Re: [AMBER] How to analyse the PCA results?

From: Daniel Roe <>
Date: Fri, 20 Jul 2018 10:20:21 -0400


On Wed, Jul 18, 2018 at 12:01 PM, Karolina Mitusińska (Markowska)
<> wrote:
> How to convert the cpu-gpu-evecs.dat diagonal matrix into a file that could
> be used by gnuplot to plot something similar to Figure 2?

You'll have to load the eigenvector file:

readdata <file> name MyEvecs ibeg 1 iend <# vecs>

Where <# vecs> is the desired number of eigenvectors to use. Then you
need to use the 'projection' command to calculate your coordinates
projection along specified eigenvectors. *IMPORTANT*: The coordinates
you project should be the same as those used to generate the
coordinate covariance matrix, so e.g. if you RMS-fit when generating
the matrix you must also RMS-fit when projecting. Also, you have to
make sure that you're using the same atoms (e.g. if you calculated the
matrix using CA atoms, you should only project with the CA atoms).

projection MyProject evecs MyEvecs out Project.dat beg 1 end <# vecs> <mask>

This will generate data sets named MyProject:X, where X corresponds to
the eigenvector. You can then use the 'hist' analysis command to
create a 2D histogram, e.g.

hist MyProject:1 MyProject:2 bins 100 out myhist.gnu

See the manual for full details on these commands. Also, you can check
out the supporting information for and for examples (including
scripts) of using cpptraj to do PCA.

Hope this helps,


> And if I finally get the "distribution of components" plot, how would I
> know, which conformation is represented by which part of the plot?
> Thanks for your help!
> Karolina Mitusińska
> PhD student
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Fri Jul 20 2018 - 07:30:03 PDT
Custom Search