Proteomics 2009 V9p1683

RESEARCH ARTICLE

The detection, correlation, and comparison of

peptide precursor and product ions from data

independent LC-MS with data dependant LC-MS/MS

Scott J. Geromanos1, Johannes P. C. Vissers2, Jeffrey C. Silva1*, Craig A. Dorschel1,Guo-Zhong Li1, Marc V. Gorenstein1, Robert H. Bateman2 and James I. Langridge2

1 Waters Corporation, Milford, MA, USA2 Waters Corporation, Manchester, UK

The detection, correlation, and comparison of peptide and product ions from a data independent LC-MS acquisition strategy with data dependent LC-MS/MS is described. The data independent mode ofacquisition differs from an LC-MS/MS data acquisition since no ion transmission window is appliedwith the first massanalyzer prior to collision induced disassociation. Alternating the energy applied tothe collision cell, between low and elevated energy, on a scan-to-scan basis, provides accurate massprecursor and associated product ion spectra from every ion above the LOD of the mass spectrometer.The method therefore provides a near 100% duty cycle, with an inherent increase in signal intensitydue to the fact that both precursor and product ion data are collected on all isotopes of every charge-state across the entire chromatographic peak width. The correlation of product to precursor ions, afterdeconvolution, is achieved by using reconstructed retention time apices and chromatographic peakshapes. Presented are the results from the comparison of a simple four protein mixture, in the pres-ence and absence of an enzymatically digested protein extract from Escherichia coli. The samples wererun in triplicate by both data dependant analysis (DDA) LC-MS/MS and data-independent, alternatescanning LC-MS. The detection and identification of precursor and product ions from the combinedDDA search results of the four protein mixture were used for comparison to all other data. Each in-dividual set of data-independent LC-MS data provides a more comprehensive set of detected ions thanthe combined peptide identifications from the DDA LC-MS/MS experiments. In the presence of thecomplex E. coli background, over 90% of the monoisotopic masses from the combined LC-MS/MSidentifications were detected at the appropriate retention time. Moreover, the fragmentation patternand number of associated elevated energy product ions in each replicate experiment was found to bevery similar to the DDA identifications. In the case of the corresponding individual DDA LC-MS/MSexperiment, 43% of the possible detectable peptides of interest were identified. The presented dataillustrates that the time-aligned data from data-independent alternate scanning LC-MS experimentsis highly comparable to the data obtained via DDA. The obtained information can therefore be effec-tively and correctly deconvolved to correlate product ions with parent precursor ions. The ability togenerate precursor-product ion tables from this information and subsequently identify the correctparent precursor peptide will be illustrated in a companion manuscript.

Received: July 3, 2008Revised: September 18, 2008

Accepted: October 1, 2008

Keywords:

Biomarker discovery / Data-independent LC-MS / Multiplexed LC-MS / Shotgunsequencing / Time-resolved mass spectrometry

Proteomics 2009, 9, 1683–1695 1683

1 Introduction

MS has evolved into a powerful tool for the analysis of pro-tein mixtures owing to its speed, sensitivity, and accuracy.

Correspondence: Dr. Johannes P. C. Vissers, Waters Corporation,Atlas Park, Simonsway, Manchester M22 5PP, UKE-mail: [email protected]: 144-161-435-4444

Abbreviations: BPI, base peak intensity; CID, collisional induceddissociation; DDA, data dependant analysis

* Current address: Cell Signaling Technology, Inc., 3 Trask Lane,Danvers, MA 01923, USA

DOI 10.1002/pmic.200800562

© 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.proteomics-journal.com

1684 S. J. Geromanos et al. Proteomics 2009, 9, 1683–1695

The technology has played a pivotal role in the postgenomicera, helping to define the functional roles of identified geneproducts and gain a deeper understanding of cellular biology.Traditional mass spectrometric approaches for the identifi-cation of peptides from enzymatically digested proteinsinclude MALDI-TOF MS for single proteins, or very simplemixtures [1, 2], and LC-MS/MS for more complex mixtures[3–5]. As sample complexity increases in terms of the abso-lute number, dynamic range, and molecular weight of theproteins present, the use of precursor mass measurementsalone, as utilized by MALDI-TOF PMF, does not providesufficient specificity to impart unambiguous protein identi-fication. In these instances, the samples are typically ana-lyzed by an electrospray LC-MS/MS approach using a datadependant acquisition (DDA) method. The major advantageof this approach is the generation of primary structuralinformation from the peptide precursor ion selected forfragmentation. The added specificity provided by the frag-ment ion information increases the quality of peptide iden-tifications from more complex protein mixtures. Despitebeing a more efficient strategy to identify proteins in com-plex matrices, there are some inherent limitations associatedwith the technique.

These limitations become evident with the desire tocategorize and quantify proteins in increasingly complexbiological matrices. A detailed understanding of the changesin the protein complement of these samples when they arestressed or compared to a perturbed biological system isincreasingly required. In order to reach this goal, there are anumber of bioanalytical challenges that must be understoodand overcome to afford such comparative analyses. Firstly,enzymatically derived peptides from proteins do not sharethe same ionization efficiency. More specifically, they illus-trate an ionization distribution that spans close to two orthree orders of magnitude, whereby the majority of the pep-tide signals are in the lower end of the distribution [6]. Sec-ondly, the concentration of the proteins present in thesecomplex samples can, and often do, illustrate an even widerdynamic range. Consequently, the majority of proteins pres-ent in a sample are at least two to three orders of magnitudelower in concentration than the most abundant protein.Hence, the majority of the tryptic peptides present in biolog-ical matrices are in the lowest intensity regime.

The latter poses one of the main analytical challengeswith respect to reproducibility in LC-MS/MS-based proteinidentification schemes. A DDA experiment is typically aserial process. The cycle starts by acquiring an MS surveyscan followed by the selection of a number of precursor ionsfor an MS/MS experiment that may or may not be at theirchromatographic apex. The selected precursor ions are seri-ally isolated for an MS/MS acquisition for an allotted periodof time, or until a certain ion current is breached. The num-ber of selected precursor ions and the allocated MS/MSacquisition time are optimized for a given sample type andcomplexity. Typically, the number of selected precursors willincrease and the MS/MS acquisition time decrease with

increasing sample complexity. This creates a fundamentalproblem for the identification of proteins over a wide dy-namic range, and/or the ability to generate any semblance ofprotein sequence coverage. That is, in a DDA experiment, anindividual precursor is subjected to MS/MS until the sum-med intensity of a given fragment ion reaches an acceptablenumber of ion counts. In the case of weak precursor ions,this can result in using valuable time in the MS/MS mode ofacquisition trying to reach a threshold that cannot beachieved. Furthermore, the allotted MS/MS acquisition timemust be limited in order to maximize the duty cycle of themass spectrometer. For example, in the instance of a 30 swide chromatographic peak, and with an MS/MS acquisitionspeed of 200 ms, only approximately 1/150 of the peak vol-ume will be sampled. If the peptide of interest is present atthe level of 1.5 fmol, and with a column flow rate of 300 nL/min and an assumed ion transfer efficiency from the liquidphase to the detector of 0.001 [7], the maximum amountavailable for detection is approximately 10 zmol. The abilityto generate a good quality MS/MS fragment ion spectrumand confident database search identification from such alimited amount is challenging.

