[gmx-users] Eigenvector and eigenvalues

Tsjerk Wassenaar tsjerkw at gmail.com
Mon Jun 10 09:19:38 CEST 2013


Hi Ankitha,

Taking the backbone for analysis doesn't give you more information than
using only C-alphas. These are extremely correlated.

Cheers,

Tsjerk



On Sun, Jun 9, 2013 at 5:28 PM, Ankita naithani <ankitanaithani at gmail.com>wrote:

> Hi Tsjerk,
>
> Thank you so much for the helpful insight. It will surely help in clearing
> some basic concepts about eigenvectors and the projections.
> Also, since in the end you mentioned about my matrix being derived only
> from C-alpha atoms which proves a limitation, would you reckon a backbone
> analysis would be better suited to gain some helpful insight into protein
> motions (Allosteric transitions, effector signalling??). It would however
> be intensively complicated to derive for the whole protein considering mine
> is a tetramer and a huge system.
>
>
> Kind regards,
>
> Ankita
>
>
> On Sat, Jun 8, 2013 at 9:18 PM, Tsjerk Wassenaar <tsjerkw at gmail.com>
> wrote:
>
> > Hi Ankita,
> >
> > It is important to use the correct names for things. The projection is
> the
> > (scalar) inner product of the configuration (coordinates) with the
> > eigenvector. This is also named the score of the configuration on the
> > selected eigenvector. With g_anaeig -proj you get the projection. Each
> > configuration of a trajectory has a projection, and you can plot them
> over
> > time, or plot the projection on one eigenvector against another (-2d).
> You
> > can also make a 3D plot of three selected eigenvectors (-3d), which is
> > written out as a PDB file. The projection is not a structure, and you
> can't
> > have a movie of structures. What you can do is extract the structures
> > corresponding to the extreme projections (-extr). Those are always
> written
> > per eigenvector. Since the eigenvectors are uncorrelated, the extremes
> are
> > not defined for combinations of them. It is also possible to filter a
> > trajectory on a set of eigenvectors. That will give a trajectory of
> > (unphysical) structures, in which the projections on the non-selected
> > eigenvectors are set to zero. This filtering is a matrix operation, using
> > the matrix of selected eigenvectors and the corresponding atoms of the
> > trajectory. Because the matrix of eigenvectors is derived from C-alpha
> > atoms only in your case, it is not possible to get anything else out of
> it.
> >
> > Hope it helps,
> >
> > Tsjerk
> >
> >
> >
> > On Sat, Jun 8, 2013 at 1:06 PM, Ankita naithani <
> ankitanaithani at gmail.com
> > >wrote:
> >
> > > Hi Tsjerk,
> > >
> > > [I am reposting it since my previous attachment was unable to proceed I
> > > guess]
> > >
> > > Thank you for the reply. I am sorry but I don't have the snapshots with
> > me
> > > as my system was formatted and I lost that information. However, I
> > > performed PCA again on Calpha atoms and this time my matrix is
> 5976X5976
> > > wherein 1992 are my total number of calpha atoms. So, yes indeed I did
> > > something foolish in the previous case to have got that result.
> However,
> > I
> > > will try to look at it and find my mistake so as to not repeat it
> again.
> > >
> > > I would be really grateful if you could help me with my another naive
> > > query. I wanted to see the movie of first few eigenvectors as
> individual
> > > and also separately. I know I can extract that with the help of
> g_anaeig
> > > but as I did PCA on only calpha atoms, all I get in the projection of
> > first
> > > eigenvector is the calpha dots. Is there any way, I can visualise the
> > > projection of whole protein on that particular eigenvector?
> > >
> > > Secondly, if I want to see the projection along first 10 eigenvectors
> and
> > > last 10 eigenvectors (slow modes and fast modes), how could I proceed
> > > because when I choose -first 1 -last 10, it gives me individual pdb
> files
> > > corresponding to the eigenvectors.
> > >
> > > Third, as can be judged by my absolute naivety, I did a 2d projection
> on
> > PC
> > > 1 and PC 2. I can visualise that plot in grace as a 2d plot. An example
> > > file is appended herein. I know each dot might correspond to a
> particular
> > > conformation in the trajectory but I really have no idea as to how to
> > > interpret the results and draw conclusions from these 2d plots and also
> > 3d
> > > pdb projections. Could you please guide me in the right direction to
> > > understand these 2d plots and infer the significant results?
> > >
> > >
> > > Kind regards,
> > >
> > > Ankita
> > >
> > >
> > > On Thu, Jun 6, 2013 at 2:49 PM, Tsjerk Wassenaar <tsjerkw at gmail.com>
> > > wrote:
> > > >
> > > >> Hi Ankita,
> > > >>
> > > >> Please provide the commands you've run and the screen output from
> > > g_covar.
> > > >>
> > > >> Cheers,
> > > >>
> > > >> Tsjerk
> > > >>
> > > >>
> > > >>
