Ok - so I tried a few tests this evening and I think the conclusion is a
possible combination of:
1) PMEMD CUDA works fine for OPLS matching sander and pmemd.
2) AMBER has NEVER implemented OPLS correctly.
3) everything is hunky dorey!!!
Sadly I think I have to go with 2 - but then I don't think anybody has ever
actually implemented OPLS correctly and 2 actually in this case implies 1 is
fine - as in pmemd.cuda implements the broken opls the same as sander and
pmemd.
Attached are the test cases. Specifically I did:
$AMBERHOME/tleap -f buildit
where buildit is:
loadamberprep opls_uni.in
loadamberparams opls_parm.dat
mol = sequence { ACE ALA ALA ALA ALA AME }
saveamberparm mol opls_test.prmtop opls_test.inpcrd
quit
Then:
$AMBERHOME/bin/sander -O -i ../mdin -o mdout.sander12 -p ../opls_test.prmtop
-c ../opls_test.inpcrd -r restrt.sander12 -inf mdinfo.sander12
$AMBERHOME/bin/pmemd -O -i ../mdin -o mdout.pmemd12 -p ../opls_test.prmtop
-c ../opls_test.inpcrd -r restrt.pmemd12 -inf mdinfo.pmemd12
$AMBERHOME/bin/pmemd.cuda_DPDP -O -i ../mdin -o mdout.pmemd.cuda_DPDP12 -p
../opls_test.prmtop -c ../opls_test.inpcrd -r restrt.pmemd.cuda_DPDP12 -inf
mdinfo.pmemd.cuda_DPDP12
This gives IDENTICAL results for all 3 codes - see attached tar file.
So... looks like we are good so far and the issue is just the off diagonal
stuff. Unless someone actually wants to go ahead and get OPLS working
correctly / figure out how to do it in leap and then try this again with
pmemd.cuda.
All the best
Ross
> -----Original Message-----
> From: Jason Swails [mailto:jason.swails.gmail.com]
> Sent: Tuesday, May 29, 2012 11:49 AM
> To: AMBER Developers Mailing List
> Subject: Re: [AMBER-Developers] [AMBER] manual prmtop file editing for
> free energy calculation
>
> On Tue, May 29, 2012 at 4:54 AM, David A Case
> <case.biomaps.rutgers.edu>wrote:
>
> > On Mon, May 28, 2012, Ross Walker wrote:
> > >
> > > Out of interest how does Leap know to use different combining rules
> for
> > > OPLS?
> >
> > The "AC" vs. "RE" flag in the non-bonded section triggers a change in
> > combining rules. (I think! check the code or run a simple test
> > calculation.)
> >
>
> This isn't going to help my case, but I don't see where this happens.
> mathop.c, line 69 (function MathOpConvertNonBondToAC). This only
> appears
> to satisfy the Amber combining rules. That function itself is only
> called
> in two places in unitio.c (and nowhere does it look like we try to
> match
> "RE" or "AC"). So unless I'm missing something, OPLS just gets the
> Amber
> combining rules inside tleap...
>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Candidate
> 352-392-4032
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed May 30 2012 - 00:00:03 PDT
