Hi Carlos,
That's correct. There's 3 test cases in the paper that show different
aspects of the algorithm's correctness:
1. Optimizing a famous function - tests if the algorithm can find a known
global minimum
2. Fitting AMBER to AMBER - ff99 -> ff99SB energies
3. Fitting AMBER to quantum - generation of the gafflipid forcefield
Best,
Robin
On Tue, Jan 24, 2017 at 12:49 PM, Carlos Simmerling <
carlos.simmerling.gmail.com> wrote:
> Hi Robin,
> just following up because I also misunderstood when I first read this
> paper, but I think I eventually understood. You didn't take the QM energies
> and recover the ff99SB parameters by optimizing the QM-MM difference, but
> instead you used the ff99SB energy profile and showed that you can find
> what parameters produce the specific MM energy profile, right? I think
> that's what you're saying above but I just wanted to make sure I
> understood.
> cheers
> carlos
>
> On Tue, Jan 24, 2017 at 2:35 PM, Robin Betz <robin.robinbetz.com> wrote:
>
> > Hi Stefano,
> >
> > In the paper I started with tetraalanine parameterized with the ff99
> > forcefield, and fit the phi and psi dihedrals to ff99SB energies.
> However,
> > ff99 has only 2 dihedral terms for phi/psi, while the ff99SB correction
> has
> > more. So I created a prmtop with those extra dihedral terms present and
> set
> > the force constants all to 1, as Paramfit doesn't care about the initial
> > value for a term.
> >
> > The only difference between ff99 and ff99SB is the phi and psi terms, so
> it
> > doesn't really matter either way. To verify, check parm99.dat (ff99) vs
> > frcmod.ff99SB and you'll see only phi and psi terms are redefined.
> >
> > This test case in the paper is more of a proof of concept of Paramfit's
> > functionality, showing that it can fit "apples to apples" and can recover
> > constants in the AMBER equation from energies generated with that same
> > equation. It's much harder to demonstrate that a program will always
> > generate the best fit to QM energies, as the AMBER equation provides a
> > Newtonian approximation of systems that are really governed by quantum
> > rules.
> >
> > Hope this helps,
> > Robin
> >
> > On Tue, Jan 24, 2017 at 11:07 AM, BOSISIO Stefano <s1374284.sms.ed.ac.uk
> >
> > wrote:
> >
> > > Dear Amber staff,
> > >
> > >
> > > I am struggling a bit to use Paramfit to reproduce the exact
> > > dihedrals'amplitudes values of ff99SB forcefield for tetraalanine as
> > > written in Paramfit paper (10.1002/jcc.23775) .
> > >
> > > In particular, I cannot understand if you started with a tetraalanine
> > > parametrized with ff99 force field ( with Phi and Psi amplitudes set
> to
> > > 1.0 ) and then fitted it to ff99SB, or if you started with ff99SB (with
> > Phi
> > > and Psi amplitudes set to 1.0 ) and fitted it to the real ff99SB.
> > >
> > >
> > > Thank you very much,
> > >
> > >
> > > Best wishes,
> > >
> > > Stefano
> > >
> > >
> > > The University of Edinburgh is a charitable body, registered in
> > > Scotland, with registration number SC005336.
> > >
> > > _______________________________________________
> > > 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
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jan 24 2017 - 15:00:03 PST
