PARALLEL REACTION MONITORING FOR HIGH RESOLUTION AND HIGH MASS ACCURACY QUANTITATIVE, TARGETED PROTEOMICS

the next reaction monitoring (PRM). In the third quadrupole of a a HR/AM parallel of one, of the a quadrupole-equipped a comparison to in terms and measurement automated data than selectivity the domain. the two our


INTRODUCTION
The most widespread protein sequencing technique is the shotgun method. Proteins are digested into peptides, chromatographically separated, and measured by mass spectrometry (MS) (1)(2)(3)(4)(5)(6)(7)(8)(9)(10). Many types of mass spectrometers are usedquadrupole ion traps (QIT), QIT hybrids such as the QLT (quadrupole linear ion trap)-Orbitrap or QLT-FT-ICR, and quadrupole time-of-flight (Q-TOF) hybridsbut, the experiments, from the MS measurement onward, are basically identical: the masses of eluting cationic peptide precursors are measured in a MS scan, and the most abundant precursors are selected in series for successive tandem MS events (MS/MS). This process, called data-dependent acquisition (DDA), continues for the duration of a chromatographic separation, and constant MS operation in this manner can generate hundreds of thousands of spectra in days. These spectra are then mapped to peptide or protein sequence databases using highly-evolved database search algorithms (11)(12)(13). Successful results can be obtained within just a few days and are nothing short of spectacular: tens of thousands of unique peptide spectral matches mapping to several thousand unique protein isoforms have become the norm. While this approach certainly can achieve ultra-high-throughput, it is unfortunately lacking in sensitivity and reproducibility. Specifically, complete coverage of specific biological pathways or functional groups is not typical (i.e., all 500 kinases, 1,400 transcription factors, etc.). Likewise, the overlap of identifications in replicate experiments is low (35-60%) (14,15). 27) to whole systems analyses (28)(29)(30). The rising popularity and promise of the SRM technique have also spawned a plethora of new analysis approaches and software tools. For example, many algorithms and software tools have been developed to expedite assay development by aiding in the selection of proteotypic peptides (31), peptide transitions, and instrument parameters (reviewed in Cham Mead, et al. (32)). Additionally, through community effort, several publicly available databases of tandem MS spectra (33)(34)(35)(36)(37)(38)(39) and validated SRM assays (40,41) are available to provide empirical guides to transition selection and assay development. Several groups have also presented elegant solutions to maximize instrument duty cycle and improve assay specificity. Picotti and colleagues (42), for instance, demonstrated use of synthetic peptide libraries to accelerate SRM assay development. Likewise, Kiyonami, et al. (43) recently introduced a strategy that boosted SRM bandwidth by restricting the number of transitions per target acquired before and after target elution. By limiting the acquisition of full transition sets to a few occasions during peptide elution, this strategy enabled simultaneous qualitative and quantitative analysis of 6,000 transitions in one hour, a substantial boon to SRM throughput.
In recent years, discovery-based proteomic methods, such as the shotgun method discussed above, have been transformed by significant advancements in instrumentation; key figures of merit, such as sensitivity, duty cycle, mass accuracy, and mass resolution have seen remarkable improvements (44).
While these developments have done little to directly curtail the reproducibility issues that the SRM method so effectively counters, the achievable depth of proteomic sampling (i.e., analytical sensitivity) within the discovery context continues to improve. Central to this evolution is the increased performance, and availability, of high resolution and accurate mass (HR/AM) instrumentation (44). Namely, developments in time-of-flight (TOF) technology (45), and the advent of the Orbitrap mass analyzer in 2005 (46)(47)(48)(49), have made fast and sensitive MS/MS scanning (50)(51)(52)(53) with <10 parts-per-million (ppm) mass measurement error routine (54). For discovery experiments, the ability to acquire MS/MS scans with high resolution and low-ppm mass errors offers several advantages, including higher confidence sequence identification (44,55), post-translational modification site localization (56), and improved quantitative accuracy. Coming from this perspective, we wondered whether the highly accurate mass by guest on May 6, 2020 https://www.mcponline.org Downloaded from measurement capabilities of today's new-generation MS instrumentation could be leveraged to provide benefits within the targeted proteomics domain.
