Molecular Simulation and Systems Biology - Systems Biology: Networks, Models, and Applications - page 85

RECENT ADVANCES IN MOLECULAR SIMULATION
Methodological Developments
There are always two major areas of methodological development in
molecular simulation: improvements in the representation of the system
(force field improvements) and improvements in how efficiently a
simulation can visit all of the important conformations of a system
(sampling improvements).
In terms of force field improvements, the next generation of bio-
molecular force fields moves beyond fixed charge models of
electrostatic interactions to ones that include polarizability. In a polar-
izable force field, the electrostatic interactions between atoms are
no longer described by a simple charge-charge interaction but rather
by one that can include both fixed and polarizable charge-charge,
charge-dipole, dipole-dipole, and higher multipolar interactions.
Recent polarizable force fields for biomolecules include variants of
the AMBER [49], CHARMm [50,51], and OPLS-AA [52] potentials
as well as the newly developed AMOEBA [53,54] model. It remains
to be seen whether the inclusion of polarizability will significantly
improve the predictive abilities of these models. Another notable
improvement in force fields is a movement away from a Fourier series
representation of the dihedral angle term toward a spline-based model
which offers more flexibility in fitting to the true quantum mechanical
potential [55].
The most significant recent development in sampling techniques has
been the wide adoption of enhanced sampling methods, including
multicanonical sampling and replica exchange simulation. Multicanonical
sampling corresponds to simulating the system with an effective
potential that generates a random walk in energy space, allowing a
broad sampling of molecular conformations [56]. The drawback of
multicanonical sampling is that the appropriate effective potential
needs to be determined by trial and error. Replica exchange simulation
allows a similar broad sampling of energies and conformations, but
achieves it by carrying out a series of coupled simulations at different
temperatures. Each simulation has one copy (or replica) of the simu-
lated system, and periodically the replicas at adjacent temperatures
are exchanged with a Metropolis Monte Carlo-like acceptance rule [57].
A schematic of the replica exchange method is shown in figure 3.9.
A common weakness of all enhanced sampling methods is that their
computational cost increases significantly with system size—a method
that works for alanine dipeptide in water may be impractical for
simulating a solvated protein.
Another recent development relevant to systems biology has
been the widespread use of approximate methods that estimate free
energies by postprocessing simulation data. In these approaches, an