knowledge-based hybrid molecular network (KHMN)

Here, we proposed a general method of a knowledge-based hybrid molecular network (KHMN) that was coconstructed by untargeted metabolomics MS/MS spectra with heterogeneous MS/MS spectra of interesting specialized metabolites as initial seeds. Using antiherbivore metabolites in maize as an example, a KHMN was generated for the rapid screening and efficient annotation of as many specialized metabolites as possible. First, a priori specialized metabolites were collected as seed metabolites. The liquid chromatography (LC)-MS/MS database of the seed metabolites was established. Knowledge (characteristic fragmentation patterns and biotransformation reaction) was fully extracted from the MS/MS spectra of seed metabolites or priori knowledge. Then, a KHMN was coconstructed using heterogeneous and homologous MS/MS spectra of the seed metabolites and untargeted ultrahigh-performance LC-HRMS (UPLC-HRMS) metabolomics data. At the same time, captured knowledge was integrated into the molecular network. Finally, both known and unknown specialized metabolites could be effectively extracted and automatically propagated to annotate by KHMN. Since we used known targeted metabolites as initial cues to trigger molecular networks, highly efficient capture of the molecular families of targeted metabolites could be achieved. Simultaneously, the integration of captured knowledge into the network increased the efficiency and accuracy of annotation propagation and benefitted the annotation of new metabolites.