AMBER Archive (2008)

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

From: FyD (fyd_at_q4md-forcefieldtools.org)
Date: Sun Jun 29 2008 - 08:18:15 CDT


Quoting Francesco Pietra <chiendarret_at_gmail.com>:

> Yes, I know those papers, perhaps not so well mastered as they
> deserve. What I am interested in is to clarify what you call a "black
> box", that is comparing - in actual cases - the result of applying
> RESP with respect to applying am1-bcc as provided by Antechamber.

My definition of a black box might not be correct. That being said,
what I meant by "black box" is a system where we do not control for
sure all the issues...

> I
> mean comparing for a demanding structure, like the tetracyclins, or
> other ones, perhaps embedded in a lipidic membrane and even made as
> ligands of proteins, just to make the system more complicate. How
> would you react if the results turn out to lie in the statistics
> expected for MD for both RESP and am1-bcc?

In the lab, we do _not_ generate charge values "on the fly" without
controlling the conformations used in the charge derivation, and we do
USE the RESP charge model. Thus, being _aware_ of the choices made in
the charge derivation before MD simulation (and in some cases being
aware of its limitations; indeed !), the charge derivation procedure
can be modified if the results (obtained after MD simulation) are not
those expected...

> The black box made transparent. There are so many approximations and
> assumptions in
> performing simulations that I would not be too surprised that there is
> such an agreement.

In this case, why developping sophisticated force fields (FF) ?

In the Kollman and co-worker strategy, the elect. interaction is the
key point for the development of AMBER FF (condidering the AMBER FF
representation quite simple). With the van der Waals (vdW)
interactions, elect. are used to generate dihedral terms. vdW + elect.
+ dihedrals are the three major FF components to generate a correct
representation of mol. conformations & inter-mol. interactions in a
condensed system.

This was published ~10-15 years ago. My understanding is that this is
still up to date as rescent FF are still based on this work, and use
even more complex charge derivation.

> I am looking for such papers in the literature.
> Curious if they do not exist. It would allow to take personal views
> without any firm support. If the papers you mention were a clear cut
> answer, why developing Antechamber with am1-bcc?

I did not develop the am1-bcc charge model, and I do not promote the
am1-bcc charge model within Antechamber. I am not qualified to answer
here. My understanding is that Antechamber also generates RESP charge
values. That being said, this is always interesting to compare charge
models. Each user can make her/his choice. The point here is to
clearly know the limitations of each model.

> It is becoming hot here, worse quite humid.

Yes...

Personnally, I do not understand how a parametrized model (i.e.
am1-bcc) can rigorously mimic a complex model (i.e. RESP) that is
controlled with a lot of difficulties. Moreover, why continuing the
parametrization of a first model to represent a second that costs
almost nothing nowadays (with new computers).

regards, Francois

-----------------------------------------------------------------------
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