Prediction Error for handling SMILES for cyclobutane stereoisomers

[lkulloa]

When inputting this SMILES string: [NH-][C@H]1CC@H=O)C1 The webtool returns this error: Prediction ERROR! Please submit an issue on GitHub with ID: ffbade4b0a82ef44195598db220bf988

The webtool works for the neutral molecule with stereochemistry specified: N[C@H]1CC@H=O)C1 Also works for the anionic molecule with stereochemistry specified with the other nitrogen deprotonated: N[C@H]1CC@H=O)C1 It also works when I don't specify stereochemistry and deprotonate the desired nitrogen. However, when I click to see the 3D structure, it had arbitrarily run the calculation with the cis diastereomer (I want the trans): [NH-]C1CC(NC(OC(C)(C)C)=O)C1

Any help you can provide for fixing this error would be much appreciated! I really like the web tool, thank you!

[NicolaiRee]

There was a convergence issue in the underlying single-point QM calculation for obtaining the CM5 charges. I ran an xTB optimization of the structure, so now you can see the results on: https://esnuel.org/MLresults/ffbade4b0a82ef44195598db220bf988

However, the ML version is not yet optimized for molecules with a formal charge. I would instead recommend you to use the xTB version by selecting "xTB" instead of "ML":