In order to obtain the highest possible sensitivity and bytaking the aforementioned into consideration, an MS1 iontransmission/isolation window around the precursor ion istypically set to 6 1.5–3 mass units, allowing for the selectionof the complete isotopic distribution of the precursor ion ofinterest with maximum sensitivity. This may not be a prob-lem for the more abundant precursor ions or peptides thatwere selected for an MS/MS acquisition near their chroma-tographic apex; however, the majority of the precursors are inlower intensity regime. In addition, more than one precursoris often present within the ion transmission window [8, 9],resulting in product ion fragmentation components that donot exclusively belong to the selected precursor. More speci-fically, Hoopmann et al. [8] found that in 17% of all MS/MSevents, more than one precursor was present in the collisioncell, whereby in a more recent study by Luethy et al. [9] thisnumber was even found to be closer to 67% of all isolatedprecursors. Although this may not be an issue if there is asignificant difference in precursor ion intensities and themore intense species are of interest. It can however make asignificant difference if the chimeric precursor ions presentare of similar intensity or similar composition [9].

Many of these limitations described have a direct effecton the lack of reproducibility, low sequence coverage, andlarge number of single peptide-based protein identificationspresent in the literature [10, 11]. As an example, in a pre-viously published independent study, six proteomics labora-tories analyzed a tryptic digest of complex biological matrix,generated from 10 000 human cells by means of LC-MS/MS[12]. In total 1757 proteins were identified of which only 52proteins were commonly identified by all laboratories. Inanother study [13], the results from three different multi-dimensional LC-MS strategies were compared with a fourth,classical source, that is, protein biochemistry and clinical


Proteomics 2009, 9, 1683–1695 1685

chemistry literature. Combining the data sets of the variousapproaches resulted in 1175 identified proteins. Interest-ingly, only 46 proteins were identified by all methods andonly 195 proteins by at least two methods. A third examplestudy involved the analysis of a glomerulus proteome [14].Here, the sample was prefractionated into 90 fractions using1D SDS-PAGE, and 2D solution phase IEF, in combinationwith SDS–PAGE. Each fraction generated was analyzed byLC-MS/MS. After validating the MS/MS data, it was deemedthat 5% of these spectra could be processed for identification.The combined search results identified 6686 proteins ofwhich 45% were comprised of single peptide identifications.A large number of tryptic peptides per protein and highsequence coverage should however be expected using a frac-tionation strategy combined with the high sensitivity of anLC-MS/MS approach.

The lack of reproducibility and the number of singlepeptide identifications is however not primarily due to theexperimental models employed, instrumentation and/orsoftware used to process and search the data, but to themethod of data acquisition. To overcome some of theseproblems, a data independent mode of acquisition wasintroduced for label free quantitative LC-MS studies [15, 16],whereby both precursor and product ion information is col-lected on all of the isotopes of all charge-states of the elutingpeptide precursor ions across the chromatographic peakwidth. Setting the MS acquisition speed in proportion to thechromatographic peak width ensures that a sufficient num-ber of data points are collected from each precursor ion toadequately measure the m/z values, retention times, andpeak volumes of all detectable ions. Other data-independentmethods have been reported, investigating the use of multi-plexed fragmentation where more than one precursor ionwas simultaneously fragmented by collisional induced dis-sociation (CID) on Fourier transform ICR mass spectro-meters [17, 18], IT mass spectrometers [19, 20], and TOFmass spectrometers [21]. Multiplexed or parallel CID, con-ducted exclusively in the source region [18, 21], gas cell [22],or a combination of both [23] have been subject of researchpapers too. All presented methods have in common that thedetected product ions have to be associated to their parentprecursor. Hoaglund-Hyzer et al. [22] have suggested andapplied the use of ion mobility gas phase separation to cor-relate simultaneously fragmented precursor ions to theirconcurrent product ions. The method applied in this studyemploys liquid phase separation and time-alignment to cor-relate precursor ions with product ions. The m/z measure-ments are converted to monoisotopic peptide masses, theintensities of all isotopes and charge states summed, and theapex retention time for each species calculated. Next, theproduct ions are time-aligned and correlated to precursorions whose apex retention time is within one-tenth of thetime associated to each precursor ions chromatographic peakwidth at half-height.

The information obtained from the chromatographic re-producibility of replicates can be used to accommodate label

and label-free relative quantification on complex biologicalmatrices [16] and has been applied for the characterization ofEscherichia coli cultured using various carbon sources [24],the response of Mycobacterium bovis to isoniazid treatment[25], identifying, validating, and measuring the absoluteconcentration of established markers in serum samples fromGaucher patients [26] and exosomes secretion by oligoden-drocytes [27]. Here, it will be demonstrated that data inde-pendently acquired LC-MS data holds the same informationin terms of detectable precursor and product ions comparedto data dependent acquired data. Also shown is how pre-cursor and product ions are correlated within and acrossexperiments of various types and how this ultimately leads toa 2–2.5-fold increase in detected components in complexbiological data sets.

2 Materials and methods

2.1 Sample preparation

50 mL of 0.5% aqueous formic acid was added to 100 mg ofcytosolic E. coli digest standard (Waters, Milford, MA, USA).A tryptic digest stock solution containing four standard pro-teins, alcohol dehydrogenase, phosphorylase B, albumin,and enolase, was prepared in 0.1% aqueous formic acid anddiluted to concentrations of 200, 200, 200, and 100 fmol/mL,respectively. Equal volumes of the E. coli digest and thestandard proteins were combined to give a sample con-centration of 0.5 mg/mL of E. coli digest and 100, 100, 100, and50 fmol/mL of alcohol dehydrogenase, phosphorylase B,albumin, and enolase, respectively. The tryptic digests of thefour proteins were also prepared in 0.1% aqueous formicacid without the presence of the E. coli digest standard at thesame concentration level of 100, 100, 100, and 50 fmol/mL,respectively. Unless stated otherwise, these solutions wereused as stocks for all the experiments described in thismanuscript.

