Re: [AMBER] parameters for Lys-sugar P2N file

From: FyD <fyd.q4md-forcefieldtools.org>
Date: Mon, 26 Mar 2012 14:47:17 +0200

Dear Urszula,

1) The starting point is which charge model you wish to use:
See http://q4md-forcefieldtools.org/REDS/popup/popkeyword.php
If you decide to use R.E.D. or R.E.D. Server
. the charge model designed to match that used by GLYCAM is RESP-C2
   and/or
. the charge model used by AMBER is RESP-A1 (when using R.E.D.-III.x)
or RESP-A1A (when using R.E.D. Server/R.E.D. IV).

this is important you are aware of the differences in the two approaches:
. RESP-C2:
   . CHELPG algo used in MEP computation,
   . a single RESP stage,
   . each hydrogen atom connected to a sp3 carbone bear a charge value
= 0; this could be achieved by using intra-molecular charge
constraint(s).
   . 1-4 electrostatic interactions are multiplied by a scaling factor
that is set to 1.
   . used for sugars/GLYCAM

. RESP-A1:
   . Connolly surface used in MEP computation,
   . two RESP stages
   . No intra-molecular charge constraint is applied to hydrogen atoms
connected to a sp3 carbon atom.
   . 1-4 electrostatic interactions are multiplied by a different
scaling factor (see "scee" in the Amber manual).
   . used for proteins and nucleic acids/AMBER

A first difficulty in your case is to choose between RESP-C2 and/or
RESP-A1 for your amino acid-sugar unit. Indeed, splitting your amino
acid-sugar into an amino-acid and a sugar parts might be difficult
considering the specific connection you have between these two
'parts'. So you will have to decide if (i) you split your systems into
two parts or not, and (ii) which charge model(s) (RESP-C2 and/or
RESP-A1) you want to use.

-a) Amber and Glycam developers have designed an approach to make the
AMBER and GLYCAM approaches compatible; i.e. to make 'co-exist' the
two approaches. So they will suggest you to use their way...

-b) We have developed a q4md-fft approach for glycopeptides; I think
this approach is far more simple than this hybrid approach -a).
Considering that your sugar part is a Glucose I think you can adopt
'our' approach -b); i.e. use RESP-A1 for the entire amino-acid-sugar
unit.
See http://www.ncbi.nlm.nih.gov/pubmed/21792425

However, here, I think the best bet for you is to follow _your_ own
approach and prove it works. You can obviously follow pieces of advice
from developers, but the best way to learn is to create your own way ;-)

> I prepared the p2n file for RED based on the Tutorial 3, however I
> am not quite sure how to specify in this case the part ...
>
> REMARK REORIENT 5 ?

if you use R.E.D. Server/Ante_R.E.D. 2.0 this should be added
automatically by Ante_R.E.D. 2.0. this idea here is only to control
the molecular orientation used in MEP computation; I suggest you to
use pairs of mol. orientations defined by sets of three atoms of
opposite order; i.e. 1 2 3 | 3 2 1 for instance.
See http://q4md-forcefieldtools.org/Tutorial/Tutorial-1.php#REORIENT

> and
>
> REMARK Atoms to be kept in the final Tripos .mol2 file (to keep some
> compatibility with
> REMARK other force field libraries already available in the AMBER
> force field topology database):
> REMARK INTRA-MCC -. | | Keep
> REMARK INTRA-MCC ... | | Keep
> REMARK INTRA-MCC .... | | Keep
> REMARK INTRA-MCC -... | | Keep

See above. In particular, if you decide to use the RESP-A1 charge
model you do not need to set the charge values of each hydrogen
connected to a sp3 carbon atom to zero.

2) then, the second important point is the conformation(s) to be
involved in MEP computation, and more generally in charge derivation;
in your case, the key points are:
-1- conformation(s) for the phi/psi dihedrals of the dipeptide backbone
  -> you could adopt an experimental/input data or 2 conformations in
the extended/helicoidal conformations
-2- conformation(s) for the dihedrals of the dipeptide side chain
  -> you could adopt an experimental/input data or the most stable one in QM
-3- conformation(s) for the omega dihedral of Glc
  -> you could adopt an experimental/input data or select the most
observed population(s)
-4- conformation(s) for the hydroxyl groups of Glc
  -> you could used geometrical constraints to prevent the formation
of intra-molecular hydrogen bonds (which are over-stabilized in
gas-phase)
    -> you can find examples of optimized geometries by QM for
alpha/beta-Glc in R.E.DD.B.

In the GLYCAM approach, weighting the conformations in charge
derivation is based on the populations observed in MD simulations.
Selection of the conformations to be involved in charge derivation is
efficient/original, but the approach remains complex, and quid of the
initial set of charge values derived before MD simulations?

