Metabolomics aims to perform unbiased analyses of all the metabolites within a biological system and reveals the relationship between metabolites and phenotype. Nontargeted metabolomics based on ultrahigh-performance liquid chromatography-high resolution tandem mass spectrometry (UPLC-HR MS/MS) is one of main tools and has been widely used in metabolomics study. Despite thousands of mass spectrometry features can be acquired with UPLC-HR MS/MS, most of these features are unannotated. The identification of metabolites in MS-based untargeted metabolomics data becomes one of the major bottlenecks. Novel deep annotation methods are urgently needed. In this study, a novel comprehensive deep annotation strategy of untargeted UPLC-HRMS/MS metabolomics data was proposed. Firstly, data acquisition and deconvolution algorithms were developed to generate the information-rich HRMS/MS data. Next, quantitative structure-retention relationship models based on hybrid machine learning algorithm were constructed to predict LC retention times by using commercial available/synthesized standard mixtures and our LC-MS database constructed based on more than 2,000 standards. For the annotation of unknown metabolites which are not included in the database, a computational-based (in silico) simulation through prior pathways was used to generate a predicted reference metabolite, and in silico MS fragmentations pattern or substructure motif was suggested. Then, multi-type molecular networking was constructed for comprehensive metabolite identification. The results show that developed strategy is very useful for both large-scale annotation of metabolome and discovery of cryptic metabolites from untargeted metabolomics data.

Deep annotation strategy of untargeted liquid chromatography-mass spectrometry metabolomics data

ACS Spring Meeting 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.