2.2 LC-MS configuration

Nanoscale LC separation of tryptic peptides was performedwith a nanoACQUITY system (Waters), equipped with aSymmetry C18 5 mm, 5 mm6300 mm precolumn and anAtlantis C18 3 mm, 15 cm675 mm analytical RP column(Waters). The samples, 1 mL full loop injection, were initiallytransferred with an aqueous 0.1% formic acid solution to theprecolumn at a flow rate of 4 mL/min for 3 min. Mobilephase A was water with 0.1% formic acid whilst mobilephase B was 0.1% formic acid in ACN. After desalting andpreconcentration, the peptides were eluted from the pre-column to the analytical column and separated with a gra-dient of 3–40% mobile phase B over 90 min at a flow rate of300 nL/min, followed by a 10 min rinse with 90% of mobilephase B. The column was re-equilibrated at initial conditionsfor 20 min. The column temperature was maintained at



357C. The lock mass compound, [Glu1]-fibrinopeptide B, wasdelivered by the auxiliary pump of the LC system at 250 nL/min at a concentration of 100 fmol/mL to the referencesprayer of the NanoLockSpray source of the mass spectrom-eter. All samples were analyzed in triplicate.

Mass spectrometric analysis of tryptic peptides was per-formed using a Q-Tof Premier mass spectrometer (Waters,Manchester, UK). Accurate mass LC-MS data were collectedin an alternating, low, and elevated energy mode of acquisi-tion (LC-MSE) [15, 16]. The spectral acquisition time in eachmode was 1.5 with an 0.1 s interscan delay. In low energy MSmode, data were collected with a constant collision energy of4 eV. In elevated energy MS mode, the collision energy wasramped from 15 to 40 eV during each 1.5 s integration. Onecycle of low and elevated energy data were acquired every3.2 s. The RF applied to the quadrupole mass analyzer wasadjusted such that ions from m/z 300 to 2000 were efficientlytransmitted; ensuring that any ions observed in the LC-MSdata less than m/z 300 were known to arise from dissocia-tions in the collision cell. For all measurements, the massspectrometer was operated in v-mode with a typical resolu-tion of at least 10 000 FWHM. All analyses were performedin positive mode ESI. The TOF analyzer of the mass spec-trometer was externally calibrated with a NaI mixture fromm/z 50 to 1990. The data were postacquisition lock masscorrected using the doubly protonated monoisotopic ion of[Glu1]-fibrinopeptide B. The reference sprayer was sampledevery 30 s.

Accurate mass LC-MS/MS DDA data were obtained asfollows. MS survey scans of 0.6 s duration with an interscandelay of 0.05 s were acquired. MS/MS data were obtained forup to three ions of charge 21, 31 or 41 detected in the surveyscan. MS/MS was obtained at a scan rate of 0.6 with 0.05 sinterscan delay and a collision energy ramp from 15 to 40 eV.A dynamic exclusion window was set to 60 s. Acquisition wasswitched from MS to MS/MS mode when the base peakintensity (BPI) exceeded a threshold of 150 counts, andreturned to the MS mode when the TIC in the MS/MSchannel exceeded 1000 counts/s or when 0.9 s (three scans)were acquired.

2.3 Data processing and protein identification

The continuum LC-MSE and DDA LC-MS/MS data wereprocessed using the default parameter settings for bothmethods of acquisition as residing in ProteinLynx GlobalSERVER version 2.3. The processed DDA data were queriedagainst a Comprehensive Microbial Resource (http://cmr.jcvi.org) E. coli K12 database (November 2007, 4403entries) appended with a one-time randomized version of thedatabase, the four spiked proteins and trypsin. The rando-mized proteins sequences serve as a decoy to validate theidentifications of the peptide. The DDA searches were con-ducted with the default search engine parameters of thesoftware. Trypsin was set as the primary digest reagent, theprecursor mass tolerance and fragment ion tolerances

equalled 25 ppm and 0.05 Da, respectively, one missedcleavage site was allowed, and CAM-cysteine as a fixed andmethionine oxidation set as a variable modification. Addi-tional protein identification reporting criteria included apeptide identification probability .95% and the presence ofa consecutive y ion series of at least three amino acids perMS/MS spectrum. A more comprehensive description of theDDA search algorithm has been described by Skilling et al.[28].

Time-aligned LC-MSE precursor and product ions wereconsidered matched to a DDA identification provided thatthe deconvoluted, protonated precursor ion mass, andretention time were within 610 ppm and 630 s, respec-tively, and that there were a minimum of three productions to match within 620 ppm. Additional data analysiswere performed with Decision Site 9.0 (Spotfire, Somer-ville, MA USA) and Microsoft Excel (Microsoft, Redmond,WA, USA).

3 Results and discussion

3.1 Alternate scanning data acquisition

The alternate scanning acquisition method (LC-MSE) isdesigned to collect high resolution, 10 000 v-mode or18 000–20 000 (FWHM) in the w-mode of instrumentoperation, accurate mass information for each detectedprecursor and any corresponding fragment ion above theLOD of the mass spectrometer. The LC-MSE data acquisi-tion mode is configured to alternate between two collisionenergy conditions. A low energy MS survey of eluting pre-cursor peptides and an elevated energy MS survey of asso-ciated product ions with no precursor ion selection appliedprior to CID. During the elevated energy MS survey, thepotential energy difference across the collision cell isramped in a linear fashion from an initial elevated energysetting to a final value, over the period of time associated toa single scan. The LC-MSE data of a proteolytic digest arecollected throughout the entire LC-MS experiment preser-ving the chromatographic profile of all the detected peptidesand their associated product ions. Product ion informationis obtained from all the isotopes and charge states of anygiven precursor peptide as they are simultaneously frag-mented. The principle of an LC-MSE acquisition has beenpreviously described in more detail [16].

The time-alignment correlation principle is illustrated inFig. 1. It is based upon the principle that the chromato-graphic behavior of all ions associated to an eluting parentpeptide precursor is similar. Namely, product ions have thesame chromatographic profile as their parent precursor ions.Selecting the appropriate scan speed ensures that the chro-matographic attributes of an ion can be accurately measured.These attributes include peak area, accurate mass, retention-time apex, peak width, and charge-state. Moreover, they pro-vide means for the elevated energy product ions to be time-


Proteomics 2009, 9, 1683–1695 1687

Figure 1. Time-alignment correlation principle of precursor andproduct ions. Chromatographic profile precursor and product ionions, (A) chromatographic profile for the precursor peptideobtained during the low energy MS experiment, (B) chromato-graphic profile for all associated fragment ions generated duringthe elevated energy MS experiment (only one extracted isotopicmass extracted of a single fragment ion is shown). (C) Chromat-ographic peak characteristics: start (a); end (b); apex retentiontime (c); width at half maximum for both precursor and asso-ciated fragment ions (d plus e). The ratio of d over e is a measurefor the chromatographic peak asymmetry.

aligned and correlated to the appropriate parent precursorion. The culmination of this process results in an inventoryof all detected precursor ions with their associated productions.

