Thanks Prof. Case,
First, sorry, I forgot to attach the file. It's going now.
Indeed, the sqm.out files have something suspicious and I am sending them
here in attachment.
I am aware of the Note.
AFAIU, the mulliken charges are read by antechamber from sqm.out files. So
if using only 3 decimals, no wonder I will run in round-off issues.
I think the modifications you suggest to increase the decimals in sqm code
should be done in the release version, instead of 3, could be at least 6. Or
am1bcc routine in antechamber should do the balance.
So I applied the modifications in $AMBERHOME/src/sqm/qm2_print_charges.f and
became much happy with the results. The net charges leaves a residue on 7th,
sometimes 6th decimal. I really would suggest to have this present in AT
1.3.
Many thanks,
Alan
On Tue, Jan 12, 2010 at 14:29, case <case.biomaps.rutgers.edu> wrote:
> On Tue, Jan 12, 2010, Alan wrote:
> >
> > I am using AT 1.3 (with sqm) in either Linux 64 (Debian 5.0 with gcc) and
> > Mac OSX (10.6.2 with intel and intel mkl).
>
> > /usr/local/amber11/bin/antechamber -i Bej.none_neutral.ideal.mol2 -fi
> mol2
> > -o Bej.none_neutral.ideal_bcc_gaff.mol2 -fo mol2 -c bcc -nc 0 -m 1 -s 2
> -df
> > 2 -at gaff -pf n # may take more than an hour to run
>
> Can you post your input mol2 file? We have not seen differences in am1-bcc
> charges between mac and linux systems, but I am sure our test molecule set
> is
> somewhat limited.
>
> > The problem is that if run such commands in my Mac, the final net charge
> is
> > 0.004 and tleap: check BEJ doesn't complain.
> >
> > But if I run on linux, net charge is 0.01 and then tleap prints:
>
> Both of these are odd. There can be round-off errors because sqm only
> prints
> Mulliken charges to three places after the decimal point. What is the
> "Total Mulliken Charge" printed out in sqm.out? Once we know how large
> your
> system is, it would be easier to know if these values could be from
> round-off
> or not.
>
> If you like, you can edit $AMBERHOME/src/sqm/qm2_print_charges.f, changing
> lines 47 and 57 to give higher precision output.
>
> You should also examine the sqm.out files to see if the geometry
> minimization
> actually completed correctly on both computers. Be sure to read the note
> on
> sqm tolerances:
>
> http://archive.ambermd.org/201001/0078.html
>
> ...good luck...dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
--
Alan Wilter Sousa da Silva, D.Sc.
PDBe group, PiMS project http://www.pims-lims.org/
EMBL - EBI, Wellcome Trust Genome Campus, Hinxton, Cambridge CB10 1SD, UK
+44 (0)1223 492 583 (office)
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jan 12 2010 - 10:30:03 PST
