2024-08-06 AMI/LW Check-in

General updates and discussions

• Conformer energy analysis

  • Problematic angles

    • r3-side: X(non-ring)-r3-O(non-ring)

    • Pavan’s ideas

      • Tried adding pairwise conformer energy targets to existing workflow. Didn’t work

      • Spurious minima method, like CC’s

    • AMI – want to make a smaller dataset to train

    • LW – does smee do better, energy wise, on structural matches? (by % of matches not number of molecules)

• ConformerRMSD filter

  • Ran into bug with atom ordering

  • Re-filtering training data removed 100+ more conformers with the proper filter (out of ~5k)

  • Not a blocker to current progress but should be fixed

• NAGL

  • LW – started putting scripts up