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

From: FyD <fyd.q4md-forcefieldtools.org>
Date: Tue, 13 Sep 2011 17:41:21 +0200

Dear Yun,

> I am worried since I saw no atom names information in the gaussian input
> file, which I used to do geometry optimization.

You only need chemical elements in a Gaussian input.
See the Ante_R.E.D. outputs.

Please, use Ante_R.E.D. 2.0 instead of Ante_R.E.D. 1.x because
Ante_R.E.D. 2.0 presents many new features;
See http://q4md-forcefieldtools.org/REDS/news.php#2

> I guess I should construct
> the gaussian input file from the xxx.pdb file generated by Ante_R.E.D., then
> do my geometry optimization, then submit the gaussian out file together with
> xxx.p2n for charge derivation. So would there be any problem by doing so?

Yes

> In addition, I have only two major conformations for my small test molecule
> after QM geometry optimizations. So I wonder if I could just construct a
> merged p2n file with pdb coordinates from the two major QM-optimized
> conformers? I don't quite understand why R.E.D. server need QM optimization
> output file.

- You need to build a P2N file with 2 conformations.
See http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#3
     http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#EXAMPLE-P2N-FILE
Do not forget the TER word between the two conformations

> Looking through the 'q4md-CD' paper, I feel it kind of counter-intuitive to
> mix GLYCAM04 geometrical parameters with atomic charges from 'Amber'
> strategy, which however, reproduced experimental data so well. But now I
> feel like that the RESP charges come hand-in-hand with the scaling factor,
> and I guess using the amber scaling factors is the major reason that your
> charge-derivation approach is successful.

   type of RESP charges +
               specific 1-4 scaling factors for non-bonding interactions +
                     specific FF dihedral parameters

> Since my molecule (the thio-glycosidic part) has no reference to look for
> experimental data to compare, what would be a simple experiment to carry out
> to validate my parameters? Or just try fitting them to QM calculations?

Yes 'fitting them to QM calculations'

> By the way, what do you mean by " 'recompute' key dihedrals" when taking
> parameters from GAFF? And which are key dihedrals?

For instance this OS-CG-S-CG (or OS-CT-S-CT) dihedral...

