Re: some recommendations for parametrizing new molecules required

From: Axel Kohlmeyer (
Date: Fri Jan 23 2009 - 09:20:50 CST

On Fri, 23 Jan 2009, CHINDEA Vlad wrote:


VC> I seems quite possible that I will not get Gaussian through our
VC> university which means that I will have to use other programs. In
VC> Paratool tutorial it is stated that GAMESS could be used but there
VC> is not much information given on how to proceed with it. Before

i would strongly recommend to first learn how to use these kind of
code correctly altogether. it is pretty straightforward, but as with
any craft, it takes some time and effort to master it and a powerful
tool in inexperienced hands can create quite a bit of damage (mostly
to the person holding it, btw). ;-)

VC> embarking on using this program I would need reassurance that it is
VC> indeed possible to use GAMESS for this purpose (is there maybe some
VC> progress on developing Paratool compatibility with GAMESS ?) and
VC> maybe some pointers on the differences between using Gaussian and
VC> GAMESS (I am guessing at least the manual input of the QM results
VC> from GAMESS in Paratool).

again, the instructions assume familiary with quantum chemistry
calculations. same as you can run amber or charmm force-field
calculations not only with amber or charmm, but quite a few other
codes (and you should get equivalent results, too), most quantum
chemistry codes can be used interchangably. since most of these
codes go way back, they all have their quirks and "slangy"
documentation, but for standard calculations, they are well tested
and the input is rather similar.

VC> Which of the 3 GAMESS (US, UK, PC/Firefly) to use ? Important
VC> criteria here would be speed, availability of support and previous
VC> work done with any of these programs for preparing NAMD input. I

parametrization requires a lot of manual work and scripting, so
be prepared to get your hands dirty. support is a different
subject. the best support is usually a colleague with experience
in that matter.

VC> have checked a little bit all programs and I only found a limited
VC> number of benchmarks which are favorable to PC/Firefly. So what is
VC> your opinion here ?

the license for gaussian prohibits publishing benchmarks
or comparing its performance to its competitors. there
are plenty other quantum chemistry codes around, that would
do the same. i've recently experienced, that for example
NWChem has excellent parallelization, even on slow networks,
when used with the right set of flags.

VC> I would appreciate a rough idea on the computation resources
VC> required. As my molecules splits naturally by cycles and the largest
VC> of them has about 80 atoms it seems to me that this will be the
VC> largest fragment on which I will have to do single-point QM. Is this

impossible to tell. please keep in mind that most quantum chemistry
calculations have very unfavorable scaling with system size, most
notably based on the number of electrons (not atoms) in the system.

VC> reasonable to calculate on a single Intel P4 2,66 Ghz processor at
VC> the required level of theory ? As I understand the threshold for

you will need quite a bit more *umph* than this sorry little machine.

VC> such calculations is usually 120 atoms, but with what computational
VC> power ? I could live with calculations taking days, but not weeks !

how about months?

VC> Is there a difference in using CHARMM or AMBER charges ? I

YES! if there weren't, people wouldn't want to develop one or
the other. whatever you parameterize, you have to parameterize
to be consistent with the rest of the force field and follow the
published 'recipes'. this is very much a "garbage in, garbage out"
kind of deal. caveat emptor!

VC> understand that they are calculated differently but once calculate
VC> does NAMD cares about this distinction (as it does for example for
VC> parameters file format: either CHARMM or X-PLOR) ? Or is there any
VC> limitation in using one or the other ?

please note that there is a difference between the functional forms,
the parameters (i.e. the force field), and the codes. NAMD contains
the functional forms for both AMBER and CHARMM style force fields
and can read AMBER and CHARMM style parameter files. in principle,
you could specify AMBER parameters also in a CHARMM format file.

VC> Sorry for the long mail and thank you for your support
VC> All the best !
VC> Vlad Chindea
VC> _________________________________________________________________
VC> Drag n? drop?Get easy photo sharing with Windows Live? Photos.

Axel Kohlmeyer
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
If you make something idiot-proof, the universe creates a better idiot.

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:52:17 CST