[gmx-users] eigenvalues

[Berk Hess]

Hi,

The cosine content is not the most important property you should look at.
Like any type of analysis, you should check the convergence.
The simplest way of doing this is cutting your trajectory in, for instance,
four parts and determining the correlation.
g_anaeig will always print the covariance matrix overlap when two eigenvector
files are provided.

Note that PCA by itself does not do any interpretation.
One can always perform PCA on any length of trajectory.
In most cases the first PC's will be the slowest converging
properties of a system. But is it up to the scientist to properly
interpret the results. In many cases there might not be a lot
of useful information. On the other hand, if you are looking
for a specific effect there might be.

Berk

> Date: Sun, 11 Jan 2009 18:51:03 +0000
> From: t.piggot at bristol.ac.uk
> To: gmx-users at gromacs.org
> Subject: Re: [gmx-users] eigenvalues
> 
> So just to make sure i got this correct, when looking at the cosine content 
> of the principal components i should look at the whole trajectory? Do i 
> need to include the initial relaxation in the first few ns of the 
> production simulation?
> 
> If there is a high cosine content for the whole trajectory is there 
> anything else to be done (if i want to look at the low frequency motions of 
> the trajectory) except for simulate for longer (i have a very large system 
> so not the favoured option!)?
> 
> As you say it seems that lots of people use PCA on short trajectories, even 
> of large systems, which to me is confusing
> 
> Thanks for any insights you can give
> 
> Tom
> 
> --On Saturday, January 10, 2009 11:49:18 +0100 Tsjerk Wassenaar 
> <tsjerkw at gmail.com> wrote:
> 
> > Hi Sanjay,
> >
> > Imagine yourself zig-zagging along a line from one place to another.
> > If you look at you're motion (and the variance), you'll find that if
> > you only look at blocks most of it is explained by the zig-zag and
> > nicely periodic (no cosine content as in Berk Hess' paper). Good, you
> > think. But if you look at the whole travel, the most important
> > contribution is the going from one place to another, and if you look
> > at you're displacement over time with respect to the mean, that will
> > give you half a cosine. The fact that results in a block do not fit a
> > cosine does not take away the fact that you're still in the process of
> > relaxation. I know it's not what you want to hear, but I've seen it
> > happen to a complex of >600 residues, where relaxation took more than
> > 30 ns. I also know that people often do PCA on short time
> > trajectories, but it's not the proper thing to do.
> >
> > Cheers,
> >
> > Tsjerk
> >
> > On Sat, Jan 10, 2009 at 7:34 AM,  <sanjay23 at iitb.ac.in> wrote:
> >> Hi Tsjerk,
> >> thanks
> >> actually my protein is quite larg (509 aa).i had divided my trajectory in
> >> different part and according to your suggestion i calculated
> >> cosine-content for all, and find that trajectory from 5to15 ns and
> >> 18to25ns having cosine value very less about 0.03 in both cases(with and
> >> without ligands), while other combination showing higher values (>0.6).so
> >> i think my system is Ist converges around 5ns and maintaining it up to 15
> >> ns after that it may be few conformational fluctuation occurring and
> >> finally it get stabilized from 18to15ns.may that part 15to18ns trajectory
> >> is transition period between to conformational fluctuation. i have also
> >> calculated temp. and pressure during whole simulation and i find that it
> >> is exact Gaussian between 5to25ns.so my confusion is whether i take my
> >> trajectory for ED analysis is from 5to15 or 18to25 or whole from 5to25,
> >> but 5to25 showing value of cosine >0.6.
> >>
> >> _______________________________________________
> >> 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
> >>
> >
> >
> >
> > --
> > Tsjerk A. Wassenaar, Ph.D.
> > Junior UD (post-doc)
> > Biomolecular NMR, Bijvoet Center
> > Utrecht University
> > Padualaan 8
> > 3584 CH Utrecht
> > The Netherlands
> > P: +31-30-2539931
> > F: +31-30-2537623
> > _______________________________________________
> > 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
> 
> 
> 
> ----------------------
> TJ Piggot
> t.piggot at bristol.ac.uk
> University of Bristol, UK.
> 
> _______________________________________________
> 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

_________________________________________________________________
Express yourself instantly with MSN Messenger! Download today it's FREE!
http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090112/95598c62/attachment.html>