilibrium uctuations of biomolecular or condensed phase systems over length scales of tens of nanometers for time scales on the order of 100 ns. Moreover, in some
exceptional cases, MD simulations have modeled processes such as protein uctuations [Freddolino et al. 2006; Sanbonmatsu, Joseph, and Tung 2005; van Gunsteren et al. 2001] or equilibrium bilayer undulations [Blood and Voth 2006; Lindahl and Edholm 2000], on somewhat longer length and time scales. Consequently it is indisputable that classical atomically detailed MD simulations have tremendously impacted our present understanding of physical chemistry, structural biology and biochemistry [Karplus and McCammon 2002], and materials science.
