From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Mon Apr 07 2014 - 12:55:08 CDT

Jim,

Sorry for my delayed reply, I was out of town for a few days.

You have brought up an important question here; unfortunately, I don't have a clear cut answer for you. Many of the characteristics of the fixed point charge model of the CHARMM force field limit its ability to accurately represent certain molecular structures. This is particularly true for molecules with complicated electronic structure, such as extended conjugation, metals, etc. I believe that the best practice is to try to design a fragmentation scheme that includes electronically coupled atoms to the best extent possible while reducing secondary contaminating interactions and heavy atom counts. You might consider posting over at the CHARMM user forums, specifically the "Parameter Discussions" forum, to see if the developers of CHARMM/CGenFF have some specific advice/insights.

Regards,
Christopher Mayne

On Apr 4, 2014, at 2:32 PM, Jim Parker wrote:

Christopher,
   I guess my concern is the "contaminating secondary interactions". The porphyrin core is key to the simulations I'm going to be running. The large aromatic ring structure shares electrons in a non-trivial manner, and I thought the point of re-parameterizing was to accurately estimate the partial charges and bond/torsion strengths that were generated by these secondary interactions. Or am I overanalyzing this system?

I did look at the heme molecule, but the inclusion of an iron atom at the core compared to my metal-free (free-base) porphyrin, should significantly affect the partial charges as well as bond strengths.

The only atom that ffTK didn't find a good geometry for water placement was the hydrogen bonded to nitrogen at the core. I suppose it would not be to terrible to use the pyrrolidine molecule (I have the parameters as part of the tutorial exercise) to get the hydrogen parameters. The others I will use the Gaussian log files generated by ffTK for the full molecule.

>From your post, I would expect that this is reasonable approach.

Cheers,
--Jim

On Fri, Apr 4, 2014 at 11:08 AM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
Jim,

ffTK positions water molecules using a geometry-based algorithm that takes into account the interaction site and atoms bonded to the interaction site atom. As you are seeing in your case, it is susceptible to secondary interactions for larger molecules, or molecules with complex 3D (i.e. non-planar) topologies. In these cases, the user may need to manually manipulate the .gau input file to resolve any secondary interactions, which is why we include the "Load GAU Files" button for users to visually inspect their Gaussian input prior to attempting the optimization calculation.

For the porphyrin case you present here, I don't believe that there is an optimal solution within the context of the cyclic structure. Our suggestion, in accordance with CHARMM and CGenFF protocols, would be to try fragment your larger structure into smaller, simpler structures, parameterize the fragments, and reconstitute your larger molecule from the fragments. There are several reasons for this suggestion; in addition to reducing contaminating secondary interactions, fragmentation reduces heavy atom counts, which is essential for some of the high level QM calculations.

You should also take a look at the toppar_all36_prot_heme.str file in the stream folder of the CHARMM36 distribution (available from Alex MacKerell's website) to see if you can use them as-is, or with minor refinements using ffTK.

Regards,
Christopher Mayne

On Apr 4, 2014, at 5:42 AM, Jim Parker wrote:

> Hello,
> I am attempting to use ffTK to parameterize a porphyrin molecule. All works well until I attempt to find the water interaction energy between a water molecule and the hydrogen on the interior pyrrole units (attached .pdb with molecule).
> The problem is that the line of interaction defined by the ffTK program causes the water to be too close to the hydrogen on the other side of the ring (.gau file attached also). I could arbitrarily rotate the line out of the plane of the molecule, which would "fix" the "atoms too close" error in Gaussian, but I don't know what was the guiding principle for choosing the original line. It appears to be the direction of closest approach to the molecule from "outside", but how critical the charges rely upon that line I don't know.
> Bottomline, I can't run as is. What guidance can you give for updating the .gau file?
>
> I'm using VMD version 1.9.2a35
> And have found the screencasts very helpful, but don't address this issue.
>
> BTW,
> ffTk is a very helpful tool, thanks for your work on this.
>
> Cheers,
> --Jim
> <TPP-opt.pdb><PORF-DON-HA.gau>