4.3.2 Utilizing
Information from Related
Proteins Families
Another important case is protein conformations which represent
an inactive form. Examples can be found in the kinase and GPCR
target classes. In these cases there may be structural data from
related proteins allowing the generation of homology models.
Additionally with large sets of related sequence data, evolutionary
analysis can suggest contacts which may be important in conforma-
tional dynamics [119–121]. These coevolution contacts could be
used in collective variables to drive molecular dynamics approaches
such as metadynamics or adaptive sampling, although there are no
successful studies of this so far. An example of coevolution contacts
alone being used to predict large open to closed functional motion
in the case of the glutamate receptor has been demonstrated by
Marcos et al. [122]. In another interesting example, coevolution
contacts were used to filter out important conformations proposed
by a very coarse and rapid protein sampling method called discrete
molecular dynamics [123]. Platforms like MELD have been used to
bias all potential contacts in a conformational search [124], which
might allow for easy searching based on coevolution contacts.4.3.3 Working with No
Knowledge of a Hidden
Conformation
A small number of MD based approaches offer general (i.e., no
knowledge of the hidden conformation required), transferable bias-
ing potentials for larger systems. One potential approach is
accelerated-MD (aMD) which adds a boost potential to all torsions
and/or the potential energy, flattening the free-energy surface and
making transitions more likely. Canonical ensemble averages and
free-energies can be recovered by reweighting a process made easier
with the more recent variant Gaussian-aMD (GaMD) [125]. aMD
has been used to find transitions between known active and inactive
conformations of Ras [126] and to find hidden conformations of
maltose binding protein [127]. These studies have used a fraction
of the simulation time of the DESRES studies discussed above, but
aMD and GaMD are much more difficult to interpret even if a
reweighted free-energy surface can be recovered.
Replica-exchange based approaches have been developing over
many years and have been demonstrated in a great many forms,
with generalized ensembles over temperature, biasing potentials,
interaction scaling, etc. These methods have recently been used to
improve orthogonal sampling for simulations where some specific
variables are being biased.
There are examples of testing combinations of Hamiltonian
exchange with aMD to evaluating its sampling power [128].
Replica-exchange methods were also combined with NMA to
raise the temperature of specific modes, which were found by
NMA without any prior knowledge [129, 130]. There is a recent
example where a variation of replica-exchange called selective
integrated tempering sampling (SITS) were applied to analyze
global conformational transitions of protein kinases [131]
(seeNote 18).Computational Study of Protein Conformational Transitions 351