From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Thu Jun 13 2013 - 15:55:52 CDT

I'm not familiar with the differences in methodology for identifying analogous structures. You might see:

J. Chem. Inf. Model. 2012, 52, 3144–3154
J. Chem. Inf. Model. 2012, 52, 3155–3168
J. Comp. Chem. 2011, 32, 2359–2368

It is clear, however, that ParamChem and SwissParam assign existing parameters from different sources. ParamChem assigns parameters from a database of molecules that have already been parameterized for CHARMM and CHARMM-compatible (i.e., CGenFF) parameters. SwissParam, however, provides vdW terms from analogous CHARMM atom types, but takes bonded terms (bonds, angles, dihedrals) from MMFF. It is my opinion that it is generally a poor idea to combine parameters from different force fields, particularly in the case of an empirical FFs with different parameterization philosophy, as they tend to be highly tuned for self consistency. Without running calculations for specific test cases, it is difficult to make a general statement about comparing and/or combining the output from these two resources.

Christopher Mayne

On Jun 13, 2013, at 11:59 AM, Peterson J wrote:

Continue talking about the analogical parameterization of the novel molecules, is it a common procedure, if not at least meaningful, to combine the parameters obtained from two different parameterizing tools for instance Paramchem and Swissparam if they use different background databases?
How similar are the parameters obtained from these tools?

-Peterson

On Tue, May 21, 2013 at 4:55 PM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
No, neither I nor ffTK will check ParamChem output for you. As a user, I can tell you that ParamChem reports a "penalty" indicating analogy to a molecule or molecular fragment with existing parameters (of implied quality). But I am not involved with developing ParamChem, so you should read the authors' papers for more information.

On May 21, 2013, at 4:38 PM, Peterson J wrote:

Thanks, but do you check if the parameters obtained are reasonable or not?

Peterson

On Tue, May 21, 2013 at 3:46 PM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
Peterson,

You might also check out ParamChem for fexofenadine. It could be a good place to get reasonable starting parameters that could be refined using ffTK.

Christopher

On May 21, 2013, at 2:24 PM, Peterson J wrote:

Thank you very much for your explanation. I will spend time looking into the CGenFF to plan a scheme before I decide my QM run.

-Peterson

On Tue, May 21, 2013 at 1:05 PM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
Please take a closer look at the ffTK documentation website. It provides links to relevant materials that are essential for understanding parameterization. ffTK does not blackbox the process of parameterization; in fact, it does the exact opposite. Therefore, it is essential that you have a general understanding of the process to effectively use ffTK and generate reasonable parameters.

If your compound of interest contains substructures that are highly analogous to existing fragments within the CGenFF standard distribution, we recommend that you retain those parameters, at least for a first pass parameterization. As I mentioned previously, if you are parameterizing from scratch, then the typing scheme is completely up to you.

We intentionally choose pyrrolidine to use as our example molecules because:
1) the molecule exists in the standard CGenFF distribution. As I mentioned previously, by using the same typing scheme, it allows us and users to directly compare our results to the existing parameters.
2) pyrrolidine is a complicated molecule, and these complications are nicely described in Vanommeslaghe et al 2010, which is why we direct users to this paper and provide a link just above the screencasts.

My advice for fexofenadine is to look through CGenFF and see what parts are already parameterized. Figure out what pieces you are missing, and design a scheme to fragment the structure into smaller pieces for parameterization. QM calculations on the full structure will be complicated and extremely computationally costly.

Regards,
Christopher Mayne

On May 21, 2013, at 11:35 AM, Peterson J wrote:

But how do you find these atom types? Does learning more about CGenFF and its atom typing scheme help a lot to understand the tutorials?
What if I'm working on a completely new chemical compound for which no published values?
I'm actually working fexofenadine and how would you advice me to find the atom types.

Thanks,
Peterson

On Tue, May 21, 2013 at 7:59 AM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
I dropped NAMD-L from the cc list; as I said, this is a VMD issue.

NG3C51 is the atom type, not the atom name. Since the tutorial shows a full parameterization from scratch, the atom types can be anything. In this case, for convenience I used the same atom typing scheme that is used by CGenFF to allow for easy comparison to the published values.

Regards,
Christopher Mayne

On May 20, 2013, at 11:48 PM, Peterson J wrote:

[https://mail.google.com/mail/u/0/images/cleardot.gif]
Hi,

I just started following the screencast tutorials available for ffTK. I'm not quite clear about renaming the added N atom from N1 to NG3C51 and all other subsequent namings for C and H atoms. How are they named? Any background information about the naming that is missing in the tutotrial?

Thanks
[https://mail.google.com/mail/u/0/images/cleardot.gif]

On Mon, May 20, 2013 at 2:32 PM, Mayne, Christopher G <cmayne2_at_illinois.edu<mailto:cmayne2_at_illinois.edu>> wrote:
Peterson,

Since you're question is primarily regarding a VMD plugin, it is not particularly appropriate to cross post on NAMD-L.

The web-based tools that you have mentioned differ in a very important way--they provide parameters based on analogy to molecules or molecular fragments that have already been parameterized and have been incorporated into the backend database. ffTK, in contrast, aides users in developing parameters directly from first principles and the workflow outlined for CGenFF. The primary advantage of the web-based services is their speed and ease of use. If you have a molecule that is highly analogous to something that has been worked out, then the results are generally pretty good. You should also note that ParamChem in particular indicates that users should assess the resulting parameters to determine if they are reasonable, and refine them if needed. In addition to a full parameterization from scratch, ffTK is suitable for conducting this refinement, and includes numerous metrics to assess parameter performance. The trade off, however, is that parameterization is a time consu!
 ming and frequently non-trivial process (we've tried to make it as streamlined as we can), and one should familiarize themselves with the principles underlying parameterization (e.g. CHARMM/CGenFF). The ffTK documentation website provides some information and links to relevant resources.

Regards,
Christopher Mayne

On May 20, 2013, at 1:03 PM, Peterson J wrote:

> Hi VMD and NAMD users,
>
> I'm very new to VMD and NAMD. I'm now in the process of parameterizing a ligand molecule. As I came across various a few web tools like paramchem, swissparam and so on. I have also seen VMD providing a plugin called Forcefield Toolkit calculating parameters using Gaussian and preparing the files for MD run using NAMD.
>
> I would like to get a few suggestion points on which one to use and the advantages and disadvantages over one another.
>
> What if I use one of the mentioned webtools instead of ffTk that use lengthy QM calculations to obtain the parameters?
>
> Thanks in advance for the suggestions.
>
> -Peterson