Once the collective variables of interest are identified, multiple
metadynamics-based solutions can be explored. Indeed, the methodological metadynamics-oriented research evolved so much in the
last 10 years and with so many fruitful contributions that it could be
now seen as a field on its own. The one here provided is just one
example to give a first guidance and also suitable if the HPC
computational resources available are rather limited or to run
with GPU support.
The approach is the metadynamics coupled with parallel tempering (PT, i.e., a sort of replica exchange in the temperature space)
[123] in the well-tempered ensemble (WTE) [124] to overcome
the usage of many close temperatures in the PT simulations when
systems with a large number of atoms need to be used. Indeed, it is
critical in a PT-metadynamics to achieve a sufficient energy overlap
between adjacent replicas and sufficient exchange rates, which need
to be carefully checked on short exploratory runs before even
moving further in the sampling or analyses of the results. In the
WTE approach, a constant bias on the energy is added to each
replica to increase the width of the energy distribution so that a
suitable exchange rate is ensured even when a lower number of
replicas are used and the separation in the temperature space is
larger.
To give a practical example, in the case of p53 DBD, after the
classical MD exploratory analyses a working hypothesis that can be
generated is that upon DNA binding or phosphorylation at Ser215
the conformational state of a distal loop, i.e., the S6-S7 loop
(residues 207-213), can be conformationally modulated
[14]. After exploring different reaction coordinates in the available
unbiased MD runs, we concluded that the combination of at least
four Cα-Cα key distances between residues of the S6-S7 loop and
its surroundings is descriptive of the conformations that the loop
assumes: Asp208-Arg156 (CV 1), Arg158-Phe212 (CV 2),
Arg209-Glu221 (CV 3), and Arg209-Glu258 (CV 4). In contrast,
CVs such as the radius of gyration are not of interest for this specific
process since there is no remarkable change in the shape of the
molecule upon opening and closing of the loop. Such as CV could
become relevant for larger conformational rearrangements when
more extended and disordered loops or entire domains change
their reciprocal orientation.
Once the CVs are selected, we also need a proper definition of
the temperatures for each replica, which should reasonably span
from low to high temperatures but where the highest temperature
should not encounter the risk to unfold our protein in the simulation time needed to reach convergence. Indeed, we want to simulate a conformational change occurring in a folded protein and not
its unfolding/folding mechanism. We can thus run unbiased simulations at high temperatures of at least some hundreds of ns to
monitor the stability of the protein architecture and identify the
236
Elena Papaleo
metadynamics-based solutions can be explored. Indeed, the methodological metadynamics-oriented research evolved so much in the
last 10 years and with so many fruitful contributions that it could be
now seen as a field on its own. The one here provided is just one
example to give a first guidance and also suitable if the HPC
computational resources available are rather limited or to run
with GPU support.
The approach is the metadynamics coupled with parallel tempering (PT, i.e., a sort of replica exchange in the temperature space)
[123] in the well-tempered ensemble (WTE) [124] to overcome
the usage of many close temperatures in the PT simulations when
systems with a large number of atoms need to be used. Indeed, it is
critical in a PT-metadynamics to achieve a sufficient energy overlap
between adjacent replicas and sufficient exchange rates, which need
to be carefully checked on short exploratory runs before even
moving further in the sampling or analyses of the results. In the
WTE approach, a constant bias on the energy is added to each
replica to increase the width of the energy distribution so that a
suitable exchange rate is ensured even when a lower number of
replicas are used and the separation in the temperature space is
larger.
To give a practical example, in the case of p53 DBD, after the
classical MD exploratory analyses a working hypothesis that can be
generated is that upon DNA binding or phosphorylation at Ser215
the conformational state of a distal loop, i.e., the S6-S7 loop
(residues 207-213), can be conformationally modulated
[14]. After exploring different reaction coordinates in the available
unbiased MD runs, we concluded that the combination of at least
four Cα-Cα key distances between residues of the S6-S7 loop and
its surroundings is descriptive of the conformations that the loop
assumes: Asp208-Arg156 (CV 1), Arg158-Phe212 (CV 2),
Arg209-Glu221 (CV 3), and Arg209-Glu258 (CV 4). In contrast,
CVs such as the radius of gyration are not of interest for this specific
process since there is no remarkable change in the shape of the
molecule upon opening and closing of the loop. Such as CV could
become relevant for larger conformational rearrangements when
more extended and disordered loops or entire domains change
their reciprocal orientation.
Once the CVs are selected, we also need a proper definition of
the temperatures for each replica, which should reasonably span
from low to high temperatures but where the highest temperature
should not encounter the risk to unfold our protein in the simulation time needed to reach convergence. Indeed, we want to simulate a conformational change occurring in a folded protein and not
its unfolding/folding mechanism. We can thus run unbiased simulations at high temperatures of at least some hundreds of ns to
monitor the stability of the protein architecture and identify the
236
Elena Papaleo