Driving our inquiry was the newly-introduced Q Exactive bench-top quadrupole-Orbitrap MS (QqOrbi) (47), which, along with quadrupole-TOF (QqTOF) instrumentation (45), possesses a geometry essentially equivalent to a QqQ, except that the third quadrupole of the QqQ is replaced by an Orbitrap (or TOF) analyzer (Figure 1A-B). The QqOrbi achieves a 12 Hz scan rate at a resolution of 17500 for both MS and MS/MS scanning, quadrupole mass filter isolation with mass windows as small as ±0.2 Th, and mass measurement errors typically <1 ppm with internal calibration and <5 ppm with external calibration (47). With these performance characteristics, we envisioned a targeted proteomics strategy where all products of a target peptide are simultaneously monitored under conditions that offer high resolution and high mass accuracy. Operation would be identical to a SRM scan except that all transitions would be co-detected and distinguished from one another, and from background, by the final mass analysis stage. We call this mode of operation parallel reaction monitoring (PRM).
The PRM technique has several potential advantages over the traditional SRM approach. First, PRM spectra would be highly specific since all potential product ions of a peptide, instead of just 3-5 transitions, are available to confirm the identity of the peptide (57,58). Second, PRM could provide a higher tolerance for co-isolated background peptides/species. Since numerous ions would be available for identification and quantitation purposes, the presence of interfering ions in a full mass spectrum would be less disruptive to overall spectral quality than interference in a narrow mass range, especially since high resolution can often separate these ions from the product of interest. Note that one could extend this concept to a multiplexed PRM scan where the product ions of several target peptides are co-mingled and detected in a single-scan (47). And third, since PRM monitors all transitions, one need not have prior knowledge of, or pre-select, target transitions before analysis. These points suggest another potential advantage of the PRM approach: elimination of much of the effort required to develop and optimize the traditional SRM assay. by guest on May 6, 2020 Given that a QqQ possesses a duty cycle approaching 100% and utilizes electron multiplier-based detection, which is inherently more sensitive than image current-based detection (Orbitrap), it is not obvious that the PRM method will afford sensitivity comparable to the current state-of-the-art SRM approach. However, we postulate that what PRM lacks in sensitivity and duty cycle might be effectively countered by the selectivity of HR/AM measurement. Here, we implement PRM on a QqOrbi system and benchmark method performance with triplicate analysis of 25 isotopically heavy-labeled synthetic peptides spanning a concentration range of 10 5 under neat and matrix-containing conditions. We assess key figures of merit, including data quality, run-to-run reproducibility/precision, dynamic range/sensitivity, and measurement accuracy/linearity. Finally, we draw a performance comparison to SRM operating on a common QqQ platform.

Materials and reagents.
Unless otherwise specified, all reagents used herein were purchased from Sigma Aldrich (St. Louis, MO). Acetonitrile was purchased from Fisher Scientific (Fair Lawn, NJ) and formic acid (>99%) from Thermo Fisher Scientific (TFS, Rockford, IL). Ultrapure water was supplied from a Barnstead Nanopure Diamond ultrapure water system (resistivity 18.2 MΩ-cm; TFS, Dubuque, Iowa).

Sample preparation.
Twenty-five heavy-labeled hypothetical tryptic human peptides ( Table 1) were synthesized by Fmoc solid-phase synthesis, purified by HPLC, and solubilized in 5% v/v CH 3 CN/water at a concentration of 5 pmol/µL ± 25% with purity >97% (HeavyPeptides AQUA QuantPro; TFS, Ulm, Germany). All   QqQ experiments were performed on a TSQ Quantum Discovery Max (TFS, Austin, TX). Each sample was analyzed in triplicate in order of increasing concentration, targeting a selected set of 14 peptides in scheduled SRM mode (see Table 1 for transitions, collision energies, and scheduling). (Redmond, WA) and is available on our website at http://www.chem.wisc.edu/~coon/software.php. by guest on May 6, 2020 Access to data in the proprietary TFS .raw file format was enabled by the XRawfile Component Object Model (COM) library (XRawfile2.dll, installed automatically with Thermo Xcalibur). ElutionProfiler used .raw files as input to generate an extracted score chromatogram (XSC) for each PRM spectrum. The spectral score was calculated based on all present, sequence-specific b-and y-ions using the following formula (equation 1): where the result of the Dirac delta functions, δ b and δ y , is 1 if the n th ion in the spectrum is a b-or y-ion with <5 ppm mass error, respectively, and 0 in all other cases. n is the product ion number (e.g., 4 for a y 4 or b 4 ion), k is the length of the peptide (number of amino acids), and 0.25 is an arbitrarily-chosen scalar to weight b-ions (not containing an isotopically heavy-labeled arginine or lysine at the c-terminus) less than equivalently numbered y-ions. XICs were generated using the summed intensity of all possible band y-product ions for a particular peptide, extracted at a ±10 ppm mass tolerance. Detection was based on the presence of product ion signals in at least 2 of 3 replicates within ±3σ min of the expected retention time, mass error within ±5 ppm, chromatographic signal-to-noise ≥3, and the presence of a combination of product ions in at least one spectrum with a score meeting or exceeding a peptide-specific threshold equal to the length of the targeted peptide (k) in all cases. was assessed for its potential to interfere in precursor and product ion measurement of the 25 peptide sequences studied here, in both light and heavy forms (50 total), considering the y-ion transitions monitored on the QqQ ( Table 1). For peptides not monitored on the QqQ, equivalent y-ions were chosen.
