[gmx-users] Normal mode eigenvalue units, nm^-2 ??

Berk Hess gmx3 at hotmail.com
Fri May 20 08:49:46 CEST 2005

>From: Marc Baaden <baaden at smplinux.de>
>Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>To: Discussion list for GROMACS users <gmx-users at gromacs.org>
>Subject: Re: [gmx-users] Normal mode eigenvalue units, nm^-2 ?? Date: Thu, 
>19 May 2005 17:21:48 +0200
> >>> "Berk Hess" said:
>  >> That should work.
>  >> But unless you simulate a very low temperature, I think this would
>  >> not make sense for a protein, since the principal components with
>  >> the largest eigenvalues are always diffusive at room temperature,
>  >> not harmonic.
>That's a very interesting aspect, I think. So what's your take then on
>quasi-harmonic analysis for room temperature simulations (be it for entropy
>estimation or comparison to NMA or ??) - should one remove the modes that
>are diffusive and continue to work on the others or is there another 
>flaw involved.

That one could do.
But usually the slowest/largest-msf modes are the ones one is
most interested in.
In proteins at room temperature the slowest are almost always
not sampled enough to obtain resonably accurate eigenvalues/msf's,
see for instance:
B. Hess. Convergence of sampling in protein simulations.
Phys. Rev. E 65, 031910 (2002)

There is an extensively studied subject in proteins which is called
the harmonic/anharmonic transition, which occurs, if I am not mistaken,
around 170 K. See for instance the work of Jeremy Smith.

>I wonder whether diffusion would show up as a mode that is not mainly
>vibrational in character (in an analysis like that done at the end of
>ref [1]). Maybe the author's statement "[..] the lowest 3 quasiharmonic
>modes are due to irreversible transitions between conformational substates
>(structural drift)[..]" reflect exactly this.

See my ref above.


Play online games with your friends with MSN Messenger 

More information about the gromacs.org_gmx-users mailing list