[Dock-fans] False positives

Scott Brozell sbrozell at scripps.edu
Thu Mar 29 17:54:39 PST 2007


Hi,

Ignoring the kinetic aspect of binding (for which DOCK produces
essentially no information) and focusing on the thermodynamic:

http://blur.compbio.ucsf.edu/pipermail/dock-fans/2007-February/000910.html
"Since the AMBER score is calculated as:
E(Complex) - [ E(Receptor) + E(Ligand) ],
it is an approximation to the binding energy.
In general, binding energies and binding affinities are correlated."

However, fortuitous cancellation of error is the bread and butter
of much computational chemistry.  In particular, to get an appreciation
of the large error bars of computed binding affinities, read this article:
@article{tirado-rives2006,
  author = "{Julian Tirado-Rives and William L. Jorgensen}",
  title = "Contribution of Conformer Focusing to the Uncertainty in Predicting Free Energies for Protein--Ligand Binding",
  journal = jmc,   volume = 49,
  pages = {5880--5884},
  year = 2006 }
@string{jmc = "J. Med. Chem."}

I apply the definition of Amber score below.

On Thu, 29 Mar 2007, Gustavo HMF Souza wrote:

> After rescore with AMBER, many ligands appears to be near active site - the
> selected_spheres, but they are not inside them like the specific ligand (as
> a positive control). Ok. Until now everything is all right, therefore, after

What kind of movable region was used ?

> AMBER these ligands does not have a negative score. All of them, including
> those that are outside the selected spheres, have almost the same score!!???
> How can I fix this ? I think that calculations is everything wrong!

I'm not sure I understand your statements.  The fact that Amber score
does not differentiate between ligands may or may not indicate an error.

> 1- Can I affirm that positive AMBER score means inhibition? Or substrates?
> If inhibitor why?

A positive AMBER score implies poor binding.
Thus, a positive AMBER score does not imply inhibition.

> 2- Why my score after dock calculations with a specific knowing ligand with
> a high Ki is lower than another with  a lower Ki? Please I need help ... If
> is the problem is not clear, Can I send some input/output problem files to
> you ?

Receptor + Ligand  <==>  Complex
K = [Complex] / ( [Receptor] [Ligand] )
-RT ln(K) = Delta G = binding affinity ~= binding energy ~~= Amber Score
Thus we hope that an empirically strong binder will have a
big negative score.

> 3- How can I interpret the data from AMBER (negative score -don`t bind;
> positive score -means bind)????????????

No, negative score implies bind; positive score implies not bind.

> 4- How can I plot the data coming from AMBER score like the tutorial
> (showing the decoys and binders)?

Plot Amber Score vs ligand using blue bars for binders and red for decoys.

> With all my best, (and urgent!)

:++=O  you got queued at the front
Scott



More information about the Dock-fans mailing list