[gmx-users] Energy minimization

[Florian Haberl]

Hi,

On Monday 10 January 2005 15:36, Andrea Bacconi wrote:
> Hallo everybody,
> I'm very new in the use of Gromacs 3.2 and I would like to ask people
> with more experience than me about the energy minimization procedure.
> I'm trying to minimize the energy of a protein I obtained by homology
> modelling (it's about 500 aa) following a paper where the authors did a
> work similar to the one I'm doing now. My problem is simple:
> in this paper authors say : "Electrostatic forces were calculated with
> the particle-mesh Ewald algorithm. Initial energy minimization was done
> with the steep descent algorithm (1000 steps) followed by conjugate
> gradient to a maximum force of 0.1 Kj mol^-1 nm^-1 ." do they mean that
> in the same em.mdp script they used both s.d and c.g one after the
> other?

You write first a em.mdp with s.d. parameters, after it has finished 
calculation you do the same with c.g. (and the results from the first run).

I only use the l-bfgs (see handbook page 130), and get good results, only 
problem of this is that it^s not yet parallelized so you need one powerful 
cpu, or more time.

> I all the tutorials I found the energy minimization was done by
> using either one or the other algorithm, or using the nstcgsteep option
> with c.g as main algorithm, so how can you use (if its possible) one
> algorithm and then the other in the same em.mdp script?
> Minimizing the protein energy using Gromacs with single precision, no
> constraints, cg and nstcgsteep gives an Fmax of 50 Kj mol^-1 nm^-1 , if
> I go below this value the system converge to machine precison but not to
> the requested Fmax (lets say 10 Kj mol^-1 nm^-1 ).

You can also decrase your steepsize or work with double precision

> My goal is to improve the reached Fmax up to the paper's level, if it's
> possible, and then run a  full molecular dinamic simulation in water.
> Thank you for your attention and help


Greetings,

Florian

-- 
-------------------------------------------------------------------------------
 Florian Haberl                            Universitaet Erlangen/
 Computer-Chemie-Centrum    Nuernberg
                                                      Naegelsbachstr. 25
                                                      D-91052 Erlangen
  Mailto: florian.haberl AT chemie.uni-erlangen.de
 -------------------------------------------------------------------------------