Can we standardize small molecule force field parameterization through SystemGenerator?

[jchodera]

@SimonBoothroyd : We're approaching a new release of OpenMM that would allow us to use the new openforcefields.generators.SystemGenerator to generate parameterized systems that include small molecules using a common interface for SMIRNOFF, GAFF, and potentially future force field residue template generator plugins. Could this play some role in the future of property evaluator, allowing us to standardize the way we implement plugins to provide other small molecule force field types?

Because SystemGenerator currently uses the OpenMM ForceField class under the hood, it's a stop-gap solution until we can more fully switch to an all-openforcefield toolkit infrastructure, but the concept of isolating all external small molecule parameter assignment engines behind a single API that permits new force fields (like CGenFF, MMFF, and others) to be "plugged in" could be useful if we can somehow share the infrastructure among projects.

[SimonBoothroyd]

Could this play some role in the future of property evaluator, allowing us to standardize the way we implement plugins to provide other small molecule force field types?

I would love for this to play a role (i.e replace the majority of the custom FF assignment code in PE) - it looks like a much cleaner / extensible solution.

I'll circle back to this once the new OpenMM release drops (or please feel free to bump this if I don't action it).

[mattwthompson]

I think an updated interpretation of this feature request is "can we use things like EspalomaTemplateGenerator with Evaluator?"

To which I figure the answer is "probably" if this is of interest - is it?