Anal Bioanal Chem DOI 10.1007/s00216-013-6856-7
ORIGINAL PAPER
Characterising and correcting batch variation in an automated direct infusion mass spectrometry (DIMS) metabolomics workflow J. A. Kirwan & D. I. Broadhurst & R. L. Davidson & M. R. Viant
Received: 2 November 2012 / Revised: 19 February 2013 / Accepted: 19 February 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Direct infusion mass spectrometry (DIMS)-based untargeted metabolomics measures many hundreds of metabolites in a single experiment. While every effort is made to reduce within-experiment analytical variation in untargeted metabolomics, unavoidable sources of measurement error are introduced. This is particularly true for large-scale multi-batch experiments, necessitating the development of robust workflows that minimise batch-to-batch variation. Here, we conducted a purpose-designed, eightbatch DIMS metabolomics study using nanoelectrospray (nESI) Fourier transform ion cyclotron resonance mass spectrometric analyses of mammalian heart extracts. First, we characterised the intrinsic analytical variation of this approach to determine whether our existing workflows are fit for purpose when applied to a multibatch investigation. Batch-to-batch variation was readily Published in the topical collection Metabolomics and Metabolite Profiling with guest editors Rainer Schuhmacher, Rudolf Krska, Roy Goodacre, and Wolfram Weckwerth. J. A. Kirwan and D. I. Broadhurst contributed equally to this work. Electronic supplementary material The online version of this article (doi:10.1007/s00216-013-6856-7) contains supplementary material, which is available to authorized users. J. A. Kirwan : M. R. Viant (*) School of Biosciences, University of Birmingham, Edgbaston, Birmingham B15 2TT, UK e-mail:
[email protected] D. I. Broadhurst Department of Medicine, University of Alberta, Edmonton, AB, Canada T6G 2EI R. L. Davidson : M. R. Viant NERC Biomolecular Analysis Facility—Metabolomics Node (NBAF-B), University of Birmingham, Edgbaston, Birmingham B15 2TT, UK
observed across the 7-day experiment, both in terms of its absolute measurement using quality control (QC) and biological replicate samples, as well as its adverse impact on our ability to discover significant metabolic information within the data. Subsequently, we developed and implemented a computational workflow that includes total-ion-current filtering, QC-robust spline batch correction and spectral cleaning, and provide conclusive evidence that this workflow reduces analytical variation and increases the proportion of significant peaks. We report an overall analytical precision of 15.9 %, measured as the median relative standard deviation (RSD) for the technical replicates of the biological samples, across eight batches and 7 days of measurements. When compared against the FDA guidelines for biomarker studies, which specify an RSD of
