Re: Minimizing and Equilibrating

From: Timothy Isgro (timisgro_at_ks.uiuc.edu)
Date: Tue Jul 18 2006 - 12:19:50 CDT

Hi Anya,

Lewyn has given some good advice. Perhaps I can expand on it a
little by giving you some concrete numbers from a few recent
simulations of a 125,000 atom system. I emphasize Lewyn's advice
that the numbers will vary depending on the starting configuration of
your system and that in general the protein should be equilibrated
until temperature, pressure, and RMSD reach stable levels. Perhaps
my numbers will give you a ballpark estimate for practicality.

1. I typically minimize until the gradient tolerance reaches single
digits. I minimized my system for 5,000 steps. Minimization is
typically not very computationally demanding compared to the dynamics
you will be running, so the difference between minimizing for 5,000
vs. 10,000 steps is likely moot, as long as your gradient tolerance
is low.

2. I equilibrated my system in the NPT ensemble for 200 ps with
proteins fixed.

3. I freed all atoms, minimized for another 5,000 steps, and ran NPT
dynamics for 5 ns. The RMSD of my protein reached a value of ~3.8
Angstroms after ~2.5 ns, and remained at that level +/- 0.3
Angstroms. This indicated to me that my protein was equilibrated.
You may wish to extend the time as far as you like to the point where
you are comfortable that the protein will not deviate significantly
from its conformation.

Also, the number of steps you will need to run for in each case will
depend on the size of your system. At 450,000 atoms, you will likely
need to run for more steps than I did, and you may not be able to
simply free your large protein as I did. Lewyn wisely suggests
decreasing constraints gradually. Again, let your system properties
be your guide.

Hope it helps,
Tim

On Jul 17, 2006, at 1:52 PM, LEWYN LI wrote:

> Anya,
>
> From my experience, the number of time steps needed for each stage
> depends on a number of things, including the size of your system,
> how relaxed/frustrated your initial structures are etc. The "rules
> of thumb" I used are:
>
> 1. Minimimize until the gradient tolerance drops below 1.0. The
> number of steps required to achieve this is system-dependent. I
> usually just minimize initially for, say, 10000 steps and
> periodically check the gradient tolerance.
>
> 2. Equilibrate the fixed protein until the temperature (and other
> quantities such as pressure if necessary) stabilizes at the desired
> value. Again, the number of steps is worked out by trial and error.
>
> 3. Equilibrate with decreasing harmonic constraints until the
> unconstrained protein is stable in terms of temperature, structure
> and other desired quantities.
>
> Other people probably have different protocols. Hope this helps!
>
>
> LEWYN
>
> On Mon, 17 Jul 2006, a-yermakova_at_northwestern.edu wrote:
>
>>
>> I am working with a protein solvated in a water box (altogether
>> 450,000atoms). As the tutorial
>> suggests, I am going to
>>
>> 1)minimize the system
>> 2)fix the proteins and equilibrate
>> 3) minimize and equilibrate again, with protein free to move.
>>
>> I am looking for suggestions regarding the number of timesteps
>> needed for each one of those steps.
>> Should there be more equilibration TS's than minimization? If so,
>> are there any suggested relative
>> proportions? Should the first and second minimizations/
>> equilibrations have the same number of
>> timesteps? How significant is this?
>>
>> Thank you very much!
>> Anya Yermakova
>>
>>
>>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:19:39 CST