[gmx-users] normalized overlap

[Lovas, Sandor]

Hi Mark,
I suspect this is a bug in the code. I have checked my inputs
(eigenval.xvg and eigenvector.trr) and they looked OK. I know that the
first and the second frames in the eigenvector.trr are for the
reference and average structures, respectively. But in this case I
think it does not matter what they are, since the eignevalues are not
zero in the eigenval.xvg filesm, so if they are read from the input or
used during the calculation they should affect the result.

Here is how I generated the files. Btw, I use gromacs-2016, but
previous versions have the same problem:

1. gmx covar -f traj.xtc -s conf.pdb  -o eigenval_tot -v eigenvec_tot;
full trejectory

2. gmx covar -f traj.xtc -s conf.pdb -e 500000 -o eigenval_5 -v
eigenvec_5 ; first half of tjhe simulation

3. gmx anaeig -v eigenvec_5.trr -eig eigenval_5.xvg  -v2
eigenvec_tot.trr -eig2 eigenval_tot.xvg; using the eigenvalues

trr version: GMX_trn_file (single precision)
Read non mass weighted reference structure with 443 atoms from
eigenvec_5.trr
Read non mass weighted average/minimum structure with 443 atoms from
eigenvec_5.trr
Read 1329 eigenvectors (for 443 atoms)

Read 1329 eigenvalues from eigenval_5.xvg
Read non mass weighted reference structure with 443 atoms from
../eigenvec_tot.trr
Read non mass weighted average/minimum structure with 443 atoms from
../eigenvec_tot.trr
Read 1329 eigenvectors (for 443 atoms)

Read 1329 eigenvalues from ../eigenval_tot.xvg
RMSD (without fit) between the two average structures: 0.093 (nm)

2 eigenvectors selected for output: 1 1329
Will compare the covariance matrices using 1329 dimensions
Trace of the two matrices: 29.8148 and 29.2895
Square root of the traces: 5.46029 and 5.41198
The overlap of the covariance matrices:
  normalized:  0.679
       shape:  0.679

4. gmx anaeig -v eigenvec_5.trr -v2 eigenvec_tot.trr ; no eigenvalues
are in the input

trr version: GMX_trn_file (single precision)
Read non mass weighted reference structure with 443 atoms from
eigenvec_5.trr
Read non mass weighted average/minimum structure with 443 atoms from
eigenvec_5.trr
Read 1329 eigenvectors (for 443 atoms)

Read non mass weighted reference structure with 443 atoms from
../eigenvec_tot.trr
Read non mass weighted average/minimum structure with 443 atoms from
../eigenvec_tot.trr
Read 1329 eigenvectors (for 443 atoms)

RMSD (without fit) between the two average structures: 0.093 (nm)

2 eigenvectors selected for output: 1 1329
Will compare the covariance matrices using 1329 dimensions
Trace of the two matrices: 29.8147 and 29.2895
Square root of the traces: 5.46029 and 5.41198
The overlap of the covariance matrices:
  normalized:  0.679
       shape:  0.679




In both cases the normalized value (0.679) are the same.

Sandor


>Hi,
>
>We don't have anything to go upon... this could be a code bug or a
usage
>bug. How did you generate your input file? Have you inspected your
input
>file to check for sanity? (e.g. with gmx dump, remembering that the
first
>frame or two are different, see gmx covar -h) What is the whole
terminal
>>output of both stages?

>Mark

>On Sat, Dec 17, 2016 at 10:30 AM Lovas, Sandor <SandorLovas at
>creighton.edu>
wrote:

> I have asked this question a few days back; I try one more time:
>
> When I am trying calculate the degrees of overlap of subsets of a
> simulation to the whole trajectory by following the explanation in
the
> manual gmx anaeig:
> "When -v, -eig, -v2 and -eig2 are given, a single number for the
> overlap between the covariance matrices is generated...."
>
> For me, it does not make any difference whether or not I use the "-
eig"
> and "-eig2" inputs, the results for normalized overlap are the same.
In
> other words, the normalized overlap is not calculated by using the
> equation given in the manual:
>
> "normalized overlap = 1 - difference/sqrt(tr(M1) + tr(M2))
>  ....
> where M1 and M2 are the two covariance matrices and tr is the trace
of
> a matrix."
>
> Sandor
> --
> ______________________________________
> Dr. Sandor Lovas, Professor
> Department of Biomedical Sciences
> Creighton University
> 2500 California Plaza,
> Omaha, NE 68178, USA
>
> Phone: 402-280-5753
> Fax:   402-280-2690
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
-- 
______________________________________
Dr. Sandor Lovas, Professor            
Department of Biomedical Sciences      
Creighton University                   
2500 California Plaza,                 
Omaha, NE 68178, USA 

Phone: 402-280-5753
Fax:   402-280-2690