AMBER Archive (2008)

Subject: Re: AMBER: Antechamber/mopac am1-bcc charge

From: Junmei Wang (junmwang_at_gmail.com)
Date: Fri Jun 27 2008 - 10:35:23 CDT

All the amber force fields including gaff are developed using RESP charges.
I don't think it is a good idea to develop force field using cheap am1-bcc
charges either. Am1-bcc is only recommended to use when ab initio
calculations are too expensive, such as in high throughput docking studies.
This is clearly stated in amber manual. I think am1-bcc is better than
gasteiger since it resembles resp to some degree. Although it is more
expensive than gasteiger, am1bcc is much cheaper and more 'general' than
resp. According to OpenEye's webpage, am1-bcc achieved the best performance
in their docking studies, much better than gasteiger...

Regards

Junmei

On Fri, Jun 27, 2008 at 4:07 AM, FyD <fyd_at_q4md-forcefieldtools.org> wrote:

> Quoting Junmei Wang <junmwang_at_gmail.com>:
>
> Hopefully, AM1-BCC can remedy some flaw of Milliken charges, since it is
>> parameterized to mimic HF/6-31G* resp charges. It seems to me that am1-bcc
>> does not do a good job for some kinds of molecules, such as sulfonamides
>> and
>> phosphates, more parameterization is needed for them.
>>
>
> Once again, I do not think 'simplicity' is the answer in a _general_
> strategy for charge derivation/general force field for biomolecules.
>
> Proposing a general/"on the fly" strategy based on some parametrisation
> (am1-bcc) to reproduce a sophisticated charge derivation (RESP with multiple
> conformations for instance) is in contradiction with the rescent force
> fields. How can a parametrized model reproduce something that is
> handled/understood/controlled with a lot of ... difficulties ?
> Indeed, RESP charges present many pitfalls; a parametrized model to 'mimic'
> RESP charges might be usefull for simple cases, but I do not think it can be
> proposed as a general strategy. However, this is clearly a convenient "black
> box" for new users.
>
> In the Kollman group strategy, _the_ RESP charge model is the basis of the
> strength of AMBER force fields. Rescent AMBER, GLYCAM, CHARMM force fields
> use even more sophisticated/modified 'RESP' charge derivation (high basis
> set, complex multi-conformation fitting approach etc).
>
> Once again, I do not understand how a parametrized model can reproduce
> another complex model that is controlled with a lot of difficulties.
>
>
> regards, Francois
>
>
>
> On Thu, Jun 26, 2008 at 3:53 PM, FyD <fyd_at_q4md-forcefieldtools.org>
>> wrote:
>>
>> Quoting Junmei Wang <junmwang_at_gmail.com>:
>>>
>>> It is true that partial charges should depend on the environment. As a
>>>
>>>> matter of fact, different HF/6-31G* RESP charges are produced for
>>>> different
>>>> conformations (of course not significant different). However, in MM
>>>> calculations using an additive force field, only one set of charges is
>>>> used.
>>>> This set of charges should be "general" and "good" for all the
>>>> conformations
>>>> sampled by MD or minimization.
>>>>
>>>> How to generate such kind of charge? In AMBER, for standard aminio acid
>>>> and
>>>> nucleic acid residues, the charges were derived with RESP using multiple
>>>> conformations. Since Mulliken charges are less conformation dependent,
>>>> one
>>>> may get such kind of "general" charges using only one conformation. From
>>>> this aspect, we say it is an advantage.
>>>>
>>>>
>>> and for sure, it is also more simple to get Mulliken charges !
>>>
>>> However, from other aspects Mulliken charges were strongly criticized...
>>> Thus, it is a reason why many rescent force fields (AMBER, GLYCAM; even
>>> CHARMM !) use 'RESP' charges (and even quite sophisticated RESP charge
>>> derivation) or charges calculated based on molecular electrostatic
