Elements and Functional Groups Currently not covered

[bannanc]

This a list of elements or functional groups that we currently don't cover with smirnoff99Frosst. I'm not necessarily suggesting we "should" or need to include anything listed here. This is a list of anything that would be assigned generic parameters given the current smirnoff99Frosst (as of 5/2017).

For the Bonds, Angles, and Torsions, I've shown images of the molecules and included parm@frosst atomtypes when available. There is a list of the molecule database we have looked at below.

Elements

These are elements that we do not include and are therefore missing all associated parameters.

• Boron (Discussion in smarty issue)
• Silicon
• Tellurium
• Selenium
• Arsenic

---

[bannanc]

Bonds

From DrugBank Database

5 molecules with the generic bond assigned

• Cl-O
• Cl=O
• O=O
• C#O
• Halogen-Halogen

From Zinc Molecule Set

One molecule with the generic bond assigned

• Cl-O

FreeSolv

No molecules with the generic bond assigned

eMolecules

A note this is not every missing molecules (that's greater than 1,000 molecules, but an idea of the functional groups that were not in the other sets

• Br-O
• I-O
• S-[O+1]
• F-O
• P-P

surprising eMolecules

• specific C=N bond
• C=C in chain of double bonds? this seems a bit odd that we miss it, but I imagine these groups may be relatively reactive and therefore not something you typically want to model in drug like molecules.

Weird chemistry in eMolecules

• There was more than one molecule with this weird aromatic ring with an extra double bond:

[bannanc]

Angles

From DrugBank Database

4 molecules with the generic angle assigned

• C-Cl=O (CA~CL~Ou)
• O=Cl=O (Ou~CL~Ou)
• O-Cl=O (O2~CL~Ou)
• F-S-F (F~Su~F)
• C-N=O (CT~N3~Ou)
• O-N=O (OH~N3~Ou)
  • I think this functional group should be a nitrate, I don't trust the nitrogen with four bonds and a radical?

From Zinc Molecule Set

No molecules with the generic angle assigned

FreeSolv

No molecules with the generic angle assigned

eMolecules

These all have weird charged nitrogens.

• S-N-O in nitro group
• O=[N+1]-C
• [O-]-[N+]=[N+]
• C-[N+]=[N+] This last one seems like it might be a representation issue, you could also right this functional group as O=N-N-[O-] with less formal charges, but I guess our SMIRKS patterns need to recognize it as the same molecule. Would be worth looking to see if GAFF can type it.

[bannanc]

Torsions

From DrugBank Molecule Set

9 molecules with the generic torsion assigned

• [C,H]-N-[nitrate N]~[nitrate O] ([H,C]~N~N2~O2)
• c:c-Cl=O ([CA,CB]~CA~CL~Ou)
• [H,C]-N-S=C (CT~N3~S~CM)
• [C,H]-C-S=O ([CT,H1]~CT~S~O)
  • more broadly torsions around X~[C,N]~S=X bonds
• O=C-[O+1]-H (O~C~Ou~HO) *c:c:,-[SA,Sa;+1]:,-c ([CA,CB]~CB~S~CB)

From Zinc Molecule Set

7 molecules with the generic torsion assigned

• C-N=P-N (CB~NB~P~NA)
• C-P=C-[C,N] (CC~P~CR~[CT,NA])
• c:c=[O+1]-C (CA~C~O~CT)
• c:c:[O+1]:c ([CB,CA]~CB~O~CA)
• N=[N+1]-C-[C,H] (NB~NB~CT~[CM,HP,CA,C])

FreeSolv

No molecules with the generic torsion assigned

eMolecules

There were only 948 molecules with "t1" in the old format, that is remarkably small for the 5million+ molecules tested. Most have the P=C or P=N bonds similar to drug bank. Below are some potentially notable additions.

• *~N-P~O N next to phosphate

• this is similar to drugbank in principle, but the neighboring phosphate like groups caught my eye

• C-N-[N+1]=N similar to the missing molecules in zinc
  

• sulfur next to phosphorous in those N-P bonds

• C-[N+]=C-S similar to above, but with sulfur, a reminder to keep new parameters general when possible

• C~P=C-[halogen]

• S-P weirdness:

• all torsions around S-S:
  

• extreme version of the N~P problem
  

Generally speaking torsions around these bonds:

• Around Nsp2-P and N=P
  • Nsp3 was covered with issue #32
• C=P
• C-P
• C=[O+]
• C-[O+]
• S-S
• P-P

[bannanc]

Molecule set:

• openforcefield/data/molecules/DrugBank_atyped.oeb 6647 molecules
  • 15/6647 molecules get generic parameters
  • Removed 192 molecules with metals
  • Removed molecules with boron (56), silicon (4), arsenic (12), selenium(14), tellurium (1), helium(1), and xenon(1).
  • Removed 7 molecules with more than 200 heavy atoms
  • Removed 11 molecules with smaller atoms (atomic number < 10) with bond order greater than 4
  • Removed 66 entries with more than one molecule, that is had a period ('.') in the SMILES string
• openforcefield/data/molecules/zinc-subset-tripos.mol2.gz 7500 molecules
  • 8/7500 molecules assigned a generic parameter
  • Removed 5 molecules with smaller atoms (atomic number < 10) with bond order greater than 4
  • This database had no metals, non-metals, or entries with multiple molecules
• MobleyLab/FreeSolv 643
  • No filtering
  • 0/643 molecules assigned a generic parameter
• eMolecules - I'm currently working on this set, it is MUCH larger (initially 8 million molecules) so there are some logistical barriers to contend with.
  • 5,689,262 after filtering for metals, metaloids, and inappropriate valency
  • GAFF cannot parameterize 357,589
  • 1,036 molecules get generic parameters

[bannanc]

I had a reminder in Slack about this article that models a ligand with Boron, potentially a source of data or temporary parameters. Saving here is a better place, it is in the Mobley Group Zotero as well https://pubs.acs.org/doi/10.1021/jacs.6b06566