2021-05-25 Topology Refactor Working Session Meeting notes

Processing the CCD

• IP – Found aa-variants.cif from PDB – 218 entries with various protonation states. Contains main chain and terminal variants.

  • This file sometimes has long codes for variants, like “his_ll_dhd1”.

    • What does the suffix mean?

      • Our default dictionary should just take the three_letter_code as the primary ID

      • We don’t know and it’s probably not necessary to know. There are hints/a longer description in the _chem_comp.name field

    • Should we always call histidine “HIS”, or should we separate it into “HIE/HID/HIP”?

      • This will affect whether we can export residue schemes that can be interpreted by existing programs

      • Same with Lysine, Asn, Gln, Asp, Glu

      • Produce two substructure dictionaries, and offer a kwarg to perceive_residues for which to use (AMBER vs. CCD)

        • Let’s just make the CCD dictionary for now. Can do AMBER later if we have time.

    • How do we handle the leaving atoms?

      • Make SMARTS both WITH all leaving atoms, and also WITHOUT any leaving atoms.

    • How do we filter out unphysical/unnecessary SMARTS?

      • Options for removing

        • Do they all have “fragment“ in the name?

        • Filter out SMARTS that have N- C- or C+

        • Filter out cif structures with FRAGMENT in the name

      • Not sure why these are in the database at all. If we need them later we can re-add them

• How do we validate the substructure library?

  • Checking for equivalence with the AMBER-derived SMARTS

  • Checking that we can label the AMBER structures' residues

  • Checking against T4

Creating feature branch

Opened

Next steps:

• Making substructure dictionary from CCD

  • Which method(s) to make public?

    • Make it all private (maybe a file _cif_to_substructure_dict.py with the classes+methods)

  • How to handle extra dependencies?

    • @requires_package decorator

  • How to package data for testing?

    • Could have a way to download release tarball from amber_ff_porting

    • Normal tests could be done on t4-protein.pdb and t4-protein.sdf (we can make the sdf)

      • Decorate these tests with pytest.mark.slow

  • IP will open a PR to bring in CIF-reading functionality, and include the current substructure dictionary.

  • JW made T4-protein.sdf using OpenEye

    • IP will review + merge

• Loading from PDB

  • Where can we get a PDB with CONECT records?

    • IP – Maybe an output from GLYCAM/Charmm-GUI?

    • JW – Otherwise we may be able to hack something using ambertools?

    • (Tried CHARMM-GUI, Chimera, GLYCAM, other tools, no luck getting CONECT records)

    • We’ll keep looking into where to get PDB files with CONECT records

• Adding tests

• Adding atom metadata

• HierarchyElements and HierarchySchemes implementation

• TypedAtom+TypedMolecule

• merge_molecules

Canonical atom ordering

Antechamber issues