[gmx-users] PCA-how much of the structure to include?

[patrick wintrode]

Dear all,

I'm performing PCA on a 20 nanosecond simulation of a
~200 aa protein. I performed PCA initially using only alpha carbons. I
then reperformed it by using the backbone for fitting, and the whole
protein for covar analysis.

This worked but when I looked at some of the motions using the -extr option in g_anaeig, the side chains look strange/distorted.

I've
noticed that most reported PCA uses just alpha carbons or backbone
atoms. Does this mean that PCA is not well suited for all-atom analysis
due to fast side chain motions? Or did I make some mistake (like not
selecting the same set of atoms for fitting and analysis)? Can someone
point me toward a paper where PCA included side chains?

Thanks.

Patrick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20081025/015a0788/attachment.html>