Re: [AMBER] again GLYCAM06 + 99SB + GAFF ?

From: Matthew Tessier <matthew.tessier.gmail.com>
Date: Wed, 21 Sep 2011 22:04:38 -0400

Yun,
Angle parameters are usually validated by crystal structure agreement for the minimum value and vibrational spectra for the vibrational constant. Torsions are more difficult so usually these are validated by showing agreement with qm rotational curves for non-parameterization molecules. Also, agreement with NMR properties typically supports the overall fit. QM torsion agreement was already done, so that leaves valence term validation and NMR observables left to test.

Matthew B. Tessier
CCRC - UGA
matthew.tessier.gmail.com
706-542-3508

Yun Shi <yunshi09.gmail.com> wrote:

>Hi Francois,
>
>As in Glycam_06g.dat, the sulfane sulfur is defined as atom type "SM", which
>has some preliminary parameters for thiosugars. But no atomic charge was
>developed for this SM atom in any molecule within the prep file. Anyways, I
>have to validate these parameters.
>
>Could you go into some detail about "follow QM data obtained using high
>theory levels"? Because I thought since we are simulating molecules in
>condensed phase, i.e., in solution, it should be better to calibrate MM
>parameters with conformational behaviors in solution instead of "in vacuum".
>And except plotting potential energy graph with respect to dihedral values,
>I am not sure how to validate angle parameters...
>
>Thanks,
>Yun
>
>On Wed, Sep 21, 2011 at 12:08 AM, FyD <fyd.q4md-forcefieldtools.org> wrote:
>
>> Dear Yun,
>>
>> > As you said in the other email, I have to do validation no matter what
>> kind
>> > of parameters I choose. So I think, at least, I need to validate with the
>> > conformational behavior of the per-OH-thiomethyl-Rhamnoside and the
>> > per-OH-O-methyl-Rhamnoside.
>>
>> If this is _Your_ choice, ok.
>>
>> From a charge fitting point of view, you might look at the errors
>> generated during the fitting step because of the use of the
>> constraints between the building blocks. You could also look at RRMS
>> value although it might not be an issue (See for instance:
>> http://www.sciencedirect.com/science/article/pii/S0010465511002086 ).
>>
>> > But it seems that I need to build a disaccharide version for each case,
>> so
>> > that I could look at average phi and psi dihedral angles, average
>> distances
>> > between atoms on adjacent sugar rings, etc. But for the thiomethyl
>> rhamnose,
>> > what I could find are some nOes between protons on different sugar rings
>> of
>> > the S-disaccharide. So how to calculate normalized nOes from AMBER,
>> > according to different mixing time?
>>
>> At a mixing time value, NOEs can be converted into distances (+/- ...;
>> these distances can be ploted versus the mixing time), and these
>> distances could be followed during MD. right? We used to do that
>> manually - NMR software do that automatically by now... I do not know
>> if the Amber suite has this feature.
>>
>> > And as you noted in the other email, the assignment of atom types for the
>> > atom linking sugar part and amino acid part could be problematic. I think
>> I
>> > will use glycam atom types for the sulfur atom and the methylene
>> attached,
>> > since the validations would be with glycam force field (whether modified
>> or
>> > not).
>>
>> The atom type for the sulfur atom is identical in AMBER and GLYCAM.
>>
>> The main atom type difference between AMBER and GLYCAM is CT versus CG
>> for the sp3 carbon atoms.
>>
>> See http://q4md-forcefieldtools.org/REDDB/projects/F-85/script1.ff
>>
>> for instance a sugar fragment:
>> # The central fragment - allows polymerisation of per-OR a(1,4)Glc (R
>> = Me, Bn, Ac or Bz)
>> # MGR = loadmol2 1-MGR.mol2
>> MGR = loadmol2 tripos8.mol2
>> set MGR name "MGR"
>> set MGR head MGR.1.C4
>> set MGR tail MGR.1.O1
>> set MGR.1 connect0 MGR.1.C4
>> set MGR.1 connect1 MGR.1.O1
>> set MGR.1 restype saccharide
>> set MGR.1 name "MGR"
>> # set FF atom types
>> set MGR.1.C1 type CG
>> set MGR.1.H1 type H2
>> set MGR.1.O1 type OS
>> set MGR.1.C2 type CG
>> set MGR.1.H2 type H1
>> set MGR.1.C3 type CG
>> set MGR.1.H3 type H1
>> set MGR.1.C4 type CG
>> set MGR.1.H4 type H1
>> set MGR.1.C5 type CG
>> set MGR.1.H5 type H1
>> set MGR.1.O5 type OS
>> set MGR.1.C6 type CG
>> set MGR.1.H61 type H1
>> set MGR.1.H62 type H1
>>
>> a peptide fragment:
>> # Succinamide fragment
>> # SCC = loadmol2 6-SCC.mol2
>> SCC = loadmol2 tripos13.mol2
>> set SCC name "SCC"
>> set SCC head SCC.1.N1
>> set SCC tail SCC.1.C4
>> set SCC.1 connect0 SCC.1.N1
>> set SCC.1 connect1 SCC.1.C4
>> set SCC.1 restype protein
>> set SCC.1 name "SCC"
>> # set FF atom types
>> set SCC.1.N1 type N
>> set SCC.1.H1 type H
>> set SCC.1.C1 type C
>> set SCC.1.O1 type O
>> set SCC.1.C2 type CT
>> set SCC.1.H21 type HC
>> set SCC.1.H22 type HC
>> set SCC.1.C3 type CT
>> set SCC.1.H31 type HC
>> set SCC.1.H32 type HC
>> set SCC.1.C4 type C
>> set SCC.1.O4 type O
>>
>> > It seems that the validation of thioglycosidic parameters in my case
>> would
>> > be very "indirect" using the S-disaccharide, but is there a better way?
>>
>> I did not check the literature. You could follow QM data obtained
>> using high theory levels.
>>
>> > By the way, I only have access to amber11, which did not come with
>> glycam04
>> > data files, which although I can get from the glycam website. But how can
>> I
>> > tell LEaP to use a specific data version such as Glycam_04k.dat (is this
>> the
>> > one you used with amber10)?
>>
>> I will ask here, and send you the GLYCAM 2004 parameter file we used.
>> Do not worry we also got lost between all these GLYCAM versions. The
>> parm94, parm96, parm98, parm99, 99SB etc... nomenclature is far more
>> simple/fully justified: one paper - one parameter file.
>>
>> I am not sure you should use this GLYCAM 2004 parameter file in your
>> case (to be checked). Cyclodextrins are particular oligosaccharides as
>> they are cyclic (may be this is the main reason why GLYCAM 2006
>> display these limitations). In general, this is advised to use the
>> last version of a force field; although better double checking what
>> people suggest.
>>
>> Personally, I would be interested in knowing if applying Amber scaling
>> factors for 1-4 non-bonding interactions to all your molecular system
>> works ;-)
>>
>> regards, Francois
>>
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>_______________________________________________
>AMBER mailing list
>AMBER.ambermd.org
>http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 21 2011 - 19:30:02 PDT
Custom Search