22 Abbreviations 23 MISSILE, metabolome identification by systematic stable isotope labeling experiments; LC-24 MS/MS, liquid chromatography-tandem mass spectrometry 25 2 ABSTRACT 26We introduce a formula-based strategy and algorithm (JUMPm) for global metabolite identification 27 and false discovery analysis in untargeted mass spectrometry-based metabolomics. JUMPm 28 determines the chemical formulas of metabolites from unlabeled and stable-isotope labeled 29 metabolome data, and derives the most likely metabolite identity by searching structure 30 databases. JUMPm also estimates the false discovery rate (FDR) with a target-decoy strategy 31 based on the octet rule of chemistry. With systematic stable isotope labeling of yeast, we identified 32 2,085 chemical formulas (10% FDR), 892 of which were assigned with metabolite structures. We 33 evaluated JUMPm with a library of synthetic standards, and found that 96% of the formulas were 34 correctly identified. We extended the method to mammalian cells with direct isotope labeling and 35 by heavy yeast spike-in. This strategy and algorithm provide a powerful a practical solution for 36 global identification of metabolites with a critical measure of confidence. 37 38 39 42 therefore considered to be direct readouts of biological activity. Many metabolites also function 43 as building blocks, signaling factors, and molecular precursors which modify and regulate cellular 44 components such as DNA, RNA, and protein. The human metabolome 1 contains conventional 45 cellular metabolites along with other chemicals derived from food, microbiota, and the 46 environment. The role of the metabolome has been increasingly appreciated in both development 47 and disease 2 . However, it is still a challenge to profile the complete metabolome due to the highly 48 diverse chemical properties of small molecules and practical limitations of analytical strategies. 49 Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is a prevalent method for 50 global metabolome profiling 3 . Combining nanoscale LC with high-resolution MS leads to the 51 detection of thousands of high-confidence metabolite features in a complex sample 4 . Numerous 52 software programs have been developed for processing large-scale datasets 5-14 . Most of these 53 programs share a common workflow, including feature detection, peak alignment, and relative 54 quantification with semi-automated identification and/or laborious manual validation of selected 55 peak features. Structural annotation of the selected features is typically achieved by searching 56 against empirical MS/MS spectral libraries such as METLIN 15,16 , HMDB 1,17 , or NIST 18 . 57Despite considerable progress in the development of software programs, identification of 58 metabolites from untargeted studies remains a daunting task. One major limitation is that spectral 59 libraries must be generated with synthetic standards. For instance, the NIST14 MS/MS database 60 contains ~14,000 empirical MS/MS spectra, making it a precious but costly resource. To...

No abstract