[gmx-users] Fwd: QMMM calculation

[Christian Burisch]

Pradip Kumar Biswas wrote:

Hi Pradip,

> Supposing that while coming back to MM from QM you care only for the 
> forces on the polar atoms as the MM part is satisfied only with them, it 
> is perhaps easily doable.

sounds good! You would need some kind of special hydrogen database for 
applying the protons. Perhaps one could use a modified ffgmx2.hdb for 
that.

> Alternately, you need to have an admixture as you have said: all-atom QM 
> part with united-atom MM part. How difficult it will be to generate 
> topology and coordinates for such a case?

Coordinates should be no problem, if one edits GROMACS' hydrogen 
database or adds the protons by an external editor like PyMol.
Concerning the topology one would have to generate some "new" all-atom 
amino acids (if these are included in the QM box) and ligands. One 
could perhaps use the "bare carbon" atom types C and CB for that, but 
I don't know if there is a proper hydrogen type.
Another disadvantage would be that you cannot use the great PRODRG 
service for automatized topology generation, since this tool "only" 
generates united-atom parameters for the GROMOS96 43A1 ff (of course).
But IIRC some half years ago there was an announcement/rumour of 
PRODRG being capable of generating OPLS-AA topologies in the near 
future. In this case one could switch to this ff completely without 
having to generate ligand topologies from scratch.

> What other problem(s) you 
> foresee?

Mmmh, personally I would try to avoid a hybrid all-/united-atom ff (I 
always try to stay as close as possible to the standard ff 
parameters). But since people seem to use e.g. united-atom lipids and 
all-atom proteins (OPLS-AA) together without problems this is perhaps 
only a fad of mine.

Perhaps Axel Kohlmeyer (who is on this list, too) could comment on the 
topic? He's been in the department next door and AFAIK they use a 
GROMOS96/CPMD interface over there. Of course I could walk over to the 
next building and ask them personally... ;-)

Regards

Christian