...
Lorenzo D’Amore compared OpenFF 2.0 performance to GAFF 2.11 with a specific focus on ring moieties: https://openforcefieldgroup.slack.com/archives/CKSHCE7SB/p1670282381642029 (powerpoint also attached here for posterity)
| View file | ||
|---|---|---|
|
| View file | ||
|---|---|---|
|
As summarised by David Mobley :
...
In general, Lorenzo concludes that Sage performs better than GAFF when all atoms are in a ring (case 1), but underperforms in all other cases (2, 3, 4).
Sage 2.1 separated some ring terms out, so Lorenzo’s analysis may not be totally applicable. However, these terms still don’t fully specify how many atoms are in the ring, so we should still attempt to split out parameters based on this and see how they perform.
| Code Block |
|---|
- a18a child parameter to separate out some ring matches from a18
- a22a child parameter to separate out some ring matches from a22 |
Parameter | SMIRKS | Notes | ||
|---|---|---|---|---|
a18 |
| General N angle | ||
a18a |
| Atom 2 is any N in a ring that is not 6-membered | ||
a22 |
| General neutral N with 2 substituents | ||
a22a |
| Neutral N with 2 substituents in a 5-membered ring |