General updates | JW – Toolkit 0.11.0 released, uses PDBFile class to load PDBs and get basic connectivity, then overlays rich chemical substructures to learn bond orders and formal charges MT – Interchange is testing an early implementation of from_openmm. Interchange in general is moving into the ParmEd space. Lots of user workflows go lots of directions involving simulation software suites, openmm’s ForceField, OpenMMForceFields, ParmEd, etc. So we’re trying to figure out where we fit in and how we can best serve our community. For example, right now people mixing FFs from different packages MUST use OpenMMForceFields or ParmEd. PE – OMMFFs wraps some tleap/ambertools stuff. GAFF is tricky since it’s not appropriately specified - Just defined as “what the source code does”, hard to get more clarification than that. MT – Yeah, anticipating problems with claims about GAFF integrity. We can derisk this by using tleap but beyond that there will always be problems. PE – The best solution is to just say “wouldn’t you rather use OpenFF?” JW – Just wanted to give a heads-up, since misguided people often end up on the OpenMM issue tracker, so I thought PEastman should know that this is coming up.
MT – Vsite question JW – We’re getting slightly different TIP5P energies+geometries PE – There are two OMM TIP5P ffs - One “normal”, and one based on CHARMM, converted by ParmEd, which uses localcoordinatesite (Tried using charmm36/tip5p.xml, differences got larger) OpenFF will try comparing serialized system contents to new FF, since they both use LocalCoordinateSite MT – We’re using localcoordinatesite currently, but I’m looking at using OutOfPlaneSite to see if we can resolve the difference. PE – Yeah, OutOfPlaneSite will be much more straightforward for debugging. MT – Agree
|