[gmx-users] still dealing with GFP-like chromophore with LINCS warnings

Anna Marabotti amarabotti at unisa.it
Thu Jun 27 15:47:36 CEST 2013

Dear gmx-users,
I'm still dealing with the problem of GFP-like chromophore. To be 
honest, with the help of another gmx user, I was able to obtain the 
topology of my chromophore, in protonated and unprotonated form, and I 
started performing the simulations. For the protonated chromophore, all 
worked well, and I obtained my simulations without many problems.

Instead, I'm experimenting problems with the protein containing the 
unprotonated form. In particular, when I started calculating the 
production run, using in the .mdp files exactly the same parameters used 
for the protonated form, I started obtaining a lot of LINCS warnings and 
production of a lot of stepXXXXX.pdb files. This happened only during 
the production run, whereas I have no errors during minimization and 
position-restrained dynamics. The minimization step seems to be OK and I 
did not further continue with it (I set emtol=500 and emstep=0.01; the 
result of minimization was:

Steepest Descents converged to Fmax < 500 in 2114 steps

Potential Energy= -6.84833590617959e+05

Maximum force=4.57025361763790e+02 on atom 2129

Norm of force=9.42624640937790e+00

that seems pretty good for me.)

I tried to overcome the problem by using longer PR-MD (100 ps instead of 
10 for NVT; 1 ns instead of 100 ps for NPT) but it doesn't work, despite 
the fact that the energies and all other factors (temperature, pressure) 
are really stable during PR-MD. I also watched the trajectory using VMD 
but, apart from parts of protein crossing periodic boundaries (a thing 
that I see regularly in my simulations), I don't see anything "strange" 
(like system exploding). Apparently, it simply stops, and that's all.

Other things that trouble me are:
1) I checked the atoms with LINCS warning, and I see that not only those 
atoms belonging to chromophore are perturbed, but also other atoms in 
the backbone of the protein (mainly CA-C-N; many of them, but not all, 
are spatially close to chromophore).
2) during these production runs I also experimented many hardware 
problems (power outages) and software problems (bad job schedulation, 
problems in core communications etc) and I don't understand if these 
problems could affect the results of my simulation in this way. I ask 
this because my first production run was stopped because of a power 
outage, and in the .log file I saw a lot of errors very early; when the 
machine turned on, I resent exactly the same dynamics, and the errors 
were far later than before (in the first run, they came across 5 ns; in 
the second run not before 20 ns).
3) Last thing: when I started the production run, I switched the 
pressure coupling from Berendsen to Parrinello-Rahman (obviously not 
giving the .edr file to grompp for the creation of the full.tpr file) 
(on the contrary, the temperature coupling was left V-rescale both in 
PR-Md and production run). Can this switch perturb the system?

My question is: do I have to suppose that the unprotonated form of 
chromophore is badly parametrized, and this can influence all the rest 
of the protein in this way? If yes, why nothing happens when I do PR-MD? 
If not, do I have simply to perform a very very very long stabilization, 
do I have to perform a deeper minimization, do I have to avoid switching 
from Berendsen to Parrinello? What do you suggest me to do?

Thanks a lot

Anna Marabotti, Ph.D.
Assistant Professor
Department of Chemistry and Biology
University of Salerno
Via Giovanni Paolo II, 132
84084 Fisciano (SA)
Phone: +39 089 969583
Fax: +39 089 969603
E-mail: amarabotti at unisa.it
Skype: annam1972
Web page: http://www.unisa.it/docenti/annamarabotti/index

"Indifference is the eighth deadly sin" (don Andrea Gallo)

More information about the gromacs.org_gmx-users mailing list