In the Amber approach, only a set of representative conformations
obtained from QM geometry optimization or from experimental data is
involved in charge derivation.

> Could you suggest me something? Do I need specify anything else there?

See above. Let us know if you need more information.

regards, Francois

PS - It looks like the dipeptide you have constructed has a beta-Gal
unit and not a beta-Glc one...
    - The conformation of the ACE capping group is weird ;-)
    - the HO6 hydroxyl group of the sugar unit points in the direction
to the NH3+- group; this type of structure is very unlikely...
    - CT1 and CT2 atom names are only to be used in RESP input
generation; i.e. in the first column of atom names and not in the
second one.
    - you might adopt more conventional atom names for the sugar atom
names in the 2nd column of atom names.


> Dear Lachele,
>
> I think dihedral force field development is performed _after_ charge
> derivation.
>
> In this case, one should (i) derive the charges and build the FF
> library(ies) for the new molecular fragment(s), define the FF atom
> types etc... (and may-be for models as well), and then (ii) perform MD
> simulation to match experimental/QM data, and/or fit MM to a QM
> profile. Indeed, non-bonded interactions are used during the dihedral
> profile fit...
>
> regards, Francois
>
>
>> There is agreement in the group that the linkage position could be
>> hard to model well. Of course, this depends on how well you need it
>> modeled, but for the things we would normally do, we would take some
>> time developing and validating, especially for something like this.
>>
>>
>> On Fri, Mar 23, 2012 at 2:01 PM, Lachele Foley (Lists)
>> <lf.list.gmail.com> wrote:
>>> My concern isn't so much developing the charges so much as is finding
>>> good parameters for the rest of the force field in the linkage
>>> environment. The anomeric center is complex to model all by itself.
>>> If you add a positively charged group so close by, that's likely to
>>> complicate things more.
>>>
>>> For the charges, especially with the tight and complex charge
>>> distribution you have there, I recommend an ensemble average. If you
>>> just use a few structures you might end up with charges that bias the
>>> structure in some manner. I think RED can be used for an ensemble of
>>> structures. Francois can help more with that. We do all ours
>>> in-house, of course. But, the results should be equivalent.
>>>
>>> Definitely validate whatever you decide to use. Find experimental
>>> j-couplings or something else that you can calculate from the
>>> structure. This is especially important in your situation. If you
>>> can't find any experimental values, then validate against appropriate
>>> quantum structures.
>>>
>>>
>>> On Fri, Mar 23, 2012 at 12:11 PM, FyD <fyd.q4md-forcefieldtools.org> wrote:
>>>> Urszula,
>>>>
>>>>> Thank you, this is a part of a peptide, attached is the whole
>>>>> peptide structure ...
>>>>
>>>> So you might start from a dipeptide:
>>>> http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#15
>>>>
>>>> R.E.D. Server can generate different fragemtnts automatically as well:
>>>> http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#25
>>>>
>>>> regards, Francois
>>>>
>>>>
>>>>> It's not really even a complete lysine, and the attachment is a bit
>>>>> unusual. The proximity of the ring oxygen, the anomeric oxygen and
>>>>> the nitrogen -- all just one carbon removed from each other -- might
>>>>> make good parameterization tricky. I'll think about it a while and
>>>>> ask around in the group. I'll get back to you.
>>>>>
>>>>>
>>>>> On Fri, Mar 23, 2012 at 11:06 AM, FyD
>>>>> <fyd.q4md-forcefieldtools.org> wrote:
>>>>>> Dear Urszula,
>>>>>>
>>>>>>> I have a peptide with a sugar bound (beta-D-Glucose), and I would
>>>>>>> like to run MD on that with a protein later. I am trying to find a
>>>>>>> easy way to
>>>>>>> generate the parameters for this part. I know that I could use a
>>>>>>> RESP-A1A charge model for the entire new residue and scaling factors
>>>>>>> for 1-4 interactions.
>>>>>>> However I have no idea how to start with that ...
>>>>>>> I was wondering if there is any tutorial for that?
>>>>>>> Could anyone suggest me something? I attached a mol2 file of the
>>>>>>> residue with sugar bound.
>>>>>>
>>>>>> Does this Lys-Glc belongs to a peptide? if yes, this likely means you
>>>>>> need a central fragment for this modified amino-acid. You might decide
>>>>>> to start from a dipeptide; i.e. ACE-AA*-NME; AA* is your modified
>>>>>> amin-acid; i.e. Lys-beta-D-Glucose.
>>>>>>
>>>>>> See http://q4md-forcefieldtools.org/Tutorial/Tutorial-3.php#15
>>>>>>
>>>>>> or, you might decide to split your molecules into two parts; a glyco
>>>>>> part and a AA part...
>>>>>>
>>>>>> regards, Francois



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Mar 26 2012 - 06:00:03 PDT
Custom Search