[gmx-users] Eessential dynamics analysis

[Tsjerk Wassenaar]

Hi Sunita,

It's a bit hard to understand exactly what you're dealing with. It
would be helpful if you could link to or attach (leightweight)
figures, illustrating what is going on. On a general note, I would
suggest to read some texts about principal component analysis, from a
statistical/mathematical point of view, rather than relying on the ED
literature only. This may already give you a feel for a proper
interpretation of your results. For one thing, you assume that you
have a multivariate normal distribution. In that case, a greater
eigenvalue would necessarily mean you have a broader distribution.
However, with distinct non-normality, you may have outliers far from
the center, which cause the eigenvalue to be larger than expected on
the ground of your distribution.
Also, your first two eigenvectors only capture 58% of your total
motility. This is not much, really. You should also inspect higher
order eigenvectors and check for (non-linear) cross-correlations
between your projections.

Hope it helps,

Tsjerk


On 1/11/07, sunita at chem.iitb.ac.in <sunita at chem.iitb.ac.in> wrote:
> Dear friends,
>
> I performed essential dynamics (ED) analysis on my peptide using g_covar
> and g_anaeig. The first two eigenvectors represent 58% of the total
> positional fluctuation. The projection of the trajectory on eigenvectors
> gives me the proj.xvg file. From which I extracted the projection along
> the first and second eigenvectors.I plotted the histgram separately for
> the projection along the first and second eigenvectors to see their
> distribution.
>
> I observed the range of distribution along first eigenvector is less than
> the second eigenvector. But in principle the range of first eigenvector
> should be greater than second (ref: Proteins, 1993, 17, 412-425) since
> first eigenvector has highest eigenvalue means higher mean square
> displacement so the spread of distibution should also be more. In contrast
> I am getting greater spread for second eigenvector.
>
> Could you please comment why I am getting so...?
>
> Thanks,
> sunita
>
>
>
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> 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