2021-09-01 AM1 studies meeting notes

What to present at FF release call?

• What do we want to hear from Bayly?

  • Tell Chris:

    • Here’s a broad overview of what we’ve done, making sure to describe our methods (oe vs antechamber, constrained vs original) and datasets (rearrangement vs. no rearrangement) and how we’re limited now by the lack of a “gold standard” method to optimize towards.

    • Here are differences in molecules between (AmberTools/original/no rearrangement) versus (OpenEye/original/no rearrangement)

      • These differences may indicate that the OE restraint scheme is harmful

    • Here are differences in molecules between (AmberTools/fully constrained/no rearrangement) versus (OpenEye/fully constrained/no rearrangement)

      • These differences indicate a difference in AM1 implementation in one program or another or a chemical perception issue

  • Ask Chris:

    • Moving forward, what is the “gold standard” for “good AM1 charges”?

• Movement score plots

  • May be better to use “RMSD” (root mean squared deviation)

Next week

• Steps forward

  • Will “gold standard” require additional computing resources

  • Which tools to implement restraints/AM1 calcs?

    • Another possible open-source AM1 implementation: