Do we need an additional atomnumber canonicalization within an equivalence class of atoms for example in cis platinium?

[gblanke02]

Is the current canonicalization process sufficient to provide a unique reproducable atom number distributionof of the ligand atoms around a metal atoms or do we need an additional canonicalization process within an equiavence class?

[schatzsc]

@gblanke02

To answer in German: "jein"

When you look at the results of the InChI webdemo with the "equivalence class" switched on, the two chlorido ligands and the ammine (NH3) groups are both correctly assigend to two different classes, Cl atoms with node labels 1+2 in class 1 and N atoms with node labels 3+4 in class 1 (doesn't really matter that there is no class 2, as the "class labels" can freely be chosen as long as they are bijective)

However, the order of the atoms within the equivalence class will depend on the position in the input molfile. Assuming that the atom with the highest priority (= lowest node label?!?) is Cl1 followed by Cl2 then the order of the two N atoms can either be N3 followed by N4 or N4 followed by N3 if moving around the four corners of the square-planar structure (I always get confused with the order since TUCAN orders by atomic number while InChI seems to canonicalize by alphabetic order of element symbol?!?)

Thus, two InChI "stereostrings" would be possible even for cisplatin:

1-2-3-4 or 1-2-4-3

Among these two, the canonical one has to be chosen, which would be 1-2-3-4

Same issue with transplatin - here the two options are:

1-3-2-4 or 1-4-2-3

Of which again the first one 1-3-2-4 is the canonical one since it gives preference to "lower node labels (= higher priority atoms) first"

But this does NOT require a new canonicalization of the atom labels themself - only the "stereostring" has to be canonicalized from a small number of options, which can be done "brute force" even for octahedral compounds, since the maximum number of different strings is 32 (as the six different ligand node labels are equivalent to a complex MABCDEF).

Assume that we have [MCl(H2O)5], which would give rise to only one equivalence class with more than one member, which is the five O atoms. If Cl node labels < O and not considering M and H, this would give Cl1 and O2-O6.

Now, if Cl1 is in the "special position" (along positive z-axis) then all of the O2 to O6 could be opposite and also the four remaining O in the equatorial plane permutated, for example:

1-2-3-4-5-6
1-6-2-3-4-5
1-5-6-2-3-4
1-4-5-6-2-3
1-3-4-5-6-2

For each pair X-Y-Y-Y-Y-X one could then also further permutate the 4xY, for example:

1-2-3-4-5-6
1-5-2-3-4-6
1-4-5-2-3-6
1-3-4-5-2-6

and so on, so among these different options also the "canonical one" must be chosen, which would be

1-2-3-4-5-6

[gblanke02]

Taking your example: “Thus, two InChI "stereostrings" would be possible even for cisplatin: 1-2-3-4 or 1-2-4-3” The molecule is already fully canonicalized on atom (number) level so the InChI “knows” whether it is . 1-2-3-4 or 1-2-4-3. Therefore I do not see the need for this additional check unless you convince me something better Von: Ulrich Schatzschneider ***@***.***> Gesendet: Donnerstag, 27. November 2025 16:00 An: IUPAC-InChI/InChI ***@***.***> Cc: ***@***.***; Mention ***@***.***> Betreff: Re: [IUPAC-InChI/InChI] Do we need an additional atomnumber canonicalization within an equivalence class of atoms for example in cis platinium? (Issue #169) [https://avatars.githubusercontent.com/u/88166560?s=20&v=4]schatzsc left a comment (IUPAC-InChI/InChI#169)<#169 (comment)> @gblanke02<https://github.com/gblanke02> To answer in German: "jein" When you look at the results of the InChI webdemo with the "equivalence class" switched on, the two chlorido ligands and the ammine (NH3) groups are both correctly assigend to two different classes, Cl atoms with node labels 3+4 in class 3 and N atoms with node labels 1+2 in class 1 (doesn't really matter that there is no class 2, as the "class labels" can freely be chosen as long as they are bijective) However, the order of the atoms within the equivalence class will depend on the position in the input molfile. Assuming that the atom with the highest priority (= lowest node label?!?) is N1 followed by N2 then the order of the two Cl atoms can either be Cl3 followed by Cl4 or Cl4 followed by Cl3 if moving around the four corners of the square-planar structure. Thus, two InChI "stereostrings" would be possible even for cisplatin: 1-2-3-4 or 1-2-4-3 Among these two, the canonical one has to be chosen, which would be 1-2-3-4 Same issue with transplatin - here the two options are: 1-3-2-4 or 1-4-2-3 Of which again the first one 1-3-2-4 is the canonical one since it gives preference to "lower node labels (= higher priority atoms) first" But this does NOT require a new canonicalization of the atom labels themself - only the "stereostring" has to be canonicalized from a small number of options, which can be done "brute force" even for octahedral compounds. Assume that we have [MCl(H2O)5], which would give rise to only one equivalence class with more than one member, which is the five O atoms. If Cl node labels > O and not considering M and H, this would give Cl6 and O1-O5. Now, if Cl6 is in the "special position" (along positive z-axis) then all of the O1 to O5 could be opposite and also the four remaining O in the equatorial plane permutated, for example: 1-2-3-4-5-6 2-1-3-4-5-6 3-1-2-4-5-6 4-1-2-3-4-6 5-1-2-3-4-6 For each pair X-Y-Y-Y-Y-X one could then also further permutate the 4xY, for example: 1-3-4-5-2-6 1-4-5-2-3-6 and so on, so among these different options also the "canonical one" must be chosen, which would be 1-2-3-4-5-6 cisplatin.png (view on web)<https://github.com/user-attachments/assets/d479fd34-3f56-4da3-9853-80def7bb5c1a> — Reply to this email directly, view it on GitHub<#169 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AT3EDCPSEKWAZBFWEXKTYJD364GVPAVCNFSM6AAAAACM2X6V2WVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKOBWGMZDGMRZGI>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>>

