An algorithm-driven intelligent mining and identification strategy for natural product mass spectrometry

Abstract

Efficiently mining and identification of new compounds from the extensive MS/MS datasets of plant extracts poses a significant challenge due to the structural diversity and compositional complexity inherent in natural products (NPs). Various data post-processing techniques have been developed to simplify the interpretation of MS/MS data; however, they often suffer from limited specificity and precision. Meanwhile, structure annotation following data post-processing is particularly time-consuming. In this study, we introduced an innovative strategy named MS-SMART, which integrates three intelligent algorithms: automatic mining of diagnostic ions, rapid filtration of alkaloids from untargeted MS/MS data, and structural recommendations for filtered components. The feasibility of this approach for rapidly discovering novel compounds was demonstrated using berberine-type alkaloids as an example. Firstly, diagnostic ions were automatically extracted and validated using available reference data. Subsequently, berberine-type compounds were filtered from raw MS/MS data. Finally, the structures of the target components were recommended using building blocks derived from berberines reported in various plants. A total of 103, 198, 60, 80 and 51 berberines were efficiently identified in diverse families and genera, including Stephaniae Epigaeae Radix, Coptidis Rhizoma, Phellodendri Chinensis Cortex, Phellodendri Amurensis Cortex and Corydalis Decumbentis Rhizoma, with 99, 169, 50, 64 and 40 new compounds identified, respectively. Among these, 8, 14, 8, 7 and 12 berberines were confirmed by reference compounds. This strategy provides a new research paradigm for the rapid discovery and identification of different types of new compounds in complex samples.