The inventory lists are generated by processing the lowand elevated energy LC-MS continuum data with a 3D peakdetection algorithm. Ion detection is essentially accom-plished by convolving a 3D LC-MS data matrix with a 2D fil-ter. An LC-MS data matrix is formed from a series of massspectral scans, sampled uniformly in time. In such a matrix,an ion appears as a 3D Gaussian peak. In the absence ofdetector noise, the location of each apex can be used to detectthe ion and to determine retention time and m/z. The pres-ence of noise gives however rise to multiple local maxima foreach ion, so a simple apex-location scheme applied to theunfiltered data will produce spurious results. To reduce, andto essentially, eliminate over counting, the 3D data matrix isconvolved with a 2D convolution filter. The width of the filterin the time dimension is matched to the chromatographic

peak width. The width of the filter in the spectral dimensionis matched to the mass spectral peak width. Ions are detectedby the presence of local maxima in the filtered data matrix.The 2D convolution filter is normalized so that an apex valuegives an approximate estimate of ion intensity. An ion isconsidered to be detected if the filtered apex intensity isabove a threshold. The location of the apex in the filtered LC-MS matrix determines the retention time and m/z ratio ofthe ion. The intensity of the ion is obtained by extracting achromatographic profile, centered on the ion’s m/z ratio. Theion’s intensity is the integrated area under the resulting 2Dchromatographic peak. The processed data are furtherreduced with a de-isotoping and charge-state reduction algo-rithm to provide a final inventory list of time-aligned mono-isotopic mass measurements. Each precursor and production is annotated with a monoisotopic mass, an intensity sumfrom all isotopes and charge states, and an apex retentiontime. Product ions are time-aligned to their parent precursorion if their apex retention time is within the time associatedto one-tenth of the precursor ions chromatographic peakwidth at half-height (typically 61 scan).

3.2 DDA and LC-MSE acquisition of a simple and

complex protein mixture

A tryptic digest of a simple protein mixture of four proteins,see experimental section, was analyzed in triplicate by bothdata directed LC-MS/MS and LC-MSE. The DDA MS surveychromatograms and the low energy LC-MSE chromatogramsfrom the three replicate injections each are illustrated inFig. S1 of Supporting Information, respectively. The profilesof the BPI chromatograms are similar between the twomodes of data acquisition, suggesting that the same peptidesare sampled reproducibly within and across the two differentexperiments. However, one distinguishing characteristicseen in the low energy precursor ion survey spectra of thetwo experiments is the consistently higher signal intensity inthe three replicate LC-MSE acquisitions. The data-directedLC-MS/MS experiments were configured to select the threemost intense ions for interrogation by MS/MS after each MSsurvey scan. Therefore, the mass spectrometer was config-ured to dedicate most of its time to the MS/MS mode ofacquisition. The MS survey scan is thus sampled less fre-quently and as such the MS signal intensity of commonprecursor ions is lower in the data directed method of acqui-sition.

A similar comparison can be made between the BPIchromatograms generated from the triplicate analyzes of thesame four proteins in the presence of a background of an E.coli lysate protein digest. The quantity of the four proteindigest mixture was identical to the study above, and thesample was analyzed in triplicate by both LC-MS/MS and LC-MSE. The BPI chromatograms of both experiments areshown in Fig. S2 of Supporting Information, and reflectsimilar trends as observed with the simple protein mixture.



3.3 Comparison of results from DDA and LC-MSE

The results from the DDA and LC-MSE acquisition are pre-sented in the heat map shown in Fig. 2 and Figs. S3a–c ofSupporting Information. The three columns on the left handside of these particular figures illustrate 153 DDA peptideidentifications to the four standard proteins. These identifi-cations represent the combined results for all peptide identi-fications from the three replicate DDA experiments. Notethat no single DDA experiment identified all of the peptideslisted. A green box and a value of one illustrates whether apeptide sequence was identified for a particular DDA injec-tion. A red box and a zero indicate that the peptide was notidentified. For an LC-MSE experiment, a green box and a onerepresent the presence of a monoisotopic precursor ion massin the LC-MSE precursor/product ion table within 610 ppmand 61 min (two times the width of a chromatographic peakat base), and with a minimum of three product ions within620 ppm, to that of an identified DDA peptide sequence. Ared box and a zero indicate that there was no precursor ion inthe LC-MSE precursor/product ion table within the chosenmatch criteria. The number of MS/MS acquisition events forthe three replicate DDA experiments were 733, 724, and 740,respectively. Comparing the m/z values and retention timemeasurements from the three replicate DDA analyzesrevealed a high degree of consistency between the acquisi-tions. Approximately 89% of the selected masses were at thesame m/z value (610 ppm) and at the same retention time(630 s). A detailed overview of the DDA detections andidentifications is provided in Table S1 of Supporting Infor-mation. Each peptide detection/identification is annotatedwith associated retention time (start time MS/MS acquisi-tion), intensity, and replication rate.

The number of detected precursor ions reported in thereplicate LC-MSE analyses was 1290, 1249, and 1219, respec-tively. Confirmation of a detected precursor ion matching to agiven peptide sequence within the LC-MSE experiments isbased solely on the presence/detection of a calculated mono-isotopic mass within 610 ppm and 1 min in retention time toa DDA identification, with a minimum of three product ions(620 ppm) to match. An intermethod 1 min time tolerancewas used to account for the fact that a DDA retention timereflects the start of the MS/MS experiment and not the actualchromatographic peak apex. Comparing the precursor ionmasses and associated retention time measurements fromthe three replicate LC-MSE experiments, using the same tol-erances as for the DDA data, also indicates a high degree ofconsistency, approximately 90%, between replicate experi-ments. Interestingly, the intersection between the DDA andLC-MSE data sets also indicated a relative high degree ofsimilarity, approximately 70%. More specifically a total of 506,514, and 517 precursor ions from the LC-MSE experimentswere found within the previously mentioned 733, 724, and740 MS/MS experiments from the DDA data based solelyupon m/z and time. The latter was achieved by using the samematching criteria as described in the previous paragraph. Adetailed overview of the LC-MSE detections and identificationsis provided in the Table S2 of Supporting Information. Eachpeptide detection/identification is annotated with the asso-ciated retention time (peak apex), intensity, charge state, andreplication rate. The results shown in Tables S1 and S2 ofSupporting Information, and the heat map representationsshown in Fig. 2 and Figs. S3a–c of Supporting Information,illustrate the commonality of the data between the acquisitionmethods. On average, 151 of the 153 possible monoisotopicpeptide masses (approximately 99%) were detected in every

Figure 2. Heat-map representation of the identified and detected peptides to alcohol dehydrogenase from three replicate LC-MS/MS andthree replicate LC-MSE and experiments in the presence and absence of 0.5 mg E. coli tryptic digest. Experiments that generated appropriateprecursor and product ion information for the corresponding peptide of the proteins of interest are indicated by a green box and a 1 andthose that did not contain the data are indicated by a red rectangle and a 0.


