[gmx-users] More EM Questions.

Erik Lindahl lindahl at stanford.edu
Mon Jul 14 22:28:00 CEST 2003

Hi Joel,

> Steepest Descents converged to 150 in 90 steps
>  Potential Energy  = -3.23481e+03
> Maximum force:  1.43325e+02
> I've written EM algorithms from scratch before and GROMACS obviously 
> is not being thorough in its energy minimization.  Sure, the potential 
> energy is being reduced overall, but it doesn't seem like EM is 
> examining the system as a whole; just a handful of atoms.

You've apparently specified in your mdp file that you only need 
convergence to within about 150 kJ/mol, and when the energy isn't 
changing more than this for a couple of steps Gromacs correctly stops.

Steepest descent is a quick and simple algorithm, but it is only 
intended to get your system to a state suitable for MD. There are 
better algorithms in the literature if you are interested in EM per se.

> Do I need to write a script that will do many EM runs in a row so that 
> I can get this system ready for some MD?
No, -3e3 kJ/mol seems fine - you should be able to do MD from there.

Remember: We don't use energy minimzation to find the lowest energy 
structure (because that will change anyway as soon as we start doing 
md), but we use it to remove collisions and other bad contacts that 
would otherwise crash the md run.

> One other thing I've been thinking about: some people have donated 
> sets of force field parameters to GROMACS.org:
> ffG43a2x.tgz (GROMOS96 modified to describe lipids)
> ffgmx_lipids.tar.gz (lipid params for use with ffgmx)
> Is this part of my EM problem?  Do I need these?

No, that's only for lipids, and it won't affect your energy 
minimazation significantly even in that case.

More information about the gromacs.org_gmx-users mailing list