Hi Marek,
So here are my 2 cents on the various ensembles one should use for heating,
equilibration, production etc. This of course comes without a warranty...
It is essential in any explicit solvent simulation to run constant pressure
at some point to equilibrate the density. Whether you continue running NPT
depends on whether you believe your system is likely to change 'shape'
during a simulation. One of the first things to note is that the combination
most people seem to be using on the list which is NPT (ntb=2,ntp=1 with
Langevin ntt=3) is about the slowest combination of settings you can use.
Hence things are always a trade off between getting more sampling or using
'supposedly' better settings. In my opinion one should typically run the
following and it is what I do for my simulations.
1) Heat system using NVT (ntb=1,ntp=0,ntt=3,gamma_ln=1.0)
It is critical to initial heat using NVT because at low temperatures the
calculation of pressure is inaccurate and can cause the barostat to
overcorrect leading to instabilities. I typically either heat to 100K or so
and then switch to NPT or I run a short (20ps or so) heating stage. The key
is not to run for too long initially in NVT (before NPT) or you will get
vacuum bubbles form in the solvent.
2) Equilibrate the system using NPT (ntb=2,ntp=1,ntt=3,gamma_ln=1.0)
Once the system is heated I then continue running Langevin dynamics at
constant pressure to allow the density to equilibrate. I typically run this
for around 1 to 5ns depending on system size, checking that the density
converges.
3) Production using NVT (ntb=1,ntp=0,ntt=1,tautp=10.0).
For production if I don't expect the system to change shape, e.g. unfold
and fold back up etc. then I run NVT but with a very weak Berendsen
thermostat. The Berendsen thermostat is not good for equilibrating the
system as it does nothing to ensure that the kinetic energy is distributed
in a Boltzmann distribution, however for an already equilibrated system the
thermostat correction on each step should be minor so this should not cause
an issue. I use a weak coupling (10ps) to approximate NVE. Setting this to
infinite would give you NVE. However, NVE can be tricky to get to work well
when running very long simulations. 2fs with shake is kind of bleeding edge
for NVE, you ideally have to tighten the shake tolerance the PME tolerances
and reduce the time step to get ideal energy conservation (although a good
test to be sure things are working properly). As such while NVT is more
expensive than NVE it is minor and when you add all the extra tolerances
needed for NVE, NVT ends up quicker. In my opinion this is a better option
than using langevin for the entire simulation as all of the issues with
simulation problems, NANs seen on the GPUs etc arise from running long ntt=3
simulations. I would even typically run NPT (if I thought my system would
change shape) using ntt=1 once the system is equilibrated.
I hope this helps.
All the best
Ross
> -----Original Message-----
> From: Marek Maly [mailto:marek.maly.ujep.cz]
> Sent: Thursday, March 17, 2011 6:25 AM
> To: AMBER Mailing List
> Subject: Re: [AMBER] NVT Vs NPT
>
> Hi Jason,
>
> which is your opinion on NVE ensamble ?
>
> I think that this question is actual especially in the connection with GPU
> calculations due to memory limitations especially in case of GTX .. cards.
>
> According to the information from this web page:
>
> http://ambermd.org/gpus/
>
> (section: "System Size Limits")
>
> NVE ensamble has the smallest memory requirements (and probably the
> simulation is fastest here as there is no
> time penalty for thermostat, barostat calculations ).
>
> There is just problem that this ensamble in general don't ensure for
> constant temperature neither for constant
> pressure although if the simulated system is pre-equilibrated using
> sufficient NPT (or NPT/NVT) run, probably
> changes on P,T during consequent NVE simulation might be so slow
> (depending on given solvent and it's concrete model implementation ),
> that for sufficient time originally adjusted values of P,T are within the
> acceptable tolerance ...
>
> I know that I may do some relevant tests by my self but I am just
> wondering if you or someone else already made
> such "benchmarks" for example for widely used TIP3P model of water ...
>
> Thanks in advance for comments,
>
> Bets wishes,
>
> Marek
>
>
>
>
>
>
>
> Dne Wed, 16 Mar 2011 17:49:39 +0100 Jason Swails
> <jason.swails.gmail.com>
> napsal/-a:
>
> > Hello,
> >
> > I'll share my opinion, which may differ from others. NPT is very useful
> > for
> > equilibrating density, and can relieve artifacts arising due to the fact
> > that the system may not have been optimally packed to begin with (which
> > is
> > the case for tleap-created systems). If you do not run with NPT at all
> > (and
> > just NVT), then you wind up bubbles appear as the water retracts into
its
> > *normal* density in the liquid phase.
> >
> > Thus, during equilibration, NPT is quite useful to maintain a constant
> > density and let the box reduce to the "proper" size for the number of
> > solvent molecules in your system. However, NPT is more expensive than
> > NVT
> > because the pressure scaling is not free. Therefore, many people use
NVT
> > after they've equilibrated at NPT. Solvent (i.e. water) is largely
> > incompressible, so NVT and NPT are, for all intents and purposes,
> > equivalent
> > (for an equilibrated system).
> >
> > Hope this helps,
> > Jason
> >
> > On Wed, Mar 16, 2011 at 12:31 AM, Mahmoud Soliman
> > <mahmoudelkot.gmail.com>wrote:
> >
> >>
> >> Dear guys,
> >> I am running MD simulation for an enzyme in a water box. Normally,
> >> during
> >> heating and equilibration of the system I impose some restraints on
> >> some
> >> residues and I use NVT during that. But in production, I do not put
> >> any
> >> restraints and I use NPT. Actually, I learn that from amber tutorials
> >> but
> >> I
> >> see some papers adopting NVT during the production phase... So can
> you
> >> guys
> >> provide me with your experience about merits/drawbacks of both
> >> protocols
> >> (NVT and NPT)...much appreciated!
> >> Best wishes
> >> Mahmoud
> >>
> >> --
> >>
> >> *************************************************
> >>
> >> Mahmoud E. Soliman
> >>
> >> Computational Chemistry & Modeling (PhD)
> >>
> >> Department of Chemistry
> >>
> >> University of Bath
> >>
> >> Bath
> >>
> >> BA2 7AY
> >>
> >> United Kingdom
> >>
> >> [1]http://people.bath.ac.uk/mess20/
> >>
> >> [2]http://www.bath.ac.uk/person/812559
> >>
> >>
> >> *********************************************
> >>
> >> Mahmoud E. Soliman
> >>
> >> Lecturer of pharmaceutical organic chemistry
> >>
> >> Pharmaceutical Organic Chemistry Dept.
> >>
> >> Faculty of pharmacy
> >>
> >> Zagazig University
> >>
> >> Zagazig
> >>
> >> Egypt
> >>
> >> **********************************************
> >>
> >> Email:
> >>
> >> [3]mess20.bath.ac.uk
> >>
> >> [4]meelkot.zu.edu.eg
> >>
> >> [5]mahmoudelkot.gmail.com
> >>
> >> References
> >>
> >> 1. http://people.bath.ac.uk/mess20/
> >> 2. http://www.bath.ac.uk/person/812559
> >> 3. mailto:mess20.bath.ac.uk
> >> 4. mailto:meelkot.zu.edu.eg
> >> 5. mailto:mahmoudelkot.gmail.com
> >> _______________________________________________
> >> AMBER mailing list
> >> AMBER.ambermd.org
> >> http://lists.ambermd.org/mailman/listinfo/amber
> >>
> >
> >
> >
>
>
> --
> Tato zpráva byla vytvořena převratným poštovním klientem Opery:
> http://www.opera.com/mail/
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Mar 17 2011 - 10:00:03 PDT
