2025-04-16 Cole Lab Check-In (Aus) Meeting Notes

DC: Issue of restraints on atoms in torsion scans - unrestrained optimisations seem reasonable

• LW: Would be happy to switch to no restraints as default, but with cutoff of 0.4 A for energy comparisons

DC: What to do next with SMEE - seems to be working well, so we should think of interesting ways to extend work.

• LW: On splitting types - BESMARTS is a very scientific codebase - likely tricky to apply as-is.

• LW: On gradient-based splitting - chat to Simon Boothroyd

FC: Idea of deliberately overfitting to see how much general types are coarse-graining

• LW: Try clustering Espaloma output

LW: On using SMEE to fit to Sage data - this was tried by Brent but not a great idea - likely too little data for this approach

LW: On Alkane torsion profiles in lipid project - Julianne is having issues with sticky torsions - could be due to drift of parameters and I wonder if they need a hard reset

• Look at butane rotations

• FC: Will upload latest fit

DC: Were thinking about test cases - might be worth adding this to our “sanity checks” for refit FF