2025-07-09 Cole Lab/OpenFF Meeting notes (Aus)

• Grant Application: Protein FFs

DC

• Planning grant with Alice Allen - looking at protein FFs and difficulties in fitting to QM. Question meaning to ask Chapin. For non-bonded terms, previously used to make sure that hydration free energies were reasonable. Is there enough data for side chains to fit non-bonded parameters for densities/ enthalpies of mixing.

  • LW: Data might be relatively spare. Haven’t ever benchmarked with AM1BCC charges to make sure they do well.

• DC: Thought ESPALOMA was better than AMBER, but not.

  • LW: Doesn’t hold protein stable.

• Grant Application: Large-Scale NAGL fit
  
  

DC

• DC: Grant application for lots of GPUs for training ESPALOMA-type NAGL model on large-scale data. Issues with pooling operations in NAGL?

  • LW: Not in release version of NAGL, but in branch.

• Latest results

FC

• LW: How different are fitted terms in the 1-4 fit

  • FC: Think they were fairly similar, but should check

• LW: For systematically fitting torsion types, this can be difficult when test set molecules are very different to training set which has been used to decide torsions. Would recommend building in explicit multiplicities from the start.

• LW: How are you decomposing torsions by contribution?

  • FC: SMEE