[gmx-users] Fwd: gmx-users Digest, Vol 79, Issue 102

Mark Abraham Mark.Abraham at anu.edu.au
Wed Nov 17 04:17:14 CET 2010

On 15/11/2010 5:56 AM, Silvia Crivelli wrote:
> We have a visualization tool that allows us to visualize the changes 
> in the energy during energy minimization.
> The areas with more intense color are those where the atoms contribute 
> the most to the total energy value.
> I just wrote a plug-in to use Gromacs for energy minimizations (or MD 
> simulations) and I need the per-atom energies
> to use our tool's energy visualization feature.

Such a tool is all very nice, but what is it good for? If you apply it 
to EM on a TIP3P water box, then you'll see that every oxygen 
contributes more to the total energy than every hydrogen, because they 
have a VDW interaction. However that was known already...

Anyway, as I mentioned in my first email, there are approaches to do the 
decomposition, but GROMACS is not engineered to make this easy, because 
that would severely compromise performance. Even if that were not true, 
making the decomposition would be vastly more work than it is worth.


> Thanks again,
> Silvia
>> Message: 3
>> Date: Sun, 14 Nov 2010 06:31:56 +1100
>> From: Mark Abraham <Mark.Abraham at anu.edu.au>
>> Subject: Re: [gmx-users] how to calculate per-atom energies
>> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>> Message-ID: <4CDEE7AC.9020305 at anu.edu.au>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>> On 14/11/2010 5:58 AM, Silvia Crivelli wrote:
>>> Hello,
>>> When I run energy minimization, I'd like to obtain the energy per atom
>>> in addition to the (potential) energy value for the entire protein.
>>> Is there a way to do this?
>> You can get the nonbonded energy per energy group, which you could
>> define to be a single atom, but you are limited to 256 of such groups,
>> and to not using PME. Otherwise, there are approaches involving a lot of
>> hacking about with .top or .tpr files and using mdrun -rerun that could
>> do this for both nonbonded and bonded.
>> However, a more important issue before doing such work is to be sure
>> about what you expect such a decomposition to tell you. A high or low
>> energy for a given atom doesn't necessarily mean anything, and even if
>> it did, the force field wasn't necessarily parametrized to produce
>> reliable per-atom energies (though it probably does do so).
>> Mark
>> ------------------------------

More information about the gromacs.org_gmx-users mailing list