Page Comparison

Lily

• Only optgeo targets were incorrect, with multiple confs

• Conformers beyond first had incorrect charges

• Atom reordering between charges

• CB: Omega changed and reordering may be an issue as well; default behavior was to reorder. We have fixed this at OE and made a test

• DM: So Omega will reorder atoms?

• CB: If you don’t set the flag, then yes it will reorder. In a new version of Omega, this flag was not set anymore.

Simon

• Expensive part is MM optimization

• LP: We use optimization because e.g. protein FF, after fitting the FF would go where no data is, and we would get weird structures

  • Solution was to adding new conformations/data and rerunning FB to fit. After about 3 cycles things worked

  • Optimization was introduced to preempt this issue, hopefully

• LP: After the parsley fit, things were a bit rushed and settings might have been baked into future fits

• CC: I see the same; it seems like the protein fits are affected by earlier fits

• SB: Where did the intial data come from for protein fits?

• LP: Used torsion drives and hessians

• CB: Energy is dominated by stiff dof, but the soft dof is important. Fitting to just energies might affect this

• Github link macro
  link https://github.com/openmm/NNPOps

• Github link macro
  link https://github.com/SimonBoothroyd/smee