[gmx-users] non-symmetric matrix and too large first values

[Bert de Groot]

medeawan at sohu.com wrote:
> 
> I have checked the first eigenvalues in .xvg files, some of them really large.
> For example, a 88 amino acid protein (PDB ID=1krn), there are 79 amino
> acid coordinates in the PDB file. 


what do you mean by that? 88 amino acids should be approx 88*10=880 atoms?

> After EM, the Max force< 1e-9, while the
> first 12 eigenvalues g_nmeig_d worked out are:
>      1     0.000438558
>      2    -0.000683771
>      3     -0.00292664
>      4      0.00597364
>      5      0.00817211
>      6      -0.0114438
>      7         1.58334
>      8         1.80916
>      9         2.04356
>     10          2.6188
>     11         2.87512
>     12         3.44095
> 



this *might* still be ok. It's been ages since I did my last NM, so I don't remember
typical values by heart. Maybe someone has an recent application and could look up
the eigenvalues for comparison?


> For a 20 nucleic acid CTCAA-GGTGT-CCCAG-CCCAC, the first 12 eigenvalues are:
>                  0                        gd_9      gd_11
> 1       3.73088e-05     -2.59139e-05     9.76281e-06
> 2       -0.000153511    3.50443e-05     -1.34763e-05
> 3       0.000574301     -0.000357439     0.00030759
> 4       -0.00113997     0.000468474     -0.000441709
> 5       -0.0285976                  0.00105487   0.000877102
> 6       0.0324488                  -0.00142717  -0.00124924
> 7       0.184117                     0.0829609  0.64291
> 8       0.297109                     0.166827   0.908421
> 9       0.617796                     0.39677    1.78765
> 10      0.68352              0.809834              1.92865
> 11      1.37551              1.12581            2.39546
> 12      1.67154              1.24509               2.93631
> 

this should certainly be investigated further. It shouldn't occur
that there is an order of magnitude spread in the first non-zero (7th)
eigenvalue. If this is really caused by the choice of the dihedral
parameter, then it's no use calculating further until you found the proper
(+consistent+complete) forcefield parameters.
Similarly, it shouldn't occur that the last 'zero' eigenvalue (6th)
is less than an order of magnitude smaller than the first nonzero one.
I can imagine that this could be partially caused by the algorithm
(in the current implementation gromacs doesn't compute the second 
derivatives analytically, but numerically), but in practice this never
was a problem before.

Could you please send me (or the list) your mdp file that you used for
energy minimisation and NM?


Bert

____________________________________________________________________________
Dr. Bert de Groot

Max Planck Institute for Biophysical Chemistry
Theoretical molecular biophysics group
Am Fassberg 11 
37077 Goettingen, Germany

tel: +49-551-2011306, fax: +49-551-2011089

email: bgroot at gwdg.de
http://www.mpibpc.gwdg.de/abteilungen/071/bgroot
____________________________________________________________________________