Re: [AMBER] paramfit multiple dihedral commbination

From: Carlos Simmerling <carlos.simmerling.gmail.com>
Date: Mon, 26 Oct 2015 15:32:35 -0400

just to follow up a bit on Adrian's correct answer, in case people search
the archives later.

It sounds like you're talking about a single rotatable bond, and several
different 4-atom combinations that describe the rotation. If that's the
case, then unless you have separate training data (from other molecules),
then you cannot fit these separately using the QM rotational profile from
this molecule. Since you are changing all of the 4-atom measures as you
rotate the bond, they are not independent and you cannot say how much of
the total energy comes for each of these components. in larger force field
development projects, additional molecules are often used to determine the
contribution of each group. if you only care about the current molecule,
then this doesn't really matter and you can fit them together. It is worth
being careful though, since the sum is what you fit to QM, the individual
terms can take on any values as long as the sum is correct, and that can
lead to numerical instability. Now if you really care about details, these
different 4-atom combinations actually can vary independently if you
distort the angles (3-atom combinations), so there are some subtle things
going on and it's not exactly as simple as saying they are not independent.

now if you didn't mean different 4-atom combinations for a rotatable bond,
but you mean the different periodicities for the same 4-atom combination,
then yes these must be fit together and your fitting program should adjust
the various amplitudes (and possibly phases) to get the best fit.



On Mon, Oct 26, 2015 at 2:50 PM, Adrian Roitberg <roitberg.ufl.edu> wrote:

> No, you do not !
>
> In fact, ALL four torsions are identical (they describe the exact same
> set of coordinate changes). That is why you rotate ONE of them (again,
> they are all the same) and since eventually amber recognizes four
> torsions, it will add them all up.
>
> If you do not divide by four, the overall sum will be too large,
>
> adrian
>
>
> On 10/26/15 2:47 PM, CHAMI F. wrote:
> > Dear Amber users,
> >
> > the torsional bond i need to fit has 4 combination (divider 4 ) of
> dihedral angles for example the case of inter-ring
> > biphenyl torsion.
> >
> > do I have to fit the 4 dihedral angles simultaneously ?
> >
> > many thanks
> > Fati
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
>
> --
> Dr. Adrian E. Roitberg
> Professor.
> Department of Chemistry
> University of Florida
> roitberg.ufl.edu
> 352-392-6972
>
>
> _______________________________________________
> 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 Mon Oct 26 2015 - 13:00:04 PDT
Custom Search