Documentation
Current limitations
Honest limits of Spectra — data coverage, identification risk, and what the product will not do.
Read this page before using Spectra in a quality, regulatory, or publication setting.
Data coverage
Spectra projects take 1H and 13C from Bruker, JEOL, Varian or JCAMP-DX, and can include HSQC, COSY, HMBC, TOCSY and NOESY in the same project. Other nuclei and undocumented vendor dialects are outside the current support matrix. A project that contains only 2D spectra still needs a 1H before a structure shortlist is a fair request.
1H alone, without formula, mass, 13C or 2D, can finish as an empty ranking with a recommended next experiment. That empty ranking is a result.
Identification risk
Verification of a proposed structure is a different risk class from open-world identification. Dereplication can miss a new analogue. Impurity identification can fail when the impurity is below the quality of the 13C.
Retrieval misses
Spectra cannot rank a molecule that is outside its search space. Novel structures need elucidation work, not only library retrieval.
ROSE versus Spectra
ROSE research numbers are not Spectra product numbers.
What Spectra does not do
Spectra does not replace chemists, guarantee an identification, run fully autonomous chemistry, or carry a regulatory certification.
Deployment and validation
Private deployment and regulated-laboratory validation are enterprise conversations, not self-service features.