From: Bryan Roessler (roessler_at_uab.edu)
Date: Tue Sep 15 2015 - 12:06:12 CDT
Thank you all for the comments. Here is the methodology I am currently
employing, could someone please comment on whether or not this is a good
way to parameterize a novel residue without direct access to Gaussian?
1. Begin with separate pdbs of the novel residues. The N-terminus and
C-terminus novel residues have an additional methyl and amino fragments
attached, respectively, in order to parameterize the peptide bond
2. Use Gaussian through pyRED to perform geometry optimization and charge
assignment using M2/ESP-A1 level of theory.
3. Load the optimized geometry into Discovery Studio and save as a .mol2
file (for CGenFF)
4. Manually assign atomtypes in the .mol2 file for CgenFF compatibility.
5. Upload to CGenFF to derive analogous forcefields.
6. Manually edit the forcefields to remove the amino and methyl atoms in
order to attach the interior peptide sequence and assign their charges to
the carboxy or N atoms.
7. Rerun CGenFF and make sure that the penalty scores for the new peptide
bonds are not high.
Using this methodology, my penalty scores went from 50-100 to a highest
penalty of 3 due to the geometry optimization. I know that Tomek mentioned
that some people then overwrite the partial charges using the Gaussian
output, but I don't think that's valid for CHARMM in this instance,
considering that we are not characterizing the water-atom interaction
through pyRED. The CGenFF analogies should be much more accurate.
*Bryan Roessler | Graduate Research Assistant*
UAB | The University of Alabama at Birmingham
Knowledge that will change your world
On Tue, Sep 15, 2015 at 11:15 AM, Bennion, Brian <bennion1_at_llnl.gov> wrote:
> I too have wondered about this apparent conundrum.
> A quick search of the gamess website at Iowa state doesn't specifically
> reveal whether the functionality below is present or not.
> Perhaps a query to their forum is in order?
> -----Original Message-----
> From: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] On
> Behalf Of Mayne, Christopher G
> Sent: Monday, September 14, 2015 7:38 PM
> To: Peter Freddolino
> Cc: <namd-l_at_ks.uiuc.edu>; Bryan Roessler
> Subject: Re: namd-l: Parameterizing a novel peptide
> It's been awhile since I've gone through the specifics, but to my
> knowledge, GAMESS cannot do dihedral scanning to construct relaxed
> potential scans. I also don't recall if GAMESS supports calculation of the
> Hessian using redundant internal coordinates, which is a required for
> optimizing bonds and angles. So, I don't think that it's just a manpower
> issue; there are some missing features.
> Christopher Mayne
> On Sep 14, 2015, at 9:08 PM, Peter Freddolino wrote:
> > Hi Christopher,
> > I've always wondered about this a bit - what is gamess missing that
> > prevents it from being used as the main qm engine in fftk? Aside from
> > writing differently formatted config files it should have all of the
> > needed features, no? I understand if this is just a manpower issue,
> > but if it is, maybe someone in the community who has need for the
> > feature could be interested to help... Best, Peter
> >> On Sep 14, 2015, at 9:48 AM, Mayne, Christopher G <cmayne2_at_illinois.edu>
> >> Bryan,
> >> I can't comment on parameterizing for AMBER, as I currently only work
> with CHARMM. In our opinion, the preferred method by far is a combination
> of the CGenFF Program (formerly ParamChem) to get a set of initial
> parameters (and to perform atom typing for you), followed by validation
> and/or refinements baesd on the penalty scores. We recognize that some
> users may not have access to Gaussian, and are attempting to address this
> limitation; however, support for an expanded set of QM software likely
> won't be available anytime soon (there are lot of software limitations to
> work around).
> >> As for your technical question, my advice is to retain as much of the
> standard protein force field parameters as possible. I would use the
> divide and conquer strategy to parameter only the missing fragments. The
> problem then becomes largely a translation issue between the CGenFF atom
> types (if that's what you've parameterized the fragments in) and the CHARMM
> atom types (the peptide backbone). Both the CGenFF and CHARMM
> methodologies use the approach to computing partial atomic charges, so they
> should be compatible.
> >> Regards,
> >> Christopher Mayne
> >>> Date: Sat, 12 Sep 2015 01:16:27 +0200
> >>> From: =?UTF-8?Q?Tomek_St=C4=99pniewski?= <tm.stepniewski_at_gmail.com>
> >>> Subject: Re: namd-l: Parameterizing a novel peptide
> >>> Hmm,
> >>> don't know anything about antechamber, most of the people I know use
> >>> ParamChem (and they get published), but overwrite the charges they
> >>> get from Gaussian, but if You don't do that I suppose nobody's going
> >>> to jail best of luck:-)
> >>> 2015-09-12 0:14 GMT+02:00 Bryan Roessler <roessler_at_uab.edu>:
> >>>> Hello,
> >>>> I know that the currently recommended way to parameterize a novel
> >>>> residues 'round these parts is to use the FFTK, unfortunately I do
> >>>> not have access to Gaussian.
> >>>> Is using antechamber and the built-in SQM MM program and GAFF
> >>>> considered 'good enough' for publication (assuming I'm using Amber
> >>>> FFs)? Or is using pyRED server with Gaussian (and the appropriate
> >>>> composite method) for Amber and/or CHARMM preferred?
> >>>> I was excited to try to create GAMESS input files with VMD, but it
> >>>> does not appear that the functionality is quite complete.
> >>>> Lastly, I have technical question regarding parameterization. My
> >>>> peptide of interest is 5 residues long: three standard amino acids
> >>>> flanked on the N and C termini by a carboxybenzyl protecting group
> >>>> and a reactive sulfur compound (both peptide bonds), respectively.
> >>>> Ideally I would like to use the standard forcefields for the
> >>>> standard amino acids and only parameterize the novel flanking
> >>>> residues. I have no problem parameterizing the novel residues as
> >>>> stand-alone compounds, but then of course I am missing the peptide
> >>>> bond parameters. How can I parameterize the peptide bonds but still
> >>>> use the standard forcefields for the 3 amino acids in the middle of
> >>>> the peptide? Do I need to parameterize the entire peptide and then
> >>>> cut and paste the peptide bond parameters? I fear that may
> unfavorably mix partial charges. I'm familiar with doing this in CGenFF but
> less so in Antechamber.
> >>>> Thanks for your help,
> >>>> Bryan
> >>>> *Bryan Roessler | Graduate Research Assistant* UAB | The University
> >>>> of Alabama at Birmingham *uab.edu/cmdb <http://uab.edu/cmdb>*
> >>>> Knowledge that will change your world
> >>> --------------------------
This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:18 CST