AW: [gmx-users] Re: Conceptual question: Why are MD production simulations usually run in the NVT ensemble?

[Florian Dommert]

 -Ursprüngliche Nachricht-
 Von: gmx-users-boun...@gromacs.org [mailto:gmx-users-
 boun...@gromacs.org] Im Auftrag von Andrew DeYoung
 Gesendet: Samstag, 14. Juli 2012 18:56
 An: gmx-users@gromacs.org
 Betreff: [gmx-users] Re: Conceptual question: Why are MD production
 simulations usually run in the NVT ensemble?
 
 Hi Justin,
 
 Thanks so much for your time.
 
 You make a very good point about NPT.  My field (although I am only a
second-
 year graduate student) is physical chemistry, and I am ignorant,
unfortunately,
 of biochemical literature.
 
 In physical chemistry, it seems that people almost always use NVT for
 production simulations, but I am not sure why, because just as in
biochemical lab
 experiments, I would expect NPT to be more realistic -- chemicals (whether
it is
 a pure liquid, like in my field, or a solvated biomolecule, like in
your field) are
 usually exposed to atmospheric pressure in experiments.  In experiments,
the
 volume is usually not fixed, I think.  So it makes sense you say that
biochemical
 simulations are typically performed in NPT.  I don't know why physical
chemistry
 experiments on, and simulations of, liquids would be any different.


Hi,

 I completely understand your question, because it also came up to me. As
Justin mentioned, NpT simulations are the common situation for biophysical
processes in cells, because it is their native environment. The minimum of
the Gibb's free energy G=H+pV determines the system state. Assuming you are
performing an equilibrium MD, G is a constant of motion. Now the task of the
integrator including the thermostat and barostat is to keep G constant. If
you perform an NVT simulation you have no control over G but only the
Helmholtz free energy H=U-TS. Hence if you are interested in a certain state
of the system like folding of a protein or free energies, which depend on G
you have to simulate the corresponding ensemble. If you want to study
thermodynamic properties such as density, conductivity, diffusion, the
choice of the ensemble depends on other reasons. If you want to study the
density, an NVT simulation is not useful, because you set the density as
input parameter. However, if you are able to extract the desired data from
an NVT ensemble, it is preferable to an NpT simulation, because you do not
have to assure that the barostat couples correctly and the sampling is
sufficient to have no bias from the barostat. This is especially important
if you are studying new compounds, where the system is not well study such
that a proper adjustment of the barostat parameters is difficult. Sometimes
it is also difficult to sample canonically in an NpT ensemble, if the
simulation time is short, the system is small or the virial theorems are not
sufficiently fulfilled for other reasons. The common protocol to equilibrate
in NpT and perform a production run in NVT is finally related to Justin's
comment regarding the parameterization. An equilibration of the system can
only be achieved, if the system is allowed to get into the corresponding
state given by the force field. If you would have a perfect force field,
you could also skip the NpT equilibration, because with the knowledge of the
experimental density you could setup your system in the correct state and
sample in an NVT ensemble. At the end in the NVT production run you only
have to assure that you sample long enough to obtain a reliable average of
dynamic properties, like diffusion or conductivity, such that they are not
biased by the thermostat.

This is my opinion to this matter and I hope it helps, but please correct
my, if I am wrong. 

/Flo
 
 I guess then this could be an issue of the field?  Are there chemists or
physicists
 reading this that can please provide advice?
 
 Best wishes,
 
 Andrew DeYoung
 Carnegie Mellon University
 
 --
 gmx-users mailing listgmx-users@gromacs.org
 http://lists.gromacs.org/mailman/listinfo/gmx-users
 * Only plain text messages are allowed!
 * Please search the archive at
 http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
 * Please don't post (un)subscribe requests to the list. Use the www
interface or
 send it to gmx-users-requ...@gromacs.org.
 * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

--
gmx-users mailing listgmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
* Only plain text messages are allowed!
* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
* Please don't post (un)subscribe requests to the list. Use the
www interface or send it to gmx-users-requ...@gromacs.org.
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

[gmx-users] Re: Conceptual question: Why are MD production simulations usually run in the NVT ensemble?

[Andrew DeYoung]

Hi,

Thanks for your time.  Based on Justin's and Florian's responses, it seems
that the choice of whether to do production simulations in NpT or NVT
depends on what you're studying.  Does the following seem like a reasonable
summary of the discussion:

-- In biophysics, lab experiments are almost always performed at constant
pressure (i.e., atmospheric pressure), so NpT is the natural choice.
Moreover, in biophysics, people are often interested in particular system
states, like the folding of a protein.  These states depend on free
energies, which in turn depend on the Gibbs free energy G.  This
necessitates the use of the NpT ensemble because it is the only ensemble (I
guess?) that allows control over G.

-- In physical chemistry studies of liquids, I speculate that lab
experiments are still almost always performed at constant pressure.
However, in these types of studies, people are less frequently interested in
particular system states; hence calculation of free energies are typically
not necessary.  But just because calculation of free energies is not
necessary does not mean that NpT cannot be used.  Nevertheless, NVT is
preferable in that it avoids the use of a barostat, which can give rise to
artifacts and whose parameters are difficult to set for new compounds.

A fellow student put forth one more possible distinction between biophysics
and physical chemistry.  He pointed out that in biophysics, one often
studies macromolecules like proteins, which undergo significant structural
changes and hence undergo large changes in volume during the simulation;
hence NVT should not be used.  In contrast, in physical chemistry, one is
more often interested in small molecules.  Small molecules do not undergo
large structural changes, so the system volume is expected to vary very
little during the simulation; hence NVT is permissible.  Also, NVT is
_perhaps_ slightly less computationally expensive than NpT, so NVT should be
used when it is permissible (i.e., when folding does not occur and when
calculation of free energies is not needed).

Do you agree with this reasoning?  Do you think that protein folding leads
to large changes in system volume, whereas in liquids the system volume is
not expected to change much?  

Thanks so much!

Andrew DeYoung
Carnegie Mellon University

-- 
gmx-users mailing listgmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
* Only plain text messages are allowed!
* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
* Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-users-requ...@gromacs.org.
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists

[gmx-users] Re: Conceptual question: Why are MD production simulations usually run in the NVT ensemble?

[Andrew DeYoung]

Hi Justin,

Thanks so much for your time.  

You make a very good point about NPT.  My field (although I am only a
second-year graduate student) is physical chemistry, and I am ignorant,
unfortunately, of biochemical literature.  

In physical chemistry, it seems that people almost always use NVT for
production simulations, but I am not sure why, because just as in
biochemical lab experiments, I would expect NPT to be more realistic --
chemicals (whether it is a pure liquid, like in my field, or a solvated
biomolecule, like in your field) are usually exposed to atmospheric
pressure in experiments.  In experiments, the volume is usually not fixed, I
think.  So it makes sense you say that biochemical simulations are typically
performed in NPT.  I don't know why physical chemistry experiments on, and
simulations of, liquids would be any different.

I guess then this could be an issue of the field?  Are there chemists or
physicists reading this that can please provide advice?

Best wishes,

Andrew DeYoung
Carnegie Mellon University

-- 
gmx-users mailing listgmx-users@gromacs.org
http://lists.gromacs.org/mailman/listinfo/gmx-users
* Only plain text messages are allowed!
* Please search the archive at 
http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
* Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-users-requ...@gromacs.org.
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists