Re: [AMBER] antechamber am1-bcc issues

From: Khavrutskii, Ilja <ikhavrutskii.ucsd.edu>
Date: Thu, 20 Aug 2009 14:04:44 +0100

Hi,

Thanks for your e-mail. RedHat Enterprise 5.x.(or so I thought) From config.h: G77=hasG77. 64 bit.
gcc version 3.4.6 20060404 (Red Hat 3.4.6-4) (Reading specs from /usr/lib/gcc/x86_64-redhat-linux/3.4.6/specs)
GNU Fortran (GCC) 3.4.6 20060404 (Red Hat 3.4.6-4).

David Case seems to have reproduced the problem anyways.

Like I said, I did not know how widespread this problem was. Note that the amber tools pass the test case with the messed-up mopac.

I too like mopac.

It was from Corina, but even xleap will get you the same structure if you do relax. Anyway it is good enough as a guess as it is. I did perturb it in xleap with distort to get all the other conformations. Do not worry I know how to do minimizations, this is how I discovered this problem in the first place.

> If you really want better charges then you should generate multiple low energy conformers of your structure and average the charges over all the structures, and this is not something to be done in antechamber.

That is exactly what I am doing, but the antechamber should at least symmetrize the charges for a given conformer.
-Ilja

Kevin Gilbert
Serena Software
Quoting "Khavrutskii, Ilja" <ikhavrutskii.ucsd.edu>:

