convergence of well tempered metadynamics using distance between two atoms as colvar

[joe_w]

Hello,
I am testing out some wt metadynamics simulations with plumed and have a rather basic question regarding convergence. Using distance between two nearby atoms in a protein as the only colvar, will the calculation be able to converge? I assume the system will start off sampling conformations with low energy barriers, but eventually, if run long enough, the bias will build up allowing sampling of all possible distances (and unfolding the protein in the process). Using standard wt meta dynamics parameters for this colvar (ie. sigma = 0.05, biasfactor=15), I presume that such a calculation would take a very long time to converge. In my case, the goal is not to sample all possible distances and unfold the protein, but rather to see the possible distances that could be attainable biologically (ie. with rather small energy barriers). I am not sure if I am on the right track with my whole thought process about this, but if anyone could provide some insights it would be much appreciated. Thanks.

[jan...@gmail.com]

hi Joe, 

this is pretty popular question and has been described in the user mailing list – just look for 'metadynamics convergence'

https://groups.google.com/forum/#!searchin/plumed-users/convergence$20metadynamics%7Csort:relevance

One additional criterion that's rarely mentioned is that your answer shouldn't depend on the initial condition: if you start your metad simulation at two very different points on your CV, the resultant free energy surface should be the about the same.