Some concern about this change affecting results
LM – Did some digging on this, have a few slidfes I can show. LW and I think we should stick with the previous /default criteria. In the longer run we thought it could be helpful to have it be a user-settable parameter but it’ll be easy for those details to get lost in the data store.
MT – Right, there isn’t machinery to support multiple MM conformers for a single mol.
LM – Yeah, there would need to be some way to record the provenance of each MM opt.
MT – So I think this is approaching a decision point where you+LW currently want this to be set to default, with an open issue to discuss user-customaization here.
MT – This sounds like having different optimization settings in different benchmarks wouldn’t have a SYSTEMATIC effect, but should rather have a RANDOM effect.
LM – For geo opts - part of the benchmarking is MM optimization from the QM opt. Sometimes this will lead to leaving the QM minimum and heading to a more global minimum. So having looser convergence criteria will likely change results, by taking fewer steps and staying closer to QM minimum.