[gmx-developers] Suggestion : annealing restraints

[Michel Cuendet]

Dear all,

Doing many independent simulations of the same system (with same 
conformation but different 50K initial velocities), I noticed that most 
of the variation between trajectories appears at the exact moment when 
initial restraints on heavy atoms are released. This seems to be a 
rather violent event for the system. Some people do several short 
simulations with decreasing restraint strength before complete release. 
In the charmm community for example, decreasing restraints step by step 
is common practice, since this is easily automated in a charmm script. 
The most gentle and most convenient way to release restraints would be 
in a continuous fashion.

A nice feature in gromacs would be a set of keywords similar those of 
the simulated annealing function, except that it would scale the 
restraint harmonic constants instead of  temperature. This way, the user 
could, in a single run, heat up the system with restraints, let it 
equilibrate, and then progressively remove the restraints. One could 
even imagine a group-based scaling, with groups defined with a 
restr_grps keyword for example.

If the usefulness of such a feature is recongnized, this should be easzy 
to implement...

Cheers,

Michel

-- 

==========================================================
Michel Cuendet, Ph.D
Molecular Modelling Group
Swiss Institute of Bioinformatics
CH-1015 Lausanne, Switzerland
http://lausanne.isb-sib.ch/~mcuendet
==========================================================