2024-04-30 BW/LW Check-in

General updates and discussion on projects

• Torsion splitting in 2.2

  • Added 54 new parameters with new code. Need

  • Started a new re-fit – haven’t done impropers

    • Includes new parameters, but not data. Does the newest QCA datasets

  • Target next FF fitting meeting

  • Next steps: generating new data

• Fragment dataset curation code

  • Tried RECAP, BRICS, ERB, XFF – ERB and XFF give smaller fragments

  • XFF first, then ERB, then fall back to RECAP/BRICS if not getting good results

  • Next steps: expand to remainder of ChEMBL, store in database – should be able to reduce number of fragments by not enumerating stars/bond splits

• Torsion shapes

  • Looked into scripts a bit

  • Still not sure how to address additional terms – MSM didn’t pan out

  • Possible ideas:

    • Same k values as existing terms

    • Loosening priors in fit

  • Possible re-fit without changing shapes, as this is likely to take a while

• hpc3

  • Previously hpc3 advised to check queue with intervals

  • Now getting advice not to