[gmx-users] RMSD

[Erik Lindahl]

On Saturday, June 21, 2003, at 08:13  PM, Kay Gottschalk wrote:

>
> Hi there,
>
> I am doing a free md of protein in water with ions, using the GROMOS96 
> FF with pressure and T- coupling and PME for electrostatics. I 
> positioned the ions with genion -random, as I read in the mailing list 
> that the positioning using the best potential is screwed up. Is that 
> so? The system equilibrates after 10-15 ps, but with a backbone RMSD 
> of ~ 0.1 nm to the original structure. Coming from NMR, that seems 
> rather high for me. Is this RMSd normal for free MD or do I have a 
> problem? What kind of problem might I have (like: start structure not 
> good or ions wrongly positioned)? Thanks for your help!
> Kay.

Hi Kay,

It's important to remember that the NMR RMSD and simulation RMSD aren't 
quite the same property.

In NMR you are always comparing an average structure (even if you have 
multiple structures, the experiment will average atomic motions) with 
the average crystal (or whatever :-). Even at equilibrium there are 
significant fluctuations over timescales 1-10ns.

In a simulation you get access to individual snapshots of "exact" atom 
positions at a single time point. Since atoms fluctuate, the individual 
frames will always have a larger RMSD than an average structure. If you 
want to compare, it is pretty common to calculate unphysical average 
structures over 1ns or so, and then calculate RMSD between this average 
and the experimental structure.

In general, the OPLS-AA/L forcefield produces better RMSDs than 
Gromos96 for proteins, but that doesn't necessarily mean it is better - 
David and I are currently comparing things like distance restraint 
violations, etc.

Cheers,

Erik