[gmx-users] how to deal with LINCS warnings despite prolonged stabilization of the system?

[Mark Abraham]

This kind of instability boils down to "if the starting configuration
and model physics are capable of running a stable simulation, you're
being too rough with it" - as
http://www.gromacs.org/Documentation/Terminology/Blowing_Up says using
different words.

Each time one changes the ensemble, that's rough. Using position
restraints can be useful while getting the temperature close to right,
but just switching them off can be too rough. Consider lowering the
force constant over several stages. Switch barostat algorithm only
when restraints have been off in the previous stage.

For that matter, each time step is rough - use a smaller one for
equilibration than for production. 0.5fs can be necessary while things
stabilize. Or turn off constraints temporarily and go even shorter.

Mark


On Wed, Jul 3, 2013 at 10:52 AM, Anna Marabotti <amarabotti at unisa.it> wrote:
> Dear all,
> I'm writing again because I really need to understand what more can I do on
> my systems that are still dealing with LINCS warnings. I minimized further
> the systems (from emtol 500 to emtol 100) and the final potential energies
> and max force were nearly identical to those I obtained when emtol was set
> to 500, in my opinion this indicates that the system has already converged
> to a minimum. I made a NVT (thermostat: V-rescale) for 100 ps and a NPT
> (thermostat: V-rescale and barostat: Berendsen) for 1 ns, and I checked the
> energies of the system: all was regular without any apparent perturbation. I
> then started the production run in NPT ensemble (thermostat: V-rescale and
> barostat: Parrinello-Rahman or Berendsen - I tried both of them) and approx.
> after 10-15 ns my system turns on LINCS WARNINGS (but never at the same
> time), and sometimes produces many stepXXXXX.pdb files, stopping the
> simulation. When I look at the energies of the system, I don't see anything
> strange apart the fact that the graph is interrupted. When I see the
> simulations with VMD, I don't see anything strange apart some crossings of
> the periodic boundaries (I don't see any system "exploding" or so on). I
> have the doubt that the cofactor could be badly parameterized, but LINCS
> WARNINGS do not involve only the cofactor, but also the protein backbone
> (the cofactor is covalently bound to protein backbone, but LINCS WARNINGS do
> not involve the parts of protein directly bound to the cofactor: would it be
> possible such an effect when only the cofactor is badly parameterized?)
> Please give me some suggestions. I read documentation in GROMACS web site,
> but I did all I saw, apart for skipping LINCS WARNING, including the
> analysis of the g_energy terms and comparison with the simulations that went
> well. I don't see any significant difference (if somebody wants to see, I
> can provide the files)
>
> Thanks a lot
> Anna
>
> --
> ______________________________________________
> Anna Marabotti, Ph.D.
> Assistant Professor
> Department of Chemistry and Biology
> University of Salerno
> Via Giovanni Paolo II, 132
> 84084 Fisciano (SA)
> Italy
> 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)
>
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/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/Support/Mailing_Lists