AMBER Archive (2007)

Subject: Re: AMBER: Parallelizin antechamber?

From: M. L. Dodson (mldodson_at_houston.rr.com)
Date: Tue Aug 14 2007 - 09:29:16 CDT


Francesco Pietra wrote:

> I wonder whether there may be other users also wishing that
> antechamber is parallelized in the future.
>
> I have already used successfully antechamber/gaff with a number
> of non-polymeric molecules, up to 100 atoms per molecule. Found
> difficult groups for antechamber, thouugh overcome. Now a big
> step: 5oo atoms, non polymeric. Furnished to antechamber
> minimized with a molecular mechanics package as a pdb
> file. Could not carry out a global space search because of too
> many variables (1500) for the MM package.
>

I'm no expert on antechamber, but I feel confident that the design
never envisioned a 500 atom "residue".

> AM1 is minimizing since 26 hours, now at cycle 404, while E is
> still decreasing by -0.15 kcal/mol per step, for a total 5%
> decrease of E (160 kcal/mole) since the beginning. DELTAX
> oscillating between 0.001 and 0.002. Of course I wish I had a
> parallelized minimizer.
>
> What makes me very curious is the temperature setting for the
> minimization, 1000 K. To overcome barriers? I was totally
> unaware of that possibility, if this is the case, for a
> conformational search through electronic calculations.
>
> A surprise was the -nc flag able to introduce the electrons
> lacking after that certain atoms (such as of sulfate ester
> groups) came to antechamber through a pdb file, thus having lost
> the parameterization of the MM package, thus coming as electron
> unpaired. The overall negative charge was reestablished though
> the -nc flag. This was a problem I was wondering how to
> surmount, until I hopeless tried -nc and the surprise
> came. Previously, I had tried unsuccessfully to inform
> antechamber of the negative charge at specific oxygen atoms
> (identified through VMD) by setting "-1" at columns 79-80 of
> pdb.
>

I've never heard of a quantum chemistry package that did not allow
the overall charge to be specified.

> Cheers
>
> francesco pietra
>
>

You know you do not have to use antechamber. Why do you not use
one of the parallelized quantum chemistry packages? Both GAMESS
and Gaussian are supported by RED in its various incarnations.
Even if you want to use antechamber, you could use, e.g., GAMESS
(which allows you to use the AM1 method) to get the structure.
Then all you need to do is use the optimized structure to get the
partial charges, a much faster calculation than geometry
optimization. One of the nice features of antechamber is that the
components can be used individually outside of the main
antechamber program itself.

But in any case, a 500 atom quantum chemistry calculation is a
very large calculation unless you use one of the linear scaling
packages. Expectations should be informed by reality.

Bud Dodson

-- 
M. L. Dodson
Email:	mldodson-at-houston-dot-rr-dot-com
Phone:	eight_three_two-five_63-386_one
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu