2021-04-12 Pulido/Wagner Meeting notes

Project planning and breaking down into 1-5 day chunks

• Scope of changes

  • Possibly a new set of functionality for things like assigning bond orders to a protein and holding hierarchy data in an openFF molecule

  • Lots of changes in topology.py

  • Probably some changes in molecule.py

  • Adding a new example showing how to load+use a protein

  • A small number of changes in other files

  • Lots of tests

• Gather requirements

  • Major users – Simon Boothroyd (maybe), Matt Thompson (might need more time to see what system/exporters need)

  • Perses developers – Dominic Rufa (Chodera lab)

  • Experimental users – Lily Wang, MoSDeF people

  • Adjacent Developers – Peter Eastman?

  • Cheminformatics expert who can advise on changes to the Molecule/Atom/Bond class – Andrew Dalke (if he joins soon enough)

    • Less coordination will be necessary if the Molecule class is compartmentalized from the Topology – for example, if the same mechanism that allows a Topology to be popoulated by a mix of AtomTypedMolecules and CheminformaticsMolecules could allow the molecule refactor to happen and just “pass through” the changed API points.

• Convert requirements into API examples

• Convert API examples into tests

• Build implementation to satisfy tests

class AtomBase(BaseModel):
    name: str

def CheminformaticsAtom(AtomBase):
    element: simtk.app.openmm.element.Element
    formal_charge: simtk.unit.Quantity

def AtomTypedAtom(AtomBase):
    atom_type: str

class TopologyAtom(AtomBase):
     ...
     topology_atom_index: int


atom1 = CheminformaticsAtom(element='C', formal_charge=0)
atom2 = AtomTypedAtom(atom_type='CA')

(somehow get these added to the same topology)
for atom in topology.topology_atoms:
    print(atom)
>>> TopologyAtom

The way that MosDeF works

• graph of elements with bonds but without bond orders/formal charges

• --(foyer atom type assignment)→

• graph of atom typed atoms (and possibly bond typed bonds) without bond orders/formal charges

  • JW – I’d guess that this is closest to what we need to support in an “AtomTypedMolecule”

• –(lookup table for FF parameters between atom types)-->

• ParmEd structure