[gmx-developers] constraints and BAR free energy

[hess at sbc.su.se]

Hi,

Strictly speaking BAR is impossible with changing constraint lengths.
Since the constrained distances are different for different lambda's,
the ensembles never overlap.

So there are two options:
1) I forget about BAR and do free energy integration
2) I use Delta_lambda*dV/dlambda for the constraint contribution

1) Has the same "approximation" for the contraints as 2) and additionally
has larger and systematic errors. So I would go for 2).

In most situations constraint lengths change very little between
adjacent lambda points, so 2) should be a very good approximation.

Berk

> Hello,
>
> I have a question about the perturbation of constraints. My understanding
> is
> that dv/dlambda is easily calculated for constraints, so that constraint
> perturbation is fine for thermodynamic integration - but that calculating
> the
> deltaV to another lambda value as required for e.g. BAR doesn't make sense
> or
> isn't practical. Here's the relevant comment in force.c:
>
> /* Notes on the foreign lambda free energy difference evaluation:
>  * Adding the potential and ekin terms that depend linearly on lambda
>  * as delta lam * dvdl to the energy differences is exact.
>  * For the constraint dvdl this is not exact, but we have no other option.
>  */
>
> So far I've always avoided perturbing constraints but I now need to
> recalculate
> some results based on topologies with a ring-bound hydrogen mutating to a
> methyl
> group. Is there an exact way to do this in gromacs using BAR?
>
> More generally I'd be concerned about the accuracy of the linear
> extrapolation
> of constraint dv/dl - depending on how you set out your transformation
> pathway I
> would imagine you could end up skipping over a lot of detail in exactly
> this
> kind of situation where a methyl group is 'pushing' out from a hydrogen.
>
> Best wishes,
>
> Floris
>
>
>
> --
> gmx-developers mailing list
> gmx-developers at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-developers
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-developers-request at gromacs.org.
>