The Amber 14 manual is your friend here. I will refer to it as necessary.
> On Apr 20, 2015, at 7:44 AM, Muthukumaran R <kumaran.bicpu.edu.in> wrote:
>
> Dear amber users,
>
>
> I have some general queries in PCA analysis. Any suggestions or guidelines
> regarding this would be very helpful.
>
> 1. I read in an article about obtaining exactly 3 eigenvectors in a PCA
> analysis. But, in practice, the user requests the number of eigenvectors to
> be generated. Is there any procedure to obtain only 3 eigenvectors as the
> author says?
The first thing you should do is RMS-fit your protein using the rmsd command. Then you need to compute the coordinate covariance matrix using the matrix command (see page 550; section 28.11 of the Amber 14 manual). Afterwards, use the “diagmatrix” analysis command to diagonalize the matrix and determine the eigenmodes (eigenvalues and eigenvectors) of the covariance matrix. These eigenmodes are what you ultimately are interested in. The eigenvalues represent the “magnitude” of each mode (i.e., how much of the covariance in the system can be explained by that mode), while the eigenvectors themselves give the “directionality” of that mode in the original basis (in this case, the original x, y, z Cartesian basis for each atom). I would suggest computing all of the eigenmodes -- you can choose to analyze only three later if you choose. But by computing all of them, you can calculate how much of the covariance is explained by whichever modes you choose to analyze.
> 2. In case we obtain the eigenvectors, is there any way to identify or
> visualize, which direction the eigenvector represents?
Note that it is very difficult to conceptualize “direction” in a 6N-dimensional space. You can use the “modes” command here -- particularly the “trajout” option -- to create a pseudo-trajectory along each principal component that you wish to analyze. The “beg” and “end” keywords can be used to specify which principal components you wish to analyze. The “pcmin” and “pcmax” keywords lets you specify the minimum and maximum projections for each mode (i.e., how “far” the trajectory bends in the positive and negative directions of each principal component).
>
> 3. Amber gives the displacement of atoms in three directions (x, y and z).
> But, how to calculate the total displacement of any atom in the direction
> of a given eigenvector?
Look at the “modes” command here -- particularly the “fluct” option -- on page 565 of the manual.
> 4. If the author talks about 3 eigenvectors and displacement in one
> direction, can it mean the x, y and z displacements? If so, how can we
> visualize the x, y and z axes
Most visualization programs will display axes if you set that option (VMD, for instance, shows the axes by default in the window displaying your molecule).
Play around with some of the different options until you understand exactly what cpptraj is doing.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Apr 20 2015 - 06:00:09 PDT
