[gmx-users] converting parmbsc0 dihedrals to RB function

[gportel at gwdg.de]

Hi,

Right, the only dihedral angles that do not allow an exact translation to
RB because of the phase are the ones involving the new parametrization for
nucleic acids. Namely, it corrects the alpha/gamma transitions to get the
populations of states right, thus avoiding the loss of helicity on the
long run. For proteins, there's no difference at all, so you can use
ffamber.

I've used the combined approach you mention without problems. The only
thing is that I do not use pdb2gmx. Rather I use amber's tools
(leap+libraries, all free in ambertools) to get a topology, and then I use
a slightly modified version of amb2gmx.pl script to get a gromacs
topology. I only convert the new parmbsc0 dihedrals to type 1 if the phase
does not allow a direct conversion to RB form. Typically you have three
entries for the same dihedral, this is no problem at the itp level, grompp
captures them all. With this approach I reproduced a torsional scan of
parmbsc0 from amber md tool using gromacs, so I feel confident about it.

Another option is to use the ffamber parameters and change the itp file
including the new ones.

I should also mention that I had problems importing the new dihedrals
using pdb2gmx. IIRC, pdb2gmx will only use one definition for dihedral if
type 1 or 3 is used. There was a new dihedral type that could be used in
the development versions, but... for a reason it did not work out for me,
don't recall exactly why. It could well be that it would have worked with
more patience from my side.

Hope this helped, somehow. You were on the right direction anyway.

best,

Guillem


Mark Abraham wrote:
> On 8/04/2010 10:32 PM, Vigneshwar Ramakrishnan wrote:
>> Dear Dr. Mark,
>>
>> Thanks very much for pointing out that GROMACS can handle multiple
>> instances of the parameters with different n for a single dihedral.
>
> I hope I'm right there. There's been discussion on such topics before.
There's some subtle difference between [ dihedrals ] and [ dihedraltypes
] which eludes my memory, but which will be in the archives.
>
>> However, given that, in the existing ffamber ports in GROMACS  the other
>> dihedral parameters have been converted to RB form, would it make a
>> difference to just import these new parameters in the non-RB form?
>
> It will make no numerical difference, but g_energy will probably report
the contribution from the two different functional forms separately.
Attempting to mix *non-bonded* interactions in this kind of way asks for
trouble, because the maths is not well-defined. Here, you're fine,
because the maths is right.
>
>> To my knowledge, the torsional term should not vary (or, vary within
>> acceptable limits) irrespective of the functional form used to calculate
>> it. But I would like to know if there could be some reason for which
>> this assumption cannot be made.
>
> The R-B form might be more efficient to evaluate than a sum of simple
dihedrals in the cases where the angle is zero or 180, which might
explain the ffamber practice.
>
> Mark
>
>> Dear Dr. Alan,
>>
>> Thanks very much for pointing out to the acpypi code. I was able to
>> better understand the conversion procedure.
>>
>> Sincerely,
>> Vignesh
>>
>> On Wed, Apr 7, 2010 at 10:04 PM, Mark Abraham <Mark.Abraham at anu.edu.au
>> <mailto:Mark.Abraham at anu.edu.au>> wrote:
>>
>>     On 7/04/2010 7:44 PM, Vigneshwar Ramakrishnan wrote:
>>
>>         Dear Users,
>>
>>         I am trying to port the new parmbsc0 forcefield
>>         (http://mmb.pcb.ub.es/PARMBSC0/frcmod.parmbsc0) into gromacs
for DNA
>>         simulations.
>>
>>         While unit conversions are sufficient to convert many of the
>>         parameters
>>         from AMBER to GROMACS format, dihedral angle conversion does not
>>         seem to
>>         be straight forward - the dihedral parameters need to be
>>         converted to
>>         the Ryckaert-Bellemans parameters.
>>
>>
>>     Why? GROMACS can probably do the non-RB form - IIRC you can
>>     implement a sum of multiple instances of 4.61 with different n.
>>
>>
>>         I went through the GROMACS 4.0
>>         manual, especially equations 4.61-4.65 to understand the
>>         procedure. The
>>         procedure involves comparing the fourier expansion of the IUPAC
>>         convention of dihedral potential (equation 4.65) with the
>>         Ryckaert-Bellemans (RB) functional of dihedral potential
>>         (equation 4.62)
>>         to get the Cn's of the RB function. However, I am not able to
>>         understand
>>         how to account for the phase angles. (Also to note, the parmbsc0
>>         forcefield contains phase angles other than 0 and 180.)
>>
>>
>>     Elegant conversion formulae require those angles to be convenient...
>>
>>     Mark
>>
>>
>>           Any advice or suggestion will be of great help.
>>
>>         Thank you,
>>



Dr. Guillem Portella
- MMB - Institute for Research in Biomedicine
- Parc Cientific de Barcelona -
http://mmb.pcb.ub.es/~gportella