Re: Problems to parametrize molecules

From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Fri Apr 17 2015 - 11:16:16 CDT

Evandro,

Reading over the message thread, it's a little difficult to follow *exactly* what is going wrong, so let's take a step back here. There are two steps to getting simulations that include small molecules to work 1) generate a proper PSF/PDB file pair and 2) acquire appropriate parameters for any missing terms. The original error report is likely due to a problem in the PSF generation, since:

FATAL ERROR: UNABLE TO FIND BOND PARAMETERS FOR CG301 CG301 (ATOMS 22 22)

 is telling you that the atom is bonded to itself. So let's address problem #1 first.

I'm unfamiliar with CHARMMGUI output for small molecules, but have extensive experience using ParamChem. When you upload a PDB or MOL2 (mol2 is preferred because it contains explicit bond orders useful in typing by analogy), it will return a CHARMM Stream file. The str file is a combination of topology and parameters. This stream file, along with a pdb of your molecule should work as expected with PSFGEN, to yield a properly formatted PSF. Make sure that you are using VMD 1.9.2, as PSFGEN used to have trouble with extended atom types like those used in CGenFF, which is fixed. PSFGEN can be a little tricky, so pay attention to your input and carefully read any output, especially warnings or errors. Here is an example of my commands:

package require psfgen
resetpsf
psfcontext reset
topology top_all36_cgenff.rtf
topology my_paramchem_output.str
segment L { first none; last none; pdb my_ligand.pdb }
coordpdb my_ligand.pdb
regenerate angles dihedrals
writepsf my_ligand.psf

Regardless of where/how you get your PSF/PDB file pair, you should always load the resulting structure into VMD and give it a thorough going over; the TopoTools commands are very useful for this. ffTK also has some atom labeling functions that I used to inspect atom names, types, charges, etc. (see ffKT -> Opt. Charges -> Input -> Label Atoms Menu).

As for acquiring appropriate parameters, problem #2, this is a much more complicated issue that is discussed elsewhere. There are a variety of places to obtain them, e.g., ParamChem, MATCH, ffTK etc; however, I strongly urge you to pay attention to penalty scores (if provided) and consider performing some form of validation or refinement (e.g., ffTK). Once you have parameters, you need to use them correctly. Errors such as:

Unable to find bond parameters for CG311 NG311 (Atoms 1 2)

is telling you that this parameter is missing. The ParamChem webserver gives you two options 1) it returns only parameters which are not part of CGenFF standard parameter set, or 2) All parameters (i.e. novel and standard) required for the molecule. My guess is that you selected option #1 and then didn't include the standard CGenFF parameter file in your NAMD configuration file.

Hopefully this information is helpful. I suggest that you revisit your work, pay attention to the errors that are emitted by PSFGEN/VMD/NAMD, and carefully think about your workflow.

Regards,
Christopher Mayne

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:21:48 CST