In the early molecular models, water was not included, but was later determined to be essential to carbohydrate structure and behavior. The solvent model types are classified as explicit solvent, in which solvent molecules are explicitly modeled using the same model type as the solute, or implicit, in which the effect of solvent is modeled by a functional form and is a function of solute configuration. Water is the most common solvent, though other solvents and mixed solvents are also used, especially to reproduce experimental solvent environments. There are several explicit water models, the most important ones being TIP3P, SPC, TIP4P, and TIP5P (41).

The implicit solvent models are based on the assumption that the major effect of solva­tion is encapsulated in its dielectric properties. In a simple protein of 2400 atoms, solvating with explicit water molecules for a cube with at least three water layers on each side in­creases the system size to 23 500 atoms. The simplest, and crudest, model simply uses a distance-dependent dielectric constant in the electrostatic term of Equation (8.1), so that the effect of the solvent is to mask the charge interaction between distant charges, assumed to have a dielectric medium between them, and to not mask at all when two charges are close to each other. The advantage of this method is that no explicit water molecules are included in the simulation and the cost of a distance-dependent dielectric constant is mini­mal, cutting the computational demand by a factor of ten or more. The drawback is that one does not model the solvation free energy correctly nor the dielectric environment inside a macromolecule.

The more sophisticated methods of modeling solvent implicitly are based on solving the Poisson equation. The most rigorous methods involve solving the partial differential equa­tions for the electrostatic potential on a grid, and are quite computationally intensive. These methods, commonly called Poisson-Boltzmann (PB) solvers (42, 43), are useful in accurate examination of electrostatic potentials around static macromolecules and are not often used for dynamics. Even though there are no explicit water molecules in a PB calculation, the computation of the electrostatic potential at each dynamics step is too costly to offset the savings. The Generalized Born approximation is used to provide a much more efficient method for parameterizing the Poisson equation which is very close to the rigorous solu­tion and provides reasonable solvation energies and other thermodynamic solvation effects (44-46). The detailed interactions with individual water molecules are missing as are the hydrodynamic effects, but for many modeling problems, a solvated simulation that repro­duces the ensemble averages of an explicitly solvated system can be performed at one fourth the computational cost. A second benefit of implicit solvent calculations is that the solvent response to solute changes is instantaneous at each step, rather than requiring many picosec­onds (thousands of steps) of equilibration of the thousands of individual water molecules in an explicit-water simulation. This solvation model is also very useful in preparing a system for fully solvated modeling and for finding probable mechanisms and structures for more detailed studies.