Re: [AMBER] validation of MMPBSA results

From: Jason Swails <>
Date: Fri, 26 Jul 2013 15:29:07 -0400

On Fri, Jul 26, 2013 at 11:52 AM, Asmita Gupta <> wrote:

> Dear users,
> I have done MMPBSA analysis on a set of RNA MD trajectories. For each of
> the trajectories, analysis has been done on 2000 and 17000 frames in order
> to check whether number of frames have any effect on the net result. The
> final DelG values that i am getting in different cases is somewhat like
> this (without running nmode)

17000 frames seems like overkill to me. If you take ~200 to 500 snapshots
from your ensemble spaced throughout the trajectory, I don't expect you to
see any difference in the free energy or the standard deviation.

> Set1: -3716.03
> Set2: -4074.06
> Set3: -3937.45
> Set4: -3945.89 (all values for 17000 frames)
> Now since each of the sets represent RNA structures with single base
> substitutions, how can i verify whether the difference in the DelG values
> is actually significant or not? e.g. in set1 and set3, the difference is
> ~200kcal/mol, is that a significant shift in the conformational free
> energy?

I don't think what you're doing is a fair comparison. Single point
energies are meaningless in force fields -- they are arbitrary for each
system (and not comparable for different systems) -- only (select) energy
differences matter.

If Set 1 and Set 2 differ in the identity of one nucleic acid, they are not
comparable (since each system has their own 'arbitrary' zero-energy value
that are unrelated to each other). It is for this reason that such studies
are always carried out using a thermodynamic cycle, typically with some
sort of model system that ideally has an easily measurable experimental
value. You need effectively a change of energy changes.

Are there any literature references available for such studies..
> I also tried using MMRISM method, but it is far too time consuming for
> 17000 frames. The values given by PB and RISM are quite different e.g.
> PB value (DelG) for set5: -3940.22
> RISM (DelG) for set5: -1432.24
> Why such a large difference is coming? It again raises the question of
> result validity...

I don't think these values are comparable. You should be comparing free
energy differences instead of single-point free energies to measure

Good luck,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Fri Jul 26 2013 - 12:30:04 PDT
Custom Search