[gmx-developers] Re: gmx-developers Digest, Vol 46, Issue 9

[hessb at mpip-mainz.mpg.de]

> Hi, all-
>
>> I agree that this might often not be a good approximation.
>> But at least the exchange code is there now (although completely
>> invisible to the user). So people can experiment with it,
>> knowing the limitations.
>> Even without the foreign lambda code, one could just call do_force
>> with a different lambda at an exchange attempt.
>> But it might be better to wait till the foreign lambda code is ready.
>
> I think it might be better to wait.  Anyone who has the sophistication
> to really beta test the code is probably willing to wait to have
> exchange in there correctly!  I'm planning on waiting, for example :)
>
>> One issue with the foreign lambda code is constraints.
>> When constraint lengths depend on lambda, one can not simply
>> determine the potential energy for a different lambda
>> than at which the actual state is.
>
> Certainly, this is the case.  There should be clear warnings if there
> are any lambda-dependent constraints, so that people don't go and try
> to predict things using it.  I try to set up simulations where
> different constraints will cancel out from the thermodynamic cycle --
> for example, when mutating a ligand from A to B, turning off the
> bonded/torsion terms to a common nonbonded substructure in both the
> ligand and protein environment.
>

The constraint part should then be extrapolated from
the dG/dlambda constraint contribution.

> I think that this is another reason one might want to also implement a
> general Trotter decomposition method and do the bondeds every 1 fs,
> and the nonbondeds every 4 fs, instead of constraints.  Then there
> won't be any such issue.
>
> Cheers,
> Michael

Hmmm.
Formally this solves the problem, but in practice it might
not have much effect. One is usually not interested in the bond
potential itself, but in other contributions that would change due to
the constraint distance changes. A typical system I often have
is a constraint between two molecules.

Anyhow, the lambda exchange code is there, but not directly visible
to the user. One can simply change the one line for the energy
difference with the real lambda energy difference when proper
foreign lamdba code is ready.
I had already put a comment in the code stating that extrapolation
is used, but one would really like to have the proper energies.

Berk.