Page Comparison

Create molecules PR

• AD – Opened

• JW – Approved

Molecule test set planning

• What were the ChEMBL failures?

  • SDF-format 2D mols with implicit Hs

    • Should these be loadable into OFFTK?

      • If they encode the a specific stereoisomer with unambiguous protonation state, then yes (even if we don’t have coordinates for the protons)

• Characteristics of a good test set:

  • 100 - 300 molecules

  • Exercises stereo differences from issue #146

  • No implicit Hs in 3D file?

    • Implicit Hs are OK as long as stereo is defined

  • Stereo in connection table vs. 3D

  • Different stereo features

    • Github link macro
      link https://github.com/openforcefield/openff-toolkit/issues/146

    • Github link macro
      link https://github.com/openforcefield/openff-toolkit/issues/725

  • Multiple components (should lead to error)

  • 2D structures? Only if they unambiguously define stereo in connection table

  • Coordinate-less structures?

  • Structures where connection table stereo contradicts 3d?

Performance in test_forcefields

• AD – The biggest timesink is the conversions to OEMols and RDMols. It’s possible to cache on toolkitwrappers, but it may be better to have a ChemicalEnvironmentMatcher object that a molecule/topology can generate. This would work because they’re immuatable once they’re in a topology.

Decision tree for loading mols

• Molecule has no coordinates (reading from SMILES, mapped SMILES, SDF or mol2 with all-zero coords) or molecule has 2D coordinates (reading from 2D SDF or mol2)

  • Hydrogens are explicit

    • Stereochemistry fully defined in connection table according to OFF definition

      • Load molecule

    • Stereochemistry incompletely defined in connection table according to OFF definition

      • UndefinedStereochemistryError unless allow_undefined_stereo=True

  • Some hydrogens are implicit, but the existence of all hydrogens is unambiguous

    • Stereochemistry fully defined in connection table according to OFF definition

      • Load molecule, with implicit hydrogens getting (0,0,0) as coords

    • Stereochemistry incompletely defined in connection table according to OFF definition

      • UndefinedStereochemistryError unless allow_undefined_stereo=True

  • Hydrogens are implicit and ambiguous

    • AmbiguousProtonationError

• Molecule has 3D coordinates (reading from 3D SDF or mol2)

  • Hydrogens are explicit

    • Stereochemistry IS defined in connection table according to OFF definition

      • 3D stereochemistry AGREES with connection table stereochemistry

        • Load molecule

      • 3D stereochemistry DISAGREES with connection table stereochemistry

        • UndefinedStereochemistryError unless allow_undefined_stereo=True. If molecule IS loaded, set stereochemistry data fields to the connection table stereochemistry.

    • Stereochemistry IS NOT defined in connection table according to OFF definition

      • Stereochemistry can be guessed from 3D coordinates

        • Load molecule, setting stereo to what’s indicated by 3D coords

  • Some hydrogens are implicit, but the existence of all hydrogens is unambiguous

    • Stereochemistry fully defined in connection table according to OFF definition

      • Load molecule, making hydrogens explicit and setting their coordinates to (0,0,0)

    • Stereochemistry incompletely defined in connection table according to OFF definition

      • UndefinedStereochemistryError unless allow_undefined_stereo=True

  • Hydrogens are implicit and ambiguous

    • AmbiguousProtonationError