Proteomics 2009, 9, 1683–1695 1689

single replicate LC-MSE experiment. Note that the detectionof the correlated DDA and LC-MSE accurate mass precursorand product ions at the appropriate retention time does notconstitute identification of the LC-MSE data at this stage. Itshould, however, also be noted that not all detected LC-MSE

features are shown in Fig. 2 and Figs. S3a–c of SupportingInformation, but only those in common with the combinedDDA results.

The results from the DDA and LC-MSE experimentsfrom the simple four protein mixture in the presence of atryptically digested cytosolic E. coli background are illustratedin the right-hand side of Fig. 2 and Figs. S3a–c of SupportingInformation. The results show a significant decrease in thenumber of identifications to the 153 possible detectable pep-tides of the four protein mixture utilizing the DDA acquisi-tion method. Previously, 137 of the 153 peptides of the fourprotein mixture were identified in each individual DDAexperiment. However, after the addition of the E. coli lysate,this number decreased drastically from 137 to 70. Thisrepresents slightly less than a 50% decrease in the number ofpeptides belonging to the four protein digest that were sam-pled. With the LC-MSE acquisition method, an average of 141of the 153 peptide monoisotopic masses of interest wasdetected, applying the aforementioned match criteria. Asummary of the results is provided in Table 1. This suggeststhat even in the presence of a very complex background, thedata independent acquisition method still detected the samepeptide precursor ions, at the same retention time. Thepresence of accurately mass measured precursor and prod-uct ions at the appropriate retention times implies that thereis a high degree of similarity between the two data sets with

respect to the presence of detectable precursor and production masses, the number of detectable product ions and theobserved fragmentation patterns. The subsequent identifica-tion of the detected peptide precursor ions and their asso-ciated time-aligned fragments will be described in detail in asubsequent manuscript that describes the databank searchalgorithm [29]. This manuscript illustrates that the sameions are detected by both acquisition methods and that mul-tiplexed data can be aligned by retention time. In addition,the S/N of LC-MSE data at both the precursor and product ionlevel is generally greater than that of the matching DDA data.

The number of MS/MS experiments for the three repli-cate DDA experiments of the four protein mixture spikedinto the enzymatically digested cytosolic E. coli backgroundwas 1586, 1589, and 1651, respectively. Interestingly, thereplication on m/z and retention time between these threeexperiments was again relatively high, approximately 83%,see Table S3 of Supporting Information. These results chal-lenge the common perception that a DDA method selectsprecursor ions for MS/MS in a serendipitous fashion. It isoften suggested and claimed that these finding are due to theinstrumentation and/or the acquisition parameters used forDDA experiments. Therefore, a comparative study has beenconducted that includes replicate DDA injections on a num-ber of different tandem mass spectrometer platforms withvarious sampling rate, scan speed and DDA characteristics.In all instances, the same LC system and on-column load ofcytosolic E. coli digest were used. It was found that the repli-cation rates, without the use of include or exclude lists, stea-died after two replicate experiments at approximately 67%and that they were instrument platform independent. The

Table 1. Fractional and replication fractional number of detected peptides by data dependent LC-MS/MS DDA and data independent LC-MSE for a four protein mixture in the absence and presence of a complex biological E. coli digest background

Protein

Alcoholdehydrogenase

Enolase Glycogenphosphorylase

Serum albumin

Maximum achievablepeptide detectionsa)

20 21 67 43

DDA LC-MSE DDA LC-MSE DDA LC-MSE DDA LC-MSE

Fractional detection four protein mixture

Inj. 1 0.90 1.00 0.90 1.00 0.88 1.00 0.93 0.99Inj. 2 0.90 1.00 0.81 1.00 0.84 1.00 0.87 0.99Inj. 3 0.90 1.00 0.81 1.00 0.84 1.00 0.82 0.97Replicating detectionsb) 0.95 1.00 0.86 1.00 0.91 1.00 0.82 0.99

Fractional detection four protein mixture in E. coli

Inj. 1 0.50 0.95 0.43 0.81 0.44 0.88 0.42 0.87Inj. 2 0.50 0.90 0.48 1.00 0.26 0.93 0.37 0.94Inj. 3 0.65 1.00 0.57 1.00 0.23 1.00 0.57 0.96Replicating detectionsb) 0.50 0.95 0.52 1.00 0.23 0.95 0.40 0.96

a) Maximum number of peptides that can be identified based on the combined results of the three replicate DDA injections.b) Fractional number based on replicating identifications (�2 out 3 replicate experiments).



latter will be the subject of a separate manuscript. The num-ber of reported precursor ions for the three replicate LC-MSE

experiments for the four protein mixture spiked into theenzymatically digested cytosolic E. coli sample was 26 902,27 015, and 25 943, respectively, see Table S4 of SupportingInformation. These numbers are significantly higher thanthe number of MS/MS experiments for the DDA LC-MS/MSanalyses since there is no precursor ion selection applied. Asa result, all detectable precursor and fragment ions are effi-ciently and continuously sampled. The replication on(M 1 H)1 and retention time between these three experi-ments was approximately 70%. A number very similar to theresults obtained for the DDA LC-MS/MS data.

The increase in sensitivity for the multiplexed acquisi-tion method is illustrated in Fig. 3 by superimposing thedistributions of the corresponding DDA and LC-MSE peptidepairs, in the presence of the digested cytosolic E. coli pro-teins. The solid red dots represent the DDA identificationsand the solid blue dots the LC-MSE detections. The grey cir-cles are E. coli peptide identifications. The size of the spot is

proportional to the log recorded precursor ions intensity. Theresults shown in Fig. 3 show an increase in precursor ionintensity of the LC-MSE ion detections over their DDA coun-ter parts. Ramos and coworkers [23] have recently reiteratedthese observations. They report that parallel fragmentationexperiments produce product ion spectra with substantiallyincreased signal intensities, attributed to the sampling ofvirtually all the ions generated by ESI. The results show thatas a consequence of the increased signal intensity of the LC-MSE generated product ion spectra, the total number ofdetectable fragment ions is increased and a more compre-hensive characterization at both the peptide and parent pro-tein level provided.

3.4 Time-alignment product/fragment ions

Precursor accurate mass and retention time are not alwayssufficiently unique to provide unambiguous peptide identi-fication. Figures 4a–h depicts deconvoluted product ionspectra and MS survey chromatograms generated by both

Figure 3. Scanning technique, i.e., DDA (red) and LC-MSE (blue), common identified peptides pairs in the presence of 0.5 mg E. coli trypticdigest. The spot size is proportional to the logarithm of the precursor ion intensity; grey circles illustrate background E. coli peptide iden-tifications.


