often a layer of subjectivity over whether a region of conformational
space is a separate conformation or a part of a larger major confor-
mation. The current authors believe that it makes sense to under-
stand conformational space with protein function as the primary
consideration.1.3 Energy Surfaces
(Sensible Projections
of Conformational
Space)
The energy surface or landscape concept was first suggested to
understand protein folding [11]; it is now often used to understand
the space which a protein’s conformation moves within. While the
energy surface concept is useful for thinking about protein confor-
mational space and folding, it is generally not sensible to attempt to
build and visualize an energy surface using all the degrees of free-
dom of a protein, or even just torsions (seeNote 1).
Many state-of-the-art techniques for studying protein dynam-
ics use free-energy surfaces (FES) as the primary means of calculat-
ing and visualizing a protein’s conformational space. Hence, we
have to try and understand protein conformational space using only
a small number of dimensions, and choosing these dimensions (for
an energy surface) is difficult but very important (seeNote 2).1.4 Kinetic
Transitions
Another sensible way to describe and analyze the motions of a
target protein system is through the kinetics of conformational
transitions. This requires some definition of conformational states
and in order that the transition probability can be converged
between them they should be in a relatively fine discretization.
The analysis of time (kinetics) with respect to protein motions,
rather than potential energy, is easier to work with and understand,
mainly because it is easier to coarse-grain and analyze the temporal
hierarchy. A kinetic network model is a sensible way to understand
and visualize an average protein drug target.2 Classification of Computational Models of Protein Transition
The field of modeling protein dynamics has grown along with the
development of the two most common approaches, molecular
dynamics (MD) and Metropolis Monte-Carlo (MMC) simulations.
MD simulates the dynamics of biomolecules using Newtonian
and statistical physics. MD simulations generally start with an
experimentally determined structure of a protein that is placed in
a box of water molecules. Each atom is assigned a velocity in
accordance with the Boltzmann distribution and small time steps
are made which modify atom positions using an integration
scheme. Time steps must be very small to ensure that the conserva-
tion of energy law is adhered to. Hence, the main drawback;
standard MD simulations can be computationally expensive and
are currently limited to timescales of single digit milliseconds [12]
(seeNote 3).Computational Study of Protein Conformational Transitions 341