[gmx-users] Problems with g_anaeig

[Mark Abraham]

On Fri, Jan 25, 2013 at 6:40 AM, James Starlight <jmsstarlight at gmail.com>wrote:

> Mark,
>
> as I told previously to Justin ( lets to consider the problems with
> g_anaeig only in that topic to avoid producing doubles ) the problem
> is not in the average structure (it looks fine)


You might have noticed I made no mention of your average structure. You
told g_covar to write it out, but that's irrelevant to the workflow. You
chose an arbitrary *reference* structure for the covariance analysis and
found problems. You have to exclude the possibility that your choice of the
reference structure is the source of the problem.


> but in the
> filter.xtc trajectory produced by g_anaeig. When I load that xtc to
> any structure in vmd I obtained dynamics picture of my protein with
> broken geometry (its look loke my protein was shrunken).


It's hard to say anything about something we can't see.

But when I
> compared eigenvectors ( produced by g__covar) with the the results of
> the same cov.analysis where there were no such problems in filter.xtc
> I obtained full coverage in the same modes ( so g_covar in both cases
> produce apropriate cov.matrix and the problem only in g_anaeig .
>

What do you mean by "coverage"?

The eigenvectors are an orthonormal basis set for the observed variation.
If the observed variation is a mess (because the structural ensemble you're
treating as a black box is actually a mess, or your reference structure is
an outlier), then the eigensystem will not lead to a sensible filter.

Since you chose not to follow my hints last time:
1) Go and look at the X-ray ensemble. Is it one? Is your arbitrary choice
of reference structure appropriate? Is there even an appropriate choice?
2) You can filter the trajectories based on the other's eigensystem. If the
eigensystems are equivalent, then the filtered trajectories should look the
same. If they don't, then your two ensembles differ.

Mark

James
>
> 2013/1/25 Mark Abraham <mark.j.abraham at gmail.com>:
> > On Tue, Jan 22, 2013 at 8:22 PM, James Starlight <jmsstarlight at gmail.com
> >wrote:
> >
> >> Dear Gromacs users!
> >>
> >>
> >> There is some bug  with g_anaeig the souce of which I could not fully
> >> understand.
> >
> >
> > Good Advice: until you can almost write a code patch to fix it, be very
> > hesitant in suggesting any software has a bug. The best people to help
> > solve the issue are often those who wrote the code, and you don't want
> them
> > annoyed with you :-)
> >
> >   I have problems when I perform PCA of X-ray data set.
> >> Below you can my workflow.
> >>
> >>
> >> g_covar -f b2ar_xray_coors.pdb  -s ref.pdb  -o PCA_eigenval.xvg -v
> >> PCA_eigenvec.trr -av PCA_average.pdb -last 8
> >> g_anaeig -v PCA_eigenvec.trr -s ref.pdb  -f b2ar_xray_coors.pdb -rmsf
> >> eigrmsfPCA.xvg -filt
> >>
> >>
> >> here b2ar_xray_coors.pdb  is the trajectory made from 10 X-ray
> >> structures of my protein (only main chain atoms are included)
> >> ref_pdb is the first frame of that trajectory
> >>
> >>
> >> As the result I've obtained reasonable eigenvalues and aigenvectors
> >> from g_covar BUT when I check filter trajectory ( produced by
> >> g_anaeig) fitted it to the ref.pdb or to the averaged structure in
> >> both cases I've obtained very distorted geometry of the  protein in
> >> thefiltered trajectory. I have no such problems in case of PCA of MD
> >> trajectory ( when  -f trajectory.trr is from the md snapshots not from
> >> x-ray structures)
> >>
> >>
> >> How it could be fixed ?
> >>
> >
> > How have you excluded the hypotheses that your reference structure is
> not a
> > valid representative of the middle of the range of variation? Or even
> that
> > the X-ray structural ensemble really is one?
> >
> > Mark
> > --
> > 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
> --
> 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
>