Umbrella samplling

[Rochika]

Hello everyone,

I am working with a porphyrin-like ligand bound in the active site of a protein. I would like to rotate this ligand by a full 2π around a specific axis by defining a torsional collective variable.

There are multiple possible rotational pathways, but I have applied geometric restraints to suppress all but one, so that the ligand can rotate only along the desired direction. Because the ligand is asymmetric, torsion values that differ by π (e.g., π/2 vs 3π/2) correspond to clearly different local environments. For this reason, I need to sample the entire 2π rotation, rather than restricting the torsion to [−π,π].

I performed the rotation using MOVINGRESTRAINT and then set up umbrella sampling with 64 windows, each corresponding to a target torsion value along the 2π path.

However, I am encountering an issue related to torsional periodicity. In particular, when I define:

• window 0 with a target torsion of 0

• window 32 with a target torsion of 3.14

I observe that configurations corresponding to +3.14 and −3.14 radians are mixed. As a result, I am unable to keep these two windows sampling distinct states, even though they should represent different physical orientations of the ligand.

My question is: how can I prevent the mixing of +π and −π when restraining a torsional CV, so that a continuous 0→2π rotation is properly enforced?
Is there a recommended way in PLUMED to handle this, for example by redefining the CV, changing its periodicity, or using an alternative variable?

Any suggestions would be greatly appreciated.

Thanks