Versions Compared

Old Version 1

changes.mady.by.user David Dotson

Saved on

compared with

New Version 2

changes.mady.by.user David Dotson

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

Goals

  • Assessment of

  • Assess the state of the problem in which torsiondrives are missing optimizations, and in some cases duplicate torsiondrives exist that refer to the same optimizations

  • Determine if these issues affect recent submissions; judge if work is needed to prevent new cases

  • Decide on a remediation approach for existing cases

Discussion topics

Item

Presenter

Notes

Describe the problem

Ben

  • BP: list of bad ids identified

    • Github link macro
      linkhttps://github.com/MolSSI/QCFractal/issues/540#issuecomment-830332247

    • DD: fortunately it looks like the most recent case is 6 to 8 months ago; it may be the case that the server issue is addressed, but now we need to assess

Remediation approach

Ben

  • TG: Biphenyl set is impacted, features the 17… torsiondrives

  • BP: querying the DB

    • OpenFF Group1 Torsions

    • OpenFF Gen 2 Torsion Set 5 Bayer

    • SMRINOFF Coverage Torsion Set 1

    • OpenFF Gen 2 Torsion Set 2 Coverage 2

    • OpenFF Gen 2 Torsion Set 2 Coverage

    • … Phenyl Benchmark

  • TG: is there a way to fix the minimum energy lookup in the client?

    • BP: the trouble is that the Torsiondrive Record is missing the optimization ids it should have. They exist in the DB, but the TD record doesn’t know that

    • BP: solution for these cases we can proceed with:

      • I will go through the 30 or so cases, find a matching Optimization for the Torsiondrive given the minimum energy recorded by that Torsiondrive

        • not too hard to do, bit of a judgment call in some cases

WBO data

Trevor

  • TG: there are cases in some old datasets where the WBO results are missing; is this fixable too?

  • BP: probably not, if it’s not present there, then it’s not anywhere

  • TG: okay, will follow-up in Slack on some cases; may be worth figuring out how widespread the problem is across the whole DB

    • e.g. does it only impact old datasets before a certain date

...