regards, Francois


> On Wed, Sep 7, 2011 at 11:57 PM, FyD <fyd.q4md-forcefieldtools.org> wrote:
>
>> Dear Yun,
>>
>> > I basically understand this approach now. But for this
>> Rha-S-CH2-CH(R)-NH-
>> > group, how should I deal with the N terminus? If I simply add an ACE cap
>> to
>> > make it Rha-S-CH2-CH(R)-NH-ACE,
>>
>> Yes, with a trans peptide bond:
>>
>> H
>> Rha-S-CH2-CH(R)-N-C-Me
>> O
>>
>> > then how to set charge constraints?
>>
>> You set a single INTRA-MCC = 0 for the ACE chemical group in the P2N file.
>>
>> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#15
>> (However in this #15 example two capping groups are added, so two
>> INTRA-MCC are set in the P2N file; In your case, you need to set a
>> single INTRA-MCC in your P2N file).
>>
>> > Should I
>> > add a Met or a Hydrogen to O3 on Rha? Because CENTRAL fragments instead
>> of
>> > TERMINNAL fragments are needed here.
>>
>> So you need two L-Rhamnoside building blocks...
>> a perOH-thiomethyl-Rhamnoside & a perOH-O-methyl-Rhamnoside
>>
>> [You could also try a different set of connecting groups between
>> S/O-hemiacetals (instead of S/O-acetals) and methoxy groups (i.e.
>> ether groups instead of hydroxyl groups).
>>
>> > Sorry our library does not have access to the Cieplak et al. paper.
>>
>> ok, I will send you the pdf file to your personal address (I think the
>> license to publish exclude the free distribution of this paper)...
>>
>> > In addition, I noted that GLYCAM06 have exactly the same charges for most
>> > atoms in the same sugar molecule with different linkage, i.e.,
>> > 2-L-alpha-Rhamnose share identical atomic charges with 3-L-alpha-Rhamnose
>> > except O2 and O3. However in amber99sb, every corresponding single atom
>> of,
>> > say TRP, would have different charges from ntTRP. I guess this may have
>> > something to do with the charge-deriving methods, as qwt=0.0005/0.001 in
>> > 99sb VS qwt=0.01 in GLYCAM.
>>
>> This has to do with the charge constraints applied during the charge
>> fitting step.
>>
>> > But then the question would be, should I set charge constraints to most
>> > atoms on the sugar ring except the anomeric carbon and derive charges
>> using
>> > RESP-C2 for the sugar part, which would allow me to avoid the use of the
>> > first Rhamnose? Or just do all charges with RESP-A1A?
>>
>> Well you are right; this is the key ;-)
>>
>> -1- I guess GLYCAM developers will suggest you to use something
>> similar to the RESP-C2 charge model; i.e. MEP computation using the
>> CHELPG algo. and a single RESP fitting stage using the qwt=.01
>> hyperbolic restraint. This means using 1.0 scaling factors for the 1-4
>> non-bonding interactions for the sugar part of your ligand and 1.2/2.0
>> for the other part. Quite complex... considering the size of your
>> ligand and considering that the Connolly surface algorithm is used in
>> MEP computation and a two RESP stage fitting approach is used for the
>> non-sugar part.
>>
>> -2- Considering that L-Rhamnose is the 6-deoxy-L-mannose, I would
>> follow the approach we described .
>> http://www.ncbi.nlm.nih.gov/pubmed/21792425/ for all your ligand and
>> more general all your molecular system (in case of a Galacto
>> configuration I would suggest you to use the GLYCAM approach -1-).
>> This means 1.2/2.0 scaling factors for all the 1-4 non-bonding
>> interactions and using the Connolly surface algorithm in MEP
>> computation and a two RESP stage fitting approach for the entire
>> ligand. Far more simple... However, this approach has obviously to be
>> validated (however as it is requested for the case -1-).
>>
>> regards, Francois
>>
>>
>>
>> > On Sat, Aug 27, 2011 at 12:16 AM, FyD <fyd.q4md-forcefieldtools.org>
>> wrote:
>> >
>> >> Dear Yun,
>> >>
>> >> - When you have a peptide bond within a molecule, R1NH-COR2 you could
>> >> split this molecule into two parts R1NH-ACE & NME-COR2 (ACE = CH3CO;
>> >> NME = NHCH3; are capping groups). You use two INTRA-MCC set to zero
>> >> for these capping groups:
>> >>
>> >> R.E.D.:
>> >> 2 FG2 fragments LEaP
>> >> R1NH-ACE + NME-COR2 --> R1NH COR2 -----> R1NH-COR2
>> >> <=> <=>
>> >> 2 INTRA-MCC = 0
>> >>
>> >>
