Re: [AMBER] MMGBSA/MMPBSA Decomposition Analysis

From: Jason Swails <jason.swails.gmail.com>
Date: Wed, 27 Feb 2013 20:53:50 -0500

On Wed, Feb 27, 2013 at 7:28 PM, Vivek Shankar Bharadwaj <
vbharadw.mymail.mines.edu> wrote:

> Hi AMBER Users,
>
> I am new to MMPB/GBSA calculations and am studying the binding of toluene,
> fumaric Acid and benzylsuccinate to an enzyme. I was able to get the
> implementation to work successfully for my system. I have some questions
> regarding the analysis.
>
> I have been able to calculate the overall binding energies using MMGBSA,
> MMPBSA and decompose the binding energies among the key residues at the
> active site.
>
> I have the following questions with regards to analyzing the results
>
> 1. I find that the total binding free energies calculated from MMGBSA give
> negative free energies of binding while the MMPBSA binding energies are
> positive (~1 kcal/mol). Why would the MMPBSA free energies show
> unfavourable binding while MMGBSA show favourable? The MMGBSA and MMPBSA
> energies also do not display the same trend for my ligands.
>

Have you applied all bug fixes to AmberTools 12? You can check if more
patches are available by doing:

cd $AMBERHOME && ./patch_amber.py --check-updates

There was a bugfix not too long ago that was a pretty substantial fix to
the PB part of MMPBSA.py.

As for why binding free energies could be positive for PBSA and negative
for GBSA, that would depend on your system. The decomposition analysis may
be able to help with that.


> 2. I understand that the ligand also has to be mentioned in the print_res
> specification for decomposition. In the output, (delta's total energy
> section) the ligand contribution to the binding energy is also reported. I
> am finding it difficult to understand how the ligand's binding energy be
> decomposed onto itself?
>

You are not decomposing the binding free energy. You are decomposing the
energy itself. Atoms with the ligand interact with each other. Therefore,
there is a "self" interaction between ligand residues.

Furthermore, PB and GB are not pairwise decomposable, since the dielectric
boundary for PB and the effective GB radii for GB are dependent on all of
the surroundings. Therefore, intra-ligand interactions will not
necessarily be the same in the bound and free states. All pairwise terms
should cancel (e.g., VDW, 1-4 VDW, 1-4 EEL, EEL, internal), but the
solvation terms --specifically ESURF/EGB or ECAVITY/EPB--will differ (that
would be sas and pol in the decomp language).

3. I find that the majority (~90%) of the binding energy (deltas of the
> total energies in the output) for the toluene binding is attributed to the
> ligand itself. [Meaning that the B.E remains undecomposed]. What could I be
> doing wrong? or is this due to the hydrophobic interactions stabilizing
> toluene? In the case of fumaric acid the ligand contribution was ~20% only.
>

It sounds like solvation effects -- you're effectively measuring the
difference between solvating a hydrophobic compound in a protein vs water.
 Virtually all ligand-receptor interactions for toluene will be weak, since
the charges on toluene should be pretty small, and vdW interactions are
short-ranged and typically pretty weak. The surface area effects here I
would expect to be pretty large, and that is a difficult value to
'decompose'.

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 Wed Feb 27 2013 - 18:00:03 PST
Custom Search