[gmx-users] Re: Turning off electrostatics between groups using custom tables?

[Mark Abraham]

On 1/09/2012 2:50 AM, Smitty wrote:
>> While for the protein-protein or ion-group ion-group interactions I provide
>> just the standard 6-12 table. The problem comes that when I run using my
>> modified table, the system will run but gradually explodes. I note that
>> when
>> I look back at my output from grompp I find a warning concerning eplison
>> and
>> sigma being used instead of C6 and C12,
>>> Ignoring warnings is a great way to waste time, and if you can't provide
>>> specific details of what it was, you're unlikely to get specific help.
> The warning is as follows:
>
> WARNING 1 [file pep_ion.top, line 1218]:
> Using sigma/epsilon based combination rules with user supplied potential
> function may produce unwanted results. I've looked into this on the mailing
> list, but it appears that so long as the forces are what I expect them to
> using the standard tables, I should be able to ignore this warning.
> Hopefully I have not misread the previous threads.
>
>>    but if I swap my custom table for
>> the sample 6-12 table (keeping the first two columns not zero) the system
>> doesn't explode and has dynamics similar to previous simulations making
>> use
>> the built-in table generation. Any ideas on where I am going wrong?
>>> Can't tell. What's the "ion group"? Does it even get close to Protein?
> The "ion group" is a set of 50 dissociated ions (25 K+, 25 Cl-). The ions
> under standard simulation conditions (non-custom interaction table) rarely
> have K+ or Cl- atoms coming within 0.5 angstroms of any atom of the protein
> group; but they do still happen. If it helps my .mdp file for the simulation
> is attached.  http://gromacs.5086.n6.nabble.com/file/n5000727/md.mdp md.mdp

1) That .mdp applies different tables to K-K, protein-protein and all 
other interactions, which isn't what you've said you've done.
2) You should follow the advice here 
http://www.gromacs.org/Documentation/Terminology/Thermostats
3) Your life will be simpler if you use energygrp_table only for the 
group-group interactions you want to change, and leave the default 
interactions normal. It will also be faster, because using tabulated 
interactions is much slower than the normal ones.

Mark