Proteomics 2009, 9, 1683–1695 1691

Figure 4. Deconvoluted product ion spectrum of tryptic albumin peptide LVNELTEFAK spectra generated by DDA (a) and the correspondingtime-aligned LC-MSE (b) product ion spectra in the four protein mixture. Spectra (d) and (e) show similar information for tryptic albuminpeptide RPCFSALTPETYVPK. Spectra (f) and (g) are as (a) and (b); however, now in the presence of 0.5 mg E. coli tryptic digest. Panes (c) and(h) show the low energy LC-MSE chromatograms and coeluting behavior of both albumin peptides in the absence (c) and presence (h) of anE. coli digest background. Fragment ion color legend: red = y ion; blue = b ion; green = immonium ion or neutral loss of NH3 or H2O;gray = not identified; magenta = precursor or fragment ion assigned to a coeluting peptide.

acquisition methods. Figure 4a shows the identified DDAproduct ions from one of the replicate injections from thefour protein mixture to tryptic peptide LVNELTEFAK fromalbumin. Figure 4b illustrates the corresponding productions in the time-aligned spectrum from one of the LC-MSE

experiments. A very similar number of detected ions isobtained from both acquisition methods by extracting theproduct ions from both spectra, see Table S5 of SupportingInformation. The DDA and LC-MSE spectra comprised 146and 189 product ions, respectively. A comparative analysis ofthe product ion lists of the two spectra with a mass precisionof 620 ppm resulted in an intersection of 26 product ions.

A more careful perusal of the LC-MSE spectrum in Fig. 4billustrates the presence of some relatively high intensitynonidentified product ions. Figure 4c shows a section of thelow energy LC-MSE chromatogram of the four protein mix-ture and indicates the presence of a second precursor ion of(M 1 H)1 1880.9124 within one-tenth of the time associatedto the chromatographic peak width at half-height of(M 1 H)1 1163.6299. As such, the LC-MSE spectrum, sinceno precursor isolation was applied, will share some productions from both precursor ions. Inspection of the DDA searchresult shown in Fig. 4d reveals that the coeluting precursorion could be identified to the tryptic peptide sequenceRPCFSALTPETYVPK from albumin. Figure 4e depicts thecorresponding time-aligned and detected product ions fromthe LC-MSE data to precursor ion (M 1 H)1 1880.9124. Notethat the time difference of the apices of the (M 1 H)1

1163.6305 and (M 1 H)1 1880.9124 precursor ions detectedby the LC-MSE acquisition method is a mere 0.04 min.Comparing the 164 product ions from the DDA spectrum ofpeptide sequence RPCFSALTPETYVPK from albumin withthe 189 product from the LC-MSE time-aligned ions fromprecursor ions (M 1 H)1 1163.6305 and (M 1 H)1

1880.9124 resulted in the detection of an additional 37 cor-responding product ions within a 620 ppm tolerance. Thisconfirms that the LC-MSE data processing algorithms werecapable of the correct detection and time-alignment of theappropriate product ions.

Figure 4f shows the DDA spectrum associated to thesame LVNELTEFAK tryptic peptide from albumin present inthe E. coli background, whereby Fig. 4g shows the detectedand time-aligned LC-MSE product ions to the precursor ionof the calculated monoisotopic mass of 1163.6309 at the ap-propriate retention time. Figure 4h illustrates a section of thelow energy LC-MSE chromatogram of the four protein mix-ture in the presence of the digested cytosolic E. coli proteins.As to be expected, the second albumin peptide RPCFSALT-PETYVPK of precursor ion mass (M 1 H)1 1880.9174 ispresent within one-tenth of the time associated to the chro-matographic peak width at half-height of the companionprecursor ion of (M 1 H)1 1163.6309. As previously stated,they will therefore also share certain product ions. Note thatthe retention times, Fig. 4c versus Fig. 4h, were most likely tobe affected due to the presence of a very large number of E.coli peptides. Also note the presence of over 14 other E. coli



peptides of varying intensity that coelute with the two albu-min peptides within a time window of approximately 6 s. Asimple perusal of Fig. 4a–h illustrates a very high degree ofsimilarity in both the fragmentation pattern and the numberof product ions detected by the two acquisition methods. Al-though the magnitude of the increase varies, it is clear thatthe product ion intensities in the LC-MSE data sets are higherthan their DDA counterparts. This is due to the acquisitionof LC-MSE data on all isotopes and charge-states across theprecursor ions chromatographic peak width and is illustratedin more detail in the next paragraph.

3.5 Relative intensity profiles of tryptic peptides from

LC-MSE data

The ability to acquire both the precursor ion and associatedfragment ion data throughout the entire peak width of alldetected peptides in a consistent and reproducible fashionenables the use of integrated peak areas as an additionalphysiochemical attribute that can be used for the characteri-zation of all identified proteins [29]. The relative relationshipof the intensity measurements of all identified tryptic pep-tides, along with the associated accurate mass measure-

ments and retention times, provide an additional dimensionof specificity for a given tryptic peptide map of a protein. Therelative intensity measurements of the tryptic peptides iden-tified to glycogen phosphorylase are shown in Fig. 5 fromboth the LC-MSE and the LC-MS/MS acquisitions. The re-producibility of these integrated peak measurements aredepicted by the error bars, which correspond to the calcu-lated intensity RSD errors. The peptides have been orderedby decreasing intensity as determined from the LC-MSE dataacquisition method. From this plot, a smooth trend of high(, = 1.0 and . = 0.666), medium (,0.666 and .0.333), andlow (, = 0.333 and .0) ionizing tryptic peptides can beobserved. Similar observations can be seen for the otherthree proteins and are shown in Figs. S4a–c of SupportingInformation. This behavior is consistent with previouslypublished data [26, 30] and can be used to facilitate theinterrogation of complex protein samples for any specificprotein of interest, once its tryptic peptide ion map has beencharacterized. For example, in cases where one is interestedin identifying low levels of the same protein one would initi-ally look for the top two or three best ionizing peptides to theprotein along with their corresponding product ion spectraand also confirm the absence of the lowest ionizing peptides.

Figure 5. Intensity profiles of the tryptic peptides identified to glycogen phosphorylase. The absolute LC-MSE (dark grey) and relative LC-MS/MS DDA (light grey) profiles, including intensity measurement errors, are shown for the 46 characterized peptides of interest.


Proteomics 2009, 9, 1683–1695 1693

These types of profiles can be generated for a panel of pro-teins and used for the investigation of a particular complexprotein sample by limiting the survey to a specific set ofproteins involved in a biologically, or phenotypically, relevantpathway [26]. The integrated DDA peak intensities for thesame peptides are also plotted with their correspondingmeasurement error. Due to the inherent more random na-ture of the LC-MS/MS acquisition method, there is no corre-lation with the observed peak intensities. This suggests thatprecursor intensity information from the tryptic peptidesobtained from DDA LC-MS/MS experiments cannot be uti-lized as quantitative features for the characterization of agiven protein.

