Re: [AMBER] On alphabetizing impropers

From: David Cerutti <dscerutti.gmail.com>
Date: Fri, 25 Mar 2016 00:21:12 -0400

I certainly agree with the point Bill is driving at. However, to keep my
code clean I'll have to adhere to exact definitions. I think that I have
determined what I need to do thanks to this discussion. I do note that the
topologies contain some impropers where the central atom is the third atom,
others where it is the second. Some impropers can be matched to their
parameter file definitions by reading the peripheral three atom types as
they appear, but others must be read in reverse order to make sense. As
Bill and Dave have pointed out, all four cases are fine--the improper
dihedral angle will be the same even, irrespective of the ordering issues.
For the purposes of parameter development, however, my goal is to tie the
impropers back to their original parameters from within the parm.dat file
so that everything can be traced back to its source. I will need to let
mdgx know that:

1.) Proper and improper dihedrals can appear with the same sequence of four
atom types, and must be treated as distinct.
2.) Impropers can be rigorously traced back to their original parameters
only by a complex process of determining the central atom, then ordering
the peripheral atoms in alphabetical order with respect to their types.
The parameter files contain a set order for these things but the
topologies, for the aforementioned reason, do not, so only in the parameter
development module will I try to dissect the improper information to this
degree.

If any of the troopers who are still following this thread have anything
else to add, or see something wrong with the above logic, I'm happy to
learn the intricacies of all this. Otherwise I just wanted to conclude the
thread so any future readers won't be left hanging.

Thanks all.


On Thu, Mar 24, 2016 at 9:35 PM, Bill Ross <ross.cgl.ucsf.edu> wrote:

> If I'm right, the only sense of order mattering is that the forces are
> different in the instant, but no matter what the shape of the energy
> well that bottoms at planarity, it is merely an artifact of the force
> field, and one shape is no more correct than another. And random
> ordering might be better scientifically as a result, except that it
> would affect exact repro of results.
>
> The only reasons I remember for fixing the order is so that test diffs
> of prmtop internal order and test energies don't diverge.
>
> Bill
>
> On 3/24/16 3:40 PM, David Cerutti wrote:
> > For Bill's question, my supposition would be that order doesn't matter so
> > long as the angles between all atoms around the improper are precisely
> 120.
> > But, even if those are the equilibria the precise positioning in a
> > simulation will not be.
> >
> > Here's what I'm doing, over the quizzical looks from my officemate. Put
> > your arm flat on a high table so your humerus is in the plane of the
> > tabletop. Your elbow is point "C", the center of the improper, your
> > shoulder is point A, your hand point B, and place a coffee cup on the
> table
> > as point D. Put your elbow at 120 degrees, so the A-C-B angle is 120.
> > Rotate your arm so that your shoulder and elbow stay in the plane of the
> > table but your hand lifts off. That's bending between A-C-B and A-C-D.
> > Now extend your elbow outwards until your hand is nearly straight. The
> > angle between the A-C-B and A-C-D planes is not changing, but visualize
> the
> > plane of B-C-D. The incline of the B-C-D plane is changing despite the
> > fact that the incline of A-C-B is not, and the A-C-D plane is, as always,
> > in the plane of the table. That's my thought experiment to convince
> myself
> > that the order of atoms in impropers matters, but perhaps only in the
> limit
> > of significant departures from planarity.
> >
> > I just spilled my coffee.
> >
> >
> > On Thu, Mar 24, 2016 at 3:24 PM, Bill Ross <ross.cgl.ucsf.edu> wrote:
> >
> >> David, why does the order matter, aside from consistency, as long as the
> >> function is planar, and possibly the central atom is in position 2 or 3?
> >> My impression is that if you randomly assigned atoms for each improper
> >> and ran a long enough simulation, the ensembles would converge. (Bearing
> >> in mind that impropers seem like a somewhat hacky way of enforcing
> >> planarity in the first place.)
> >>
> >> Regards,
> >> Bill
> >>
> >> On 3/24/16 4:23 AM, David A Case wrote:
> >>>> 2.) Can anyone confirm that, for any atom types A, B, and D, an
> >> improper A
> >>>>> B C D is equivalent to an improper D B C A and also B D C A, or any
> >> other
> >>>>> permutation that leaves C in position 3?
> >>> Above is wrong: the fact that the order of A,B and D matters is the
> >> origin of
> >>> requiring them to be an alphabetical order. The only invariant thing
> >> you can
> >>> do with a torsion (proper or improper) is to reverse it: ABCD == DCBA.
> >> All
> >>> other permutations are different.
> >>
> >> _______________________________________________
> >> 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 Thu Mar 24 2016 - 21:30:03 PDT
Custom Search