> > > >> On Thu, Jun 6, 2013 at 3:44 PM, Ankita naithani <
> > > ankitanaithani at gmail.com
> > > >> >wrote:
> > > >>
> > > >> > Hi,
> > > >> >
> > > >> > I wanted to know about the eigenvectors and eigenvalues. I
> recently
> > > >> > performed the principal component analysis (only the backbone into
> > > >> > consideration) on a trajectory of 2000 residues. I obtained 15641
> > > >> > eigenvectors and 17928 eigenvalues. There is a difference in the
> > > number,
> > > >> > which I am not quite sure off (perhaps that has to do with
> > eigenvalue
> > > >> for
> > > >> > each eigenvector, and the eigenvector has 3 co-ordinates x,y,z. I
> > > know I
> > > >> > may be wrong completely but since there are 15641 eigenvectors,
> then
> > > >> > shouldn't there by only 15641 eigenvalues for those eigenvectors)
> > > >> >
> > > >> > My second problem lies that I am trying to extract information or
> > say
> > > >> RMSF
> > > >> > values for the first 10 eigenvectors (10 slowest modes) and the
> last
> > > 10
> > > >> > eigenvectors (fastest modes) and there I face segmentation fault.
> I
> > > can
> > > >> get
> > > >> > information on the first 10 but when I try last 10 with the
> command
> > as
> > > >> > follows:
> > > >> >
> > > >> > g_anaeig -v eigenvec.trr -eig eigenval.xvg -rmsf eig_rmsf.xvg
> -first
> > > >> 15631
> > > >> >
> > > >> > [where -first should have the starting eigenvector ]
> > > >> >
> > > >> > (I choose, 15631 since I want the plot for last 10 but I get
> > > >> segmentation
> > > >> > fault, which is also to do with the fact that the eigenval.xvg has
> > > 17928
> > > >> > values. This number doesn't match and so, maybe the plot is from
> > > >> > 15631-17928). This has further confused me about the slowest and
> > > fastest
> > > >> > modes (and somehow I do need information on the first 10 slowest
> > modes
> > > >> and
> > > >> > 10 fastest modes). In a broad way, the slowest modes would be the
> > ones
> > > >> with
> > > >> > high eigenvalues say, first 10 eigenvectors in eigenval.xvg would
> > give
> > > >> the
> > > >> > slowest 10 modes and the last 10 in eigenval.xvg should give the
> > > >> fastest 10
> > > >> > modes.
> > > >> >
> > > >> > Here, again I feel quite confused because eigenval.xvg has 17928
> > > entries
> > > >> > and in the legend, it says that x axis is the eigenvector index
> and
> > y
> > > >> axis
> > > >> > is eigenvalue index so it leaves me quite perplexed about the
> > problem.
> > > >> >
> > > >> > I am sorry for this extremely long and confusing post, but any
> help
> > in
> > > >> this
> > > >> > regard would be really beneficial.
> > > >> >
> > > >> >
> > > >> > Kind regards,
> > > >> >
> > > >> > Ankita
> > > >> >
> > > >> > --
> > > >> > Ankita Naithani
> > > >> > --
> > > >> > gmx-users mailing list    gmx-users at gromacs.org
> > > >> > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > >> > * Please search the archive at
> > > >> > http://www.gromacs.org/Support/Mailing_Lists/Search before
> posting!
> > > >> > * Please don't post (un)subscribe requests to the list. Use the
> > > >> > www interface or send it to gmx-users-request at gromacs.org.
> > > >> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > > >> >
> > > >>
> > > >>
> > > >>
> > > >> --
> > > >> Tsjerk A. Wassenaar, Ph.D.
> > > >> --
> > > >> gmx-users mailing list    gmx-users at gromacs.org
> > > >> http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > >> * Please search the archive at
> > > >> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > > >> * Please don't post (un)subscribe requests to the list. Use the
> > > >> www interface or send it to gmx-users-request at gromacs.org.
> > > >> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > > >>
> > > >
> > > >
> > > >
> > > > --
> > > > Ankita Naithani
> > > >
> > >
> > >
> > >
> > > --
> > > Ankita Naithani
> > >
> > >
> > >
> > > --
> > > Ankita Naithani
> > > --
> > > gmx-users mailing list    gmx-users at gromacs.org
> > > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > > * Please search the archive at
> > > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > > * Please don't post (un)subscribe requests to the list. Use the
> > > www interface or send it to gmx-users-request at gromacs.org.
> > > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> > >
> >
> >
> >
> > --
> > Tsjerk A. Wassenaar, Ph.D.
> > --
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://lists.gromacs.org/mailman/listinfo/gmx-users
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> > * Please don't post (un)subscribe requests to the list. Use the
> > www interface or send it to gmx-users-request at gromacs.org.
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
>
>
>
> --
> Ankita Naithani
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>



-- 
Tsjerk A. Wassenaar, Ph.D.



More information about the gromacs.org_gmx-users mailing list