2025-08-06 Cole Lab Check-In (Aus) Meeting Notes

Data-Generated Torsion Types

FC

• LW: What threshold were you using for torsion specificity levels?

  • FC: Can’t remember (EDIT: 10)

• LW: Tried something similar but used types which were too specific (using pretty much all of the available descriptors). Issues with generalisation.

Scientific input on YAMMBS PR

• LW: If metrics haven’t changed much since original PR, I’m happy with changes.

  • FC: No major changes since then

• FC: Also, approach of freezing torsion atoms possibly problematic. Another reason to implement geomeTRIC?

  • LW: Would be good to have both as options.

Fitting to OMOL25

LW

• LW: Hard as there are no SMILES – need to be guessed.