[gmx-users] Re: stepsize too small ... but potential energy negative!

[Anna Marabotti]

Dear Justin,

you are definitely true. I tested with Gromacs 4.0.7 a couple of other
proteins that were correctly minimized with the same machine and the
previous version of Gromacs, and both of them stopped with the
neighborsearching error. Then, I found a way to switch to version 4.0.5 on
the same machine, and I minimized pretty well my model...
Now the big difficulty will be saying to the administrators of the system
that hosts my simulations that their installation is faulty...and I fear you
will not able to help me...;-)

Thank you so much for all your invaluable help and time you spent on my
problem. I also want to thank Luca Mollica for his precious support during
these days.

Best regards
Anna

-----Messaggio originale-----
Da: Justin A. Lemkul [mailto:jalemkul at vt.edu] 
Inviato: martedì 25 maggio 2010 15.28
A: Anna Marabotti
Oggetto: Re: R: coordinates and topologies of file "stepsize too small"


Hi Anna,

I strongly suspect the installation is faulty.  I see no other explanation.

Setting a box is easy: "editconf -c -bt dodecahedron."  It may appear
"strange" 
because most visualization software defaults to a triclinic rendering, but
you 
can use "trjconv -pbc mol -ur compact" to get the correct box appearance.
In 
the case of 1WBE, I used ~2000 fewer water molecules than your original
system.

-Justin

Anna Marabotti wrote:
> Dear Justin,
> this is probably the key of my problem! GROMACS 4.0.7 has been just
> installed on the machine I'm currently using, and this is my first
> calculation using it (before, I was using version 4.0.5.). I'm not able to
> answer to your question about compilers because I'm only a user of that
> machine, but now I will try to repeat calculations using some systems that
I
> successfully used with version 4.0.5, and will see what happens, and if
the
> problem will repeat, I will contact the sysadmin to check with him the
> installation!
> Thank you very much for your help, I will keep gmx-users updated with the
> (hopefully!) end of this story!
> 
> Let me ask you just a further little question: do you use some particular
> parameters when you solvate the system in a non-cubic box? I tried
sometimes
> in the past to use the dodecahedral box, but I obtained "strange" results,
> and therefore I preferred to switch to the cubic one, although there are
> much more water molecules to manage. Have you got some suggestions?
> 
> Many thanks again and best regards
> Anna
> 
> -----Messaggio originale-----
> Da: Justin A. Lemkul [mailto:jalemkul at vt.edu] 
> Inviato: martedì 25 maggio 2010 14.55
> A: Anna Marabotti
> Oggetto: Re: coordinates and topologies of file "stepsize too small"
> 
> 
> Hi Anna,
> 
> I must admit I'm stumped.  I was able to minimize the 1WBE structure just
> fine, 
> both in vacuo (in a 10-nm cubic box) and in solution with a counterion 
> (dodecahedral solvent box, 1.0-nm solute-box distance).  Both procedures 
> proceeded without problems, so I doubt that anything is wrong with your
> template 
> structure.
> 
> What compilers did you use to build Gromacs?  Sometimes weird results
happen
> 
> from buggy compilers, but aside from your Gromacs installation being 
> nonfunctional, I don't know what's going on.  Have you successfully worked
> with 
> other systems with this particular Gromacs installation?
> 
> -Justin
>