2025.11.19 Cole Lab Check-In Meeting Notes

Pre-Meeting Discussion

• DC: Anyone worked with osmotic coefficients for FF testing?

  • DC: In one of Paul Robustelli’s papers

  • DC: To do with whether ions cross an membrane or not

  • AA: What’s it useful for?

    • DC: Useful as a test

  • CC: CHARMM required NB fixes to get that right

  • FC: See work from Shirts lab

Bespoke fitting progress

• AA: Why fragment with MLP?

  • FC: Purely for comparison to bespokefit1

• AA: Why are fits variable?

  • FC: Likely due to generation of samples with MD

• CC: Could tune regularisation by minimising variability of fits

• JH: Might also be worth comparing Bespokefit1 and 2 on conformer energies and forces

SMARTS Matching

• New bespokefit makes whole-mol SMARTS patterns, but this is really slow because of symmetries. Rachel Pirie (Cole lab alum and now at NextMove) would be interested in implementing better performance in RDKit. Do we want to do that?

  • JW – We may be able to find a workaround with some more searching through rdkit/cleverness. But in general just having this work more performantly out of the box would be pretty nice.