From: Mayne, Christopher G (
Date: Wed May 28 2014 - 08:17:56 CDT


My suggestion would be to pick your favorite as a test molecule, compare to the CHARMM NA parameter set, run through ParamChem, and refine using ffTK. If you can successfully do this once, then you should have a better idea what you're up against. With regard to the 100+ number -- if your "hypothetical modifications" are amenable to a fragmentation scheme you may just need one (or a few) common nucleobase core(s) and can decorate it with functional groups for which parameters already exist in CGenFF. You may have to refine the linkages, but that is a far cry from a complete parameterization effort.

Christopher Mayne

PS Aaron, sorry for the double send. I forgot to cc vmd-l

"There aren't 100+ common nucleobases in nature."

I'd like to evaluate many small hypothetical modifications. GAAMP would be
great but they tell me that I don't have authorization to get the necessary
passcode, which is a shame. As far as I am aware the ffTK also requires the
use of Gaussian for geometry optimization/water interaction/charge
optimization etc. steps.

As it turns out, my lab does have access to Gaussian, so this may not be a
limitation any longer but if there's a way that smaller changes can be
evaluated in a more automated fashion, I'm all ears!


On Tue, May 27, 2014 at 7:11 PM, Kenno Vanommeslaeghe <<>> wrote:

> On 05/27/2014 10:19 AM, JC Gumbart wrote:
>> If you want to err on the side of speed, then something like ParamChem
>> might work: It should be noted, however,
>> that since nucleobases are already in the charmm36 force field, using
>> ParamChem (which pulls from the CGenFF) is going to be far from ideal.
> Very good advice. Though in the original question, "nucelobases mostly"
> and "100+, not well parameterized in any force field" are at odds. There
> aren't 100+ common nucleobases in nature. The ones that are are well
> supported by the CHARMM Nucleic Acid (NA) force field, and modest
> modifications may be performed with ffTK (or GAAMP). But if you're talking
> about bigger changes, you're quickly drifting outside the coverage of the
> NA FF and into the realm of general heterocycles, which are better covered
> by CGenFF. In that case, the recommendation would be to first get an
> initial guess from<> , then refine it with ffTK (or GAAMP)
> where necessary.
> However, it does presently rely on Gaussian, so if you donít have access
>> anywhere, you might be out of luck.
> Q-Chem now has relaxed PES capability. We've been planning to make a
> Q-Chem variant of our procedure, which we could then present to the ffTK
> developers, but there are always more pressing things getting in the way.
> Of course Q-Chem, just like Gaussian, is commercial software...

Aaron Larsen, Ph.D.
Harvard University Department of Chemistry and Chemical Biology
Harvard Medical School Department of Genetics
Mobile: 617-319-3782
FAX: 617-643-3328