Yes, fair enough--Berendsen is bad, I'm with Bernie (Brooks)! A very weak
Berendsen thermostat is what was recommended to me for simulations where
diffusion properties and NVT are needed simultaneously, by one of our
respected colleagues no less. And, I agree--formally, it's bad, but with
weak thermocoupling, in practice the effects are negligible. As I said, a
weak Langevin thermostat is also an option here. Just get the system to
equilibrium with a typical amount of thermocoupling, then roll on with very
weak thermocoupling to measure kinetic properties, or roll with NVE and
keep an eye on the overall system temperature.
On Tue, Nov 22, 2016 at 6:28 PM, Adrian Roitberg <roitberg.ufl.edu> wrote:
>
>
> On 11/22/16 3:42 PM, David Cerutti wrote:
> > The thing you need to worry about is whether the thermostat will affect
> the
> > diffusivity of the water, as well as how the water model itself behaves.
> > Any state properties like an RDF, mean structural characteristics, or
> > potentials of mean force will be unaffected by your choice of thermostat,
> > but rates of any processes will be.
> With all due respect, the statement right above is not true. Berendsen
> thermostats do not produce canonical ensembles, so they will change the
> equilibrium properties. The effect might be system dependent, but in
> principle, they are not good choices.
> > As Gustavo suggested, the safest thing
> > to do for thermostating when kinetic properties is to use no thermostat
> at
> > all (NVE), but this carries its own risks, as there is often strain in
> the
> > protein (potential energy) which will become kinetic energy as it is
> slowly
> > released--this takes hundreds of ns in reality, even though energy
> > minimization and a few ns of constant pressure dynamics will relieve the
> > bulk of the problem. A good compromise for our case may be to use a very
> > weak Berendsen thermostat (ntt = 1, set tautp to a high value like
> 1000.0)
> > or a weak Langevin thermostat (set gamma_ln to a low value like 0.1). In
> > order to test your setup, you can run a box of water at 298K at see what
> > the diffusion constant comes out to. For TIP3P the diffusion should come
> > out to about 5.0 (way too high), but that's what the model does--may not
> be
> > what you want if you need kinetic properties of water. A water model
> like
> > SPC/E, TIP4P-Ew, or SPC/Eb will have a much more reasonable diffusion
> > constant around 2.4.
> >
> > Dave
> >
> >
> > On Tue, Nov 22, 2016 at 5:09 PM, Geoffrey Gray <gmgray2.mail.usf.edu>
> wrote:
> >
> >> Occupancy is another useful measure for how frequently water is found in
> >> different parts of the system. You can do this using the volmap tool in
> >> VMD.
> >>
> >> Best,
> >>
> >> Geoffrey
> >>
> >> -----Original Message-----
> >> From: Karolina Mitusińska (Markowska) [mailto:markowska.kar.gmail.com]
> >> Sent: Tuesday, November 22, 2016 3:06 PM
> >> To: AMBER Mailing List <amber.ambermd.org>
> >> Subject: Re: [AMBER] DIfferences between thermostats
> >>
> >> Thanks Gustavo and Geoffrey!
> >>
> >> I've ran 50ns simulations with different thermostat and in NVE
> conditions.
> >> Is there a clever way to compare the water behaviour during these
> different
> >> simulations?
> >> I was thinking about RDF function - could you advise something else?
> >>
> >> Thank you for your help!
> >> Best regards
> >>
> >> 2016-11-22 21:05 GMT+01:00 Karolina Mitusińska <mitusinska.gmail.com>:
> >>
> >>> Thanks Gustavo and Geoffrey!
> >>>
> >>> I've ran 50ns simulations with different thermostat and in NVE
> >> conditions.
> >>> Is there a clever way to compare the water behaviour during these
> >>> different simulations?
> >>> I was thinking about RDF function - could you advise something else?
> >>>
> >>> Thank you for your help!
> >>> Best regards
> >>>
> >>> 2016-11-22 20:38 GMT+01:00 Geoffrey Gray <gmgray2.mail.usf.edu>:
> >>>
> >>>> Depending on the properties you are interested in will determine your
> >>>> thermostat. Both the Andersen and the Langevin thermostats sample
> >>>> canonical
> >>>> distributions, so either will give correct structural configurations
> and
> >>>> thermodynamic properties. However, time-dependent properties (such as
> >>>> diffusion) are not correct, because the thermostats are stochastic and
> >>>> include random fictitious collisions. If you are interested in
> transport
> >>>> properties, such as diffusion coefficients, then a deterministic
> >>>> thermostat,
> >>>> such as the Nose-Hoover, is recommended.
> >>>>
> >>>> Hope this helps.
> >>>>
> >>>>
> >>>> -----Original Message-----
> >>>> From: Karolina Mitusińska (Markowska) [mailto:markowska.kar.gmail.com
> ]
> >>>> Sent: Tuesday, November 22, 2016 2:31 PM
> >>>> To: AMBER Mailing List <amber.ambermd.org>
> >>>> Subject: [AMBER] DIfferences between thermostats
> >>>>
> >>>> Dear Amber Users,
> >>>>
> >>>> I would like to analyse a flow of water molecules through a protein.
> The
> >>>> problem is, I'm not sure which thermostat to choose to have the
> >> _correct_
> >>>> behaviour of water molecules. For example if my system have two
> tunnels
> >>>> and
> >>>> one is more hydrophobic than the secnd one - which thermostat should I
> >>>> choose?
> >>>>
> >>>> Sorry for such stupid question. I've spend whole day reading papers
> >> about
> >>>> different thermostats and yet I only know that there could be some
> >>>> differences in water molecules behaviour, but I still don't understand
> >>>> the
> >>>> reason of these differences.
> >>>> Could you help me, please?
> >>>>
> >>>> Best regards.
> >>>> Karolina Mitusińska
> >>>> PhD student
> >>>> _______________________________________________
> >>>> 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
> >>>>
> >>>
> >> _______________________________________________
> >> 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
> >>
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
>
> --
> Dr. Adrian E. Roitberg
> University of Florida Research Foundation Professor.
> Department of Chemistry
> University of Florida
> roitberg.ufl.edu
> 352-392-6972
>
>
> _______________________________________________
> 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 Tue Nov 22 2016 - 16:00:02 PST
