Re: [AMBER] interaction energy

From: Aust, Susanne <>
Date: Mon, 14 Mar 2011 14:32:01 +0100

thanks again to Jason, Thomas and Bill,
I'm waiting for the new AmberTools release.


-----Ursprüngliche Nachricht-----
Von: Jason Swails []
Gesendet: 10 March 2011 16:25
An: AMBER Mailing List
Betreff: Re: [AMBER] interaction energy


On Thu, Mar 10, 2011 at 3:24 AM, Aust, Susanne <> wrote:

> Dear Thomas,
> thank you very much for the fast answer and the long literature list!
> I calculate the average structure over the equilibrated part of the QM/MM
> MD, so I hope, I have a realistic ensemble.

It really doesn't matter how long you equilibrated your system for -- you
are still using only a single structure that cannot possibly represent an
entire ensemble. Moreover, average structures tend to be very warped and
adopt conformations of very high strain, especially if your system is
flexible. Consider a methyl group with 3 hydrogens attached. These
hydrogens are generally rotationally degenerate, and since the average
structure averages over the rotation of this group, your *average* structure
will see all 3 methyl hydrogens directly on top of one another (indeed, very
close to the methyl carbon they're all attached to), even after an RMS fit.
This alone will cause your energy to blow up.

In general, the energy of an average structure is not even close to the
average energy of the structures used to create that average. You would
really need to use MM/PBSA on many of the structures of your ensemble to get
even a remotely reliable binding energy. As Bill mentioned, will
be released with AmberTools with the capability of performing QM/MM-GBSA

Hope this helps,

> I try to calculate MM-GBSA and PBSA, but I got positive values for PBtot
> (for example 60.41 kcal and GBtot 2.02 kcal/mol) When I search for the
> problem in the mailing list, I could find following thread:
> Dear Venessa,
> > I also had similar kind of problem in past, but in my
> > case I have an Zn ion and PB is not very good in
> > dealing with ions.
> > Other thing was ion was free floating (not bound to
> > anything), therefore during calculation surface is
> > changing all the time which also gives huge error.
> > I hope this help. Or you may want to read papers about
> > how PB works
> > Thanks
> I think the best way is to use MM-PBSA or GBSA, but I don't know, where is
> the failure. Can I use the QM/MM-MD for the MM-GBSA calculations?
> I restraint the md simulation relatively restrictive, by reason of time,
> could there be the failure?
> Yes I know, that the Khandelwal et. al (2005)paper is due to methods not
> the best, but my boss like QM/MM calculation and Khandelwal got an r² of 0.9
> with this approach. So we try to go this way, but with the Amber forcefield.
> Thanks a lot!
> schöne Grüße aus dem frühlingshaften Deutschland :-)
> Susanne
> -----Ursprüngliche Nachricht-----
> Von: []
> Gesendet: 10 March 2011 10:40
> An: AMBER Mailing List
> Betreff: Re: [AMBER] interaction energy
> Hi,
> >> I generate a average structures after md .
> isnt that a way of undoing what you gained from doing the MD, namely a
> realistic conformational ensemble? A single structure, even if it were the
> global energy minimum would give you an interaction energy at best, but
> you need a free energy to compare to Ki values.
> >> I could'nt use MM-GBSA or PBSA due to zinc in the active site.
> I believe it would be possible to incorporate Zn in MMPBSA. All you need
> is a reasonable ion radius. What went wrong when you tried?
> >> Is it possible to calculate a single QM/MM interaction energy for this
> >> type of complex in Amber ( this was done in the paper, from which I have
> >> the workflow, but they worked with the Tripos forcefield)?
> You could do a QM/MM minimization of your average structure and compare
> the final energies for the complexes to their Ki, but any correlation
> would probably be accidental (especially if you throw away the water
> around the complexes). There is an enormous amount of literature out there
> on protein-ligand interactions and why it is an exceptionally hard problem
> to tackle, maybe the paper you refer to is not an optimal starting point
> for what you want to do...
> Check out e.g.:
> Michel, J. Essex, J. (2008) Hit identification and binding mode
> predictions by rigorous free energy simulations. \emph{J. Med. Chem.},
> \textbf{51}, 6654--6664.
> Jorgensen, W. (2004) The many roles of computation in drug discovery.
> \emph{Science}, \textbf{303}, 1813--1818.
> Gilson, M. Zhou, H. (2007) Calculation of protein-ligand binding
> affinities. \emph{Ann. Rev. Biophys. Biomol. Struct.}, \textbf{36},
> 21--42.
> deAzevedo, W. Dias, R. (2009) Computational methods for calculation
> of ligand-binding affinity. \emph{Curr. Drug Targ.}, \textbf{9},
> 1031--1039.
> Klebe, G. (2006) Virtual ligand screening: Strategies, perspectives and
> limitations. \emph{Drug Disc. Today}, \textbf{11}, 580--594.
> Gohlke, H. Klebe, G. (2002) Approaches to the description and
> prediction of the binding affinity of small-molecule ligands to
> macromolecular receptors. \emph{Angew. Chem., Int. Ed.}, \textbf{41},
> 2644--2676.
> Kind Regards,
> Thomas
> Dr. Thomas Steinbrecher
> formerly at the
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
> _______________________________________________
> AMBER mailing list
> _______________________________________________
> AMBER mailing list

Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
AMBER mailing list
AMBER mailing list
Received on Mon Mar 14 2011 - 07:00:04 PDT
Custom Search