[schatzsc]

@gblanke02

“Thus, two InChI "stereostrings" would be possible even for cisplatin:

1-2-3-4 or 1-2-4-3”

The molecule is already fully canonicalized on atom (number) level so the InChI “knows” whether it is . 1-2-3-4 or 1-2-4-3. Therefore I do not see the need for this additional check unless you convince me something better

Put the following molfile in the InChI webdemo:

  ChemDraw11272517312D

  5  4  0  0  0  0  0  0  0  0999 V2000
    0.0000    0.0000    0.0000 Pt  0  0  0  0  0  0  0  0  0  0  0  0
    0.0000   -0.8250    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    0.8250    0.0000    0.0000 Cl  0  0  0  0  0  0  0  0  0  0  0  0
   -0.8250    0.0000    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    0.0000    0.8250    0.0000 Cl  0  0  0  0  0  0  0  0  0  0  0  0
  1  2  1  0      
  3  1  1  0      
  1  4  1  0      
  1  5  1  0      
M  END

For the index you get from bottom counterclockwise: 2-4-5-3 which canonicalized to 3-4-2-1.

If you move the highest priority (= lowest node label) to a special position, like bottom and enumerate counter-clockwise you get

1-3-4-2

However, if you exchange the coordinates of the two Cl atoms to

  ChemDraw11272517312D

  5  4  0  0  0  0  0  0  0  0999 V2000
    0.0000    0.0000    0.0000 Pt  0  0  0  0  0  0  0  0  0  0  0  0
    0.0000   -0.8250    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    0.0000    0.8250    0.0000 Cl  0  0  0  0  0  0  0  0  0  0  0  0
   -0.8250    0.0000    0.0000 N   0  0  0  0  0  0  0  0  0  0  0  0
    0.8250    0.0000    0.0000 Cl  0  0  0  0  0  0  0  0  0  0  0  0
  1  2  1  0      
  3  1  1  0      
  1  4  1  0      
  1  5  1  0      
M  END

you get non-canonicalized 2-4-3-5 and canonicalized 3-4-1-2

If you move the highest priority atom to bottom left you get

1-2-3-4

This is because the xyz coordinates remain attached to a certain atom and if you exchange them, also the order changes.

[schatzsc]

@gblanke02

As long as you only consider the atom labels, this does not matter, but if the coordinates are additionally evaluated and used to rotate a molecule to a standard orientation that matters.

Within an equivalence class, you always need to do a 2nd canonicalization step with regard to the position in the stereostring.

[gblanke02]

Taken fom above

Molfile 1 with related InChI and indices (original atom number,canonical number, equivalence class taken from the Web Demo

ChemDraw11272517312D

5 4 0 0 0 0 0 0 0 0999 V2000
0.0000 0.0000 0.0000 Pt 0 0 0 0 0 0 0 0 0 0 0 0
0.0000 -0.8250 0.0000 N 0 0 0 0 0 0 0 0 0 0 0 0
0.8250 0.0000 0.0000 Cl 0 0 0 0 0 0 0 0 0 0 0 0
-0.8250 0.0000 0.0000 N 0 0 0 0 0 0 0 0 0 0 0 0
0.0000 0.8250 0.0000 Cl 0 0 0 0 0 0 0 0 0 0 0 0
1 2 1 0
3 1 1 0
1 4 1 0
1 5 1 0
M END

InChI=1/Cl2H4N2Pt/c1-5(2,3)4/h3-4H2

311 233

15

521 443

Molfile 2 with related InChI and indices (original atom number,canonical number, equivalence class taken from the Web Demo

ChemDraw11272517312D

5 4 0 0 0 0 0 0 0 0999 V2000
0.0000 0.0000 0.0000 Pt 0 0 0 0 0 0 0 0 0 0 0 0
0.0000 -0.8250 0.0000 N 0 0 0 0 0 0 0 0 0 0 0 0
0.0000 0.8250 0.0000 Cl 0 0 0 0 0 0 0 0 0 0 0 0
-0.8250 0.0000 0.0000 N 0 0 0 0 0 0 0 0 0 0 0 0
0.8250 0.0000 0.0000 Cl 0 0 0 0 0 0 0 0 0 0 0 0
1 2 1 0
3 1 1 0
1 4 1 0
1 5 1 0
M END

InChI=1/Cl2H4N2Pt/c1-5(2,3)4/h3-4H2

521 233

15

311 443

The canonicalized numbers depend on the position of the C atom in the molfile but it should be independent from the order in the molfile. @schatzsc that looks ike an incomplete canonicalization to me as you describe it as well.