>>> potential(s); _NOT_ Mulliken charges.
>>>
>>> I am not sure 'simplicity' is the answer here...
>>>
>>> regards, Francois
>>>
>>>
>>>
>>> On Thu, Jun 26, 2008 at 12:29 PM, Francesco Pietra <
>>> chiendarret_at_gmail.com
>>>
>>>> >
>>>> wrote:
>>>>
>>>> On Thu, Jun 26, 2008 at 4:33 PM, Junmei Wang <junmwang_at_gmail.com>
>>>> wrote:
>>>>
>>>>> > Hi,
>>>>> > The bcc parameters are same and the final am1-bcc charges depend on
>>>>> which
>>>>> > am1 package to be used. However, I don't think the difference of
>>>>> charges
>>>>> of
>>>>> > Cl- is that significant. One advantage of am1-bcc over resp is that
>>>>> it
>>>>> is
>>>>> > less variant for different conformations since am1-mulliken charges
>>>>> are
>>>>> > pretty conformation independent.
>>>>>
>>>>> That lesser dependence of partial charges from the conformational
>>>>> status (if this is what you really meant) is interesting. However - as
>>>>> a devil's advocate - is that really an advantage? I wish you reassure
>>>>> me that it is an advantage. I always thought that it is normal for the
>>>>> partial charges to depend on the environment, be that also the
>>>>> conformational status itself. We are in the hot season, may be I can't
>>>>> rightly grasp what you wrote. Presently, I see that independence from
>>>>> the conformation as illusory, thus a disadvantage Sincerely, however,
>>>>> I hope I am wrong.
>>>>>
>>>>> francesco pietra
>>>>>
>>>>> >I think bcc parameters were derived using
>>>>> > the mopac package.
>>>>> >
>>>>> > Best
>>>>> >
>>>>> > Junmei
>>>>> >
>>>>> >
>>>>> > On Thu, Jun 26, 2008 at 1:21 AM, Dong Xu <d1xu_at_ucsd.edu> wrote:
>>>>> >>
>>>>> >> Hi Junmei,
>>>>> >>
>>>>> >> Thanks for your response. I installed ambertools and tested am1-bcc
>>>>> with
>>>>> >> mopac6. When comparing the output files of am1-bcc charge
>>>>> calculations
>>>>> from
>>>>> >> divcon and mopac for the same molecule, it seems they came to
>>>>> different
>>>>> >> optimized geometries and final electronic energies: -34543.95016086
>>>>> EV
>>>>> >> (divcon) and -34945.38044 EV (mopac). The charges are thus different
>>>>> as
>>>>> >> well, e.g. a Cl atom has -0.00595 (divcon) and 0.012900 (mopac). So
>>>>> my
>>>>> >> questions are, How to reduce this type of variations in am1-bcc
>>>>> charge
>>>>> >> calculation and how much effect does it have on the protein-ligand
>>>>> >> electrostatic interaction energy estimation?
>>>>> >>
>>>>> >> Thanks,
>>>>> >>
>>>>> >> -DX
>>>>> >>
>>>>> >> On Wed, Jun 25, 2008 at 3:44 PM, Junmei Wang <junmwang_at_gmail.com>
>>>>> wrote:
>>>>> >>>
>>>>> >>> I would like to recommend you to use antechamber in AmberTools.
>>>>> Mopac6
>>>>> is
>>>>> >>> included in the package and you simply specify "-c bcc" to
>>>>> calculate
>>>>> am1-bcc
>>>>> >>> charges. For example antechamber -fi mol2 -fo mol2 -i input.mol2 -o
>>>>> >>> output.mol2 -c bcc
>>>>> >>>
>>>>> >>> Best
>>>>> >>>
>>>>> >>> Junmei
>>>>> >>>
>>>>> >>> On Wed, Jun 25, 2008 at 5:09 PM, Dong Xu <d1xu_at_ucsd.edu> wrote:
>>>>> >>>>
>>>>> >>>> Hi,
>>>>> >>>>
>>>>> >>>> I downloaded and compiled antechamber-1.27 and noticed that divcon
>>>>> is
>>>>> >>>> replaced by mopac6. Could anyone let me know the procedure and
>>>>> command
>>>>> to
>>>>> >>>> calculate am1-bcc charges using antechamber and mopac?
>>>>> >>>>
>>>>> >>>> Thanks,
>>>>> >>>>
>>>>> >>>> -DX
>>>>>
>>>>>
>>>>
>>> -----------------------------------------------------------------------
>>> The AMBER Mail Reflector
>>> To post, send mail to amber_at_scripps.edu
>>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>>> to majordomo_at_scripps.edu
>>>
>>>
>>
>
>
> F.-Y. Dupradeau
> ---
> http://q4md-forcefieldtools.org/FyD/
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo_at_scripps.edu