Re: ffTK Without Using Gaussian

From: JC Gumbart (gumbart_at_physics.gatech.edu)
Date: Tue Mar 20 2018 - 14:37:35 CDT

Thanks for the very thorough response Fotis!

I would just add that an institution being banned doesn’t preclude you from using it when available at a supercomputing center, for example. I think you just have to sign some papers that you won’t compete with Gaussian, you won’t help others to compete with them, and that you won’t even think about publishing a benchmark.

Best,
JC

> On Mar 20, 2018, at 1:59 PM, Fotis Baltoumas <fbaltoumas_at_biol.uoa.gr> wrote:
>
> Hello Alex,
>
> Unfortunately, fftk depends heavily on Gaussian, since all required QM steps are done with it.
>
> However, the vanilla version of CGenFF parametrization and optimization, which also utilizes QM, actually uses a CHARMM module (MOLVIB) for some of the QM steps rather than Gaussian (CHARMM the MD simulation suite, that is, not CHARMM the force field). Note that from version c40 and onwards, CHARMM in its non-parallel version ("charmm") is given free of charge. So in principle, you could use CHARMM itself for the parametrization, following the tutorials in the CGenFF page: http://dogmans.umaryland.edu/~kenno/tutorial/
> The downside, of course, is that you need to become familiar both with the CHARMM scripting syntax and with the underlying mechanics of CGenFF and CHARMM parametrization (or find someone with expertise to help you). Plus, you won't have all those nice, automated perks that fftk offers.
>
>
> Alternatively, you can look to some of the other CHARMM parametrization services that are available out there. A really good option is GAAMP (http://gaamp.lcrc.anl.gov/ ) which will do some QM optimization for you. However, it is not open to the public and you first need to get in contact with the team behind it to obtain access.
>
> Other, freely available options include MATCH (http://brooks.chem.lsa.umich.edu/index.php?matchserver=submit ) and SwissParam (http://www.swissparam.ch/ ). However, both have their limitations; MATCH does not use QM itself and therefore may also need optimization. As for SwissParam, although it performs some optimization, it is generally recommended for docking or simple energy calculations rather than full MD simulations. Plus, the whole procedure they use (mixing the Merck forcefield with CHARMM22) may not be fully compatible with CHARMM36.
>
> As to whether these QM steps are necessary or not depends, to some degree, on how you have set your initial topology/parameters for your molecule. Have you made them yourself, using the same procedure as the fftk tutorial, or have you gotten them from CGenFF/ParamChem?
>
> If the latter applies, first check what penalties CGenFF has assigned to the charges and the bonds/angles/dihedrals. If most (preferrably all) penalties are low enough (10 or lower are preferred, but if you have 15-20 it could still be fine), you could just use the initial parameter scheme "as-is". If only a few bonded parameters have high penalties, you could try fixing them yourself through analogy with other similar molecules, or through trial and error (assuming you have anything on your molecule to act as reference).
>
> If the overall quality of your molecule's parametrization is low, I'm afraid you WILL need QM to get meaningful results.
>
>
> Hope I helped,
> Fotis Baltoumas
>
> On 03/20/2018 05:49 PM, Alex Saad-Falcon wrote:
>> The force-field toolkit uses Gaussian for several of the steps. However, my institution is banned from using Gaussian. Can I safely skip these steps, or are they necessary? Are there any other ways to parameterize a small organic molecule for NAMD?
>>
>> Thanks,
>>
>> Alex
>
> --
> *******************************************
> Fotis A. Baltoumas
> Bioinformatics Postgraduate Programme
> Section of Cell Biology and Biophysics
> Department of Biology, National & Kapodistrian University of Athens
> Panepistimiopolis, Athens 157 01, GREECE
> --------------------------------------
>
> email : fbaltoumas_at_biol.uoa.gr
> http://biophysics.biol.uoa.gr
> http://bioinformatics.biol.uoa.gr
> *******************************************
>

This archive was generated by hypermail 2.1.6 : Sun Sep 15 2019 - 23:19:31 CDT