2020-11-09 Sprint Planning Meeting notes

DH

• Attended+presented at German cheminformatics conference. Need to analyze Vytas’s data.

• Pharma benchmarking call.

• Working on writing benchmarking paper.

MT

• Wrapped up/finished lots of PRs/issues.

• Units meeting – Figuring out goalposts and which package meets our needs

  • (General) – How can we convince VU to adopt Pint? What’s the cost of spending time/energy on this? What are the best and worst outcomes?

DD

• Led benchmarks meeting. Had previously decided that some datasets will be public (about 1400 compounds per partner). Each partner will have private, proprietary datasets

  • Call for Public Compounds

• Worked on software proposal, need to finish that up.

  • Will use openff-benchmark as the home for this. Will lay out issues and ask for help from JH, JW, DH.

• Prepared some QCA datasets. One for Jessica Maat that I had promised a while ago. Another for John Chodera.

• Coordinated other QCA activity.

• DH – Action on dataset that Thomas Fox sent to Mobley?

  • DD – Yes, I’ll redirect that to Gary Tresadern and myself, and have those added to public datasets.

  • DH – I’ve looked at the compounds, it’s SMILES w/o hydrogens and w/o stereo

  • DD – We’ll need better resolution on those compounds. I’ll follow up with TF and DM.

• Need to figure out plans intelligent exploration subgroup – Looking for feedback on how scheduling/attendance will be

  • MT – I think it should keep going in some capacity. Not sure about pacing, could solicit more feedback.

  • DD – I’ll put out a poll for participants and decide on a path forward.

JW

• Worked on fixing Zenodo authorship – Webhooks are still tied to Chodera. Got login for info@openff and will move Zenodo hooks over that way.

• Contacted conda-forge folks on gitter. Probably made us look like clowns.

• Played with conf. gen CLI. Working on improving performance re: stereo and partial charges, output deduplication

• Some back-and-forth on issues and PRs.

• Started Toolkit Showcase example PR.

• TG discussion – Directions for authorship. Need to talk with him and Mobley about how to get on papers.

  • DD – Spoke with him a few weeks ago about this. Agree that we should find a paper-destined direction for his efforts.

PB

• Worked on WBOs and fitting. Some progress on separating trends based on chemical series. Iterating on some sets – will do Trevor’s next (currently working on Lim+Mobley mols)

• Worked on preparing some datasets involving protomer/tautomer enumeration. JH is thinking about ligand expo set

• Talked with Hyesu about getting involved in QM benchmark, expect to hear from her this week.

• Q: How to filter benchmark molecules? Using RDKit fingerprint. I’d like to exclude benchmarking molecules from my testing set.

  • JW – Would recommend doing graph isomorphism checks. Should not compare bond order, formal charge, or stereochemistry.

DH – Thinking about PLBenchmark repo. Looked at doing a release, I find it overwhelming to keep track of the data – Should it go in GH release assets? Git tree? Zenodo assets?.

• DH – Would like to be able to track the dataset changes between releases

• DH – Currently datasets are in the 100MB → 1 GB range

• DD – Would be good to make a policy document on this.

• (General) – Goals:

  • Stable link/URL

  • Track changes – Not necessarily line by line, but see when a file changes or is added/deleted

  • Unlimited storage

  • Explorable in browser

• (General) – Options

  • Github release assets

    • Does get a stable URL

    • Doesn’t track changes

  • Zenodo datasets

    • Does get a stable URL

    • Doesn’t track changes

  • Github git tree

    • Does track changes

    • Explorable in broser

    • 100GB storage

    • Stable-ish link

    • Could compress individual files, so diffs would still indicate when a file changed/was added/removed.

  • Git LFS?

    • MT – Looks liek ti compresses the contents and just displays hashes

  • Google drive link

Sprint planning

Begin at 8:45 Pacific