You need to look at the scores that paramchem.org gives you.  If there are large penalties associated with the various parameters then you need to either dive in deeper and learn more or try amber and the gaff forcefield for your small molecule.

Brian J Bennion and Stéphane Abel,

Thank you for kind response. I have gone thorugh the parametrization procedure as mentioned in CHARMM cgenff tutorial. What the charmm-cgenff does not recognize (high penalty scores) are - 2 partcular atoms and their corresponding partial charges(2), bonds (3) angles(3), and dihedrals (2). Should I do a MP2 optimization of the block which does contain the unparametrized part and assign MP2 charges (little bit modified as CHARMM guidelines) and some scans of bonds, angles, dihedrals and feed the parameters into the parameter file generated by cgenff? Please help.

Definitely Sir. I'm attaching the paramchem output file.

Judging by the atom types and connectivity, I am piecing together that you have a 6-membered heteroaromatic ring with a thiocarbonyl that is causing the problems?  That would explain the high penalties.  The only comparable model compound we have in the CGenFF parameter set is a 5-membered ring with very different elements in the ring.  Hence the high penalties; the analogy there is not great.  The good news is that the penalties are not that horrible (in the grand scheme of things) and they are localized to the thiocarbonyl group.  So yes, you would need some QM and some refinement.

For the bonds, we typically look at vibrational frequencies (e.g. MOLVIB module in CHARMM), for angles we look at 3-point scans, and dihedrals are scans in 15-degree increments.  There is a CGenFF tutorial online that goes over all of that.  The charges can be refined by targeting QM water interaction energies, e.g. single water molecules placed at configurations that would arise via hydrogen bonding and/or lone pair interaction sites.  Adjust the charges until you get good agreement with the QM-optimized distance and energy values.

Sirs,

I have performed few MP2 single point calculations placing water molecules in different positions. I got the interaction energies with water. Now while a recalculate the interaction energy with mm partial charges and CHARMM force field should I calculate electrostatic interactions with all of the atoms or only the atom nearest to water? Should I take all of the atoms of water molecule (H, O, H) too?

My next question is how should I calculate lennard jonnes potential contribution while calculating water interaction energy with mm parameters?

The interaction energy is the total interaction energy, but you should not be targeting a simple single-point energy.  The water interactions should be optimized at the MP2/6-31G* level and that energy recorded, otherwise it is simply an arbitrary placement of water.  Then, you should perform an MM energy minimization to see if you arrive at the same result.  If not, tune the charges accordingly.

The LJ component of the energy is a product of the atom types, which you almost certainly should not be changing at this point, unless it is impossible to get good agreement.  If you have to introduce new atom types, the amount of work you have to do becomes orders of magnitude larger.

Justin Lemkul,

Thank You very much sir. I'm trying to get the parameters following the way you suggested (running md in NAMD). I have also gone through CHARMM cgenff tutorials in details. Only problem is I don't have a CHARMM license yet to run the scripts.

Hi, 

There is never a way around parameterizing. With these sites you are going guess work which maybe very far from reality. You may end up working for months and generate results that are meaningless. So what you need to do at this stage is to do QM calculations which may take a while but it will save you a lot of work later on. 

From your other post, since you have novel base pairs I am guessing your molecule of interest would be within 50 Atoms. If you don't want or can't do QM calculations, you can try ATB here: http://compbio.biosci.uq.edu.au/atb/

Read its manual carefully - it works best for 20-30 atom systems. It will do the QM calculations for you and give you parameters you can trust. (maybe - but its the best you have).

Good Luck

Best,

/A

Note that ATB is for use with GROMOS force fields, which are not as good as the CHARMM force field for nucleic acids.

Hey Justin, 

He mentioned in the other post that he was using NAMD. NAMD has the ability to work with gromacs forcefield. Also - on the overall - better to have properly worked out parameters with QM and coupled with gromacs then to be way-off with guess work and coupled with CHARMM Forcefield. DNA interactions are studied under GROMACS, but yes I do agree with you that CHARMM parameters are better for Nucleic Acids especially with CHARMM36.

Best,

/A

It is good that you have used the paramchem website.  The penalties show you where you need to concentrate your efforts.  If garbage goes into the simulation you can't expect that garbage will not also come out of the simulation.

I highly recommend you learn how to use FFTK which will allow you parameterize your non standard residue.  Use the Charmm forums to discuss the specifics of your residue with the experts there.

Dear Sirs,

                I am learning a lot from your answers. I have calculated an initial set of parameters with basic level of validation for my residue. My molecule is an unnatural base terminated by methyl group from sugar-phosphate backbone. Now I want to link my molecule with backbone again. The sugar and phosphate part is parametrized in CHARMM. It seems that I need to manually edit the topology and parameter files of CHARMM to link my unnatural base with sugar-phosphate backbone. This is the problem troubling me most.