Re: Rule of thumb for equilibration

From: Aron Broom (broomsday_at_gmail.com)
Date: Mon Jul 28 2014 - 09:59:04 CDT

First, energy minimization isn't the same as what is typically referred to
as equilibrating. Generally, you'll only find a fairly local energy
minimum with that technique. So the real purpose of energy minimization is
just to ensure your system does not explode or go hurtling off down some
unreasonable trajectory because you started with too much potential energy.
But, for energy minimization, 100 is possibly a little low, though it's
very system dependent, and is much more about when the energy plateaus and
additional minimization does nothing. The NAMD output shows how the energy
is changing with each step, if it's still going down, you might want to run
more, like 1000 or 10000.

Equilibration usually means steps of actual MD that are done initially, but
later discarded from any analysis. The general rule of thumb here is
similar to the energy minimization, in that you go for long enough that
some critical property stops changing. You could use the potential energy,
as in energy minimization, but also the RMSD over time of your
macromolecule (if you are doing something with a protein or nucleic acid)
is often used. There are other properties that people use, and it's
probably very dependent on what you are looking for.

Additionally, what one very often wants to do, unless resources are
extremely constrained, is to run the simulation at least twice, with a
different random seed for the thermostat (assuming you are using a Langevin
thermostat) and with slightly different starting coordinates (so if it took
5ns of MD to reach your equilibration criterion, then run another 5ns, and
that will give you two slightly different starting structures). This way
you can be more confident that whatever your conclusions from the actual
production MD aren't just a fluke.

On Mon, Jul 28, 2014 at 10:45 AM, Stephan Grein <
grein_at_informatik.uni-frankfurt.de> wrote:

> Dear NAMD users,
>
> is there a rule of thumb how long one should equilibrate, i. e. when we
> coax the system to the desired simulation/productive period of the MD
> simulation?
> For no reason I use a minimize #steps of 100 as suggested in the NAMD
> tutorial.
>
> Best,
> Stephan
>
>

```--
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo
```

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:40 CST