Analysis of untargeted MS-based metabolomics data: the metaMS package for R

Introduction Untargeted metabolomics data, often recorded using mass spectrometric techniques, provide a wealth of data on the presence and abundance of metabolites in biological samples. The extraction of relevant information can be difficult, and many software platforms have been proposed. One of the most popular tools for analysing LCMS and GCMS data is XCMS, written in the R language. An add-on to XCMS, developed specifically in the context of untargeted metabolomics, is metaMS, providing facilities for building in-house databases of chemical standards geared towards specific organisms or groups of metabolites, automatic annotation, and quantification. MetaMS, like XCMS, is publicly available from the Bioconductor repository. Methods For LCMS, the main part of the metaMS pipeline is similar to the XCMS pipeline, consisting of peak picking, grouping, and alignment. The additions from metaMS focus on improved annotation using in-house databases, an m/z and intensity-dependent mass accuracy window and an explicit definition of minimal support for annotation. The outcome is a matrix summarizing for all samples the intensities of the aligned peaks. The GCMS pipeline is different, working on so-called pseudospectra rather than individual peaks; here, the output is a relative intensity measure for chemical compounds rather than individual peaks (Wehrens et al., 2014). The compounds may be annotated (when there is a match with the database), or labelled as Unknowns. Results A web-based pipeline has been built using the metaMS package, which now is in daily use by the metabolomics platform at FEM. Processing hundreds of samples takes only a couple of hours on a regular four-core linux desktop computer. The generated tables can be immediately be used for subsequent statistical analysis. A common application is Quality Control: a score plot from a Principal Component Analysis can be inspected to see whether the quality control samples, often repeated injections of a pooled sample cluster together, are in the middle of the samples, and do not show a trend with injection order. Conclusions Open-source software like metaMS provides ultimate control over data processing, which is of utmost importance when analysing data as complex as GCMS or LCMS data. In addition, the large user base of the underlying XCMS package guarantees rapid adaptation to new developments, timely bug reporting and on-line user feedback. MetaMS provides a top layer over XCMS, specifically geared to untargeted metabolomics. Novel Aspect The novel aspects of metaMS are found at several levels: at the most abstract level the functionality of XCMS is extended and geared towards untargeted metabolomics data. At more detailed levels this includes tools for setting up in-house databases, doing annotation in a principled way, and a completely new approach to handle GCMS metabolomics data. R. Wehrens, G. Weingart and F. Mattivi: J. Chrom. B DOI: 10.1016/j.jchromb.2014.02.05