[gmx-users] crash in gromacs-4.0.2 using vsites and 2fs t.s.

[gportel at gwdg.de]

Hi everyone,

As I wrote yesterday, I'm having problems running stable simulations with
vsites and constraints in all bonds in different time-steps using
gmx-4.0.2. As long as the dna strand does not cross pbc, the simulations
run fine (the longest I did was 5ns). The dna topology was generated with
pdb2gmx, several details on how the simulations were carried out are
listed in the quoted mails below. I also using -dlb auto, just in case it
matters.

I reproduced the crash in two different machines, different starting and
simulations conditions (rectangular or octahedric boxes, Berendsen and
v-rescale, the two options for periodic_molecule). As I reported before,
it happens when pbc are crossed, this time I'm really sure: I resolved the
steps right before the crash time step by time step, plus the distance
separating the 1-4 interactions (the first warning I get before the crash)
is around the length of the corresponding box vector.

In my first round of attempts I did not observe any crash in these systems
in NPE simulations (before I said NVE, but I do have pressure control)
regardless of pbc crossings. I'm running now tests in this ensemble to see
if I can also reproduce the 'non-crashing' behavior. Problem is, at the
moment, that the dna is cooling down and it does not move much from the
box center :-(, so after 4ns it did not cross any box boundary.


Does anybody have a clue of what can be the problem? Or how to debug it in
a more efficient manner? I'm trying to give the information I suspect
could be useful to narrow it down. I can provide more details, data, input
files, etc...

Thanks for your help!

Guillem

Justin A. Lemkul wrote:
> >
> >
> > gportel at gwdg.de wrote:
>> >>
>> >> Hi everyone,
>> >>
>> >> I'm having problems running an md simulation (with gmx-4.0.2) using
vsites
>> >> and a time step of 4fs. I did generate the tpr with pdb2gmx. The
problem
>> >> seems to occur when the molecule crosses pbc, since the first sign
of the
>> >> simulation not working fine after ~3ns is
>> >>
>> >> "
>> >> Warning: 1-4 interaction between 6 and 8 at distance 6.122 which is
larger
>> >> than the 1-4 table size 2.400 nm
>> >> These are ignored for the rest of the simulation
>> >> This usually means your system is exploding,
>> >> if not, you should increase table-extension in your mdp file or with
user tables increase the table size
>> >> "
>> >>
> >
> > This is a commonly reported problem.  See here:
> >
> > http://wiki.gromacs.org/index.php/blowing_up
> >

Thanks, I was aware of this entry in the wiki.

> > Also search the list archives; you will pull up several hundred posts
describing the problems other users have faced and how they overcame them.
> >

> > Other things that would be helpful to know if you are still having
problems: What does your system contain?  Did it adequately minimize? What
kind of equilibration procedure did you perform?

It's dna, ions and water. After I placed the appropriate entries in the
ddb and bonded itp, I was checking whether I could go for 4fs, and I
prepared different runs with different time steps. The set-ups that made
the simulation crash were the ones where the dna had the time/chance to
cross pbc. If this did not occur, the simulation went fine (i.e. did not
explode). Also, in the mean time I run a simulation in NVE with the same
settings and I did no observe the crash after 5ns, despite having jumps
over the pbc in the last few hundreds of ps.

I energy minimized the system with steepest descend, repeated a couple of
times. They went fine after some initial lincs warnings. I admit I did not
equilibrate much, however the crash happens after more than 3ns... let's
consider the first 2ns some sort of tempering.

Although my nstxtcout was set low enough to resolve the frames just before
the crash, I strongly suspect it has to do with crossing the pbc.... I
will run a couple of simulations with output freq. 1 and double check.


> >
> > <snip>
> >
>> >> Tcoupl                   = v-rescale
>> >> tc_grps                  = system
>> >> tau_t                    = 0.1
>> >> ref_t                    = 300.00
> >
> > Maybe you want "Protein Non-Protein" instead of "system" here?  But
that's just general advice, without any knowledge of what's in your system.
> >


Right, although I don't think this is the problem. Just wanted to see if
the sim. would run fine for some ns before production runs. Just occured
to me now, I will also try with the old berendsen, just to double check,
since I did not have any problem in NVE.

> > -Justin

Thanks for your time!

Guillem

> >
>> >> Pcoupl                   = Berendsen
>> >> Pcoupltype               = isotropic
>> >> tau_p                    = 1.0
>> >> compressibility          = 4.5e-5
>> >> ref_p                    = 1.0
>> >> constraints              = all-bonds
>> >> constraint-algorithm     = Lincs
>> >> lincs-order              =  6
>> >> lincs-iter               = 2
>> >> lincs-warnangle          = 30
>> >>
>> >>
>> >> I've tried, perhaps naively, using both options for
periodic_molecule with
>> >> the same results.
>> >>
>> >> I guess it's hard to tell what did go wrong, but perhaps somebody
has an
>> >> idea.. I thought of submitting a bugzilla, but maybe there are some
more
>> >> tests I could do to pin point the problem before. Any ideas?
>> >>
>> >>
>> >> All the best,
>> >>
>> >>
>> >> Guillem
>> >>
>> >> Dr. Guillem Portella
>> >> MMB - Institute for Research in Biomedicine
>> >> Parc Cientific de Barcelona
>> >> http://mmb.pcb.ub.es/~gportella
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >> _______________________________________________
>> >> gmx-users mailing list    gmx-users at gromacs.org
>> >> http://www.gromacs.org/mailman/listinfo/gmx-users
>> >> Please search the archive at http://www.gromacs.org/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/mailing_lists/users.php
>> >>
> >