2024-11-04 Science team meeting

Project updates

BW

• Past week

  • Refiltering industry dataset

    • LW: Is this meant to be the required way people do it (re: Github PR)?

      • BW: No just recommended/example

    • LW: When would we want to re-filter it?

      • BW: If packages update or if we want to add new datasets

    • LW: Should we over-engineer it and put dates/numbers on it now? Might be good to keep some standards up to date

    • (Everyone): How to handle outliers?

      • Unclear, kind of intersects with other discussion of bad QCA records. Table until tomorrow’s meeting?

      • Also would we want these versions to be a “release” dataset

  • qcsubmit and bespokefit PRs

  • lipidmaps dataset

  • test organometallics dataset

    • BW: filtering for >10 atoms, q <4, periods in smiles

• Next week

  • Run benchmarks

  • lipidmaps dataset

  • organometallics dataset

Project

updates

AMI

• Past week

  • Sage 2.2.1 + S data + TM data/splits result

    • P angle not improved

    • (Conclusion): re-fit with P angle frozen, lower priority for now, could investigate splitting angle in future

  • DDX error analysis

    • Got it down to 5%

    • Need to delete records from dataset and resubmit them with a new guess

    • Probably can’t do it through QCA-DS

  • AIMNet lit search

• Next week

  • More on DDX dataset--re-submit stragglers with new guess, start optimizations

  • Look into NAGL architecture

  • Look into testing dataset

LW

• Past week:

  • CSIRO talk

  • Evaluator on NRP!

  • Interchange packing and simulation (+Evaluator)

• Next week:

  • Protein stuff?

  • Evaluator (virtual sites)

  • Interchange follow-ups