The way the standard deviations are calculated depend on whether you used a
single trajectory or multiple trajectory approach. Given that your
standard deviations are not that large (and actually get smaller for
composite terms like G gas and G binding), this means that you're using a
single trajectory.
On Mon, Jul 9, 2012 at 9:35 AM, Massimiliano Porrini <M.Porrini.ed.ac.uk>wrote:
> Dear Amber developers,
>
> I am dealing with some MMGBSA calculations and their statistical
> uncertainties and I came across this very useful and explanatory email
> from J. Swails in the Amber archives.
>
> However I am wondering how the standard deviations on the DeltaG_gas
> and DeltaG_solv are handled by MMGB(PB)SA.py.
>
> For instance in one my calculations I obtained the following results:
>
>
> ****************************************************************************************
> Differences (Complex - Receptor - Ligand):
> Energy Component Average Std. Dev. Std. Err. of
> Mean
>
> -------------------------------------------------------------------------------
> VDWAALS -183.0991 11.5396
> 0.1632
> EEL -1166.7511 55.0115
> 0.7779
> EGB 1213.1014 54.1049
> 0.7651
> ESURF -25.3833 1.3550
> 0.0192
>
> DELTA G gas -1349.8502 62.1130
> 0.8783
> DELTA G solv 1187.7181 53.0433
> 0.7501
>
>
> DELTA G binding = -162.1321 +/- 11.7599
> 0.1663
>
>
> ****************************************************************************************
>
> If I understood correctly, the standard deviations of each
> contribution (VDWAALS,
> EEL, EGB and ESURF) are worked out as Jason Swails explained.
> But, for instance, how is the standard deviation
> of "DELTA G gas" derived? Being that just the summation of VDWAALS and EEL?
>
It is calculated by adding VDWAALS and EEL for every frame, so there is a
"DELTA G gas" for each frame (that is made up of the sum). The standard
deviation is taken of this population. Note that the G gas is really
composed of the internal potential terms (including the 1-4 non-bonded
interactions) as well, but they are only included in that value if printed
to the output file.
>
> And consequently, how is the standard deviation derived for "DELTA G
> binding", being just
> the summation of DELTA G gas and DELTA G solv ?
> In this case the SD is drastically reduced: from ~62 and ~53 to ~12 .
>
The DELTA G is calculated the same way. That is, the "complex - receptor -
ligand" value is calculated for each frame, and the average and standard
deviation is taken of this population.
By calculating the standard deviation this way, the correlation in the data
is fully included (thereby reducing the standard deviation).
HTH,
Jason
--
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/amber
Received on Mon Jul 09 2012 - 09:30:03 PDT
