Re: [AMBER] problem with MM-PBSA

From: case <>
Date: Tue, 28 Dec 2010 09:19:22 -0500

On Tue, Dec 28, 2010, Maryam Hamzehee wrote:
> I have calculated binding free energy for my complex of ligand-receptor
> and its mutant form (one of the residues in ligand was mutated to ALA)
> using the MM-PBSA method (Perl version) implemented in AMBER. I
> ignored the contribution of entropy in the calculations. Because it is
> computationally and apart from this when comparison of binding energy
> is desired entropy can be neglected.

This is not correct. A big portion of the total entropy of binding (the
solvent portion) is included in the implicit solvent model. The "entropy"
part of MM-PBSA is the solute configurational entropy, which is only a part of
the total entropy.

So, it is not correct to compare MM-PBSA "without entropy" to a ligand binding
energy. There is, in fact, no simple way to get enthalpy and entropy
contributions to binding from MM-PBSA: only the total free energy is

[Jason: maybe we/you should change references to "entropy" in mmpbsa_py to
"solute entropy" or something similar?]

> The residue that I mutated to Ala, was Tyr and it has been proven
> experimentally that this residue is so important in interaction with
> receptor and it is one of hot spots in binding site of interaction
> by providing stacking with phenyl alanine of receptor. But
> when I mutated it to Ala and the binding energy was calculated,
> it was observed that the value of binding energy was smaller than
> wild type more likely to negative values than the wild type and
> accordingly resulted to that improved affinity toward receptor by
> this substitution in spite of our expectation that it is highly
> unfavourable substitution.

A lot depends on the magnitudes of the differences you are talking about. It
is rarely helpful to use MM-PBSA for a single comparison, since (even under
favorable circumstances) errors arising from using an implicit solvent model
can be several kcal/mol. Leaving out the solute entropy further complicates
the situation. MM-PBSA (in my view) is best viewed as an analysis tool, to
help understand more rigorous calculations (or experiments) by suggesting
what sorts of interactions should be considered in understanding ligand
binding. There are circumstances where it can be used as a quantitative tool
for prediction, but these usually involve lots of comparisons (say in virtual
screeening) where one is looking more for trends than for individual results.

...good luck...dac

AMBER mailing list
Received on Tue Dec 28 2010 - 06:30:05 PST
Custom Search