>> >> - When you have a disaccharide R'-O-R", you could split it into two
>> >> monosaccharides by using an INTER-MCC set to zero between the methyl
>> >> group of the methylglycoside (building-block 1) and the chosen
>> >> hydroxyl group belonging to the second monosaccharide unit:
>> >>
>> >> R.E.D.:
>> >> R'O-Me HO-R" 2 FG1 fragments LEaP
>> >> <-----> ---> R'O R" -----> R'-O-R"
>> >> 1 INTERMCC = 0
>> >>
>> >> by analogy, you could test:
>> >>
>> >> R'S-Me HS-R"
>> >> <-----> ---------------> R'-S-R"
>> >> 1 INTERMCC = 0
>> >>
>> >> With these simples rules you can split your red-blue-red-green
>> >> molecule into 4 building blocks.
>> >>
>> >> These have been defined some time ago; see
>> >> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php
>> >> &
>> >> Cieplak et al. Application of the multimolecule and
>> >> multiconformational RESP methodology to biopolymers: Charge derivation
>> >> for DNA, RNA, and proteins. J. Comput. Chem. 1995, 16, 1357-1377.
>> >>
>> >> With R.E.D. Server/R.E.D. IV you can generate all these fragments into
>> >> a single R.E.D. job by using the corresponding P2N files and by
>> >> defining the correct constraints during the charge fitting step.
>> >> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php
>> >> & http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#29
>> >>
>> >> regards, Francois
>> >>
>> >>
>> >> Quoting Yun Shi <yunshi09.gmail.com>:
>> >>
>> >> > Sorry that I should have attached this model molecule I want to study
>> >> > earlier.
>> >> >
>> >> > As you can see, I divide this ligand into four parts, and the
>> parameters
>> >> > (atomic charges, bonds, angles, dihedrals, impropers) of the red
>> >> (terminal)
>> >> > sugar and green ASP can be obtained from GLYCAM and 99SB respectively.
>> >> >
>> >> > But I have to include the glycosidic oxygen between two sugar rings in
>> >> the
>> >> > blue thio sugar, as this glysicidic oxygen is defined in the preceding
>> >> sugar
>> >> > (the blue thio sugar) instead of the succeeding sugar according to
>> >> GLYCAM.
>> >> >
>> >> > I plan to use the approach -2- you mentioned (so that I don't need
>> >> generate
>> >> > many conformations), but considering my specific case, I wonder why
>> not
>> >> > using
>> >> >
>> >> > Me-O3-L-RhamnoS-Me + MeS-CH2CH(R)NH-COMe
>> >> > <----------------------> <===========>
>> >> > INTER-MCC=0 INTRA-MCC=0
>> >> >
>> >> > , and then remove the Me - Me and MeS - COMe respectively?
>> >> >
>> >> > Since the L-Rha and TRP in my case carry only one modified site
>> compared
>> >> to
>> >> > their standard counterparts, should I include charge constraints (keep
>> >> the
>> >> > atomic charges) for atoms far away from the modified sites? Such as C5
>> >> and
>> >> > C6 in the thio sugar, and the entire indole ring of the modified TRP?
>> >> >
>> >> > Thank you so much!
>> >> >
>> >> > Yun
>> >> >
>> >> >
>> >> > On Thu, Aug 25, 2011 at 11:37 PM, FyD <fyd.q4md-forcefieldtools.org>
>> >> wrote:
>> >> >
>> >> >> Yun,
>> >> >>
>> >> >> > Sorry, I just read the tutorial, that is, "Central fragment of a
>> xxx".
>> >> >> >
>> >> >> > So for the modified (C-terminal reduced) amino acid, should I add a
>> >> MeCO
>> >> >> at
>> >> >> > the N-terminus and MeS at the reduced C-terminus? (considering the
>> >> >> reduced
>> >> >> > C-terminus is used to link glycosidic sulfur atom).
>> >> >> >
>> >> >> > And for the sugar molecule, should I add a Me cap to 3-hydroxyl
>> group
>> >> in
>> >> >> > addition to a methyl group attached to the glycosidic sulfur? (This
>> >> sugar
>> >> >> > molecule would fall into the category of central fragment of my
>> >> ligand)
>> >> >>
>> >> >> ups I overlooked your problem ;-) You do not have any cysteine
>> >> residue...
>> >> >>
>> >> >> Let's re-do the explanations with your terminal molecule/pseudo
>> >> amino-acid:
>> >> >>
>> >> >> -1- you consider the following molecule
>> >> >>
>> >> >> Thio-L-Rhamno-CH2CH(R)NH-COMe ; R = CH2-Ph
>> >> >> <==>
>> >> >> INTRA-MCC
>> >> >>
>> >> >> You only have to set up an intra-molecular charge constraint
>> >> >> (INTRA-MCC keyword) for the COMe capping group (see <==> above).
>> >> >> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#15
>> >> >>
>> >> >> R.E.D. will generate a sm fragment (mol2 file) for you.
>> >> >>
>> >> >> -2- you split your molecule into two building blocks.
>> >> >>
>> >> >> L-RhamnoS-Me + HS-CH2CH(R)NH-COMe
>> >> >> <-------> <==>
>> >> >> INTER-MCC INTRA-MCC
>> >> >>
>> >> >> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#16
>> >> >> or http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#17
>> >> >>
>> >> >> You set up an intra-molecular charge constraint (INTRA-MCC keyword)
>> >> >> for the COMe capping group.
>> >> >> You set up an inter-molecular charge constraint (INTER-MCC keyword)
>> >> >> between the methyl group of L-RhamnoS-Me and the thiol group of your
>> >> >> pseudo terminal amino-acid.
>> >> >>
>> >> >> The approach -2- is once again more complex but more flexible...
>> >> >>
>> >> >> R.E.D. will generate a FG fragment (mol2 file) in the Mol_MM
>> directory.
>> >> >>
>> >> >> Sorry for misunderstanding your problem in my first email.
>> >> >>
>> >> >> regards, Francois
>> >> >>
>> >> >>
>> >> >> > On Thu, Aug 25, 2011 at 9:58 AM, Yun Shi <yunshi09.gmail.com>
>> wrote:
>> >> >> >
>> >> >> >> Thank you very much!
>> >> >> >>
>> >> >> >> I looked at the CD project, and I saw when FFTopDB were
>> constructed,
>> >> the
>> >> >> >> 1-metylated glucose was used to derive RESP charges. So in my
>> >> >> >> thio-glycopeptide case, should I use the sugar molecule with a
>> methyl
>> >> >> group
>> >> >> >> attached to the glycosidic sulfur as well?
>> >> >> >>
>> >> >> >> What about the modified (C-terminal reduced) amino acid? Add one
>> >> more
>> >> >> >> methyl group to the modified C-terminal and acetate to the
>> >> N-terminal?
>> >> >> >>
>> >> >> >> But in the end, what is deposited in the FFTopDB is the residue
>> >> without
>> >> >> the
>> >> >> >> methyl or acetate cap, right?
>> >> >> >>
>> >> >> >> Yun
>> >> >> >>
>> >> >> >>
>> >> >> >>
>> >> >> >> On Wed, Aug 24, 2011 at 11:38 PM, FyD <
>> fyd.q4md-forcefieldtools.org
>> >> >> >wrote:
>> >> >> >>
>> >> >> >>> Dear Yun Shi,
>> >> >> >>>
>> >> >> >>> > I am trying to understand how this works.
>> >> >> >>>
>> >> >> >>> If you look at the data available in the "F-85" R.E.DD.B.
>> project,
>> >> you
>> >> >> >>> will find a x/tLEaP script to construct the CD-based
>> glycopeptides
>> >> as
>> >> >> >>> well as a frcmod file for missing force field parameters with
>> >> comments.
>> >> >> >>> http://q4md-forcefieldtools.org/REDDB/projects/F-85/script1.ff
>> >> >> >>> http://q4md-forcefieldtools.org/REDDB/projects/F-85/script3.ff
>> >> >> >>>
>> >> >> >>> > So instead of combining individual residues in a building-block
>> >> >> manner,
>> >> >> >>> as
>> >> >> >>> > in the assignment of atomic charges for proteins with
>> amber99sb,
>> >> it
>> >> >> is
>> >> >> >>> > recommended to consider the ligand as a holistic molecule when
>> >> >> >>> calculating
>> >> >> >>> > the RESP charge?
>> >> >> >>>
>> >> >> >>> 'recommended'? ;-) ... Personally, I use most of the time the
>> >> building
>> >> >> >>> block approach whatever if the target 'big' molecule is a ligand
>> or
>> >> a
>> >> >> >>> nucleic acid/protein/polysacharide.
>> >> >> >>>
>> >> >> >>> > I am curious that if I could do things in a building-block
>> >> >> >>> > manner since it can potentially decrease a lot of computational
>> >> time
>> >> >> for
>> >> >> >>> > geometry optimization.
>> >> >> >>>
>> >> >> >>> The building-block approach has many advantages:
>> >> >> >>> - it potentially "decreases a lot of computational time for
>> geometry
>> >> >> >>> optimization" as you said.
>> >> >> >>> - it allows rigorously defining the conformation of each
>> >> >> >>> building-block and not to use a conformation more or less
>> randomly
>> >> >> >>> chosen.
>> >> >> >>> - it allows avoiding interactions between charges group during
>> >> >> >>> geometry optimization in gas phase.
>> >> >> >>> - it allows the construction of analogs for the target molecule.
>> >> >> >>> - it allows the construction of oligomers/polymers for the target
>> >> >> >>> molecule.
>> >> >> >>>
>> >> >> >>> However, it also has disadvantages:
