Constructing a mass measurement error surface to improve automatic annotations in liquid chromatography/mass spectrometry based metabolomics

• Published in: Rapid communications in mass spectrometry 2013-11, Vol.27 (21), p.2425-2431
• Main Authors: Shahaf, Nir, Franceschi, Pietro, Arapitsas, Panagiotis, Rogachev, Ilana, Vrhovsek, Urska, Wehrens, Ron
• Format: Article
• Language: English
• Subjects: accuracy; Annotations; Chromatography, Liquid - methods; Construction; Error analysis; Errors; Mass spectrometry; Mass Spectrometry - methods; Mathematical models; Metabolomics - methods; Models, Statistical; molecules; precision; recalibration; Reproducibility of Results; sample; Searching
• ISSN: 0951-4198; 1097-0231
• DOI: 10.1002/rcm.6705

RATIONALE Estimation of mass measurement accuracy is an elementary step in the application of mass spectroscopy (MS) data towards metabolite annotations and has been addressed several times in the past. However, the reproducibility of mass measurements over a diverse set of analytes and in variable operating conditions, which are common in high‐throughput metabolomics studies, has, to the best of our knowledge, not been addressed so far. METHODS A method to automatically extract mass measurement errors from a large data set of measurements made on a quadrupole time‐of‐flight (QTOF) MS instrument has been developed. The size of the data processed in this study has enabled us to use a statistical data driven approach to build a model which reliably predicts the confidence interval of the absolute mass measurement error based on individual ion peak conditions in a fast, high‐throughput manner. RESULTS We show that our model predictions are reproducible in external datasets generated in similar, but not identical conditions, and have demonstrated the advantage of our approach over the common practice of fixed mass measurement error limits. CONCLUSIONS Outlined is an approach which can promote a more rational use of MS technology by automatically evaluating the absolute mass measurement error based on the individual peak conditions. The immediate application of our method is integration in high‐throughput peak annotation pipelines for database searches. Copyright © 2013 John Wiley & Sons, Ltd.