Physically impossible ring stereo
A drawing can show ring stereo that no real molecule can have: a hydrogen pointing into a small cage, or a small ring fused trans. InChI encodes such a drawing anyway and gives it its own identifier.
A hydrogen pointing into the cage
9-(Methylsulfonyl)-3-oxa-9-azabicyclo[3.3.1]nonan-7-amine. In a small bicycle like this both bridgehead hydrogens point outwards. The drawing on the left has both bridgehead wedges the same way, so one hydrogen would point into the cage. That molecule cannot exist; the chemist meant the one on the right.
As drawn: one H inside
A chiral InChI (/m0) for a drawing of an impossible molecule. InChI warns,
but the warning does not say what is wrong:
Warning (Ambiguous stereo: center(s))
One wedge flipped: both H outside
The real compound: achiral, and the amine carbon becomes a stereocentre (endo/exo). The two drawings of one compound get two different InChIs.
The check: in or out
Look along the line through the two bridgeheads. From one end the three bridges run clockwise, from the other end anticlockwise. So when both bridgehead substituents point out, the two bridgeheads have mirror-image local configurations. Equal configurations mean one points in.
"In" exists only in large cages, around bicyclo[4.4.4] (about 12 bridge atoms). The proposal flags it up to 8 bridge atoms: bicyclo[2.2.1], [2.2.2], [3.2.1], [3.3.1] and similar. The limits are literature values and need checking.
Small rings fused trans
A ring fusion is the same test with a zero-atom bridge. A cyclopropane fused to a cyclohexane must be cis; trans would need a ring of at least 7.
cis-Bicyclo[4.1.0]heptane
The real compound (norcarane).
trans-Bicyclo[4.1.0]heptane
Cannot exist, yet gets its own InChI, without any warning.
| Smaller ring | Trans-fused partner |
|---|---|
| 3 | impossible below 7 (trans-bicyclo[5.1.0] is isolable) |
| 4 | impossible below 5 (trans-bicyclo[3.2.0] is strained but known) |
| 5 or more | allowed |
Small rings and multiple bonds
Related limits, mostly warnings rather than stereo:
| Feature | Smallest isolable ring | In InChI |
|---|---|---|
| E (trans) double bond | 8 | already handled: ring E/Z is ignored below 8 |
| allene | 9 | to add to allene support |
| triple bond | 8 | not stereo; warning only |
| bridgehead double bond (Bredt's rule) | about 7 bridge atoms | not stereo; warning only |
Also considered: atropisomer axes tied back by a short tether that cannot hold the drawn twist (lower priority), and cages with several bridgehead pairs (the same test per pair; anything harder is left to 3D validators).
What would change
InChI would warn, e.g. "Impossible ring stereo", and name the atoms. The InChI itself stays as it is: InChI identifies what was drawn, not what was meant. Fixing the drawing is up to the person or program that made it.
Your view counts: is a warning enough, are the size limits right, which cases are missing? Use the Comment buttons on this page or the general discussion below.
How to use it
Nothing to switch on yet. Today InChI answers the impossible drawing with a general warning only:
$ inchi-1 bicyclononane_in_out.mol -STDIO -AuxNone -Key InChI=1S/C8H16N2O3S/c1-14(11,12)10-7-2-6(9)3-8(10)5-13-4-7/h6-8H,2-5,9H2,1H3/t7-,8-/m0/s1 InChIKey=KQSRGMGLOXLLJM-YUMQZZPRSA-N Warning (Ambiguous stereo: center(s)) (log)
With the proposal, the log would also say what is wrong and name the bridgehead atoms; the InChI would stay the same.