2025-06-18 Cole Lab Check-In Meeting Notes

Datasets for Zenodo

Jennifer Clark

• The two charge datasets will be merged into one

• No specific filtering required

Latest NAGL charge model results

Charlie Adams

• Ash charge model in excellent agreement with RESP ELF10

• ESPALOMA to be dropped from paper

SMEE-SPICE valence training and Bespokefit-SMEE Updates

Finlay Clark

• CC: Maybe 1-4 fit trapped in local minima - more favourable global minimum possible

• JH: Metadynamics on grid of torsions might become problematic as molecules get large

• DC: Is it fine just to use the model from the last minibatch?

  • FC: Think so - test loss doesn’t fluctuate too much.

  • JH: This is what we do with the charge models