Implement restraints referenced to center of geometries of atom selections

This is just a placeholder to remind us that we will need to implement variants of the Harmonic and FlatBottom restraints that use the centers of geometry of DSL-specified atom selections in order to carry out host-guest (and like T4 lysozyme) binding free energies for the paper.

We can use CustomCentroidBondForce for this purpose. For example, Harmonic would use something like this

# Harmonic restraint between two groups of atoms
num_groups = 2
force = openmm.CustomCentroidBondForce(num_groups, "(K/2)*distance(g1,g2)^2")
force.addGlobalParameter("K", K)
force.addGroup(particles1); # create group 0
force.addGroup(particles2); # create group 1
force.addBond([0,1], []); # add bond between groups 0 and 1
system.addForce(force)

[jchodera]

Ah sorry, unless you meant that the restraint object can take a DSL string during initialization. I think can add this easily.

Yes, this is what I meant! Since the ligand atom indices will be different for each ligand, I think it would be a nightmare without the ability to at least specify the ligand residue name.

Question: am I correct in thinking that the standard state correction in this case should be computed very similarly to the standard Harmonic restraint (i.e. integrate shell volume and compute log ratio with standard volume)?

Yes, as I understand it, the corrections would be identical to the case where only a single atom is chosen for receptor and ligand.

[davidlmobley]

I’m just commenting so I can follow this discussion. Thanks for working on this.