@IAlibay what is the urgency of this? I'd like to triage it into an upcoming milestone, so any sense of when you'd like to see this enhancement would help!

@atravitz is driving this; will work closely with @jthorton to develop our approach.

Here are some more thoughts based on the last GUFE meeting on the design of these components.

Based on the need for an ExplicitSolvationComponent we identified a general ExplicitRepeatedComponent that be created which represents any molecule in the system that is repeated like a solvent or a membrane. This class would then have the following attributes:

ExplicitRepeatedComponent:

• smiles of the repeated unit molecule (needed if we are to parameterise using Sage to ensure correct bonding info if loaded from PDB file)
• some representation of the positions and atoms (rdkit molecule?)
• need to be able to load from a PDB/pdbx file
• convert to pdb/pdbx and openmm topology for system creation
• flag to consider it a small molecule or not (this flags if we should treat the unit with the small molecule force field (solvation calculation) or not (membrane))

From this we can then create an ExplicitSolvationComponent which adds optional box vectors to the class and a MembraneComponent though I am not sure what extra attributes that component would need over the ExplicitRepeatedComponent.

From the need to be loaded from a pdb it might make sense if it inherits from the protein component are there any draw backs to this and are there any glaring issues with this design?

One thing I would point out is that from a user perspective this is a bit more awkward, simulating an explicitly solvated protein-ligand complex now requires you to extract the protein, solvent, and small molecule into different files before loading them into the components and creating the system. Does this seem reasonable @atravitz, @IAlibay @hannahbaumann ?

From an initial read, I don't think this would work.

Could you maybe present this at a meeting for further discussion please?

A simpler option for the solvated membrane component case is to just sneak this all into a ProteinCompoent and add box vectors to the model see the example here then when defining the system the solvent component would just be missed out and no solvation would be done on the system. This however will not help the solvation free energy case which will still need some kind of explicit solvation component.

Another related issue that might help with design work OpenFreeEnergy/openfe#911

We are currently gathering user stories from folks on this to help guide the design of the implementation. We want to get these user stories as part of our v1.3 cycle, even if the implementation lands in a future release.