User feedback on structure of GROMACS files

[mattwthompson]

Under construction

• Is it safe to assume two isomorphic molecules are parametrized identically? Yes
  • Without this, each i.e. solvent molecule would need its own [ molecule ] directive, which scales poorly and is space- and time-inefficient
  • However, co-parametrizing isomorphic molecules is weird and it's sensible to force workarounds to enable that
• Any need to have molecules exported to .itp files? Nice to have, not necessary
• User pain points with existing workflows (ParmEd + gmx tools + AmberTools/tleap + ? + OpenMM + ?)
  • Some important parameters (1-4 scaling factors, bond constraints) are reliably mangled or get reset to default values in OpenMM -> ParmEd conversion pathways
  • Non-bonded parameters can get mangled if atom or atom types (forget which) have identical names in different molecules

[mattwthompson]

I met briefly with @hannahbaumann who gave the following feedback and described one of her workflows:

1. "Isomorphic molecules get the same parameters" seems like a reasonable assumption
2. ITP file writer might be useful in same cases, i.e. a user manually manipulating a ligand (or doing various alchemical things in free energy calculations)

Her "separated topologies" workflow is as follows:

• Construct monolithic topology files for complexes A and B
• Use some scripts to combine ligands in each complex into a one monolithic file with A and B (one of which is probably only dummy, non-interacting atoms)
• Use restraints to keep dummy ligand in place

The key take-away here is that topology manipulation/access (i.e. getting the index of the ligand molecule in either complex) is needed for this workflow to run smoothly. I think OpenFF's topology objects should work well for this since molecules are first-class citizens and everything's index is accessible. The dummy atoms/ligands will probably require some hacking, however.