Known issues and bugs (@striketeam)

Discussion link: https://openforcefieldgroup.slack.com/archives/CKSHCE7SB/p1611592728000200

Brief summary: Large discrepancies with experiment for the COVID Moonshot on compounds containing a chloro group on aromatic scaffolds

• https://www.nist.gov/srd/refprop

Discussion link:

https://openforcefieldgroup.slack.com/archives/CKSHCE7SB/p1596485145006600

Brief summary:

High preference to non-planar amides using Parsley 1.2.0

• https://openforcefield.atlassian.net/wiki/spaces/FF/pages/826474497

Discussion link:

https://openforcefieldgroup.slack.com/archives/C01BN0C5AKF/p1623935171007300

https://openforcefieldgroup.slack.com/archives/C01BN0C5AKF/p1623935450009200

Brief summary:

RMSD analysis of the QM minimum energy conformer and MM optimized conformer with 1.3.0

Observations of Thomas Fox:

• CC bond in oxirane 0.12A too long
  ("C1CO1")

• CC bond between the two C=O in quadratic acid 0.1A too long
  ("Oc1c(O)c(=O)c1=O")

• sulfoximide S=N bond 0.18A too long
  (Eg. "CCCCS(=N)(=O)CCC(N)C(O)=O")

• 2-vinyl-furan ("C=CC1=CC=CO1") is not planar

• aromatic nitro groups not in plane with ring (Eg. "C1=CC=C(C=C1)[N+](=O)[O-]")

• aromatic thioethers out of plane (Eg. "CSC1=CC=CC=C1")

• ring puckering of spiro-pyrrolidine
  (Eg. "C1CNCC12NC3=CC=CC=C3C(=O)N2")

• aryl-methoxy groups not planar (45°)
  (Eg. "COC1=CC=CC=C1")
  
  Disclaimer: added some smiles as a starting point/reference in case of general chemical space observations, may not be the exact molecules Thomas Fox is referring to when you see Eg..
  
  

No action needed

Problem could not be reproduced with publically available molecules

Open

QM in green above (from PhalkEtOH dataset)

Open

1.2.0 in magenta above

1.3.0 in cyan above

Discussion link:

• From Josh Horton’s lengthy MM run

Discussion link:

t49 "*~[#7a]:[#6a:3]~*" in Sage 2.1.- seems to be handled by later t84; should t49 be deleted?

(raised by Paul Labute from CCG in a private email)

@Alexandra McIsaac

Completed

t49 is not redundant

t123 "[*:1]~[#15:2]-[#6:3]-[*:4]" in Sage 2.1.0 seems entirely contained in t123a and t124 - should t123 be deleted?  The V1 value is suspicious too.

@Alexandra McIsaac

Completed

t123 is redundant (at least for the molecules in the training set)

t164 "[*:1]~[#7:2]=[#15:3]~[*:4]" in Sage 2.1.0 seems suspicious with V2 = -0.9671 which encourages 90 degrees.  This can't be true for a true double bond between P and N (e.g.,C-P=N-C).  This rule seems intended to cover CN=P(C)(C)C  only.  Is this true?  If so, then #15X4 may be more appropriate.

@Brent Westbrook Only covered by 3 conformations of 1 molecule:

Completed

t164 needs way more training data

t129 "* [*:1]-[#8X2r5:2]-;@[#7X2r5:3]~[*:4]:" in Sage 2.1.0 has a suspicious V = -19.907.  Is this correct?

@Brent Westbrook

Completed

Reasonable for the training set, but may need more coverage in Sage 2.2

A rather more serious problem is with phosphorus parameters:

t159 "[*:1]-[#8X2:2]-[#15:3]~[*:4]"  9.3828   0 -1.8998   0  0.4283   0  0.0000   0  0.0000  0.0000 
t160 "[#8X2:1]-[#15:2]-[#8X2:3]-[#6X4:4]"  8.2041   0 -1.5697   0 -0.7592   0  0.0000   0  0.0000  0.0000 # t160

Are the phosphorus parameters are supposed to handle phosphines or just phosphates?

@Brent Westbrook t159 is covered 142 times in the sage 2.1.0 dataset, t160 is covered 74 times.

To do: phosphine vs phosphates question

t65 "[*:1]-[#6X4:2]-[#7X3$(*~[#8X1]):3]~[#8X1:4]" 

should this be nitro [#7X3$:3](~[#8X1])~[#8X1:4]" like other similar rules or is nitroso also intended? If nitroso is intended then all you need is [*:1]-[#6X4:2]-[#7X3:3]~[#8X1:4]

As of Sage 2.1.0, t65 is trained on largely nitro groups but also additional N-O groups. However, it’s only benchmarked on nitro groups in the industry dataset, and its intention pretty clearly seems to be nitro applications. Changing it to the suggested SMARTS may be best here.

t77-t79 amide rules appear to encourage cis amides