Attendees

Piet, Tom Goddard, Scooter, Greg, Eric, Elaine

August 1, 2024

Agenda

• Selina work in progress

Discussion Notes

• Sequence databases
  • Scooter accidentally deleted BLAST and AlphaFold and ESMFold k-mer sequence databases.
  • Databases are not backed up and need to be rebuilt.
  • Cron update scripts rebuilt PDB and NR databases.
  • Tom restored AlphaFold and ESMFold databases.
  • Scooter writing a script to update Uniref databases which were 10 years out of date.

• Selina work in progress
  • Working on CLIP organ segmentation method.
  • Tested on UCSF PHI data, 72 data sets, only 35 usable.
  • Also tried unreleased newer nnUnet organ segmentation from John's Hopkins. Works slightly better than CLIP.
  • Analyzed different kinds of errors: under segmentation, over segmentation, mixing organs.
  • Ground truth established by eye. Selina says boundaries are visually clear.
  • Selina made prototype web-site showing results.
  • Scooter suggests next goal is write short paper showing CLIP and nnUnet analysis results.
  • Also want to make a ChimeraX bundle to call a web service to run CLIP or nnUnet.

• Ion classification
  • Responding to manuscript reviewers of MIC paper at Nature Communications.
  • Reviewers wanted a "no ion" class.
  • Working on making an "open bubble" Python module to calculate molecular neighborhood fingerprinting used by MIC.
  • Have made a ChimeraX tool to run it.
  • Will run MIC from ChimeraX on local computer using PyTorch since CPU speed is adequate.

Action Items