@@ -106,7 +106,7 @@ To run a simulation with gromacs+plumed you just need to add a -plumed flag
...
@@ -106,7 +106,7 @@ To run a simulation with gromacs+plumed you just need to add a -plumed flag
Here plumed.dat is the name of the plumed input file. Notice that PLUMED will write
Here plumed.dat is the name of the plumed input file. Notice that PLUMED will write
information in the md.log that could be useful to verify if the simulation has been set up properly.
information in the md.log that could be useful to verify if the simulation has been set up properly.
\subsubsection munster-exercise-1 Exercise 1
\subsubsection munster-exercise-0 Exercise 0
In this exercise, we will run a plain molecular dynamics simulation and monitor the \f$\Phi\f$ and \f$\Psi\f$ dihedral angles
In this exercise, we will run a plain molecular dynamics simulation and monitor the \f$\Phi\f$ and \f$\Psi\f$ dihedral angles
on the fly.
on the fly.
...
@@ -398,7 +398,10 @@ you can bias them using metadynamics. In this way, a time dependent,
...
@@ -398,7 +398,10 @@ you can bias them using metadynamics. In this way, a time dependent,
adaptive potential will be constructed that tends to disfavor visited configurations
adaptive potential will be constructed that tends to disfavor visited configurations
in the collective-variable space. The bias is usually built as a sum of
in the collective-variable space. The bias is usually built as a sum of
Gaussian deposited in the already visited states.
Gaussian deposited in the already visited states.
A sample input follows
\subsubsection munster-exercise-1 Exercise 1
Now run a metadynamics simulation with the following input
\verbatim
\verbatim
phi: TORSION ATOMS=5,7,9,15
phi: TORSION ATOMS=5,7,9,15
psi: TORSION ATOMS=7,9,15,17
psi: TORSION ATOMS=7,9,15,17
...
@@ -416,13 +419,6 @@ keywords sets the range of the collective variables and tell PLUMED to keep the
...
@@ -416,13 +419,6 @@ keywords sets the range of the collective variables and tell PLUMED to keep the
potential stored on a grid. This affects speed but, in principle, not the accuracy of the calculation.
potential stored on a grid. This affects speed but, in principle, not the accuracy of the calculation.
You can try to remove those keywords and see the difference.
You can try to remove those keywords and see the difference.
If you run a simulation with this input, PLUMED will produce an additional
`HILLS` file, containing as list of the deposited Gaussians. Notice that
the apparent height written in the `HILLS` file is not the same height that you
specified with the `HEIGHT` keyword, but is scaled by a factor \f$\gamma/(\gamma-1)\f$,
where \f$\gamma\f$ is the biasfactor. This is a technical issue and is related to the
fact that well-tempered metadynamics is only partially compensating the free-energy landscape.
Now, run a metadynamics simulations and check the explored collective variable space.
Now, run a metadynamics simulations and check the explored collective variable space.
Results from a 200ps (100000 steps) trajectory in
Results from a 200ps (100000 steps) trajectory in
vacuum are shown in Figure \ref munster-ala-traj-metad.
vacuum are shown in Figure \ref munster-ala-traj-metad.
...
@@ -430,22 +426,12 @@ vacuum are shown in Figure \ref munster-ala-traj-metad.
...
@@ -430,22 +426,12 @@ vacuum are shown in Figure \ref munster-ala-traj-metad.
\anchor munster-ala-traj-metad
\anchor munster-ala-traj-metad
\image html munster-ala-traj-metad.png "(phi,psi) scatter plot obtained with metadynamics and two different values of the biasfactor. Simulation performed in vacuum."
\image html munster-ala-traj-metad.png "(phi,psi) scatter plot obtained with metadynamics and two different values of the biasfactor. Simulation performed in vacuum."
As you can see, exploration is greatly enhanced.
As you can see, exploration is greatly enhanced.
Notice that the explored ensemble can be tuned using the biasfactor \f$\gamma\f$.
Notice that the explored ensemble can be tuned using the biasfactor \f$\gamma\f$.
Larger \f$\gamma\f$ implies that the system will explore states with higher free energy.
Larger \f$\gamma\f$ implies that the system will explore states with higher free energy.
As a rule of thumb, if you expect a barrier of the order of \f$\Delta G^*\f$,
As a rule of thumb, if you expect a barrier of the order of \f$\Delta G^*\f$,
a reasonable choice for the biasfactor is \f$\gamma\approx\frac{\Delta G}{2k_BT}\f$.
a reasonable choice for the biasfactor is \f$\gamma\approx\frac{\Delta G}{2k_BT}\f$.
One can then exploit the
fact that the deposited bias potential has a relationship with the underlying free-energy landscape
and compute the latter. To this aim, you can use the command line sum_hills tool:
\verbatim
> plumed sum_hills --hills HILLS
\endverbatim
(see \ref sum_hills).
Finally, notice that \ref METAD potential depends on the previously visited trajectories.
Finally, notice that \ref METAD potential depends on the previously visited trajectories.
As such, when you restart a previous simulation, it should read the previously deposited
As such, when you restart a previous simulation, it should read the previously deposited
HILLS file. This is automatically triggered by the \ref RESTART keyword.
HILLS file. This is automatically triggered by the \ref RESTART keyword.
