[gmx-users] ED analysis: help on cosine content and overlap of the fluctuations

Ran Friedman r.friedman at bioc.unizh.ch
Fri Apr 13 13:44:19 CEST 2007

Dear Caterina, GMX users,

It's very difficult to decide whether a simulation converged, at least
for macromolecules. There are a lot of studies in the literature about
it, but there's no concrete answer. I'd try to ask when the results can
be useful, and not when the simulation (or PC) converged - and that
depends on your system and what you are trying to model. One hint is
that a PC with a high cosine content is probably not going to be very
useful if you try to simulate any kind of structural transitions.


Caterina Arcangeli wrote:
> Dear all,
> I performed essential dynamics (ED) analysis on my protein (253 amino
> acids). The simulation (10ns) achieves stability in the RMSD after
> 6.0ns, but a convergence analysis based on cluster analysis (as
> described by Daura, 1999) indicates that the conformational sampling
> reaches a stable value only for the last 1.0ns.
> So, I've performed ED on both 6-10 ns (M1) and 9-10ns (M2) time interval
> using g_covar.
> To check if the principal modes are well defined, I've calulated the
> ovelap of the sampling between the first and second half of the
> trajectories (I've splitted M1 and M2 into two halves) and the cosine
> content of the first eigenvectors:
> - the normalized overlaps are 0.499 (M1) and 0.434 (M2);
> - the subspace overlaps (rmsip) for the first 10 eigenvectors are 0.748
> (M1) and 0.686 (M2);
> - the cosine content of the first eigenvectors is 0.828 (pc1) for M1 and
> 0.125 (pc1) for M2.
> So, apparently, the two method are inconsistent:
> a higher rmsip value is observed for M1 (according to Amadei, a rmsip
> value > 0.7 shows a reasonable convergence) wheras the lower cosine
> content is obtained for M2 (according to Hess, a value < 0.5 indicates
> jumping between clusters and not a merely random diffusion).
> My questions are: which time window better describe the dynamics of my
> protein? From where come out this inconsistency? What I'm wrong?
> Thank to all.
> Caterina
> P.S. I aware that 10 ns are in general insufficient to describe the
> thermodynamics of a protein but I want just to have an idea of some
> structural properties of my protein.
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php

Ran Friedman
Postdoctoral Fellow
Computational Structural Biology Group (A. Caflisch)
Department of Biochemistry
University of Zurich
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
Tel. +41-44-6355593
Email: r.friedman at bioc.unizh.ch
Skype: ran.friedman

More information about the gromacs.org_gmx-users mailing list