Molecular dynamics simulations implicit solvation model

Schaefer, M., Bartels, C., and Karplus, M. (1998). Solution conformations and thermodynamics of structured peptides Molecular dynamics simulation with an implicit solvation model./. Mol. Biol. 284, 835-848. 

Over the recent years implicit solvent models have undergone a transition to relatively mature methodology that is now widely employed in molecular dynamics simulations and related applications. Most popular are implicit solvent models based on a decomposition of the solvation free energy into electrostatic and nonpolar components. The electrostatic free energy is typically obtained according to a continuum electrostatics model that is described by Poisson theory or by the more approximate but much more efficient Generalized Born formalism. 

Likewise, Monte Carlo (MC) methods can tolerate implicit solvent models better than can molecular dynamics methods, where such forces are important. Sharpes has discussed a variety of implicit solvation models available for molecular simulation of macromolecules. An introduction to solvation models in general has been reviewed in this book series. ... 

An accurate description of the aqueous environment is essential for atom-level biomolec-ular simulations, but may become very expensive computationally. An imphcit model replaces the discrete water molecules by an infinite continuum medium with some of the dielectric and hydrophobic properties of water. The continuum implicit solvent models have several advantages over the explicit water representation, especially in molecular dynamics simulations (e.g., they are often less expensive, and generally scale better on parallel machines they correspond to instantaneous solvent dielectric response the continuum model corresponds to solvation in an infinite volume of solvent, there are no artifacts of periodic boundary conditions estimating free energies of solvated structares is much more straightforward than with explicit water models). Despite the fact that the methodology represents an approximation at a fundamental level, it has in many cases been successful in calculating various macromolecular properties (Case et al. 2005). 

In contrast, there will be many cases where continuum solvent models are less useful. These include situations where one of the goals of the simulation is to obtain a detailed picture of solvent structure, or where there is evidence that a particular structural feature of the solvent is playing a key role (for example, a specific water-macromolecule hydrogen bond). In these situations, however, explicit representation of some water combined with implicit solvation may suffice. Another example is when molecular dynamics simulations are used to study kinetic, or time-dependent phenomena. The absence of the frictional effects of solvent will lead to overestimation of rates. In addition, more subtle time-dependent effects arising from the solvent will be missing from continuum models. Continuum solvent models are in effect frilly adiabatic, in the sense that for any instantaneous macromolecular conformation, the solvent is taken to be completely relaxed. For electrostatic effects, this implies instantaneous dielectric and ionic double layer relaxation rates, and for the hydrophobic effect, instantaneous structural rearrangement. An exception would be dielectric models that involve a frequency-dependent dielectric. Nevertheless, continuum solvent models should be used with caution in studying the time dependence of macromolecular processes. 

Presently, only the molecular dynamics approach suffers from a computational bottleneck [58-60]. This stems from the inclusion of thousands of solvent molecules in simulation. By using implicit solvation potentials, in which solvent degrees of freedom are averaged out, the computational problem is eliminated. It is presently an open question whether a potential without explicit solvent can approximate the true potential sufficiently well to qualify as a sound protein folding theory [61]. A toy model study claims that it cannot [62], but like many other negative results, it is of relatively little use as it is based on numerous assumptions, none of which are true in all-atom representations. 

Going beyond the mean-field level, several "hybrid" approaches are now being explored in MD simulations. Examples include a recent model [70] in which the immediate hydration of the solute is modeled explicitly by a layer of water molecules, and the GB model is used to treat the bulk continuum solvent outside the explicit simulation volume. A similar idea was recently found very effective in the context of replica-exchange simulations [71]. An explicit ion/implicit water (PB) solvation model for molecular dynamics of nucleic acids has recently been tested [72]. 

Semiclassical solvation models depend on parameters. Implicit models mostly depend on how the cavity that contains the solute is built, and large parameterization is required to handle a variety of different solvents. In this respect, the work of Marenich et al. for the polarizable continuum model (PCM)" has gone a long way in this direction, in the opinion of this author. Explicit models also depend on parameters that enter the definition of the individual model. However, the quality of the results obtained when employing explicit solvation models with QM methods also depends on the quality of the sampling of the solvent configuration space. This is usually accomplished with a molecular dynamics (MD) simulation, which does not need to use the same polarizable solvation model. The quality of the MD simulation will influence the reliability of the following QM results. 

---