As previously mentioned, product ion spectra for everydetected precursor are obtained at their chromatographicapex for optimum sensitivity, which is a direct con-sequence of the successful monitoring of every peptideacross its chromatographic elution. Compared to conven-tional tandem DDA LC-MS/MS, where individual peptidesare selected and fragmented serially, LC-MSE allows datafrom multiple peptide species to be collected simulta-neously, capturing all of the precursor and fragment ioninformation without bias, and potentially with higherthroughput. Masselon et al. [18, 31] have demonstratedhow accurate mass instruments can be used to increasethe throughput of peptide identifications using a multi-plexed approach as the basis for faster and more sensitivepeptide identifications in LC-MS-based experiments. AnLC-MSE experiment integrates the ability to measure bothaccurate mass and retention time, whilst generating prod-uct ion spectra having consistently higher intensity thanthat of the corresponding LC-MS/MS product ion spectra.Also noted by Masselon et al. [18, 31], it is expected thatnew statistical approaches, and more sophisticated scoringsystems, will evolve to properly manage this type of multi-plexed data. Using the inherent intensity profile informa-tion of the detected precursors and their associated productions obtained from this method, it is expected that lowabundance proteins can be more effectively identified incomplex mixtures.

In a typical database search, however, the mass accu-racy of the detected precursor and fragments as well as thenumber of matching fragments are some of the criteriaused to determine a positive peptide identification. As thecomplexity of the data increases, more stringent criteria areneeded to improve the accuracy and specificity of thesearch results [32]. The reproducibility of the mass andintensity measurements of the peptides, and their asso-ciated product ions, in conjunction with an additional di-mension of information (time) provides a higher degree ofspecificity and selectivity for conducting proteomic studies.An hierarchical database search strategy has been devel-oped that incorporates these and other attributes to effi-ciently process LC-MSE data, providing both high sensitiv-ity and specificity, and is the subject of a further manu-script [29].

4 Conclusions

An LC-MS data acquisition method comparison is describedto illustrate that the information content of data independentacquisition LC-MS experiments is comparable to that ofDDA LC-MS/MS acquisitions. The utilized data-independentLC-MSE method in this paper could be considered to be aparallel acquisition approach, resulting in a vastly improvedmass spectrometer duty cycle. Unlike traditional datadependant MS/MS approaches, the method does not requirereal time decisions to be made on which precursor ions toselect, such as MS/MS switching thresholds or the recogni-tion and subsequent selection of specific charge states forfragmentation. As a consequence, partial sampling of chro-matographic peaks does not occur, eliminating some of thedrawbacks associated with current data directed LC-MS/MSapproaches. The LC-MSE method acquires precursor andproduct ion data on all charge-states of an eluting peptideacross its entire chromatographic peak width, providingmore comprehensive precursor and product ion spectra.Moreover, with a data independent acquisition, the combi-nation of a high-peak capacity chromatographic separationwith high sampling-rate orthogonal acceleration TOF massspectrometer provides a rapid and parallel approach for gen-erating peptide precursor and product ion detections on alleluting species across the chromatographic peak profiles.This would not have been afforded by a more traditional datadependent approach because of the inherent undersamplingof the method. A data independent approach is thereforebelieved to be more suited for relative and absolute quantifi-cation, in both label-free and stable isotope labeled quantita-tive proteomics experiments. Furthermore, a data independ-ent acquisition method is likely to be more reproducibleacross instrument platforms when analyzing similar sam-ples. The unbiased and reproducible nature of these experi-ments will, in the authors’ opinion, promote the increaseduse of data-independent, parallel methods, for the analysis ofcomplex biological samples.

Low energy, parent precursor, and elevated-energy,product ions share the same chromatographic profile andapex retention-time, which provides the capacity to correlatethem and provide additional specificity to the experiment.However, precursor ions will and do experience coelution toa degree. This degree of coelution is acknowledged andaddressed during the processing, which is in contrast todata directed/dependant LC-MS/MS experiments, where thecoincident fragmentation of precursor ions in the collisioncell is typically not addressed. Despite this, certain coelutingproduct ions, the elevated-energy ions that cannot be exclu-sively assigned to a precursor ion, will initially be sharedbetween multiple precursors. However, with the affordedmass measurement accuracy on both precursor and prod-ucts ions and the subsequent database search strategyemployed, these additional product ions present within thespectrum have a very minor effect on the identification ofthe peptides.



Mass spectrometric identification information is cur-rently populated into proteomic repositories such as Pepti-deAtlas, The Global Proteome Machine and PRIDE, sup-porting the concept of using peptide database libraries as ameans to identify and/or validate the presence of peptidesand proteins from the compilation of empirically deriveddata. The information ultimately gained from LC-MSE

experiments, i.e., the identification of peptides and proteins,including all the precursor ions and product ions matched toevery protein, each with its monoisotopic mass, retention-time, charge-states, and summed intensities, allows for thecontinuous accumulation of both the qualitative and quanti-tative information in these types of repositories, with thebenefit of applying specific (bio) analytical reproducibilitymeasures. The qualitative and quantitative results obtainedfrom these studies will serve to further advance the under-standing of proteomics and will be used to address importantbiological questions.

The authors would like to acknowledge the valuable con-tributions of Timothy Riley throughout the development of thiswork. We also thank Martha Stapels, Michael Nold and JoanneConnolly for their ideas and contributions throughout the editingof this manuscript.

The authors have declared no conflict of interest.

5 References

[1] Strupat, K., Karas, M., Hillenkamp, F., Eckerskorn, C., Lott-speich, F., Matrix-assisted laser desorption ionization massspectrometry of proteins electroblotted after polyacrylamidegel electrophoresis. Anal. Chem. 1994, 66, 464–470.

[2] Mann, M., Hojrup, P., Roepstorff, P., Use of mass spectromet-ric molecular weight information to identify proteins insequence databases. Biol. Mass Spectrom. 1993, 22, 338–345.

[3] McCormack, A. L., Schieltz, D., Eng, J. K., Goode, B. et al.,Direct analysis and identification of proteins in mixtures byLC/MS/MS and database searching at the low femtomolelevel. Anal. Chem. 1997, 69, 767–776.

[4] Washburn, M. P., Wolters, D., Yates, III, J. R., Large scaleanalysis of the yeast proteome via multidimensional proteinidentification technology, Nat. Biotech. 2001, 19, 242–247.

[5] Peng, J., Elias, J. E., Thoreen, C. C., Licklider, L. J., Gygi, S. P.,Evaluation of multidimensional chromatography coupledwith tandem mass spectrometry (LC/LC-MS/MS) for large-scale protein analysis: The yeast proteome. J. Proteome Res.2003, 2, 43–50.

