[MMTK] Energy minimization

Konrad Hinsen research at khinsen.fastmail.net
Wed Oct 12 16:23:08 UTC 2011

On 11 Oct, 2011, at 20:34 , Новиков Глеб wrote:

> I wounder to know about algoritms of energy minimization presented in the MMTK. What exactly type of that algoritms is most suitable for the common proteins ( 2000-5000 atoms ) ?  I've tried to use the example script for energy minimization

MMTK has two minimizers, one implementing steepest descent (more or less the same as running MD at zero temperature), and the other one implementing the conjugate gradient algorithm. Steepest descent is most suitable when you are far from a minimum, whereas conjugate gradient is best when you are already close to a minimum, close enough for the potential energy surface to be approximately harmonic on the length scale you are exploring.

> What average number of steps I should define for my proteins for obtaining well minimized universe  ? Also I wounder to know about minimization of membrane proteins? Does this
> procedure must be conducted in the explicit membrane environment or the simulation in the vacuum would be reasonable?
That's hard to answer without knowing why you want to do energy minimization on a membrane protein.

> Finally small qustion about scripting :) If I save my universe in pdb after minimization e.g
> universe = InfiniteUniverse(Amber94ForceField())
> universe.protein = Protein('bala1')
> minimizer = ConjugateGradientMinimizer(universe,
>                                        actions=[StandardLogOutput(50)])
> minimizer(convergence = 1.e-3, steps = 10000)
> universe.writeToFile('insulin_with_h.pdb')
> does I obtain minimized structure of my protein?

Yes. Note however that the numerical precision of the atom positions in a PDB file is very small. If you minimize well, save to PDB, and then load back from PDB, you can be quite far from your minimum due to the loss of precision. If you want to store your configuration for loading it back into MMTK, you should use MMTK.save and MMTK.load, which store the full available precision.

Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: research AT khinsen DOT fastmail DOT net

More information about the mmtk mailing list