Update Dataset Tracking

Project Board; Slides

• Running PR 440: Chodera tmQM

  • Still moving

• PR449: In scientific review

• JW Agrees with everything JC says

QDS handling of non-QCSubmit dataset.

Scaffold Submission PR is completed and approved

• Reliant on PR to update QCFractal version on QDS, in current iteration

Dataset archival project

Complete!

MolSSI Info / Align Priorities on MolSSI Asks

Notes July 8th meeting

New from last QCAUM meeting:

• New release QCF v6.2!

• Ben offered VTech resources 96 cores + 768 GB. He has lots of CPU time in my allocation that I don’t use.

• Issue with large dataset views, Ben suggested saving without trajectories:

• ij = ds.create_view( description=f"Full {ds_name} {ds_type} dataset", provenance={}, include=['**'], exclude=["wavefunction", "trajectory"], include_children=True )

However include_children=True overwrites this open, so it turns out that include_children=False will greatly reduce the file size (217 GB to 900 MB for the Industry Benchmarking Dataset v1.2)

• Built in QCFractal error cycling is available. Ben has enabled for us.

Requests:

• Consider running pr449-1200 records on MolSSI resources?

• Ben is making some changes to TorsionDrive records on a new branch. Can we test?

  • JW: Yes, we can make a draft PR, alter the devtools/basic.yaml and psi4.yaml a -pip section with the git link (not exactly a link, ask JW).

  • Watch CI on this pr: https://github.com/openforcefield/openff-qcsubmit/pull/347

Old Issue of the Week

Conformer generation should fall back to RDKit ETKDG on Omega failures (Closed!)

• John suggests that if Omega fails in generating initial conformers, RDKit should be the fallback.

  • JW: This complicates provenance

• Should this be a QCSubmit ticket?

  • JW: Yes, migrating to QCSubmit, using examples from QDS PR #2 we found that RDKit is already used as a fallback.

Bonus: Missing chemistry to (potentially) cover post-release-1 (Not addressed this week)

• [#8]~[#35]: O-Br single bonds are present in GAFF2 but not present in our current datasets. We could port in a placeholder value from GAFF2, but there are no molecules with this chemistry in our current datasets.

  • Still not addressed

• [#7X3]~[#7X3]~([#8])~[#8]: Nitroamines

  • Addressed with smirks="[#6,#7,#8:1]-[#7X3:2](~[#8X1])~[#8X1:3]"

• [#6:1]~[#6:2]=[#15:3]~[#6:4]: C=P double bond (potentially with adjacent singles)

  • Still not addressed

AI summary

QDS/QCSubmit Meeting Summary (July 15, 2025)

Current Projects and Progress

• Jennifer is working on converting QC Elemental molecules to RD Kit molecules to enable sorting based on fingerprints and connectivity

• This capability will help with better train-test splits of data and is of interest to Chris as well

• For molecules failing to assess in TMQM, Jennifer plans to use MoleAssembler to build complexes of interest

Data Set Issues

• Many calculations showed SCF convergence failures after 500 iterations

• 14 out of 30 errored structures had incorrect charges reported in CCD files

• Jennifer is developing a method to predict oxidation states, which she'll present to Richard on Thursday

• For the remaining 16 structures with issues, Jennifer plans to implement a tiered optimization approach with loose tolerances initially

Completed and Ongoing Tasks

• Scaffold submission PR is completed and approved

• Jennifer needs to submit a PR with the new QC Fractal version

• Data set archival is complete but waiting for James and Lily to return before closing

• Using include_children=false reduced dataset size from 217GB to 900MB

QC Fractal and Testing

• Ben contacted Jennifer about testing OpenFF code against a development branch of QC Portal

• Jeffrey demonstrated how to set up CI testing with a different version of QC Fractal

• For an old issue regarding conformer generation, they discovered RD Kit is already implemented as a fallback for Omega failures