AMBER Archive (2009)

Subject: RE: [AMBER] PCA analysis: the first vector is the movement of whole molecule.

From: Hannes Loeffler (
Date: Fri Jul 31 2009 - 02:27:54 CDT

Ok, I see that you don't really understand what you are doing.

The '*' mask means ALL atoms. Unless this selection is not just the
molecule(s) you are interested in you are obviously trying to fix the
wrong part of your system. So make sure that you use here the same mask
as in the 'matrix' command.

Why simulation length matters? PCA is a statistical method as you are
calculating the variance-covariance matrix as an _average over time_.
This should make it clear to you that you will need sufficient sampling
before you can rely on your results.

Here are two sample input files for computing the
eigenvectors/eigenmodes and the projection of the eigenvectors onto the
trajectory for a protein with 100 residues. The examples are just meant
to guide you. So adapt it to your problem.

-------------------------- ------------------------
rms first :1-100_at_CA

# PCA analysis
matrix covar name PCA :1-100_at_CA
analyze matrix PCA vecs 25 out evec.pev
-------------------------- ------------------------

-------------------------- ------------------------
# centering and imaging not necessary if selection is just one molecule
center :1-100 mass origin
image origin center

rms first :1-100_at_CA

projection :1-100_at_CA modes evec.pev out proj.ppj beg 1 end 25
-------------------------- ------------------------

Hope that helps,

AMBER mailing list