[gmx-users] Implemetation of the CHARMM27 ff into gromacs : managment of dihedral terms

[Mark Abraham]

> Hello everybody,
>
> I am quite interested by the implementation of the CHARMM27 force field
> into Gromacs. I have read few messages  on the Gromacs mailing list and
> have downloaded the CHARMM conversion on this web page:
>
> http://www.gromacs.org/old/topologies/force_fields.php

I also recommend the contribution I made here
http://www.gromacs.org/contributed_by_users/task,doc_details/gid,59/
 which I believe to be based on a a more sound conversion technique.
However the residue topologies, etc. available from the source you cite
are the only known source of these.

> So, I have some questions about this translation.
> - The dihedral terms (proper angles) initially treated by a periodic
> function have been converted to Ryckeart-Bellemans potential. Why this
> choice? And how the conversion has been done?

As described on my above link, and in more detail in my scripts, CHARMM
sometimes uses two periodic functions for a given dihedral angle with
different multiplicity. GROMACS does not permit the use of two such
functions, however. Fortuitously, the combinations of multiplicity and
phase shift are such that in each case you can use cosine expansions of
cos (n * phi - delta) = sum_i a_i (cos phi)^i (see standard trigonometry
works) to express these as a single Ryckaert-Bellemans function. I haven't
verified the other implementation, but my algorithm is clear in the
scripts I supply and it has been verified against the CHARMM
implementation of CHARMM27.

> -  In a same time, improper angles initially treated by harmonic terms
> have been converted to periodic terms. Why? I'm not sure to understand
> this choice.

Sorry, don't recall anything here - can you give an example?

> Additionally, I have seen that the Urey-Bradley constants were ewritten in
> the ffcharmmbon.itp file. I guess they are parsed but not really used by
> Gromacs or the implemtation of the potential has been achieved?

As you can see in the GROMACS manual, U-B potentials are implemented - but
there was an erroneously-retained fatal warning in versions prior to 3.3
that you can comment out successfully. The author of the implementation of
CHARMM27 that you cite implemented U-B using normal angle functions plus
distance restraints, if I recall correctly, which required that you use a
compatible .mdp file and it was all a bit messy. My scripts use the true
U-B potentials and again have been verified against the CHARMM
implementation of CHARMM27.

Mark