This repository contains the code, data and presentation of my talk at the
SWEMSA conference 2019 in Erding, Germany. The talk is
focused on LC-MS/MS data analysis with xcms and mentions also recent
developments in the RforMassSpectrometry
The online version of the talk can be accessed here.
To cite the talk: 10.5281/zenodo.3499649.
Bioconductor is an open-software, open-development software project for the analysis of high throughput data in genomics and molecular biology [1] in the context of the rich statistical programming environment offered by the R project. Bioconductor enables the rapid creation of workflows combining multiple data types and analysis methods provided by any of its over 1,500 software packages. The xcms package [2] is a well established Bioconductor package and one of the standard toolboxes for the preprocessing of untargeted metabolomics data. The MSnbase package [3], primarily designed for proteomics data analysis, provides basic infrastructure for import and handling of mass spectrometry (MS) data.
We recently improved the MS data handling capabilities of MSnbase by introducing an on-disk data mode that, in contrast to keeping the full MS data in memory, holds only a subset of the data in memory, importing MS peak data (m/z and intensity value pairs) from the original MS data files on demand. The resulting reduction in memory demand enables the analysis also of very large experiments, even on standard hardware. High performance is achieved by making extensive use of the random-access capabilities of indexed mzML, mzXML and CDF files and by parallelizing all operations on a per-file basis. Because peak data are now no longer residing in memory, they can also not be directly manipulated anymore. We thus implemented an approach employing a lazy evaluation strategy, that adds manipulation operations (such as centroiding or base line correction) to a processing queue and applies them to the peak data upon request (i.e. each time the user accesses MS peak data).
In addition to above described improvements in MSnbase, we rewrote big parts of xcms to reuse objects and functionality from the MSnbase and thus better integrate xcms into the Bioconductor framework. xcms gained native support for MS level > 1 data handling and inherits all data processing, sub-setting and filtering methodology from MSnbase. As a consequence, access to raw data and extraction of ion chromatograms (EICs) was simplified and is now possible at any stage during data preprocessing. Because settings for preprocessing algorithms, such as centWave-based chromatographic peak detection, are highly data set specific, we implemented functions to to aid in evaluation of parameters for the various methods. It is for example now possible to perform peak detection directly on EICs (Figure 1) which enables the fine-tuning of peak detection settings on selected signals.
The most recent improvements in xcms comprise tools to work with LC-MS/MS experiments. For experiments employing data dependent acquisition (DDA), MS2 spectra can be identified and extracted for each detected chromatographic peak or feature thus facilitating an improved compound annotation. For data independent acquisition (DIA) data such as SWATH data, xcms allows to perform chromatographic peak detection on MS2 data and reconstruct MS2 spectra for MS1 chromatographic peaks by matching MS2 chromatographic peaks to MS1 chromatographic peaks based on the isolation window and peak shape correlation.
While xcms’ lack of a graphical user interface might seem disadvantageous at first, its command line usage enables the definition of analysis scripts which, as a consequence, ensure reproducibility of analyses, even more so in combination with rmarkdown which allows to create reproducible analysis workflows and, more importantly, human readable analysis reports.
Over and above, we improved the MS data handling infrastructure in R/Bioconductor enabling also the analysis of large experiments and, by re-using the capabilities of MSnbase in xcms, allowing also an efficient preprocessing of very large metabolomics data sets. The recent changes simplified metabolomics data handling in R and added support for the analysis of LC-MS/MS data.
Acknowledgements
The author thanks Laurent Gatto, Michael Witting, Jan Stanstrup and Steffen Neumann for their contributions, suggestions and comments.
- Huber W, Carey VJ, Gentleman R, Anders S, Carlson M, Carvalho BS, et al. Orchestrating high-throughput genomic analysis with Bioconductor. Nat Meth. 2015;12:115–21.
- Smith CA, Want EJ, O'Maille G, Abagyan R, Siuzdak G. XCMS: processing mass spectrometry data for metabolite profiling using nonlinear peak alignment, matching, and identification. Anal. Chem. 2006;78:779–87.
- Gatto L, Lilley KS. MSnbase-an R/Bioconductor package for isobaric tagged mass spectrometry data visualization, processing and quantitation. Bioinformatics. 2012;28:288–9.
Author Dr. Johannes Rainer Institute for Biomedicine, Eurac Research Via Luigi Galvani 31, I-39100 Bolzano Italy