Re: AMBER: TI Approach - mass perturbation

From: David A. Case <>
Date: Mon, 10 Dec 2007 23:13:08 -0800

On Mon, Dec 10, 2007, Ilyas Yildirim wrote:
> My question is regarding the Thermodynamic Integration implementation
> in AMBER 9. In the amber mailing list, Petr Kulh√°ne kwrote that in a TI
> calculation, mass changes are not reflected in perturbation calculations.

Note that this comment has to do with earlier versions of Amber, which carried
out TI calculations in a different fashion.

For Amber 9, please see the last paragraph on p. 153 of the Users' Manual.
In classical mechanics, the Boltzmann distribution of configurations (and
hence all averages like <DV/DL>) are independent of the masses. We
arbitrarily decide to use the masses in the prmtop file for the first group.
(This is enforced in mdread.f.)

This does mean that the actual trajectories that you get on going from 0->1
will differ from those on going from 1->0, but any converged statistical
averages will be the same. Note that, following classical dynamics, we don't
even bother to calculate kinetic energy terms, since they should cancel in any
legitimate thermodynamic cycle.

> Does that mean that I cannot
> do any mass perturbation (like a perturbation from oxygen to nitrogen) in
> TI Approach?

No: you certainly can to transformations that include changes in masses. As
always, you have to subtract two sides of a thermodynamic cycle (such as a
transformation in solution vs. one in an enzyme, or a transfomation in
gas-phase vs. one in solution) in order to get a physically meaningful number.

[Amber 10 will have the ability to carry out thermodynamic integration on the
masses for path-integral dynamics. Here, since one is taking quantum
(nuclear) dynamics into consideration, there will be a non-trivial kinetic
energy contribution to the partition function, and this can be used to
estimate kinetic isotope effects and equilibrium isotope partitioning. Or,
it could be used as part of an "alchemy" change (such as N to O, etc.)]

...hope this helps....dac

The AMBER Mail Reflector
To post, send mail to
To unsubscribe, send "unsubscribe amber" to
Received on Wed Dec 12 2007 - 06:07:26 PST
Custom Search