[gmx-users] suitable parameter settings for intermolecular non-bonded interaction

[Mark Abraham]

On 31/03/2010 11:44 PM, vedat durmaz wrote:
> thanks to both of you for your helpful hints. actually, i'm not using
> the LIE method as available with the gromacs package, but planning to
> implement the model myself using energies from gromacs runs.

Sure... it's still only an algorithm that generates a fairly coarse 
approximation.

> be that as it may, one thing remained unclear (see below please).
>
>
>   Mark Abraham schrieb:
>> On 31/03/2010 9:15 PM, vedat durmaz wrote:
>>> hello out there,
>>>
>>> i'm trying to simulate a receptor ligand system in order to extract
>>> intermolecular non-bonded interaction energies (vdw/coulomb) and apply
>>> the linear interaction energy model.
>>
>> Agreed on Sander's point, here.
>>
>>> which parameter settings do you recommend for highly accurate results?
>>> up to now, i had chosen (copied from somewhere :)
>>
>> Well, write it down, because you'll need to defend your choices later :-)
>>
>>> rlist                = 1.2
>>> vdwtype              = shift
>>> rvdw                 = 1.0
>>> rvdw-switch          = 0.9
>>> coulombtype          = PME-Switch ; PME
>>> rcoulomb             = 1.1
>>>
>>> resulting in LJ-14 = 0, and Coulomb-14 = 0 (which i "intuitionally"
>>> prefer to LJ-SR and Coulomb-SR).
>>
>> I've no idea what you mean by this.
>
>
> what i mean is that after having applied "g_energy" to the energy output
> file in order to extract non-bonded energies, all "Coulomb-14" and
> "LJ-14" terms shape up to be equal to zero (as if they haven't been
> calculated),

Well that means they weren't calculated. Whether that's reasonable 
depends on your forcefield.

>  but "Coulomb-SR" and "LJ-SR" are plausible decimal numbers
> like for example -48.1234. though, at the moment i "tend" to operate on
> *-14 energies. how can i tell gromacs to compute these terms?

*Tend*? Your usage of that word makes no scientific sense that I can 
see. What on earth is more significant about non-bonded interactions 
between atoms separated by 3 bonds than other non-bonded interactions? :-)

Mark

>>> i have a box of 8.1857 x 8.1857 x 8.1857 nm. Is it helpful to increase
>>> rcoulomb and rvdw up to e.g. 8.0 and rvdw-switch up to 4.0 or/and should
>>> i take vdwtype=cut-off instead of shift?? what's the difference? how can
>>> i get my LJ-14/Coulomb-14 terms? can anyone provide a well-working
>>> combination of parameter settings for non-bonded interactions within a
>>> system as described above?
>>
>> As I wrote earlier today, haphazardly varying parameters is not what
>> you want to do. Merely increasing these cut-offs may feel like "more
>> accuracy", but computing a more expensive approximation to a
>> theoretical model is not necessarily any good, because the parameters
>> you are using were developed for some different cut-off. It does not
>> follow that the parameters are transferable to other contexts. You
>> should aim to use a model physics for which the parameters were
>> developed, or for which they were subsequently proved to be effective.
>>
>>> that's quite a lot questions, i know.
>>>
>>> gromacs says: "Refer to the primary literature for your of choice for
>>> the recommended settings". However, i couldn't find any recommondations
>>> for ffamber99 (ffamber99sb is what i use) in the pdf documents provided
>>> with the amber port's download.
>>
>> They're not primary literature. Refereed publications tend to
>> accompany the introduction of forcefields, and they are mandatory
>> reading. AMBER (port)'s webpage will tell you what to read. Also
>> search for subsequent work by the same authors, for they are probably
>> using best practice for their own forcefield...
>>
>> Mark