Code

Data

(currently on google drive, should be released on github:https://drive.google.com/drive/folders/1A8ncO30eaS1vE1czGl3lwaIkjXgMyc4G?usp=sharing )

Content

• CB: valid benchmark set:

  • what’s the protein? → identity (EC number)

  • what are the ligands? → smiles

  • what are the activities?

  • SB: state information (temperature, pressure, ion concentration, …)

  • everything else is an interpretation. (methods, ff, poses, charges, ….)

• Stages

  1. above data

  2. + structures (PDB + poses)

  3. + partial charges, FF parameters

  4. method, method parameters

• Ligands:

  • Structure as sdf file

    • coordinates

    • partial charges?

    • activity?

    • reference?

  • Charges (CB: if you want to evaluate e.g. another pose, we want to keep the charges constant)

• Protein:

  • Structure as pdb file (protein.pdb + all other crystal molecules in <find_a_name>.pdb)

    • generated with gromacs gmx pdb2gmx

    • <find_a_name> : ‘water+other’, ‘water+cofactors’, ‘other’?

• Hybrid:

  • hybrid struct based on ligand A

  • hybrid struct based on ligand B

• Problems with current version:

  • partial charges

  • boxes (dimensions, number of mols)

  • position of waters and ions

• Split up data into essential input (coordinates in sdf + pdb) and results (detailed topologies (exact ff version and files), MD engine version, …)?
  Essential input: A file which cannot be generated from the other files in the release data (except if it takes very long/to many resources to generate it)

• Experimental data → OpenFF-wide centralized generalized storage?

• conda env export / gromacs exact version

Directory structure

...