[AMBER] PCA

From: Sofia Vasilakaki <svasilak.chem.uoa.gr>
Date: Thu, 7 Apr 2016 21:59:38 +0300

Hi to everyone,

It is the first time I am performing PCA, so excuse my ignorance...

I am using a common script for this:
matrix covar name covmat out covmat.dat :1-704&!.H=
diagmatrix covmat out evecs.dat vecs 57
projection modes evecs.dat out pca.dat beg 1 end 57 :1-704&!.H=

1. When the diagmatrix is performing the following warning appears. What
does this mean? How the eigenvectors are generated from the covariance
matrix?

Warning: In matrix 'covmat', # of frames 1534 is less than # of columns
16998.
Warning: The max # of non-zero eigenvalues will be 1534

2.Is there any rule for choosing the number of vecs in the diagmatrix
command? Should it be a multiple of 3 (x,y,z)?

3.I used "hist pca.dat:2 bins 200 .." which run smoothly while "hist
pca.dat:2,*,*,*,200 out .." returned
Warning: Data set 'pca.dat:2' not found.
Error: Dataset pca.dat:2 not found.
Error: Could not setup analysis [hist]

Why? I used "readdata pca.dat" before "hist".

4.I am trying to plot gnu files which I got from "hist" and "kde" command,
but the graphs I am getting with splot are not right (a line appears in
the middle of the graph). Could you please explain the format of the gnu
files and maybe provide me with correct gnuplot command to plot them. For
example, the pca.his.gnu looks like
set pm3d map corners2color c1
set ytics 1.000, 1.000
set ytics("HIST1" 1.000)
set xlabel "pca.dat:2"
set ylabel ""
set yrange [ 0.000: 3.000]
set xrange [-272.601: 398.945]
splot "-" with pm3d title "pca.hist2.gnu"
-269.277 1.000 0.0002
-269.277 2.000 0

-265.953 1.000 0.0000
-265.953 2.000 0 .... and so on.

5.How you calculate the % of total motion of one mode? Is there any way to
calculate an entropy estimation for the top PCs?

Sorry for this long e-mail.

Thank you very much for you help in advance.

Kind Regards,
Sofia









_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Apr 07 2016 - 12:00:04 PDT
Custom Search