Computational Approaches in Peptide and Protein Design: An Overview - Peptide and Protein Design for Biopharmaceutical Applications - page 24

increase the efficiency of sampling of conformational space, one can ele-
vate the temperature for a short period and then resume the simulation at
the desired temperature, allowing time for reequilibration. This technique,
called simulated annealing , is useful for overcoming activation energy
barriers between local minima. Another approach generates several ensem-
bles (in the Monte Carlo approach) or several trajectories (in molecular
dynamics) in parallel at different temperatures with periodic exchange
between ensembles or trajectories (replicas) [56]. This method, called
replica exchange , rather efficiently avoids situations where Monte Carlo
or molecular dynamics runs are trapped in local minima. Replica exchange
was applied to flexible peptides, such as enkephalin [56], but also for
simulation of protein folding in the case of the small model proteins [57].
Generally, both the Monte Carlo and the molecular dynamics
approaches have been widely used in many computational studies of
peptides (see, for example, our earlier reviews [3,4]) and proteins (e.g. a
recent review [26]). As an example, one can mention the Folding@home
approach; the application of molecular dynamics sampling to small proteins
[58,59]. This tool specifically aims to reproduce the dynamical process of
protein folding from any extended 3D structure to the native fold. Current
conformational resources and molecular dynamics algorithms are not suffi-
cient to generate trajectories long enough to observe this kind of folding
transition. However, even in very short trajectories there is a very small
chance to generate a native-like structure as a result of stochastic crossing of
the corresponding barriers. If a very large number of short molecular
dynamics trajectories are generated independently and their results are
pooled, the number of observed near-native structures may be large enough
to be interpreted as an ensemble of structures in the nearest vicinity of the
native one. The Folding@home method facilitates running hundreds of
thousands of independent short trajectories of the same protein on a world-
wide distributed grid of computers and then pools the results. In this way, it
was possible to fold, for instance, a 36-residue protein from the villin
headpiece with a root mean square (RMS) distance between atoms of the
calculated and experimental structures of 1.7 ˚ [60]. It is noteworthy,
however, that the combination of molecular dynamics trajectories collected
by Folding@home can be regarded as data on conformational sampling
rather than a 'real' trajectory of protein folding.
Sampling of ligand orientations within the receptor In fact, almost all
methods of sampling the conformational state of peptides, such as sys-
tematic search, Monte Carlo and molecular dynamics are also applicable
to the sampling of possible orientations (poses, binding modes) of a small