Processing methods for differential analysis of LC/MS profile data

Abstract

Background: Liquid chromatography coupled to mass spectrometry (LC/MS) has been widely used in proteomics and metabolomics research. In this context, the technology has been increasingly used for differential profiling, i.e. broad screening of biomolecular components across multiple samples in order to elucidate the observed phenotypes and discover biomarkers. One of the major challenges in this domain remains development of better solutions for processing of LC/MS data.

Results: We present a software package MZmine that enables differential LC/MS analysis of metabolomics data. This software is a toolbox containing methods for all data processing stages preceding differential analysis: spectral filtering, peak detection, alignment and normalization. Specifically, we developed and implemented a new recursive peak search algorithm and a secondary peak picking method for improving already aligned results, as well as a normalization tool that uses multiple internal standards. Visualization tools enable comparative viewing of data across multiple samples. Peak lists can be exported into other data analysis programs. The toolbox has already been utilized in a wide range of applications. We demonstrate its utility on an example of metabolic profiling of Catharanthus roseus cell cultures.

Conclusion: The software is freely available under the GNU General Public License and it can be obtained from the project web page at: http://mzmine.sourceforge.net/.

Figure 1

MZmine graphical user interface: (A) List of imported raw data files. (B) Total ion chromatogram (TIC) for selected file. (C) Mass spectrum for selected retention time for the same file. (D) Peak list for the same file, with listed m/z values, retention times, and intensities. (E) 2d map of the same file, with retention time on x-axis and m/z on y-axis. (F) Zoomed-in spectra for a different file. (G) Peak alignment matrix for all files listed. (H) Available alignment results, e.g. for different normalizations. The spectra shown in the GUI are from lipidomic profiling of mouse white adipose tissue using Quattro Micro (Waters, Inc.) triple quadrupole mass spectrometer.

Figure 2

The two-step peak picking process used by the two available peak picking methods: (A) This plot is a zoom-in to a small part of a spectrum. In the first step, one-dimensional spectral peaks are detected in each spectrum alone. Green dots over the spectrum show the locations of detected spectral peaks. (B) This plot is a zoom-in to a small fragment of two-dimensional view of raw data. Black lines show two-dimensional peaks created by connecting successive spectral peaks. Peak height is calculated as the highest intensity among these data points, while the peak area corresponds to the sum of the intensities.

Figure 3

(A) Total ion chromatograms from two representative samples from Catharanthus roseus cell cultures. (B) Log-ratio plot view, comparing mean intensities of detected peaks between two selected groups of samples from Catharanthus roseus (10 elicited vs. 10 controls).

Figure 4

Data analysis of Catharanthus roseus metabolite profile data, using the 2175 detected peaks as variables. (A) Principal components analysis shows differences between the elicited and control strains. Subsequent factor analysis revealed the clustering of the elicited group is largely due to the tabersonine and ajmalicine. (B) Comparison of intensity distribution between elicited and control groups for internal standard (Vincamine), Ajmalicine, and Tabersonine.