[gmx-users] dihedral restraint MD

[Mark Abraham]

On 13/06/2012 2:12 AM, Lin Huang wrote:
> Hello Everyone,
>
> I was willing to conduct umbrella sampling on the dihedrals derived from the
> sidechains of a ion channel. So i tried dihedral restraint to achieve the
> goal. However I encountered an awkward problem, hope someone can help me
> with that.
> The dihedrals of the initial conformation I used in my MD is -60 and 153
> degree, respectively for phi and psi. And then I restricted both of them to
> 180 degree and conducted the dihedral restraint MD for 2ns. However the
> dihedral analysis of my final result was quite awkward, since they kept on
> fluctuating just around their initial value during this 2ns MD process. At
> first, I considered maybe it was caused by the small restraint force
> constant, which was set as 100 in my .mdp file. Nevertheless, after I
> changed it to 500, no better result was given out. I am wondering whether
> the energy barrier is too high for them to cross, so in the 2ns short MD
> process, they have to move around their initial configuration.

That is very likely the case.

>   If this is
> the key to my problem, how can I solve this problem?

You will need to work out how to construct your initial conditions to be 
close to those you wish to simulate later. If they involve substantial 
rearrangement of your protein and/or membrane, then there is no way you 
can see it happen in a few ns of simulation. If you cannot source such a 
structure from outside, then one possible approach is to take your 
protein to a vacuum environment, use freeze groups on most of it and 
allow only the part you think needs to rearrange to move. Now using 
dihedral restraints to move things in the desired direction is easier 
(but you still might need to do several stages), because you can 
simulate much longer and there are no solvent degrees of freedom to 
rearrange. Then you will need to take that resulting structure and do a 
thorough re-equilibration in the condensed phase.

Mark