MMSAT: Automated Quantification of Metabolites in Selected Reaction

Nov 23, 2011 - Without analytical standards, such metabolites will go undetected by conventional data analysis methods. Furthermore, a single SRM meth...
0 downloads 7 Views 582KB Size
Technical Note pubs.acs.org/ac

MMSAT: Automated Quantification of Metabolites in Selected Reaction Monitoring Experiments Jason W. H. Wong,* Hazem J. Abuhusain, Kerrie L. McDonald, and Anthony S. Don Lowy Cancer Research Centre, Prince of Wales Clinical School, Faculty of Medicine, University of New South Wales, Sydney, NSW, 2052, Australia S Supporting Information *

ABSTRACT: Selected reaction monitoring (SRM) is a mass spectrometrybased approach commonly used to increase analytical sensitivity and selectively for specific compounds in complex metabolomic samples. While the goal of welldesigned SRM methods is to monitor for unique precursor−product ion pairs, in practice this is not always possible due to the diversity of the metabome and the resolution limits of mass spectrometers that are capable of SRM. Isobaric or nearisobaric precursor ions with different chromatographic properties but identical product ions often arise in complex samples. Without analytical standards, such metabolites will go undetected by conventional data analysis methods. Furthermore, a single SRM method may include simultaneous monitoring of tens to hundreds of different metabolites across multiple samples making quantification of all detected ions a challenging task. To facilitate the analysis of SRM data from complex metabolomic samples, we have developed the Metabolite Mass Spectrometry Analysis Tool (MMSAT). MMSAT is a web-based tool that objectively quantifies every metabolite peak detected in a set of samples and aligns peaks across multiple samples to enable quantitative comparison of each metabolite between samples. The analysis incorporates quantification of multiple peaks/ions that have different chromatographic retention times but are detected within a single SRM transition. We compare the performance of MMSAT against existing tools using a human glioblastoma tissue extract and illustrate its ability to automatically quantify multiple precursors within each of three different transitions. The Webinterface and source code is avaliable at http://www.cancerresearch.unsw.edu.au/crcweb.nsf/page/MMSAT.

L

monitored within each LC−MS/MS run for tens to hundreds of samples, data analysis becomes challenging. To our knowledge, in most laboratories, SRM-based metabolomic data sets are analyzed using proprietary vendor specific software (such as XCalibur, Thermo Fisher Scientific and MassHunter, Agilent). These proprietary software tools generally require users to manually define each metabolite being monitored in the SRM method, followed by validation of the retention time for each transition based on available standards. When there are hundreds of transitions, this can be a laborious task. In particular, without appropriate standards, in transitions where metabolites with identical or near-identical precursor and product ions are present, it can be difficult to assign the appropriate peak to each respective metabolite. It can be a greater problem when a user is unaware of the presence of isobaric metabolites that are only present in a subset of samples. A number of nonproprietary, vendor independent tools are also available for the analysis of MS-based metabolomics data. Most commonly used open-source tools include XCMS2,2 MZmine2,3 and msinspect.4 These tools were designed for the analysis of untargeted metabolomics (XCMS2 and MZmine2) or proteomics data (msinspect) and are only able to analyze full-scan

iquid/gas chromatography coupled with mass spectrometry (LC/GC−MS) is perhaps the most widely used highthroughput method for the study of metabolomes of a given organism. Despite the increasing sensitivity of modern mass spectrometers, the complexity of metabolomes often necessitates targeted metabolomic analysis to detect specific metabolites or metabolite classes. While a number of experimental approaches can be applied for targeted analyses, such as chemical derivatization of specific metabolite functional groups, the use of triple quadrupole (QQQ) mass analyzers has been widely implemented to provide highly sensitive targeted profiling of metabolomes.1 QQQ mass analysers are able to operate in various selective ion-scanning modes such as precursor ion scanning and selected reaction monitoring (SRM). In SRM mode, a precursor with specific mass giving rise to a single specific product ion(s) is screened and quantified. A major goal in targetted MS-based metabolomic experiments is to design SRM methods with transitions that are able to distinguish and isolate metabolites based on unique precursor and product ions. However, because of the diversity of the metabolome, this is sometimes difficult or even impossible for metabolites that have identical or nearidentical precursor and product masses. For example, lipids that differ only with respect to the position of a double bond are common in complex biological mixtures. LC or GC separation prior to MS detection is crucial in differentiating these metabolites. Furthermore, when tens to hundreds of transitions are © 2011 American Chemical Society

Received: October 7, 2011 Accepted: November 23, 2011 Published: November 23, 2011 470

dx.doi.org/10.1021/ac2026578 | Anal. Chem. 2012, 84, 470−474

Analytical Chemistry

Technical Note

of the product ion at each scan such that

LC−MS data sets without further software customization. For the analysis of SRM data, Skyline5 and MRMer6 are two recent proteomics-centric tools developed. Skyline is designed as a package to aid the development and refinement of SRM methods for proteomics-based data and is unsuitable for metabolomic data. MRMer is able to process metabolomicsbased SRM data, however, in its native form the tool is userinteractive and only able to process one sample at a time. Furthermore, MRMer is unable to automatically quantify multiple isomers arising from a single transition. In this work we present a new software platform, Metabolite Mass Spectrometry Analysis Tool (MMSAT), for automated quantification of metabolites from SRM experiments. We designed the software such that it can (1) be used independent of any MS instrument and is thus compatible with mzXML converted data from major mass spectrometer vendors (including Thermo Scientific, Waters, Agilent, and AB Sciex), (2) automatically detect and quantify all metabolites present across all SRM transitions in an unbiased manner, such that no prior knowledge of metabolites are required, (3) output quantitative data in tab delimited format to facilitate downstream statistical analysis and visualization using packages such as Excel or R, and (4) provide a web accessible interface to maximize accessibilty of the software to the metabolomics community. To illustrate the utility of the sofware, we use MMSAT to analyze and quantify lipids from a glioblastoma lipid extract detected in SRM mode. To validate the output, the chromatography peak areas are compared to those obtained with XCalibur and MRMer. Finally, using three different examples, we demonstrate the advantage of MMSAT in enabling the unbiased detection and quantification of metabolites.

N

i(t ) =

∑ an n=1

where i is the sum of the ion intensity, a, of N detected product ions monitored at retention time t. Once the XIC has been constructed, peak detection is performed for each XIC. The peak detection algorithm employs a combination of smoothing and optional denoising using the Savitzky−Golay smoothing8 and wavelet shrinkage denoising9 algorithms, respectively, to aid in the location of all significant local maxima within each transition. Using the smoothed signal, i′(t), the algorithm proceeds to find all local maxima above a user-defined intensity threshold such that for each peak (local maximum), t*

i′(t *) > i′(t )

t ∈ t* ± 2

(1)

Once a peak has been found, the algorithm locates the base of each peak where

t start* = t

where

t < t * , i′(t − 1) > i′(t )

t end* = t

where

t > t * , i′(t + 1) > i′(t )

(2)

The area under the curve (AUC) for each peak is finally estimated using the trapezoidal rule such that



EXPERIMENTAL SECTION Algorithm Description. The outline of the MMSAT algorithm is summarized in Figure 1. MMSAT accepts MS

AUCt = *

∫t



1 2

t end*

i(t ) dt

start*

t end*

∑ t = t start*

T (i(t + 1) + i(t )) (3)

where T is the retention time difference between each two scans of a SRM transition. Once the AUC of all peaks across all transitions for all samples have been calculated, the peaks are aligned and compiled. This is performed by iterative pairwise comparison of peaks between a masterlist of peaks where the list is first initialized using one of the samples. To determine whether two peaks are derived from the same metabolite, the peak must come from the same transition and the retention time difference between the two peaks must be less than a userdefined tolerance. Once all samples have been aligned, the reported retention time of each peak is averaged over all samples. Algorithm Implementation. MMSAT is designed as a C++ application with a python wrapper to provide a web-interface (Apache/mod_python) as well as to compile quantitative data across multiple samples. Both the web-interface as well as the source code can be accessed at http://www.cancerresearch. unsw.edu.au/crcweb.nsf/page/MMSAT. A number of parameters are available to the user to allow a degree of control over the algorithm. This includes the product ion mass tolerance, peak detection intensity cutoff, and optional denoising of the XIC for peak detection. Denoising of the XIC is particularly useful for low-abundance mass spectral data. Once the data set has been analyzed, MMSAT outputs the summarized results in a tab-delimited file other with other that can be downloaded and further analyzed or visualized using software such as Excel or R. Preparation of Human Brain Tissue Extracts. Six freshfrozen glioblastoma (GBM) tissue samples were obtained from

Figure 1. Schematic diagram of the MMSAT algorithm.

data in mzXML format parsed using a customized parsing library from the Global Proteome Machine.7 Multiple mzXML files may be submitted as a zipped file. Once a mzXML file has been parsed, the algorithm proceeds to determine the transitions present in the SRM experiment based on the scans present in the file. For each SRM transition, an extracted ion chromatogram (XIC) is constructed based on the intensity 471

dx.doi.org/10.1021/ac2026578 | Anal. Chem. 2012, 84, 470−474

Analytical Chemistry

Technical Note

Table 1. Quantitative Comparison of Selected Metabolites Quantified from a SRM Experiment Using MMSAT, XCalibur, and MRMera metabolite Sphingosine dhSphingosine Sphingosine-1-phosphate C16 Cer C18:1 Cer C18 Cer/C20:4 Cer/C18:1 dhCer C20 Cer C22 Cer C24:1 Cer C24 Cer C24:1 dhCer C16 HexCer/C22:1 C1P C18:1 HexCer C18 HexCer/C24:1 C1P C20 HexCer/C26:1 C1P C16 Sulfatide C18 Sulfatide C24:1 HexCer C24 HexCer/C23:1 HexCer OH C16 LacCer/C22:1 Sulfatide C18 LacCer/C24:1 Sulfatide C18 LacCer/C24:1 Sulfatide C24 LacCer

product (m/z)

precursor (m/z)

MMSAT (AUC)

XCalibur (AUC)

MRMer (AUC)

MMSAT (RT)

XCalibur (RT)

MRMer (RT)

300.2 302.5 380.1 538.6 564.6 566.7 594.7 622.7 648.8 650.8 650.9 700.7 726.7 728.7 756.7 780.5 808.6 810.8 812.7 862.6 890.6 890.6 974.6

264.1 284.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 632.9 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1 264.1

63 893 67 969 51 789 3 270 652 25 263 9 163 730 250 105 282 037 4 342 532 235 917 65 216 446 252 9 210 5 277 945 255 379 5 317 131 328 12 178 705 3 680 183 712 891 455 617 1 411 681 31 497

63 572 49 987 50 868 3 69 307 38972 9 200 226 249 923 281 946 4 43 371 149 860 69 448 447 385 11 537 5 276 116 256 840 0 131 65 12 213 849 3 810 806 712 770 459 851 1 426 474 31 376

88 889 113 063 68 194 3 271 744 135 687 9 348 982 299 643 285 617 4 425 478 421 987 780 096 461 121 67 283 5 289 399 286 813 23 669 676 366 12 231 821 5 146 047 742 021 NQ 1 905 147 60 338

3.59 4.07 5.31 16.86 17.17 17.94 19.25 20.84 21.19 22.73 21.86 16.04 16.40 16.94 17.99 16.21 17.12 19.53 20.79 15.65 16.48 19.84 19.93

3.63 4.07 5.36 16.87 17.14 17.97 19.34 20.92 21.30 22.98 21.99 16.05 16.40 16.98 18.01 16.23 17.16 19.60 20.94 15.69 16.55 19.88 20.09

3.63 4.07 5.36 16.87 14.91 17.97 19.34 20.92 21.30 22.95 21.37 16.05 14.30 16.98 18.01 16.23 18.63 19.60 20.94 15.69 NQ 19.88 20.09

a

dh = dihydro; Cer = Ceramide; C1P = cCeramide-1-phosphate; Hex = hexosyl; Lac = lactosyl. The area under curve (AUC) and retention time (RT) of each metabolite is shown. For several transitions there is more than one possible lipid structure indicated, and the actual lipid, as determined by RT, is shown in bold. In the case of C18 LactCer, MRMer does not quantify (NQ) the peak as it is designed to quantify only the most abundant peak within each transition, which in this case is C24:1 sulfatide.



the Steve Waugh Brain Tumour Bank, with ethics approval from the South Eastern Sydney Illawarra Area Health Service− Northern Section Human Research Ethics Committee. The glioblastoma tissue was pulverized on dry ice, and lipids were extracted into an ethyl acetate/isopropanol/water mixture (60:30:10 v/v/v), as described previously.10 Each sample was spiked with internal standards immediately prior to lipid extraction: 250 pmol each of d18:1/12:0 glucosylceramide, d18:1/12:0 lactosylceramide, d18:1/12:0 sulfatide, d18:1/12:0 sphingomyelin, and 50 pmol each of d18:1/17:0 ceramide, d17:1 sphingosine, and d17:1 sphingosine 1-phosphate. Final extracts were solubilized in 0.2 mL of methanol/water (80:20 v/v) containing 0.2% (v/v) formic acid and 1 mM ammonium formate. Analysis of Tissue Extracts. Glioblastoma extracts were analyzed on a Thermo Quantum Access triple quadrupole mass spectrometer operating in positive ion mode, coupled to a 3 mm × 1.5 mm XDB-C8 LC column (Agilent) on an Accela UPLC system (Thermo). Lipids were separated using a binary gradient program, at a flow rate of 0.5 mL/min: mobile phase A was 0.2% formic acid/2 mM ammonium formate in water; and mobile phase B was 0.2% formic acid/1 mM ammonium formate in methanol. Total HPLC time was 30 min per sample: for the first 14 min, a gradient starting at 20:80 A/B and increasing to 100% B was run. The mobile phase was held at 100% B for 14 min, after which the column was re-equilibrated at 20:80 A/B for 2 min. Over the period 7.5−30 min, a list of 69 SRM events encompassing at least 83 different ceramide, hexosylceramide, lactosylceramide, and sulfatide species were monitored. The scan time for each single event was 0.025 s.

RESULTS

Comparison of Performance of MMSAT with Existing Software. To demonstrate that MMSAT accurately quantifies peaks from SRM experiments, we applied MMSAT to analyze SRM data acquired from six human glioblastoma extracts. A list of 69 SRM events was designed to detect different molecular species of ceramides, hexosylceramides, lactosylceramides, and sulfatides. The data set was converted to mzXML format using ReAdW and zipped before being submitted to MMSAT. Default parameters (product ion tolerance = 0.5 Da, peak detection cutoff = 1 000 and no XIC denoising) were used in the analysis. For comparison, quantification was also performed using the XCalibur (Thermo Fisher Scientific) as well as MRMer for one of the six samples. Results in Table 1 and Figure 2 show that all three programs perform very similarly in their ability to quantify known metabolites in the sample (R2 > 0.99 and paired t test p > 0.1 for MMSAT versus both XCalibur and MRMer). This demonstrates that the quantitative accuracy of MMSAT is comparable with existing software. Identification of Multiple Metabolites from a Single SRM Transition. To further demonstrate the utility of MMSAT and in particular the importance of its ability to quantify multiple metabolites with identical or near-identical precursor and product ion mass, we describe three examples where this occurs in a lipid extract of glioblastoma tissue. Table 2 shows the quantified metabolites that will be referred to in the following examples (the full output from MMSAT analysis of the six glioblastoma samples are shown in Supplemental Table 1 in the Supporting Information). From the output, it can be seen that the number of possible molecular species is greater 472

dx.doi.org/10.1021/ac2026578 | Anal. Chem. 2012, 84, 470−474

Analytical Chemistry

Technical Note

Figure 2. Comparison of abundance of lipids detected by MMSAT against XCalibur and MRMer. Values are as described in Table 1 and represent area under curve (AUC) values.

than the number of transitions because a single transition can detect more than one distinct lipid species, even when the optimal fragment ion is used for both lipids. This arises because most QQQ instrument can only achieve unit mass accuracy and generally lack accurate mass resolution. In the case of m/z 890.6, there are clearly two distinct peaks (Figure 3A). On the basis of their retention time relative to external standards and other lipid species detected in the analysis, the earlier eluting species is d18:1/18:0 lactosylceramide, with a predicted m/z of 890.6563, while the later eluting is d18:1/24:1 sulfatide (i.e., galactosylceramide sulfate), with a predicted m/z of 890.6385. A mass difference of only 0.0178 can not be resolved on a QQQ or ion trap instrument, which are generally the preferred instruments for this type of targeted metabolomic analysis. The requirement for the peak extraction software to list all detected peaks across all samples is therefore important in this type of analysis, but surprisingly there is no such software tool currently available for SRM lists. In another example, three peaks are observed for the m/z 840.8 event. One of these, with a retention time of 21.2 min, has the same retention time as the major m/z 838.8 peak (see row 3 of Table 2), indicating that it is a 2+ m/z isotope of this lipid [d18:1/26:1 hexosylceramide]. The other two peaks are distinct forms of hexosylceramide with very similar masses

Figure 3. Extracted ion chromatograms for single SRM events that yield multiple relevant peaks. Chromatograms shown in parts A and B are derived from a single glioblastoma tissue extract; chromatogram shown in part C is derived from a mixture of synthetic ceramide standards, at a concentration of 10 pmol on column. (A) SRM event with precursor m/z 890.6 and product m/z 264.1 (collision energy 40 eV) yields two major peaks corresponding to d18:1/18:0 lactosylceramide and d18:1/24:1 sulfatide. (B) Precursor m/z 840.8 with product m/z 264.1 (collision energy 35 eV) yields two different species of hexosylceramide that are both abundant in brain. (C) Double bond isomers d18:0/24:1 and d18:1/24:0 ceramide are both detected when monitoring for precursor m/z 650.9 and product m/z 632.9 (collision energy 17 eV).

