Page Comparison

Project updates

BW

• Last week

  • More benchmarks

    • v3 besmarts

    • null-looser-priors protein ff

    • null-0.0.3

  • QCA workers

    1x Hessian datasets still going (protomers)

  • More benchmarks

    • v3 besmarts

  • QCA workers

  • Second lipidmaps dataset

  • Organometallics dataset

  • yammbs/YDS torsions

  • Protein fits

    • lipid-maps-medium

      • Wrapping up soonish

    • lipid-maps-large (on virginia tech)

      • ~1% finished in 3 days across 1 node, 24 cores/job, 90 GB memory/job. Large ones seem to be using ~70 GB RAM

      • Molecules may be too large

  • Organometallics dataset

    • Waiting until QCA workers free up to merge

  • Protein fits

    • check weights and resubmit if they aren’t equal

    • BW: 47% Opt, 26% SM TDs, 27% proteins

  • nagl2

    • combine opt datasets into one

      • AMI came back and did this

    • basic dataset from combined dataset

      • same props as mlpepper

      • save wavefunction

    • Protein fit benchmarks

  Next week

  • • • deprioritised in favour of organometallics

  • Organometallics dataset

    • Will merge and hopefully get data for ML TM meeting

    • JCl and I met earlier to transition

  • Protein benchmarks

    • Still getting NaNs

    • LW: when did it crash? should be in *.out file

    • goes up to 156399 ps

  • nagl2

    • JCl: the README or gen needed a description for the NAGL2 dataset

    • (Gen): the script updates the description later in file

• Next week

  • Off-boarding:

    • Meeting with JE is exit interview

    • Meeting with JW (Fri) is account closure

    • curato, lipoma, benchmarking, vflib2, qcaide

    • make sure dfs9 is rw-able for everyone in the group

      • LW: only checked subdirectory

    • big datasets should be uploaded to zenodo or google drive

      • fragment database/s

    • Projects

      • Torsion multiplicity

      • Torsion profile target in ForceBalance

        • restrained-torsion-profile:

          Github link macro
          link https://github.com/

      openforcefield

      • • ntBre/

      proteinbenchmark

      • • forcebalance/

      blob/main/proteinbenchmark/benchmark_targets.py

      LW: in benchmark_targets.py – would recommend setting traj_length to 5 microseconds (in a new target) instead of 10, for purposes of speed. We should be able to extend the trajectory fairly easily later

      DCole NAGL repo:

      • OpenMM exception: NaN coordinate

      • LW: restart again? Otherwise I’ll see if I can visualize anything obviously wrong in VMD

        • tree/restrained-torsion-profile

      • FF fits for Trevor

        • Nothing further since v3

      • Smee fits

        • reran with scaling factors, fcs still went to 0

        • /dfs9/dmobley-lab/bwestbr1/

        proteinbenchmark/output/results/gb3-test-5us-null-looser-priors-tip3p/setup

      • /data/homezvol0/bwestbr1/mambaforge

    • 3 replicates to start with

Project updates

JCl

Past week:

• Working with JW on CMILES issue in existing datasets

  • In RDKit, some CMILES turn out as radicals but don’t result in errors. However, will error in OpenEye. Trying to add an error to catch this scenario from RDKit

Radicals with SMILES is a question mark – will we have to think about this with organometallics project?

More in future work

CMILES Issue: [H][NH+][H][H] has not implicit hydrogens in RDKit, but found that also doesn’t have all “explicit” hydrogens in graph. This is resolved with a test.

• First sage dataset done, needs to finish review

Next week:

• Working more on CMILES issue: Suspended until the new year. Need to go through GitHub issues

• Dataset release stuff

Project updates

LW

• Past week:

• Mostly off on leave!

• Evaluator on Kubernetes – passed off to Matt

Liquid simulation validation

• thermostats are not regulating temperatures

Next

• Finish Sage

• Industry Benchmarking dataset

• CIF parsers:

  • gemmi, other cif parsers

    • https://gemmi.readthedocs.io/en/latest/cif.html

    • https://mmcif.wwpdb.org/docs/sw-examples/python/html/index.html

  • Does BW have statistics for CCD?

    • BW: some of the filtering code in my repo could be used for this

    • Set up a pair programming session

    • Jeff is also a good resource

    • CIF also has bond orders, so possible to assemble rdkit molecule

  • QCA-DS

    • Uploading to Zenodo

Project updates

LW

• Past week:

  • Looking into Evaluator liquid simulations

  • Working with Matt on Kubernetes

  • Protein benchmarks

  Picking up everything that happened on leave

• Next week: