2021-04-19 Meeting notes

PLBenchmarks demo

David H

• DH: document on Confluence

• targets.yml is the index of targets used by the code

  • for each target directory, there is a ligands.yml that includes experimental affinity data

  • target.yml

  • See this map:

  • 00_data, 01_protein/crd and 02_ligands/<lig>/crd must be defined; at the moment top files manually curated

• PLBenchmarks/metadata.py has some validation functions for checking internal consistency of index/top files with input data

• 03_hybrid is generated by workflow

• If there are charges in the SDF file, workflow will use those charges

• DD: will consider whether we can start using openff-system for gromacs parameterization

• workflow3_solvate.py can generate many numbered replicates

  • the outputs of these would be used for interfacing with FAH

• step 4 generates inputs for SGE, SLURM, ready for energy minimization

• step 5 submits to queuing system

• step 6 checks simulations for some basic issues

• step 7 analyzes results

  • could be 30mins per target, about 10 seconds per replicate edge

• Existing results live in benchmarkpl:

• Random components

  • initial velocities of atoms

  • placement of solvent, ions

• DH: you will likely need to install git-lfs in order to clone the full PLBenchmarks repo datasets