during the simulation through classical analyzes such as the
root mean square deviation, the root mean square fluctuation,
and radius of gyration. Moreover, it is also important to analyze
the bias per se over the simulation. For example, for the meth-
ods that add a potential energy to the original potential energy,
an important point is to analyze the evolution of the added
restraint potential energy over time and to compare the values
with respect to the literature. With a poorly parameterized bias,
the simulation can produce improbable processes.
- In the current aMD implementation, the modified potential is
added to all the system, with no possibility to be limited to
some dihedrals.
- It is recommended to use the dual boost if the solvent is
described explicitly and only on the torsion terms in the case
of implicit solvent.
8.αcannot be set to zero. Whenα¼0, the modified potential is
flat and the system experiences a random walk.
- As it was observed in previous studies of aMD, the value of 0.2
forαD and 0.16 forαP seems to work well for proteins.
Following the suggestions, for lower boost, value between
0.15 and 0.19 can be used instead of 0.2 forαD.
- Instead of considering the protein backbone of the trajectory,
alpha carbons or atoms of the binding site constitute an alter-
native that can be also used for PCA depending on the studied
problem.
- Many statistical packages provide PCA. Bio3D package in R
has been employed here for its practicality since it supports
both the PCA and diverse relevant graphic plot for PCA result
analysis. The cpptraj module supports the PCA but supposes to
make the graphic plots with another software. An example of
cpptraj script is given in the reweighting section of the support-
ing information.
- In the literature, several criteria have been developed to deter-
mine the number of principal components to be analyzed.
Experience shows that 3–5 dimensions are often sufficient to
capture over 60–70% of the total variance; but it is not always
the case, especially when the two first components do not
express a high percentage of variance, which leads to consider
a relatively important number of components. The Cattell
scree criterion involves plotting the scree plot as in Fig.2 and
considers the number of meaningful principal components as
to be the inflection point between the steep slope and a leveling
off. In our case, this number is 3. It is recommended to select
the number of component to be analyzed. The famous Kaiser
criterion, which considers all principal components having an
eigenvalue larger than 1, is not applicable when using a
Enhanced Molecular Dynamics 423
