# Asparagine: SMILES versus InChI input
Checked 2026-10-09 with RDKit 2026.03.6.

The reported discrepancy is reproducible. The original 20-pair comparison parsed PubChem's isomeric `SMILES` field using `Chem.MolFromSmiles`. It did not parse InChI, canonicalize tautomers, or perform a pH-dependent standardization. The original asparagine result is correct for that input graph.

Source records: [L-asparagine, CID 6267](https://pubchem.ncbi.nlm.nih.gov/compound/6267) and [D-asparagine, CID 439600](https://pubchem.ncbi.nlm.nih.gov/compound/439600), reused from the saved PubChem snapshot.

## Observed structures
RDKit canonical isomeric SMILES after parsing:

| Input | L-asparagine | D-asparagine |
|---|---|---|
| PubChem SMILES | `NC(=O)C[C@H](N)C(=O)O` | `NC(=O)C[C@@H](N)C(=O)O` |
| PubChem Standard InChI | `N=C(O)C[C@H](N)C(=O)O` | `N=C(O)C[C@@H](N)C(=O)O` |

The side-chain amide, –C(=O)NH2, becomes an imidic-acid tautomer, –C(OH)=NH, on InChI reconstruction. Standard InChI normalizes recognized mobile-hydrogen tautomerism: its `(H2,6,7)` layer here does not preserve the original amide hydrogen/bond assignment as a fixed-H representation would. Parsing it reconstructs a representative graph; it need not recover the source SMILES tautomer. This does not imply that the reconstructed tautomer predominates in solution.

Both routes retain L = S and D = R at the alpha carbon. Inverting that tetrahedral center reproduces the opposite enantiomer within each route. Both forms regenerate the corresponding source Standard InChI. Formula remains C4H8N2O3, formal charge zero, and molecular weight 132.12 at PubChem precision. RDKit values differ only by floating-point summation noise (~3e-14), not chemical mass.

## Fingerprint comparison
L-versus-D Tanimoto similarity, not SMILES-versus-InChI similarity:

| Fingerprint | SMILES input | InChI input |
|---|---:|---:|
| Morgan, chirality off | 1.000000 | 1.000000 |
| Morgan, chirality on | 0.619048 | 0.636364 |
| Atom-pair, chirality off | 1.000000 | 1.000000 |
| Atom-pair, chirality on | 0.657895 | 0.650000 |

Settings match the original calculation: 2,048-bit fingerprints; Morgan radius 2; atom-pair count simulation disabled; otherwise RDKit generator defaults. Tanimoto is shared set bits divided by the union of set bits. Changes in hydrogen placement and bond order alter encoded atom environments and bit overlap, even though the formula and enantiomer relationship are unchanged. Exact scores depend on representation, fingerprint settings, and software version.

For a different question—L-asparagine from SMILES versus L-asparagine from InChI—the Morgan similarity is 0.590909 with chirality either off or on. Atom-pair similarities are 0.488889 (off) and 0.477273 (on). Thus disabling chirality does not make these fingerprints tautomer-invariant.

## Standardization check and recommendation
Default `rdMolStandardize.TautomerEnumerator().Canonicalize` removed the alpha-carbon stereo annotation in this test, yielding `NC(=O)CC(N)C(=O)O` for both enantiomers. This would invalidate a chirality comparison.

For this pair, explicitly setting `SetRemoveSp3Stereo(False)` before canonicalization made both input routes converge to the amide structures while retaining the verified S/R assignments. This is a checked result for asparagine, not a universal guarantee for all tautomeric systems. Any standardization workflow must revalidate stereochemistry.

For reproducibility, retain the original SMILES-based results and specify the input field. If comparing mixed input formats, use a documented common tautomer/protonation policy and validate stereocenters after processing. Canonical SMILES serialization alone is not tautomer canonicalization. Neither score establishes equal biological activity.

## Plot caption
`asparagine_input_comparison.png`: One deterministic L/D-asparagine pair, evaluated using two input routes with fixed fingerprint settings. Circles show SMILES input and squares show InChI input. Each point is one calculated similarity; no replicates or uncertainty estimates are implied. Chirality-off scores remain 1 within each route; chirality-on scores change slightly with the reconstructed tautomer.