(Figure 3B), and again the mass difference is too small to be differentiated on a triple quadrupole mass analyzer. A third example is the instance of lipids that differ only in the position of a double bond and therefore have identical mass but

Table 2. Selected Metabolites Quantified by MMSAT Across 6 Glioblastoma Samples Where More than One Metabolite Is Present within Each of the Respective (650.9 m/z → 632.9 m/z, 840.8 → 264.1 m/z, and 890.6 m/z → 264.1 m/z) Transitionsa m/z

RT (mins)

650.9 650.9 838.8 840.8 840.8 840.8 890.6 890.6

21.86 22.77 21.08 19.73 21.08 22.66 16.48 19.84

transition 650.9 650.9 838.8 840.8 840.8 840.8 890.6 890.6

→ → → → → → → →

[632.4−633.4] [632.4−633.4] [263.6−264.6] [263.6−264.6] [263.6−264.6] [263.6−264.6] [263.6−264.6] [263.6−264.6]

sample 1

sample 2

sample 3

sample 4

sample 5

sample 6

65 216 161 559 2 777 307 129 834 157 030 228 974 455 617 1 411 681

26 883 69 611 1 551 124 155 710 145 949 55 541 311 694 4 843 553

59 748 166 050 1 475 209 248 641 111 553 33 915 140 204 3 774 722

19 386 962 022 9 431 75 861 155 805

16 608 390 284 189 031 65 548 7 994 9 414 137 564 917 288

23 758 214 762 197 824 151 495 6 159 89 860 349 630

a The 838.8 m/z → 264.1 m/z transition is also shown as the peak at 21.08 min is the monoisotopic ion of the peak with the identical retention time in the 840.8 → 264.1 m/z transition.

473

dx.doi.org/10.1021/ac2026578 | Anal. Chem. 2012, 84, 470−474

Analytical Chemistry

Technical Note

relying on available standards and conventional data analysis methods. Such metabolites can be identified based on chromatographic retention time or further downstream experiments.

differing column retention times: d18:1/24:0 ceramide (C24:0 ceramide) is much more abundant in biological samples than d18:0/24:1 ceramide (more commonly referred to as C24:1 dihydroceramide), but both are present and can be detected. Although ceramides yield an m/z 264.1 fragment and dihydroceramides yield an m/z 266.1 fragment, C24:1 dihydroceramide is often detected using a water loss event. In our experience as well as published methods, monitoring water loss provides the greatest sensitivity for this relatively low abundance lipid11 and in our case was found to be the only effective method for its detection. However, C24:0 ceramide is also detected in the same SRM event and as a larger peak (Figure 3C). The use of automated software that selects the most abundant peak for each event would make it very difficult to quantify C24:1 dihydroceramide, unless we were to employ a less sensitive transition.



ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■ ■

AUTHOR INFORMATION Corresponding Author *E-mail: [email protected]. Fax: (+612) 9385 1510. ACKNOWLEDGMENTS The authors would like to thank Dr. Russell Rickford for critical reading for the manuscript. Mass spectrometric results were obtained at the Bioanalytical Mass Spectrometry Facility within the Mark Wainwright Analytical Centre of the University of New South Wales. This work was undertaken using infrastructure provided by the NSW Government coinvestment in the National Collaborative Research Infrastructure Scheme. Subsidized access to this facility is gratefully acknowledged. J.W.H.W. and A.S.D. are supported by the Cancer Institute NSW Early Career Fellowships (Grants 10/ECF/2-44 and 08/ ECF/1-03, respectively). H.A. is supported by the Prince of Wales Clinical School Ph.D. Scholarship. K.M. is supported by the Cancer Institute NSW Career Development Award and the Cure For Life Foundation.



