- Contemporary messages sorted: [ by date ] [ by thread ] [ by subject ] [ by author ] [ by messages with attachments ]

From: Jason Swails <jason.swails.gmail.com>

Date: Wed, 16 Feb 2011 13:10:46 -0500

Hello,

To add onto what Professor Case already said: sander/pmemd explicitly use

the exact cosine expressions for dihedral terms that is found in the manual

and every article describing the Amber force field. NAB and NMODE, on the

other hand, takes a different approach.

If the phase angle is 0 or 180 (0 or pi), then you can analytically expand

the cosine of any periodicity in terms of the cosine of the angle itself

using the multiple-angle formulae. For example, cos(2x) = 2 * cos^2(x) -

1. These formulae are trivially derivable using Euler's equation (e^(i * x)

= cos(x) + i sin(x)) and can allow for faster and more stable evaluation of

these dihedral terms (in addition to making derivatives much easier to

take). Whether the value is 0 or pi simply determines what sign the term is

supposed to have (as can be verified with Euler's equation, again). The

dihedral formulae cannot be so simply reduced for phase angles not equal to

0 or pi, so NAB/NMODE defaults back to the way sander does it for these

cases (which are atypical). Up until the release of ff99SBildn, a

periodicity of 5 was never encountered in Amber force fields, so the nab

architects never bothered deriving/implementing that case -- they stuck with

periodicities of 1, 2, 3, 4, and 6. As Professor Case mentioned, the

periodicity 5 case was recently derived and implemented (and fixed up) in

the development version, and should be released as a patch after we've been

able to run it through some tests.

What this means is that sander and pmemd can handle an arbitrary periodicity

regardless of what phase angle is assigned. However, if a phase angle is

assigned a value of 0 or pi, then NAB and NMODE can only accept values of 1,

2, 3, 4, (in the future 5,) and 6 as allowable periodicity values. The

tolerance, meaning how much the periodicity is allowed to differ from 0 or

pi in terms of radians, is 0.01 (which is ~0.6 degrees).

The reason for all of this possibly useless information is to demonstrate a

temporary workaround for this limitation in NAB and NMODE -- adjust the

phase slightly (by at least 0.6 degrees) for terms where the periodicity is

equal to 5 until a fix comes out. I can't guarantee it'll work, but based

on the limited code investigation I did I think it will.

Good luck!

Jason

On Wed, Feb 16, 2011 at 12:14 PM, case <case.biomaps.rutgers.edu> wrote:

*> On Wed, Feb 16, 2011, serena_leone.libero.it wrote:
*

*> >
*

*> > I would like to know if Sander in Amber10 or Amber11 can handle dihedral
*

*> > potentials described with additional coefficients in a frcmod file (i.e.,
*

*> V5
*

*> > besides V1, V2, V3 and V4).
*

*>
*

*> As far as I can see, there is no problem with sander (or with pmemd). The
*

*> nmode and nab codes need an update, which we are working on.
*

*>
*

*> ...dac
*

*>
*

*>
*

*> _______________________________________________
*

*> AMBER mailing list
*

*> AMBER.ambermd.org
*

*> http://lists.ambermd.org/mailman/listinfo/amber
*

*>
*

Date: Wed, 16 Feb 2011 13:10:46 -0500

Hello,

To add onto what Professor Case already said: sander/pmemd explicitly use

the exact cosine expressions for dihedral terms that is found in the manual

and every article describing the Amber force field. NAB and NMODE, on the

other hand, takes a different approach.

If the phase angle is 0 or 180 (0 or pi), then you can analytically expand

the cosine of any periodicity in terms of the cosine of the angle itself

using the multiple-angle formulae. For example, cos(2x) = 2 * cos^2(x) -

1. These formulae are trivially derivable using Euler's equation (e^(i * x)

= cos(x) + i sin(x)) and can allow for faster and more stable evaluation of

these dihedral terms (in addition to making derivatives much easier to

take). Whether the value is 0 or pi simply determines what sign the term is

supposed to have (as can be verified with Euler's equation, again). The

dihedral formulae cannot be so simply reduced for phase angles not equal to

0 or pi, so NAB/NMODE defaults back to the way sander does it for these

cases (which are atypical). Up until the release of ff99SBildn, a

periodicity of 5 was never encountered in Amber force fields, so the nab

architects never bothered deriving/implementing that case -- they stuck with

periodicities of 1, 2, 3, 4, and 6. As Professor Case mentioned, the

periodicity 5 case was recently derived and implemented (and fixed up) in

the development version, and should be released as a patch after we've been

able to run it through some tests.

What this means is that sander and pmemd can handle an arbitrary periodicity

regardless of what phase angle is assigned. However, if a phase angle is

assigned a value of 0 or pi, then NAB and NMODE can only accept values of 1,

2, 3, 4, (in the future 5,) and 6 as allowable periodicity values. The

tolerance, meaning how much the periodicity is allowed to differ from 0 or

pi in terms of radians, is 0.01 (which is ~0.6 degrees).

The reason for all of this possibly useless information is to demonstrate a

temporary workaround for this limitation in NAB and NMODE -- adjust the

phase slightly (by at least 0.6 degrees) for terms where the periodicity is

equal to 5 until a fix comes out. I can't guarantee it'll work, but based

on the limited code investigation I did I think it will.

Good luck!

Jason

On Wed, Feb 16, 2011 at 12:14 PM, case <case.biomaps.rutgers.edu> wrote:

-- Jason M. Swails Quantum Theory Project, University of Florida Ph.D. Candidate 352-392-4032 _______________________________________________ AMBER mailing list AMBER.ambermd.org http://lists.ambermd.org/mailman/listinfo/amberReceived on Wed Feb 16 2011 - 10:30:04 PST

Custom Search