AMBER Archive (2009)

Subject: Re: [AMBER] GBSA: dielectric symmetry

From: David A. Case (case_at_biomaps.rutgers.edu)
Date: Thu Mar 05 2009 - 18:05:51 CST

On Fri, Mar 06, 2009, Giacomo Bastianelli wrote:
>
> I have been using MM_GBSA to calculate free energy of binding of protein
> complexes and I have tried to optimize the internal and external
> dielectric for maximizing the correlation with experimental data.
> What I observed is that if I calculate the free energy with int-diel = 1
> and ext-diel = 80 I obtain different results from the case of int-diel = 2
> and ext-diel = 40. This is surprising since I was expecting the same value
> because the symmetry in the GB formula: (1/int-diel - 1/ext-diel).
> Am I missing something?

Among other things, (1/1 - 1/80) does not equal (1/2 - 1/40). So you
should certainly expect (and find) different results.

[As an aside: it seems rather dangerous to "optimize the internal and
external dielectric to maximize correlation with experimental data".
You can easily fool yourself here: if you do enough different
calcualtions (even at "random"), one will fit the data better than the
others. Remember that you want the right answer for the right reason.]

...dac

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber