2025-03-06 JAC/LW check-in

For CCD dataset of 154 molecules, determining oxidation state appears to be more involved, so previously discussed moving forward with a spectrum of multiplicities, and the results can inform on our method and allow us to move forward.

Post processing is still required in two categories:

• Ferrocene groups often don’t have hydrogens in these structures. We have a script to resolve that for 9 of 9 structures

• 14 structures are missing coordinates for 1 to 3 atoms. Not resolved

  • Can we freeze some atoms and use UFF for the ones that are missing?

    • LW: Are they hydrogens?

    • JC: Unfortunately not

    • LW: I suspect there is but if you don’t find an RDKit method, we should discard the geometries

    • JC: Sounds good, I’ll post in slack to get other takes

Submitting QCPortal datasets without QCSubmit

It seems that the most straightforward way would be to write a functions to:

• Create a json file from a QCPortal dataset to provide the same outcome as QCSubmit serialization

• Create a function to import that json file into a QCPortal dataset (and extra metadata?)
  Most should be covered by:
  dataset.to_dict()
  Molecule.to_qcshcema()
  Spec.to_qcschema() # (separate into protocols, optimize_procedure, and qc_specifications?)

• LW: would start by looking at submit()

•  

The odd ones out:

"protocols": {}
"optimization_procedure": geometric?

• Should be combined with "qc_specifications" for QCPortal

"filtered_molecules" : filters and molecules that didn't make valid QCSchema molecules

Ok this still seems like more than what we want to do now, so summarize notes in Zenhub epic for revamping QCSubmit

BTW: Added epic to start putting ideas into QCSchema updates