Atefeh Rafiei1, Lekha Sleno. 1. Université du Québec à Montréal, Chemistry Department/Pharmaqam, Montreal, QC, Canada.
Abstract
RATIONALE: Data analysis is a key step in mass spectrometry based untargeted metabolomics, starting with the generation of generic peak lists from raw liquid chromatography/mass spectrometry (LC/MS) data. Due to the use of various algorithms by different workflows, the results of different peak-picking strategies often differ widely. METHODS: Raw LC/HRMS data from two types of biological samples (bile and urine), as well as a standard mixture of 84 metabolites, were processed with four peak-picking softwares: Peakview®, Markerview™, MetabolitePilot™ and XCMS Online. The overlaps between the results of each peak-generating method were then investigated. To gauge the relevance of peak lists, a database search using the METLIN online database was performed to determine which features had accurate masses matching known metabolites as well as a secondary filtering based on MS/MS spectral matching. RESULTS: In this study, only a small proportion of all peaks (less than 10%) were common to all four software programs. Comparison of database searching results showed peaks found uniquely by one workflow have less chance of being found in the METLIN metabolomics database and are even less likely to be confirmed by MS/MS. CONCLUSIONS: It was shown that the performance of peak-generating workflows has a direct impact on untargeted metabolomics results. As it was demonstrated that the peaks found in more than one peak detection workflow have higher potential to be identified by accurate mass as well as MS/MS spectrum matching, it is suggested to use the overlap of different peak-picking workflows as preliminary peak lists for more rugged statistical analysis in global metabolomics investigations.
RATIONALE: Data analysis is a key step in mass spectrometry based untargeted metabolomics, starting with the generation of generic peak lists from raw liquid chromatography/mass spectrometry (LC/MS) data. Due to the use of various algorithms by different workflows, the results of different peak-picking strategies often differ widely. METHODS: Raw LC/HRMS data from two types of biological samples (bile and urine), as well as a standard mixture of 84 metabolites, were processed with four peak-picking softwares: Peakview®, Markerview™, MetabolitePilot™ and XCMS Online. The overlaps between the results of each peak-generating method were then investigated. To gauge the relevance of peak lists, a database search using the METLIN online database was performed to determine which features had accurate masses matching known metabolites as well as a secondary filtering based on MS/MS spectral matching. RESULTS: In this study, only a small proportion of all peaks (less than 10%) were common to all four software programs. Comparison of database searching results showed peaks found uniquely by one workflow have less chance of being found in the METLIN metabolomics database and are even less likely to be confirmed by MS/MS. CONCLUSIONS: It was shown that the performance of peak-generating workflows has a direct impact on untargeted metabolomics results. As it was demonstrated that the peaks found in more than one peak detection workflow have higher potential to be identified by accurate mass as well as MS/MS spectrum matching, it is suggested to use the overlap of different peak-picking workflows as preliminary peak lists for more rugged statistical analysis in global metabolomics investigations.
Authors: Sakda Khoomrung; Intawat Nookaew; Partho Sen; Thorunn A Olafsdottir; Josefine Persson; Thomas Moritz; Peter Andersen; Ali M Harandi; Jens Nielsen Journal: J Proteome Res Date: 2019-11-01 Impact factor: 4.466
Authors: Padma Maruvada; Johanna W Lampe; David S Wishart; Dinesh Barupal; Deirdra N Chester; Dylan Dodd; Yannick Djoumbou-Feunang; Pieter C Dorrestein; Lars O Dragsted; John Draper; Linda C Duffy; Johanna T Dwyer; Nancy J Emenaker; Oliver Fiehn; Robert E Gerszten; Frank B Hu; Robert W Karp; David M Klurfeld; Maren R Laughlin; A Roger Little; Christopher J Lynch; Steven C Moore; Holly L Nicastro; Diane M O'Brien; José M Ordovás; Stavroula K Osganian; Mary Playdon; Ross Prentice; Daniel Raftery; Nichole Reisdorph; Helen M Roche; Sharon A Ross; Shengmin Sang; Augustin Scalbert; Pothur R Srinivas; Steven H Zeisel Journal: Adv Nutr Date: 2020-03-01 Impact factor: 11.567
Authors: Kelly H Telu; Xinjian Yan; William E Wallace; Stephen E Stein; Yamil Simón-Manso Journal: Rapid Commun Mass Spectrom Date: 2016-03-15 Impact factor: 2.419