DISCUSSION Although the vendor software such as XCalibur can be used to identify and quantify more than one metabolite within a single SRM event, this approach becomes very laborious and timeconsuming for a list of 69 events such as we have used herein. Using XCalibur each metabolite of interest must be defined prior to data processing. Using MMSAT, all peaks are first identified and assigned to a particular SRM event. The major or multiple metabolite peaks detected across multiple samples can be readily identified on the basis of peak areas and the number of samples in which they are detected. This approach provides a more objective overview of the data and the ability to view all peaks in a single spreadsheet is important for assigning structures to peaks and avoiding incorrect peak assignment, particularly where synthetic standards are not available for a particular metabolite. In these instances, the identity of a given metabolite must be determined on the basis of its column retention time relative to other metabolites and related standards and/or further MS characterization. Even though it has been demonstrated here that MMSAT accurately quantifies peaks in comparison to XCalibur and MRMer, it currently does not provide native ability in visualization and validation of quantified AUC values. In this case, it is recommended that a tool such as XCalibur or MRMer be used initially alongside MMSAT for validation of peak assignments and AUC accuracy in specific data sets. Nevertheless, due to the decreased complexity of SRM mass spectral data compared with full scan or precusor ion scanning experiments, few parameters are generally required for accurate peak selection and quantification, and we therefore believe that MMSAT can be reliably used as a web-based tool to expedite the analysis of metabolomics SRM data.



REFERENCES

(1) Rabinowitz, J. D.; Lu, W.; Bennett, B. D. J. Chromatogr., B 2008, 871, 236−242. (2) Benton, H. P.; Wong, D. M.; Trauger, S. A.; Siuzdak, G. XCMS2: processing tandem mass spectrometry data for metabolite identification and structural characterization. Anal. Chem. 2008, 80. 6382−6389. (3) Pluskal, T.; Castillo, S.; A. Villar-Briones., Oresic, M. MZmine 2: Modular framework for processing, visualizing, and analyzing mass spectrometry-based molecular profile data. BMC Bioinf. 2010, 11. (4) Bellew, M.; Coram, M.; Fitzgibbon, M.; Igra, M.; Randolph, T.; Wang, P.; May, D.; Eng, J.; Fang, R.; Lin, C.; Chen, J.; Goodlett, D.; Whiteaker, J.; Paulovich, A.; McIntosh, M. A suite of algorithms for the comprehensive analysis of complex protein mixtures using highresolution LC-MS. Bioinformatics 2006, 22. 1902-1909. (5) MacLean, B.; Tomazela, D. M.; Shulman, N.; Chambers, M.; Finney, G. L.; Frewen, B.; Kern, R.; Tabb, D. L.; Liebler, D. C.; MacCoss, M. J. Skyline: an open source document editor for creating and analyzing targeted proteomics experiments. Bioinformatics 2010, 26. 966-968. (6) Martin, D. B.; Holzman, T.; May, D.; Peterson, A.; Eastham, A.; Eng, J.; McIntosh, M. MRMer, an interactive open source and crossplatform system for data extraction and visualization of multiple reaction monitoring experiments. Mol. Cell. Proteomics 2008, 7. 22702278. (7) Craig, R.; Cortens, J. P.; Beavis, R. C. Open source system for analyzing, validating, and storing protein identification data. J. Proteome Res. 2004, 3. 1234-1242. (8) Savitzky, A.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627. (9) Donoho, D. L. IEEE Trans. Inform. Theory 1995, 41, 613−627. (10) Bielawski, J.; Szulc, Z. M.; Hannun, Y. A.; Bielawska, A. Simultaneous quantitative analysis of bioactive sphingolipids by highperformance liquid chromatography-tandem mass spectrometry. Methods 2006, 39. 82-91. (11) Bielawski, J.; Pierce, J. S.; Snider, J.; Rembiesa, B.; Szulc, Z. M.; Bielawska, A. Comprehensive quantitative analysis of bioactive sphingolipids by high-performance liquid chromatography-tandem mass spectrometry. Methods Mol. Biol. 2009, 579. 443-467.



CONCLUSIONS Data analysis is a critically important aspect of the development of metabolomics. As mass spectrometers continue to improve in sensitivity, targeted mass spectrometry-based metabolomic experiments will become increasingly complex. MMSAT is a simple to use Web-based tool designed for the analysis of data sets containing many samples with multiple SRM events. It has been designed to be nonvendor specific, accepting mzXML files as the input. In particular, it is able to automatically quantify different metabolites arising from a single precursor-product ion transition. This enables the quantification of metabolites for which a researcher would not otherwise have found when 474

dx.doi.org/10.1021/ac2026578 | Anal. Chem. 2012, 84, 470−474