NMR predictions are not measured spectra
Use predicted proton and carbon shifts as working hypotheses. Preserve atom mapping and understand the model’s scope.
What prediction can help with
A shift model can help compare candidate structures and discuss possible assignments before or alongside measurements. It generates values from learned relationships and molecular input. It has not physically measured your sample. Label the output as a prediction even when the resulting table looks like experimental data.
Mapping is part of the result
The input structure, geometry and displayed atoms must remain aligned. A valid numerical prediction becomes misleading if assigned to the wrong atom. Explicit and implicit hydrogens can affect that mapping.
Cheemly NMR supplies model-predicted proton and carbon chemical shifts. Unsupported structures or an unhealthy worker should produce an honest limitation, not invented ppm values.
Avoid transferring a benchmark to your sample
A published model error belongs to its evaluation dataset and protocol. It is neither your molecule’s known error nor a separately measured Cheemly accuracy figure. Compare the same atoms and nuclei, with attention to conditions and chemical environments.
The current output does not establish calibrated uncertainty, splitting, J couplings or experimental line shapes. Use it to guide further evaluation, not to replace a laboratory record.