[6] Ishihama, Y., Oda, Y., Tabata, T., Sato, T. et al., Exponentiallymodified protein abundance index (emPAI) for estimation ofabsolute protein amount in proteomics by the number ofsequenced peptides per protein. Mol. Cell. Proteomics 2005,4, 1265–1272.

[7] Geromanos, S., Freckleton, G., Tempst, P., Tuning of an electro-spray ionization source for maximum peptide-ion transmissioninto a mass spectrometer. Anal. Chem. 2000, 72, 777–790.

[8] Hoopmann, M. R., Finney, G. L., MacCoss, M. J., High-speeddata reduction, feature detection, and MS/MS spectrumquality assessment of shotgun proteomics data sets usinghigh-resolution mass spectrometry. Anal. Chem. 2007, 79,5620–5632.

[9] Luethy, R., Kessner, D. E., Katz, J. E., MacLean, B. et al., Pre-cursor-ion mass re-estimation improves peptide identifica-tion on hybrid instruments. J. Proteome Res. 2008, 7, 4031–4039

[10] Carr, S., Aebersold, R., Baldwin, M., Burlingame, A. et al.,The need for guidelines in publication of peptide and proteinidentification data: Working Group on publication guide-lines for peptide and protein identification data, Mol. Cell.Proteomics 2004, 3, 531–533.

[11] Wilkins, M. R., Appel, R. D., Van Eyk, J. E., Chung, M. C. M. etal., Guidelines for the next 10 years of proteomics, Prote-omics 2006, 6, 4–8.

[12] Chamrad, D., Meyer, H. E., Valid data from large-scale pro-teomics studies, Nat. Methods 2005, 2, 647–648.

[13] Anderson, N. L., Polanski, M., Pieper, R., Gatlin, T. et al., TheHuman plasma proteome – A nonredundant list developedby combination of four separate sources, Mol. Cell Prote-omics 2004, 3, 311–326.

[14] Miyamoto, M., Yoshida, Y., Taguchi, I., Nagasaka, Y. et al., In-depth proteomic profiling of the normal human kidney glo-merulus using two-dimensional protein prefractionation incombination with liquid chromatography-tandem massspectrometry. J. Proteome Res. 2007, 6, 3680–3690.

[15] Bateman, R. H., Carruthers, R., Hoyes, J. B., Jones, C. et al., Anovel precursor ion discovery method on a hybrid quadru-pole orthogonal acceleration time-of-Flight (Q-TOF) massspectrometer for studying protein phosphorylation. J. Am.Soc. Mass Spectrom. 2002, 13, 792–803.

[16] Silva, J. C., Denny, R., Dorschel, C. A., Gorenstein, M. V. etal., Quantitative proteomic analysis by accurate mass reten-tion time pairs. Anal. Chem. 2005, 77, 2187–2200.

[17] Williams, E. R., Loh, S. Y., McLafferty, F. W., Cody, R. B.,Hadamard transform measurement of tandem fourier-transform mass spectra. Anal. Chem. 1990, 62, 698–703.

[18] Masselon, C., Anderson, G. A., Harkewicz, R., Bruce, J. E. etal., Accurate mass multiplexed tandem mass spectrometryfor high-throughput polypeptide identification from mix-tures. Anal. Chem. 2000, 72, 1918–1924.

[19] Wilson, J., Vachet, R. W., Multiplexed MS/MS in a Quadru-pole Ion Trap Mass Spectrometer. Anal. Chem. 2004, 76,7346–7353.

[20] Venable, J. D., Dong, M.-Q., Wohlschlegel, J., Dillin, A.,Yates, III, J. R., Automated approach for quantitative analy-sis of complex peptide mixtures from tandem mass spectra.Nat. Methods 2004, 1, 39–45.

[21] Purvine, S., Eppel, J.-T., Yi, E. C., Goodlett, D. R., Shotguncollision-induced dissociation of peptides using a time offlight mass analyzer. Proteomics 2003, 3, 847–850.

[22] Hoaglund-Hyzer, C. S., Li, J., Clemmer, D. E., Mobility label-ing for parallel CID of ion mixtures. Anal. Chem. 2000, 72,2737–2740.

[23] Ramos, A. A., Yang, H., Rosen, L. E., Yao, X., Tandem parallelfragmentation of peptides for mass spectrometry. Anal.Chem. 2006, 78, 6391–6397.


Proteomics 2009, 9, 1683–1695 1695

[24] Silva, J. C., Denny, R., Dorschel, C. A., Gorenstein, M. V. etal., Simultaneous qualitative and quantitative analysis of theE. coli proteome: A sweet tale. Mol. Cell. Proteomics 2006, 5,589–607.

[25] Hughes, M., Silva, J. C., Geromanos, S. J., Townsend, C. A.,Quantitative proteomic analysis of drug-induced changes inmycobacteria. J. Proteome Res. 2006, 5, 54–63.

[26] Vissers, J. P. C., Langridge, J. I., Aerts, J. M. F. G., Analysisand quantification diagnostic serum markers and proteinsignatures for gaucher disease. Mol. Cell. Proteomics 2007,5, 755–766.

[27] Kraemer-Albers, E.-M., Bretz, N., Tenzer, S., Winterstein, C.et al., Oligodendrocytes secrete exosomes containing majormyelin and stress-protective proteins: Trophic support foraxons? Proteomics Clin. Appl. 2007, 1, 1446–1461.

[28] Skilling, J., Denny, R., Richardson, K., Young, P. et al., Prob-Seq–A fragmentation model for interpretation of electro-

spray tandem mass spectrometry data. Comp. Funct.Genom. 2004, 5, 61–68.

[29] Li, G.-Z., Vissers, J. P. C., Silva, J. C., Golick, D. et al., Data-base searching and accounting of multiplexed precursorand product ion spectra from the data independent analysisof simple and complex peptide mixtures. Proteomics, 2009,9, 1696–1719.

[30] Silva, J. C., Gorenstein, M. V., Li, G.-Z., Vissers, J. P. C., Ger-omanos, S. J., Absolute quantification of proteins byLCMSE: A virtue of parallel MS acquisition. Mol. Cell. Prote-omics 2006, 5, 144–156.

[31] Masselon, C., Pasa-Tolic, L., Lee, S.-W., Li, L. et al., Identifi-cation of tryptic peptides from large databases using multi-plexed tandem mass spectrometry: Simulations andexperimental results. Proteomics 2003, 3, 1279–1286.

[32] Zubarev, R., Mann, M., On the proper use of mass accuracyin proteomics. Mol. Cell. Proteomics 2006, 6, 377–381.


Documents

Proteomics 2009 V9p1683