> The attached should make it easier, I noticed that the pdb format is deformed
> in the post.
>
> So with the catechol.pdb in place and after modifying that mopac.sh I would
> run
>
>
> 1. antechamber -i catechol.pdb -fi pdb -o oh2bnz.mol2 -fo mol2 -c bcc -s 2
> -at gaff -ek "AM1 ANALYT GRAD LET GNORM=1.D-8 SCFCRT=1.D-6 XYZ"
> 2. antechamber -i mopac.pdb -fi pdb -o oh2bnz.mol2 -fo mol2 -c bcc -s 2 -at
> gaff -ek "AM1 ANALYT GRAD LET GNORM=1.D-8 SCFCRT=1.D-6 XYZ"
> 3. repeat line 2 until you are happy (look in mopac.out for clues :)
>
>
> The first step produces mopac.pdb that has the optimized coordinates, the
> second step overwrites it with an even better optimized coordinates.
>
> There is still an big issue that remains to be addressed and that is the
> symmetrization of partial charges on pseudo-equivalent atoms.
> In the catechol example charges on atoms C1=C2, C3=C8, C4=C6, O5=O7, H9=H10,
> H11=H14 and H12=H13. This must be the case for true am1bcc charges.
>
> Antechamber fails to symmetrize the charges and I have to write a script to
> do that. The solution is complex (yet possible) if you want to automate it,
> but for me for now I can just specify which atoms are equivalent and the
> script will take care of the symmetrization. For example since C1=C2 their
> charge should be (qC1+qC2)/2 etc.
>
> Good luck,
> -Ilja
>
>
>
> On 8/20/09 4:28 AM, "Alan" <alanwilter.gmail.com> wrote:
>
> Hi Ilja,
>
> I am concern about it and I want to run some tests with mopac7-1.14 for Mac
> OSX.
>
> Would you please send me your inputs and outputs files (in private) so
> I can check things here. Testing with amber mopac I got the same
> problem your reported but for mopac7 I am not sure what to see because
> I didn't get any pdb output.
>
> If you could give more details about how did you test it with your own
> compiled version of mopac7 that I would appreciate.
>
> Many thanks in advance,
>
> Alan
>
> On Tue, Aug 18, 2009 at 22:00, Khavrutskii, Ilja<ikhavrutskii.ucsd.edu>
> wrote:
> > Comments regarding Antechamber.
> >
> >
> > 1. I did not notice any symmetrization of charges on atoms that are
> equivalent like (H-atoms of methyl groups, benzene molecule etc.)
> > 2. Geometry optimization is not working properly in standard Amber10
> installation.
> >
> > I do not know if I am the only one experiencing these problems or not, and
> I will expand a bit more on problem 2 below:
> >
> > I was somewhat dissatisfied with the apparent inability of antechamber to
> get converged geometries for small molecules, like catechol for example. I am
> using the Amber10 standard installation on RH Linux box. Since antechamber is
> running mopac under the hood, and some time ago I used that program stand
> alone, I gave it a try to see if I can improve upon that PRECISE keyword. To
> my great surprise the mopac that is compiled with Amber (using Intel
> compilers) was not able to optimize a reasonable guess at all.
> >
> > The exact command I used was:
> >
> > antechamber -i catechol.pdb -fi pdb -o oh2bnz.mol2 -fo mol2 -c bcc -s 2 -at
> gaff -ek "AM1 ANALYT GRAD LET GNORM=1.D-8 SCFCRT=1.D-6 XYZ"
> >
> > This generates an mopac.in file that looks like this:
> >
> > AM1 ANALYT GRAD LET GNORM=1.D-8 SCFCRT=1.D-6 XYZ CHARGE=0
> > created by wmopcrt() for mopac
> >
> > C -0.0170 1 1.3780 1 0.0100 1
> > C 1.1690 1 2.0890 1 0.0020 1
> > C 2.3790 1 1.4200 1 -0.0130 1
> > C 2.4050 1 0.0350 1 -0.0210 1
> > O 3.5940 1 -0.6250 1 -0.0350 1
> > C 1.2110 1 -0.6800 1 -0.0130 1
> > O 1.2320 1 -2.0400 1 -0.0200 1
> > C 0.0020 1 -0.0040 1 0.0020 1
> > H -0.9600 1 1.9040 1 0.0260 1
> > H 1.1500 1 3.1690 1 0.0080 1
> > H 3.3040 1 1.9780 1 -0.0190 1
> > H 3.9460 1 -0.8200 1 0.8440 1
> > H 1.2460 1 -2.4380 1 0.8610 1
> > H -0.9260 1 -0.5570 1 0.0080 1
> >
> > The output looks like a big mess to me:
> >
> > SCF CRITERION = 0.1000E-05
> > CYCLE: 1 TIME: 0.06 TIME LEFT: 3599.9 GRAD.:999999.999
> HEAT:-54.21049
> > CYCLE: 2 TIME: 0.05 TIME LEFT: 3599.9 GRAD.:999999.999
> HEAT:-54.21157
> > CYCLE: 3 TIME: 0.02 TIME LEFT: 3599.8 GRAD.: 198.479
> HEAT:-21.66359
> > GRADIENTS OF OLD GEOMETRY, GNORM= 198.479378
> > -15.499537 57.227984 122.266395 1.431796 -3.530504 23.028728
> > -35.279682 -50.180782 -38.433267 -25.461497 -3.680921 6.443132
> > 15.413123 -45.601920 -37.698058 -13.430117 50.503021 -4.669441
> > 34.211819 41.404541 11.990522 -34.282666 39.516530 3.591553
> > -30.807113 -11.363384 -6.427677 -2.833937 3.217770 -3.782121
