[WIP] Adding network edges

[IAlibay]

Work done in this PR:

• [ ] Updated edge yaml format to solve #92
• [ ] Added actual MSTs generated using the Perses mapper + geometric score

TODO:

• [x] Add star maps (working with @bobym to get a list of cognate ligands)
• [ ] Add mutli-star maps?
• [ ] Other scorers / mappers?

[ijpulidos]

From running a single iteration with perses using these edges I'm facing the following problems with eight (8) of the edges:

cdk8/out_18617587_22.stderr:ValueError: could not convert string to float: ' '
eg5/out_18617580_2.stderr:openmmtools.mcmc.IntegratorMoveError: Potential energy is NaN after 20 attempts of integration with move LangevinDynamicsMove
hif2a/out_18617578_12.stderr:openff.toolkit.utils.exceptions.MoleculeParseError: Unable to read molecule from file: ligands.sdf
shp2/out_18617589_13.stderr:perses.rjmc.atom_mapping.InvalidMappingException: Atom mapping contains no mapped atoms
shp2/out_18617589_20.stderr:perses.rjmc.atom_mapping.InvalidMappingException: Atom mapping contains no mapped atoms
shp2/out_18617589_6.stderr:perses.rjmc.atom_mapping.InvalidMappingException: Atom mapping contains no mapped atoms
shp2/out_18617589_8.stderr:openmmtools.mcmc.IntegratorMoveError: Potential energy is NaN after 20 attempts of integration with move LangevinDynamicsMove

I'll be digging into what's making these error out.

[ijpulidos]

Oh, I just realized that the edges information has atom mapping as key, while I'm not sure this is problematic, we probably want to avoid using spaces here, so we want keys to be atom_mapping instead.

[dotsdl]

@ijpulidos : can probably finish out this PR next week after review of currently-running benchmarks.

[dotsdl]

From @hannahbaumann: failures in edges may be stemming from poorly-equilibrated starting systems. Can be investigated post-merge of the PR.

[dotsdl]

Any updates on this front? Is anything now blocking this PR?

[IAlibay]

I believe it's still in QA last I saw @dotsdl