[gmx-users] free energy calculation

[David Mobley]

Hi,

> You are probably starting your runs from a non equilibrated structure.

I don't think this is necessarily true: For example, you could be
seeing the crash you describe on the first step of equilibration...
However, you might want to make usre you have thoroughly *minimized*
at each lambda value.

Generally, this means you have large forces for some reason. One
possible problem is failure to minimize. Another would be failure to
use soft core interactions, or (sometimes) a bad choice of soft core
parameters (I particularly recommend the parameters I list here:
http://www.dillgroup.ucsf.edu/group/wiki/index.php?title=Free_Energy:_Tutorial&printable=yes
which are fairly carefully selected.

> Assume, the system changes via a slow growth run.
> The atoms may rearrange due to changes in their potentials.
> Therefore it may be problematic, if you start from the same structure
> for different lambdas. Try simulating a one way slow growth and pick the
> snapshots close to your lambda value from the trajectory and start your
> "quasi-FEP" with those.

I don't agree with this. I think it's best to start with the same
structure at different lambdas, otherwise you can introduce
hysteresis. Regardless, if you minimize sufficiently, independently,
at each lambda value, you should be OK, assuming your other settings
are all OK. This always works fine for me.

David


> Paolo Cerri wrote:
> > Dear all,
> >
> > I'm trying to calculate free energy differences using the standard
> > method of thermodynamic integration.I calculate dG/dlambda for different
> > values of lambda and then integrate.
> >
> > When decoupling LJ interactions i have LINCS WARNING:
> >
> > Step 1, time 0.002 (ps)  LINCS WARNING
> > relative constraint deviation after LINCS:
> > max 0.007366 (between atoms 7 and 8) rms 0.002535
> > bonds that rotated more than 30 degrees:
> >  atom 1 atom 2  angle  previous, current, constraint length
> >       7      8   50.1    0.1010   0.1017      0.1010
> >       9     10   65.8    0.1090   0.1085      0.1090
> >      17     18   32.1    0.1010   0.1011      0.1010
> >      17     19   32.0    0.1010   0.1011      0.1010
> > [...]
> > and eventually the simulation stops (during the equlibration run at
> > lambda >= 0.9)
> >
> > I obtain the following message:
> >
> > -------------------------------------------------
> > Program mdrun, VERSION 3.3.1
> > Source code file: clincs.c, line: 559
> >
> > Fatal error:
> > Too many LINCS warnings (10001) - aborting to avoid logfile runaway.
> > This normally happens when your system is not sufficiently
> > equilibrated,or if you are changing lambda too fast in free energy
> > simulations.If you know what you are doing you can adjust the lincs
> > warning threshold
> > in your mdp file, but normally it is better to fix the problem.
> > --------------------------------------------------
> >
> > I remark that in my simulations lambda is not changed continuously
> > during simulation, but i perform several simulations at different fixed
> > lambda values.
> >
> > I change only LJ interactions (having previusly turned of Coulomb
> > interactions)
> > I use  Amber forcefield  ffamber99
> >
> > thanks in advance
> >
> > Paolo
> >
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > 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
> >
> > .
> >
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> 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
>