Percepta: Predictions differ unexpectedly between two apparently identical structures

Description
Users enter two structures that appear the same visually (e.g., from different drawing tools or suppliers), but Percepta predictions (logP, solubility, pKa, etc.) differ. Closer inspection often reveals differences in protonation state, tautomers, stereochemistry, or hidden charges.

Solution

1. Examine structural details carefully
   • Check for:
     • Explicit vs implicit hydrogens.
     • Formal charges on atoms.
     • Different tautomeric forms (keto vs enol, amide vs imidic acid).
     • Missing or extra stereo centers.
2. Normalize ionization state where appropriate
   • Decide which canonical form should be used for predictions (e.g., neutral free base, free acid).
   • In ChemSketch or another editor, adjust protonation and charges to that canonical form for both structures.
3. Align tautomeric form
   • If one structure is drawn as a different tautomer:
     • Redraw it so both inputs represent the same tautomeric form.
   • Re‑run Percepta predictions and compare again.
4. Check stereochemistry
   • Ensure R/S and E/Z definitions are the same if relevant to predictions.
   • Remove ambiguous wedges if the intent is to predict for a racemate or unspecified isomer.
5. Re‑run predictions with harmonized input
   • Once you confirm the structures are truly identical at the connectivity, charge, tautomer, and stereo levels, rerun Percepta.
   • The results should now align within expected numeric noise.
6. Record best practices
   • Use a single, canonical structure source for a given compound family when possible.
   • Note in internal SOPs that apparent visual similarity does not guarantee identical cheminformatics representation.