A standard deviation (I would think correctly that's the SD of and
not from the mean i.e. the standard error) implies that you assume a
Gaussian distribution with a shape determined by SD. So actually you
are comparing two probability distributions. My first thought would be
that their overlap would give you a hint as to how significant the
difference (another probability) is. Of course, all your statistics is
estimated from what you already have and one typically assumes that this
is representative for the property under study,,,
Has DELTA S been calculated to the "same" quality as DELTA G? I would
think that the entropy is much harder to catch than the energy.
On Fri, 14 Sep 2012 10:28:07 +0200
George Tzotzos <gtzotzos.me.com> wrote:
> For a large number of DELTA G binding (Complex - Receptor - Ligand)
> the standard deviation is > 2 when DELTA G is ~30kcal/mol
>
> Translating DELTA G to dissociation constant, a 2kcal/mol difference
> (either way) would mean dissociation constants differing by two
> orders of magnitude from the mean.
>
> To put it in a different way. If DELTA G binding for ligand 1 to
> receptor A is 30kcal/mol and DELTA G binding of ligand 2 to the same
> receptor is 28kcal/mol with a standard deviation of 2, can one infer
> that the receptor has higher binding affinity for ligand 1?
>
> Furthermore, if one factors in DELTA S with SD 2 (or more)
> interpretation of the results becomes much more tenuous.
>
> And last, but not least. I have obtained standard deviations varying
> between 2 and 5 for nmode DELTA S calculations involving different
> ligands with the same receptor but the same number of frame readings.
> Is this to be expected?
--
Scanned by iCritical.
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 14 2012 - 02:00:04 PDT
