[gmx-users] another pull-code question

chris.neale at utoronto.ca chris.neale at utoronto.ca
Fri Jul 2 17:07:03 CEST 2010

Sounds correct to me. One can use any method to generate the starting  
conformations. Some methods may yield equilibrium starting structures  
faster than other methods, but they are all ok in principle.  
Nevertheless, as with any simulations, one must ensure that care is  
taken to maintain equilibrium and measure convergence (unless a  
non-equilibrium approach is explicitly desired).

I also agree that people should be reading your paper  -- although  
maybe they are and I just misinterpreted some of the comments that I  
read. Perhaps it would be less work for you if you simply put in a  
line like "this tutorial is very helpful, but if you try to save an  
hour and skip reading the associated paper then things are likely to  
go terribly wrong somewhere!" .. and if you already have that line,  
then perhaps a larger font ;)


-- original message --

I definitely agree.  I think the general point (which I know I've been a bit
off-hand about explaining) is that, if properly applied (key caveat!), just
about any combination of pull method + pull geometry can be used to  
conduct SMD.
   You're quite right that just trying to shoot some molecule across  
your system
can have terrible effects for complicated environments.  What I've  
always meant
is that people shouldn't feel constrained to using only the settings I have in
the tutorial, since they are not the most versatile.

I think I'll add some discussion on this topic to my tutorial, which seems to
have become quite popular.  I guess I was hoping that people who used the
tutorial would refer to my paper (from which the tutorial is derived), wherein
we demonstrate that our results were not influenced by the pull rate  
(with both
k/2 and k/10).  Maybe I should make such things more obvious :)


chris.neale at utoronto.ca wrote:
> Hi guys,
> I just wanted to chime in on the idea that "how you perform the  
> initial pulling to get structures isn't all that important" because  
> I strongly disagree and think that it is, in general, immensely  
> important.
> OK, it's actually not important if we're talking about pulling two  
> Na+ ions, or two entirely rigid molecules in water, because you are  
> not going to affect their conformations by pulling too quickly.  
> However, let's take the other extreme and say that one wants to  
> generate initial conformations of something in a bilayer or an ion  
> in a membrane channel. In these cases, initial pulling that is too  
> fast can push the bilayer, ligand, or protein so far out of its  
> equilibrium conformation that your US will not sample enough  
> simulation time to re-equilibrate.
> To sum-up my point of view, the initial pulling simulations should  
> be done as slowly as you can afford, and if you want to have some  
> confidence that your rate is "slow enough", then you should probably  
> run at rate=k and rate=k/10 and check that the starting structures  
> that you get are not separated from each other by huge barriers  
> resulting from some non-equilibrium response at rate=k that was not  
> observed at rate=k/10. Even if you don't go through some procedure  
> like the one above, there is still no need to take risks by limiting  
> your setup pulling simulation to 1 hour of simulation when you're  
> going to then spend the next 2 months doing US (garbage in = garbage  
> out).
> Chris.

More information about the gromacs.org_gmx-users mailing list