Fritillaria, a dual-purpose medicinal and edible herb of the Liliaceae family, contains bioactive alkaloids as its key therapeutic components. Precise characterization of these alkaloids is critical for quality control and pharmacological mechanism studies. However, conventional analytical approaches face significant challenges due to the inherent complexity of plant metabolite matrices and the ubiquitous presence of structural isomers, resulting in low identification accuracy and difficulties in detecting unknown compounds. First, a novel dimension-enhanced approach was developed by integrating solid-phase extraction (SPE) with liquid chromatography/ion mobility-quadrupole time-of-flight mass spectrometry (LC/IM-QTOF-MS) to acquire four-dimensional structural information (tR, MS, MS/MS, and CCS). Subsequently, collision cross-section (CCS) values were predicted using machine learning, and a multidimensional database for Fritillaria alkaloids, namely FasMID, was established. In the analysis of prediction tools, ALLCCS/CCSbase demonstrated superior accuracy and applicability. Leveraging FasMID, which contained information on 248 Fritillaria alkaloids (including m/z, CCS, and MS/MS data), an integrated multi-dimensional matching strategy was developed for annotating both known and unknown chemical structures. Automated identification of known Fritillaria alkaloids was achieved by processing data with UNIFI and searching against FasMID, elevating the identification accuracy from a baseline of 12 % with traditional methods to 70 % with optimized approach. Furthermore, feature-based molecular networking (FBMN) combined with multidimensional characteristics of Fritillaria alkaloids was utilized to annotate unknown components, facilitating the exploration of novel chemical structures. In future research, prediction accuracy can be improved by expanding sample sources and optimizing deep learning algorithms, while extending this analytical strategy to other complex systems.
Keywords: Collision cross section; Fritillaria alkaloids; Molecular networking; Multidimensional information database.
Copyright © 2024. Published by Elsevier Ltd.