>> >> >> >>> - it is complex to set up when one starts, but R.E.D. has been
>> >> >> >>> designed for this approach.
>> >> >> >>> - errors during the charge fitting step are introduced when using
>> >> the
>> >> >> >>> building-block approach; these errors have to be minimized by
>> >> >> >>> correctly selecting the connecting groups between the different
>> >> >> >>> building-blocks. The statistics module available in R.E.D.
>> >> >> >>> Server/R.E.D. IV also helps to localize/minimize these errors.
>> >> >> >>>
>> >> >> >>> > And when it comes to geometrical parameters, we should use
>> GLYCAM
>> >> for
>> >> >> >>> sugar
>> >> >> >>> > part, 99SB for standard amino acids, and GAFF for organic part?
>> >> >> >>>
>> >> >> >>> Yes
>> >> >> >>>
>> >> >> >>> - We only select 'obvious' missing force field parameters from
>> GAFF
>> >> >> >>> (we recompute key dihedrals), and when used we always rationalize
>> >> >> >>> these force field parameters as it was done in the Cornell at al.
>> >> >> >>> force field.
>> >> >> >>>
>> >> >> >>> - In this work, we used Amber scaling factor values for 1-4
>> >> >> >>> non-bonding interactions for all the glycopeptide molecular
>> systems;
>> >> >> >>> i.e. we did not split the system into a peptide and a sugar
>> parts.
>> >> >> >>>
>> >> >> >>> > BTW, could you tell me how to generate multiple conformations
>> with
>> >> >> >>> geometry
>> >> >> >>> > optimization from Gaussian 09?
>> >> >> >>>
>> >> >> >>> You could do a conformational search - although if the
>> >> building-block
>> >> >> >>> approach is used the conformational search is quite simplified...
>> >> >> >>> We also often modify a key dihedral to look for lowest
>> >> minimum/minima.
>> >> >> >>>
>> >> >> >>> To create a P2N file with multiple conformations, see:
>> >> >> >>> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#3
>> >> >> >>>
>> >> >>
>> >>
>> http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#EXAMPLE-P2N-FILE
>> >> >> >>>
>> >> >> >>> To create a QM file with multiple conformations to be used in the
>> >> Mode
>> >> >> >>> 2 of R.E.D. (see
>> >> >> >>> http://q4md-forcefieldtools.org/REDS/popup/popredmodes.php),
>> simply
>> >> >> >>> concatenate the different QM outputs into a single file.
>> >> >> >>>
>> >> >> >>> regards, Francois
>> >> >> >>>
>> >> >> >>> >> Dear Yun,
>> >> >> >>> >>
>> >> >> >>> >> > Is it technically possible to do it due diligence in the
>> first
>> >> >> place?
>> >> >> >>> >> That
>> >> >> >>> >> > is, cut the molecule into three parts as I mentioned before,
>> >> use
>> >> >> >>> GLYCAM
>> >> >> >>> >> for
>> >> >> >>> >> > the sugar part, 99SB for the Thr, and GAFF for modified Phe
>> and
>> >> >> the
>> >> >> >>> >> > thio-glycosidic linkage. And may I then link these parts
>> >> together
>> >> >> >>> using
>> >> >> >>> >> LEaP
>> >> >> >>> >> > ?
>> >> >> >>> >>
>> >> >> >>> >> Concerning the use of GLYCAM + GAFF + Amber99SB you might be
>> >> >> >>> >> interested by looking at the following paper:
>> >> >> >>> >> http://www.ncbi.nlm.nih.gov/pubmed/21792425
>> >> >> >>> >> & its corresponding R.E.DD.B. project .
>> >> >> >>> >> http://q4md-forcefieldtools.org/REDDB/projects/F-85/ + its
>> LEaP
>> >> >> >>> script:
>> >> >> >>> >>
>> http://q4md-forcefieldtools.org/REDDB/projects/F-85/script1.ff
>> >> >> >>> >>
>> >> >> >>> >> This work is about cyclodextrin based-glycopeptide and 1-4
>> >> >> non-bonding
>> >> >> >>> >> interactions in GLYCAM & Amber99SB.
>> >> >> >>> >>
>> >> >> >>> >> Your structure is not a cyclodextrin but this work describe
>> (i)
>> >> how
>> >> >> to
>> >> >> >>> >> derive charges and build force field libraries for new
>> fragments
>> >> by
>> >> >> >>> >> using R.E.D. IV and (ii) proposes new directions concerning
>> the
>> >> >> >>> >> treatment of 1-4 non-bonding interactions in the context of
>> >> >> >>> >> glycopeptides.
>> >> >> >>> >>
>> >> >> >>> >> Finally, in the LEaP script you will find examples how to
>> connect
>> >> >> >>> >> organic, amino-acid and monosaccharide units...
>> >> >> >>> >>
>> >> >> >>> >> regards, Francois



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Sep 13 2011 - 09:00:03 PDT
Custom Search