Stereochemistry about openff-fragmenter HOT 6 CLOSED

I presume we currently want to use the canonical isomeric SMILES string as our way of judging whether fragments are unique. Later on, we may relax that.

from openff-fragmenter.

For the incoming molecules to fragment - should we enforce isomeric SMILES? If it's not isomeric, should we generate an isomeric SMILES before we generate conformations of the molecule to fragment?
In general I need to add a cleanup step for incoming molecules - generating an isomeric SMILES can be part of that step. However, shouldn't we also enumerate enantiomers if several stereocenters exist? The fragments can be different for different enantiomers.

from openff-fragmenter.

To clarify here:

• For incoming molecules, we need one stage that expands ambiguous stereochemistry and enumerates likely protonation/tautomeric states. At this point, all molecules will then have specific stereochemistry and explicit hydrogens.
• We should expand the JSON generated to specify both explicit-hydrogen canonical isomeric SMILES and non-isomeric SMILES so that either can be used as a key for indexing.
• The resulting molecules can then be distributed in parallel to another function, which fragments each of these molecules and passes fragments on to the next stage.

from openff-fragmenter.

We should expand the JSON generated to specify both explicit-hydrogen canonical isomeric SMILES and non-isomeric SMILES so that either can be used as a key for indexing.

@jchodera, The reason we want to expand the JSON to include explicit-hydrogen canonical SMILES is to avoid ambiguity for charged states. However, when the charged states are expanded, explicit H are added where needed.So is the explicit-hydrogen SMILES redundant?

Example for a positively charged Imatinib:
Cc1ccc(cc1Nc2nccc(n2)c3ccc[nH+]c3)NC(=O)c4ccc(cc4)CN5CCN(CC5)C
This is what it looks like with explicit hydrogens:
[H]c1c(c(c([n+](c1[H])[H])[H])c2c(c(nc(n2)N([H])c3c(c(c(c(c3C([H])([H])[H])[H])[H])N([H])C(=O)c4c(c(c(c(c4[H])[H])C([H])([H])N5C(C(N(C(C5([H])[H])([H])[H])C([H])([H])[H])([H])[H])([H])[H])[H])[H])[H])[H])[H])[H]

Currently each fragment has 5 SMILES associated with it:

1. Explicit hydrogen index tagged SMILES (canonical and isomeric)
2. canonical isomeric SMILES
3. canonical isomeric explicit hydrogen SMILES
4. canonical SMILES
5. canonical explicit hydrogen SMILES.

Once the database starts getting larger, this might become too large to maintain.

from openff-fragmenter.

I think all five will be useful! But it's possible 3 and 5 are redundant if the tautomeric and charge state are both encoded by 2 and 4. This is probably a question for @bannanc or @cbayly13

from openff-fragmenter.

Stereoisomer (cis/trans, R/S) enumeration was addressed with #10.

from openff-fragmenter.