[gmx-users] regarding 2002 post showing protein unfolding with long timesteps

[David van der Spoel]

chris.neale at utoronto.ca wrote:
> David van der Spoel, can you please comment on this post of your from 2002:
> 
> "long timesteps combined with dummies give me unfolding proteins while 
> without
> dummies and 2 fs the systems are stable. These are long-time effects (i.e.
> after a couple of ns things start to happen). I have to do more test
> before I can conclude anything."
> 
> reference: http://www.gromacs.org/pipermail/gmx-users/2002-July/002171.html
> 
> I searched your name with the work timestep and looked through the ~200 
> titles of posts, but couldn't find anything.
> 
> I am currently considering a 5fs timestep and, even though it is nicely 
> described in the gromacs 4 paper, I wanted to ensure that whatever 
> problem you might have been experiencing is now solved. Alternatively, 
> if it was a particular type of system that appears to be refractory to 
> the 5fs timestep options (properly instituted with dummies, etc) then 
> that information would also be valuable to me.
> 
> My usage will be for free energy calculations of a protein in a lipid 
> bilayer so obtaining information on the dynamic processes is not 
> important to me here.
> 
> Thanks,
> Chris.

I don't recall this to be honest, but obviously there are a lot of 
factors playing in. The most important one is the force field of course, 
but also the treatment of cutoffs. Many problems of instabilities go 
away when one uses PME instead of cutoffs. I think that many people use 
at least 4 fs time steps routinely, and the integration algorithm is 
fine with that. If you want to know more than you should probably test 
it explicitly.



-- 
David van der Spoel, Ph.D., Professor of Biology
Molec. Biophys. group, Dept. of Cell & Molec. Biol., Uppsala University.
Box 596, 75124 Uppsala, Sweden. Phone:	+46184714205. Fax: +4618511755.
spoel at xray.bmc.uu.se	spoel at gromacs.org   http://folding.bmc.uu.se