AMBER Archive (2007)

Subject: RE: AMBER: NVT/NPT equilibration issue in GPCR-membrane simulation

From: Akshay Patny (akshay17_at_olemiss.edu)
Date: Fri Jan 26 2007 - 17:27:56 CST


Dear Ross

Thanks a lot for your detailed explanation and several useful suggestions.
I will use k=10 for protien backbone restraints and try using some of
numbers you have suggested to see if simulations are stable.

Thanks a lot again, Best Regards

Akshay
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
Akshay Patny
PhD Candidate (5th Yr.), Computational Chemistry
Department of Medicinal Chemistry, School of Pharmacy
The University of Mississippi
805 College Hill Rd, # 9, Oxford, MS 38655
E-mail: akshay17_at_olemiss.edu
Phone (O): (662)-915-1286,(M): (662)-801-5496
= = = = = = = = = = = = = = = = = = = = = = = = = = = = = =
-----Original Message-----
From: owner-amber_at_scripps.edu [mailto:owner-amber_at_scripps.edu] On
Behalf Of Ross Walker
Sent: Friday, January 26, 2007 5:11 PM
To: amber_at_scripps.edu
Subject: RE: AMBER: NVT/NPT equilibration issue in GPCR-membrane simulation

Dear Akshay

> I have done minimization of the system stepwise: first minimizing only
> water/ions, then water/ions and lipid, and then everything with
> harmonic restraints on protein backbone (k = 50).

A value of 50 is probably slightly too large, 10 or less would be more
appropriate but if you simulations are stable then you probably don't have
to worry.

> 1. First raising the temp of the system to desired level (my case: 300
> K) gradually using NVT ensemble (I understand that this step (NVT
> equilibration before NPT equilibration) is very important to take care
> of the voids that are created by deleting lipids etc. and spaces at
> the edges of boundaries, which is especially true in my case).

Your understanding is kind of back to front here. The reason you can't do
constant pressure from the beginning is that the calculation of pressure is
inaccurate at low temperatures. This can result in very large over
corrections by the barostat and cause yous system to explode. However, you
want to get an NPT simulation running as soon as possible to correct the
density that initially is too low due to the creation of voids and buffer
space etc around the solvent when building the system. Thus you should
typically run a short amount of NVT simulation until you reach a suitable
temperature. 100K+ is probably okay. Then you can switch to constant
pressure and equilibrate out these voids etc. If you run NVT for too long
(sometimes as little as 100ps) you can get very large vacuum bubbles form
and these will take a very long time to equilibrate out at constant
pressure.

So I would recommend. Minimization -> 20ps heat 100K NVT -> 100ps heat to
300K NPT -> continue at NPT until the density has equilibrated. At this
point you can then either switch back to constant volume or remain with
constant pressure. This will depend on what you are studying and whether you
want ro run without a thermostat. If you don't want a thermostat for your
production run then you have to use constant volume as there is no NPE
ensemble.

The numbers above are in now way rigid. The best option is to do things in
small stages and look at the trajectory to see if anything is going astray.
Also plot things like RMSD, potential energy, temperature and density to see
if they are converging on sensible values. Any kind of spikes will indicate
a problem.

> B. the large size of the gpcr-membrane system i.e. 63,000 atoms
>
> QUESTION 1: I am thinking of some 300 PS to raise the temperature from
> 0 - 300 K using (600,000 steps with a time step of 2 using SHAKE). IS
> THIS REASONABLE OR SHOULD THE LENGTH OF THIS NVT EQUILIBRATION BE MORE
> OR LESS, AND HOW MUCH??

In terms of heating you can really do this as quickly as you can get away
with. Doing it to slowly certainly won't hurt so if you can spare the
computer time then choose a long heating cycle. 300ps should easily be long
enough. But I have regularly heated 70K+ atom systems in 20ps or less
before, especially with the langevin thermostat so I wouldn't worry here.

In terms of how long to run NPT simulations this is really long enough that
the density has converged. I.e if the density changes by less than 5% or so
for a doubling of the run length then you are probably good.

All the best
Ross

/\
\/
|\oss Walker

| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross_at_rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu