From: Vincent Leroux (vincent.leroux_at_loria.fr)
Date: Wed Jun 13 2012 - 10:08:18 CDT

Hi,

Sorry if I give an answer too long, but since I cannot understand where
your problem is exactly...

If you want to prepare a small molecule by hand for use with CGenFF,
identify the appropriate groups in the CGenFF topology file and match
the atom and residue names. Comments in the parameter file may also help
identifying the correct atom types. Then use psfgen as you would for a
protein, and save the pdb with the psf at the same time. Eventually
correct the partial charges with the bond charge increment method to
make sure the total charge is what you expect. You probably will have to
edit a lot of text by hand there is a risk you introduce a small mistake
and not notice it if you are not careful enough... Plus you need some
experience to get to the correct analogies.

https://www.paramchem.org is much easier and gives consistent results.
You should expect no miracle if you submit a compound with groups that
are unusual or known to cause trouble with parameterization. Most often,
the parameters you get, directly blended from CGenFF data (they are not
optimized, this may happen some time... at least you get comments with
"penalty scores" giving you an idea of how reliable the data is), are
quite solid and give stable results when injected in a protein-ligand
system.

Now, for the "long names problem". This seems to be a recent CHARMM
standard (seriously, no one wants to enforce the "PDB standard" and its
logic rooted in the punch-card Fortran era... even the PDB database
itself does not) that is not totally transparently implemented into NAMD
yet (VMD seems to handle it properly, provided the psf file is loaded
before the pdb coordinates, as should always be done). Upon using
psfgen, you should *not* touch the pdb that is saved with the psf, only
make sure the pdb given as input to psfgen (or the CHARMM program, as an
alternative) is properly recognized (if not, you will know). ParamChem
does all that dirty work for you - since you just have to provide a
correctly-protonated mol2 file and wait a short time. You may use psfgen
only to merge your ligand back into a more complex system, which is very
straightforward.

Now that you have a consistent pdb/psf couple with the long names (to be
used with CGenFF parameters and probably additional ParamChem-guessed
analogies), in order to make sure there will be no problems later with
NAMD, look at the 1st line in the psf file: the tag "NAMD" should be
present, with "PSF" and "CMAP" (if you use a CMAP version of CHARMM22).
Correct manually if needed.

Regards
VL

On 13/06/2012 12:33, Ajasja Ljubetič wrote:
> Dear All,
>
> the CGenFF <http://mackerell.umaryland.edu/~kenno/cgenff/> is very
> useful for parametrizing small molecules. But since it contains so much
> data, the atom names are longer than four characters. This causes havoc
> with many programs that truncate atom names to 3 or 4 characters (Well
> actually the PDB standard
> <http://www.wwpdb.org/documentation/format32/sect9.html#ATOM> allows
> only four chars for the atom name). My question is:
>
> How can I prepare a PDB file with for use with psfgen? Will psfgen work
> with parameter sets, that contain atoms with long names?
>
> There was a related question
> <http://www.ks.uiuc.edu/Research/namd/mailing_list/namd-l.2011-2012/0264.html>
> on the namd mailing list a while ago, but nobody answered, so I
> apologize, if I missed something obvious.
>
> Best Regards,
> Ajasja