Dear David,
> I know it is quite an involved issue but if this is really the opinion
> of the AMBER developers then it might be prudent to add a
> remark to this
Okay, here is my $0.02... In my opinion....
You have hit upon a very hot topic here. Originally when I wrote the DNA
tutorial the langevin dynamics temperature regulation was fairly new and not
much was known about its behaviour. It was known, however, that overall
temperature equilibration was better.
However, since then we have become much more familiar with the method and
the implications of using it. Essentially while ntt=3 equilibrates your
temperature better as a method of maintaining an equilibrium temperature it
is probably not so good. This is what Dave was referring to when he was
talking about using it with explicit solvent. Essentially it would appear
that the best approach, in terms of obtaining accurate dynamics, would be to
equilibrate your system initially with ntt=3, since it equilibrates
temperature very well, and then, in my opinion, switch to ntt=1 for the
production phase. Alternatively assuming your system is well equilibrated
you should be able to switch off the thermostat all together (ntt=0) and
just run your explicit solvent simulation with no thermostat. This is true
since the force field is conservative, hence you should be able to maintain
constant energy for a very long time. However, errors in the integrator (the
fact you make a timestep approximation) and errors involved with using a
cutoff will eventually lead to you bleeding energy over time. Hence I
believe a weak thermostat is always a good idea as it corrects for these
errors...
Anyway, back to the point at hand... Ntt=3.
The problems with using ntt=3 in a production calculation, is that it alters
the dynamics of your system, essentially the short term dynamics (fast
dynamics) are radically altered and so if you are interested in obtaining
information about the fast dynamics of your system during the production
phase you cannot use ntt=3 since it essentially corrupts this information.
Berendsen (ntt=1) does not do this.
In terms of the long term dynamics ntt=3 also effects these but where such
transitions are not directly dependent on the fast dynamics, i.e. something
like the A-DNA to B-DNA conversion, then ntt=3 can actually serve to
increase the speed of these dynamics. In other words you can cover more
phase space in less time. However, the langevin dynamics while not effecting
the actual transition does effect the speed of the transition and as such if
you want information regarding timescales of structural interconversions you
again need to use ntt=1. Hence if you are just interested in going from a
high energy structure to a lower one quickly ntt=3 is probably the best
method, but if you want to compute things like time correlation functions
then you need accurate fast dynamical information so you need avoid ntt=3.
However, in the case of the DNA tutorial it was fortuitous in the use of
ntt=3 since it allows the A-DNA to B-DNA conversion to occur much more
quickly than it would using a berendsen thermostat (or none at all).
Essentially the langevin dynamics allowed the energy barrier to be crossed
much quicker by increasing the sampling of higher energy configurations.
Hence for the A-DNA to B-DNA test case it was probably the correct choice.
However, this does not mean it would be the correct choice for all explicit
solvent simulations...
Hence what I will probably do is simply add a few sentences discussing the
above to the tutorial to try to make the situation clearer.
I hope it makes sense, please let me know if what I have stated makes sense
to you, or if it sounds like complete gibberish... Anyway, if it is clear
let me know as I will then adjust the tutorial to match.
All the best
Ross
/\
\/
|\oss Walker
| Department of Molecular Biology TPC15 |
| The Scripps Research Institute |
| Tel:- +1 858 784 8889 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk/ | PGP Key available on request |
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Fri Dec 03 2004 - 07:53:00 PST
