MS/MS spectra can provide sub-class level information such as sugars or fatty acids or carnitine. This function takes that sub-class information for partially annotated compounds using a MS/MS spectra interpretation. You can train a machine learning model using the publicly available mass spectral data to obtain sub-class information, and there are great online tools out there that can predict sub-classes/sub-structures for your MS/MS spectra.
We are taking the NIST Hybrid Search as an example approach to compute classes for MS/MS spectra as shown in this paper by Ivana (LINK) . The Hybrid Search works great for some chemical classes such as acyl-carnitines, but may give trouble for other classes such as polyphenols because of the diversity of chemical compounds in later categories. Nonetheless, it is worth exploring NIST Hybrid Search provided Chemical Class as set-definitions for unknown metabolites in non-targeted metabolomics.
More is coming!