Molecular Simulation and Systems Biology - Systems Biology: Networks, Models, and Applications - page 88

a fixed charge water model that is hoped to be superior for simula-
tions that employ Ewald techniques for the treatment of long-ranged
electrostatic interactions in systems studied with periodic boundary
conditions. This model was developed by finding the force field
parameters that give the best agreement with experimental data for
the temperature dependence of the density and heat of vaporization
of pure water. These observables are very sensitive to the strength of
the water-water intermolecular hydrogen bond. Clearly it is important
to have an accurate representation of hydrogen bonding, since it is
so important to the structure and stability of biological molecules.
Notably, even though it was not used to determine the parameters
of the water-water interaction, the temperature dependence of the
self-diffusion coefficient predicted by the model agrees remarkably
well with experimental measurements (figure 3.11). Other water
models have had difficulty reproducing this very important kinetic
observable.
Force fields for biomolecular simulations have also been carefully
assessed with respect to their ability to predict hydration free energies.
Hydration free energies are a rather sensitive measure of the strength
of interaction between a solute and water. These free energies can be
accurately measured by a variety of experimental techniques. The
various amino acid side chains exhibit very different hydration free
energies and these differences are a key factor in determining the
folding rates and stabilities of proteins. In a folded protein, hydrophilic
amino acid side chains are mostly in contact with water, whereas
hydrophobic amino acid side chains are mostly in contact with other
hydrophobic groups.
Recent work by Shirts et al. [66] demonstrated the ability of classical
models to reproduce experimentally determined hydration free ener-
gies. The computation of hydration free energies by simulation is a
computationally demanding task. Simulations that are long enough
to produce adequately precise measurements of the hydration free
energy for the comparison and assessment of force field accuracy
required the Folding@Home [67] distributed cluster. This novel com-
putational system makes use of tens of thousands (now over 100,000)
of contributed computers that run a specially developed screen
saver that is capable of performing simulations on small peptides and
biomolecules. The hydration free energy results (figure 3.12) are illu-
minating in that, although they are not accurate in an absolute sense,
most force fields do very well at reproducing the relative solvation
free energies of the various amino acid side-chain analogs. This is
remarkable because this observable was not used to develop the force
fields in the first place. Although improvements are possible, the
force fields are probably adequate for many purposes related to
structure prediction and, perhaps, also ligand binding.