[AMBER] Regarding principal component analysis using ptraj

From: Arvind Marathe <bahuroopi.gmail.com>
Date: Fri, 28 Aug 2009 07:54:40 +0100

Hello list,

I am trying to do a principal component analysis of an explicit
solvent 12-mer DNA simulation trajectory. I want to find the principal
components (eigenvectors and eigenvalues) of the DNA motion. Following
is my ptraj input script:

-----------------------------------------------------------------------
trajin dna-12_mer.mdcrd.gz

strip :WAT
strip :Na+
strip :H
strip :H1
strip :H2
strip :H4
strip :H5
strip :HA
strip :HC
strip :HO

rms first
matrix covar name cov_mat out cov_mat.dat
analyze matrix cov_mat out eigen_vectors.dat

go
-----------------------------------------------------------------------

The DNA has 486 non-hydrogen atoms and 760 atoms including hydrogens.
I wanted to exclude all the water molecules, ions and hydrogen atoms
from the analysis. I found the mask for all the hydrogen atoms using
the ambmask command. I would have expected the file cov_mat.dat to
contain a 486*3 x 486*3 matrix. However, it contains a 760*3 x 760*3
matrix. Why are the hydrogen atoms not getting excluded? Is something
wrong with the ptraj input file?

The second question is regarding the file eigen_vectors.dat. Shouldn't
the eigen_vectors.dat file also have a 760*3 x 760*3 matrix if the
cov_mat is 760*3 x 760*3? I am getting a matrix of 654 * 7 in the
eigen_vectors.dat file. The two header lines give:

-----------------------------------------------------------------------
 Eigenvector file: COVAR
 2280 2280
-----------------------------------------------------------------------

I assume the input matrix is being read correctly. So why am i getting
this size of 654 * 7?

Finally, how do i get the eigenvalues themselves? Does one of the 7
columns in the file eigen_vectors.dat give the eigenvalues? But if
that is the case, shouldn't they be in descending order? None of the
columns in that file are in descending order.

There are no errors in the ptraj log. I have a feeling i have
misunderstood something rather badly.

Thanks in advance for any help.

Regards,
arvind

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 02 2009 - 23:05:52 PDT
Custom Search