> > -21.071254 2.388208 -4.099834 9.610558 -17.402015 -12.587598
> > GRADIENTS OF NEW GEOMETRY, GNORM=*************
> > ************************************************************************
> > ************************************************************************
> > ************************************************************************
> > ************************************************************************
> > ************************************************************************
> > ************************************************************************
> >
> >
> >
> > CALCULATION ABANDONED AT THIS POINT!
> >
> >
> > SMALL CHANGES IN INTERNAL COORDINATES ARE
> > CAUSING A LARGE CHANGE IN THE DISTANCE BETWEEN
> > CHEMICALLY-BOUND ATOMS. THE GEOMETRY OPTIMIZATION
> > PROCEDURE WOULD LIKELY PRODUCE INCORRECT RESULTS
> >
> >
> > It is clear that the installed version of mopac that comes with Amber10 is
> not compiled properly.
> >
> > If I follow the instructions for a stand alone mopac:
> >
> > http://sourceforge.net/projects/mopac7
> > And get f2c (which seems to be important) to compile it with:
> >
> > http://www.webmo.net/support/f2c_linux.html
> >
> > Then everything seems to get back to normal and I can optimize geometry to
> my satisfaction.
> >
> > In pdb format:
> >
> > ATOM 1 C1 CAT 1 -0.691 -1.620 0.126
> > ATOM 2 C2 CAT 1 0.704 -1.620 0.126
> > ATOM 3 C3 CAT 1 1.409 -0.418 0.126
> > ATOM 4 C4 CAT 1 0.706 0.790 0.124
> > ATOM 5 O5 CAT 1 1.318 2.026 0.119
> > ATOM 6 C6 CAT 1 -0.706 0.793 0.128
> > ATOM 7 O7 CAT 1 -1.447 1.951 0.128
> > ATOM 8 C8 CAT 1 -1.403 -0.422 0.127
> > ATOM 9 H9 CAT 1 -1.239 -2.573 0.126
> > ATOM 10 H10 CAT 1 1.252 -2.573 0.124
> > ATOM 11 H11 CAT 1 2.508 -0.414 0.125
> > ATOM 12 H12 CAT 1 2.273 1.885 0.137
> > ATOM 13 H13 CAT 1 -0.832 2.700 0.136
> > ATOM 14 H14 CAT 1 -2.502 -0.409 0.127
> >
> >
> > This kind of things get me really worried. Then David mentioned bond type
> issue...
> >
> > I would like to see any fixes to these problems if they are widespread.
> >
> > Thanks,
> >
> > -Ilja
> >
> >
> >
> >
> >
> >
> > On 8/7/09 2:19 PM, "David Mobley" <dmobley.gmail.com> wrote:
> >
> > All,
> >
> > I keep having issues with partial charges coming from Antechamber AM1-
> > BCC. In addition to the bug I mentioned previously relating to the -j
> > option resulting in incorrect charges when it should have no effect,
> > I'm also consistently finding that Antechamber generates substantially
> > different "AM1-BCC" partial charges than other implementations of AM1-
> > BCC. For a set of roughly 20 molecules, I obtained partial charges
> > from Christopher Bayly that he computed with his own, internal Merck
> > AM1-BCC implementation, and I also computed them with OpenEye's
> > Quacpac python toolkit. We then computed mean differences between
> > these charge sets. The Bayly AM1-BCC and OpenEye AM1-BCC
> > implementations agree extremely well, but Antechamber AM1-BCC charges
> > disagree with both Bayly and OpenEye partial charges reasonably
> > frequently.
> >
> > The differences are nontrivial -- they can make up to several kcal/mol
> > difference in computed hydration or binding free energies.
> >
> > I am attaching just one example of profound disagreement. The 2-
> > nitrothiophene.mol2 file contains Bayly AM1-BCC partial charges (sybyl
> > atom naming); 2-nitrothiophene_ante_amber.mol2 contains antechamber
> > AM1-BCC partial charges (amber atom naming), and 2-
> > nitrothiophene_oechem_amber.mol2 contains OpenEye AM1-BCC partial
> > charges.
> >
> > In this example, the Antechamber AM1-BCC charges differ from the
> > others by more than 0.3 e on the nitrogen alone.
> >
> > Anyway, I just wanted to put this out there to advise people. It
> > appears I will be ditching the Antechamber AM1-BCC implementation
> > until some of these problems are straightened out.
> >
> > Thanks.
> > David Mobley, Ph.D.
> > Assistant Professor of Chemistry
> > University of New Orleans
> > New Orleans, LA 70148
> > dlmobley.uno.edu
> > Phone 504-383-3662
> > Fax 504-280-6860
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Alan Wilter S. da Silva, D.Sc. - CCPN Research Associate
> Department of Biochemistry, University of Cambridge.
> 80 Tennis Court Road, Cambridge CB2 1GA, UK.
> >>http://www.bio.cam.ac.uk/~awd28<<
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
>





_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 20 2009 - 06:11:35 PDT
Custom Search