AMBER Archive (2006)

Subject: Re: AMBER: How to evaluate the binding energy of a ligand?

From: David Mobley (dmobley_at_gmail.com)
Date: Tue Apr 11 2006 - 14:27:13 CDT

Alfredo,

> When trying to calculate the binding energy as explained in Tutorial 2 I found
> that (after running the dinamic over the ligand (not shown)) the
> binding energy
> is positive, indicating that the complex is not stabilized. This is not true
> since Autodock finds a negative energy component for the affinity and also the
> binding occurs experimentally. Which may be my problem??,

There could be one of any number of problems. First, it's important to
realize that what you are trying to do here is to calculate the
binding enthalpy, which is not the same as the binding free energy.
Also, this is an absolute approach which is generally hard to make
work, basically because you are subtracting two very large numbers
(about -9000 kcal/mol) with fairly large uncertainties (you're getting
RMS fluctuations of about 100 kcal/mol) and trying to get a value for
the difference that is accurate to about 5-10 kcal/mol. In my
experience one would typically need to run several hundred nanoseconds
at least to get the error bars for this down small enough for results
to be reasonable (I've looked at this for some binding free energy
calculations I was doing, and I was getting an uncertainty of about 10
kcal/mol with 5 ns of data). My *guess* as to what is going on is that
the uncertainties in your results are simply swamping out the binding
energy. Maybe the folks who do MM-PBSA will disagree with me on this
though. However, I would just say that proper error analysis
(including an estimate of the autocorrelation time) is key to figuring
out whether this is going on.

There are, of course, other possibilities:
(1) Your starting pose is incorrect, so the computed unfavorable
binding energy is actually correct (telling you that the ligand
doesn't bind in that orientation)
(2) The bound structure of the protein is not the same as the unbound
structure, so the protein is sterically clashing with the ligand
(since you're using the unbound structure) and hasn't had time to
deform to accomodate it.
(3) Binding is somehow entropically driven, and since you're only
trying to compute the enthalpic part of the binding energy, you're
missing the part that acutally drives binding.

Again, my suspicion is the uncertainty issue is what's going on, but
you'll have to figure that out for yourself. There is also a
substantial literature on doing binding free energy calculations, but
at present I would say that these are mostly still under development,
and hard to do properly.

Best wishes,
David Mobley
UCSF

> Thanks
> Best wishes
>
> Alfredo Quevedo
> Dpto. Farmacia - Fac. Ciencias Quimicas
> Universidad Nacional de Córdoba
> Argentina
>
>
> ----------------------------------------------------------------
> This message was sent using IMP, the Internet Messaging Program.
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu