AMBER Archive (2009)
Subject: Re: [AMBER] Re: antechamber am1-bcc issues
Date: Sat Aug 08 2009 - 08:55:40 CDT
Hi Dave and Dave,
The problem is not in the AM1 charges or the bond correction parameters but in
the assignment of the bond types for the nitro group. If you run antechamber and
ask for the output in antechamber format you see that both NO bonds are assigned
a bond type of 2, but there is no parameter for 23 02 31 (N double bond O), and
so you set the correction term to zero and get back the original AM1 charges. If
you use a bond type of 9, a delocalized bond, for the nitro group, then the bond
correction term is -.1500 (230931 in the Bayley table), and when you use this
correction factor you get the OEChem results. Since antechamber uses a
delocalized nitrogen atom type it probably should use a delocalized bond type.
In this case I think antechamber is wrong and OEChem is correct.
It would be most useful to have a suite of molecules with assigned atom and
bond types that correspond to the test suite used by Bayley et al in
parameterizing the AM1BCC method. I wrote an email to the list last month
pointing out that Antechamber did not reproduce the structure given in the
Bayley paper (which is the only case I know of) and I have not had any response
to that email. Even the Bayley example raises questions about atom typing - the
nitrogen adjacent to the double bond in the six membered ring is types as a
normal nitrogen instead of a delocalized nitrogen - that need to be discussed if
you expect to get the same result from all the programs implementing AM1BCC.
Kevin E. Gilbert
Quoting case <case_at_biomaps.rutgers.edu>:
> On Fri, Aug 07, 2009, David Mobley wrote:
> > 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.
> Thanks for sending the example. Let's look at the nitrogen charge you
> > In this example, the Antechamber AM1-BCC charges differ from the
> > others by more than 0.3 e on the nitrogen alone.
> There are two possible origins for this. First, the AM1 Mulliken charge
> might be different. We get 0.613 for the nitrogen Mulliken charge. I don't
> know what the other programs give, but that would be interesting to check.
> (For completeness, our minimized AM1 energy is 31.74.)
> Second, and more likely, is that the BCC corrections are the source of
> the difference. Here are the corrections that Junmei's am1bcc program
> bond at1 at2 pre-charge correction post-charge
> 0 C1 N1 -0.4880 0.6130 -0.0452 -0.5332 0.6582
> 1 C1 C2 -0.5332 -0.0680 0.0000 -0.5332 -0.0680
> 2 C1 S1 -0.5332 0.7320 0.2699 -0.2633 0.4621
> 3 N1 O1 0.6582 -0.3680 0.0000 0.6582 -0.3680
> 4 N1 O2 0.6582 -0.3480 0.0000 0.6582 -0.3480
> 5 C2 C3 -0.0680 -0.1690 0.0000 -0.0680 -0.1690
> 6 C2 H1 -0.0680 0.1850 0.0000 -0.0680 0.1850
> 7 C4 S1 -0.4420 0.4621 0.2699 -0.1721 0.1922
> 8 C3 C4 -0.1690 -0.1721 0.0000 -0.1690 -0.1721
> 9 C3 H2 -0.1690 0.1680 0.0000 -0.1690 0.1680
> 10 C4 H3 -0.1721 0.1850 0.0000 -0.1721 0.1850
> Note that there is only a small correction (from the C1-N1 bond). We don't
> apply any corrections for the N1-O1 or the N1-O2 bond -- that seems a little
> odd(??), and may indicate that there is a difference between our table of
> corrections and those that Chris or openeye are using.
> Of course, at this point, I should (and will!) go check the above bond
> corrections against the original papers. But I'm travelling right now,
> without access to the needed papers. So maybe Junmei or someone on the list
> can continue this check.
> [as an aside: you can get the bond correction output above by setting the
> "debug" flag in am1bcc.c. I've attached a patch file against the current
> release that creates the above format, which is more readable than what is
> there now....]
AMBER mailing list