Intact confounder peptides were considered in charge states from 1-5 (monoisotopic mass only, isotopes were not considered). Confounder product ions were considered in charge states from 1 to one less the Theoretical comparison of SRM and PRM. The process of targeting a peptide with SRM involves two stages of quadrupole mass filtering with tight tolerances for both members of a precursorproduct ion transition. Since all product ion transitions targeted for a given precursor peptide (usually 3 to 5) are required to simultaneously elute, the likelihood of mistaking a non-target peptide or background ion for the targeted peptide is a rare occurrence; hence, SRM is considered to be a highly specific assay.
The proposed PRM method, however, involves only one stage of quadrupole mass filtering (of the precursor of interest) prior to mass analysis in an Orbitrap. The Orbitrap, however, by nature of its high resolution and high mass accuracy should more effectively separate ions of interest from background ions than the electron multiplier-based detection used in a QqQ. Thus, to motivate our experiments, we asked how PRM compares theoretically to SRM in terms of specificity. In other words, can the selectivity of Orbitrap HR/AM mass analysis make up for use of only one stage of mass filtering?
To answer this question, we digested the human proteome with trypsin in silico to yield over 20 million unique peptide sequences (see Experimental Procedures for details). For our calculations, we considered these peptides to be potential confounders in the measurement of the precursor and product ions of the 25 isotopically heavy-labeled peptides targeted in this study, as well as their 25 corresponding unlabeled peptides (50 total). For each unique confounder peptide, considered in charge states from 1 to 5, we further generated all possible b, y, a, b/y/awater, and b/y/aammonia product ions, internal fragments, and immonium ions in charges ranging from 1 to one less the precursor charge state. We then asked how often, depending on the amount of evidence required by the assay and the mass analyzer employed, the numerous ions generated by the confounder population resulted in indistinguishable interference in the measurement of one of our target peptides. The results of these queries are summarized in Figure 1C.
First, we consider a query requiring the least evidence of the targeted peptide, a SIM experiment.
In SIM, an intact target peptide ion is isolated in Q1 and mass analyzed without further transformation.
We begin our theoretical calculations with SIM to explore the effect of mass accuracy/resolution alone on selectivity. Assuming Q1 isolation widths of ±0.5 and 1 Th, and mass errors less than ±250 and 5 ppm, for the QqQ and QqOrbi, respectively, highly accurate mass analysis improves the chances of correctly identifying the target peptide (96 ± 3%) by exclusion of a large number of spurious peptides (7151 ± 1447, on average). Still, such an experiment, however, would only produce an unambiguous identification ~1% of the time ( Figure 1C). Note, these calculations assume no upfront chromatographic separations and that all genome-predicted peptides are translated and detectable (i.e., the worst-case scenario).
Though it is obvious that high accuracy mass measurements increase specificity, it is less clear that this benefit persists in reaction monitoring experiments. To test this, we considered the number of intact peptides that could potentially be co-isolated with a given target peptide (assuming ±1 and ±0.5 Th Q1 isolation windows centered on the target peptide m/z for the QqOrbi and QqQ, respectively) and then generate at least one product ion (of any type) with a m/z within ±5 ppm (QqOrbi) or ±250 ppm (QqQ) of a y-ion transition from the target peptide. As shown in Figure 1C, QqQ SRM assays, the likelihood of correctly identifying a target at low resolution and unit mass accuracy is still less than 1% for the 50 peptides considered here (0.6 ± 0.3%). Again, these calculations do not consider the significant benefit that is achieved by chromatographic separation and model a worst-case scenario. Even still, the detection of three y-ion transitions at high mass accuracy (i.e., <5 ppm mass error) provides nearly unambiguous target confirmation (~1.6 ± 1.1 potential confounders) from the background of the entire human peptidome.
by guest on May 6, 2020 These theoretical calculations confirm our guiding supposition that PRM can provide greater routine specificity as compared to conventional SRM. Whether this result implies greater overall performance in targeted, quantitative proteomics studies compared to SRM, however, depends on several additional, but important, factors: whether both analyzers, one image current-based and the other electron multiplier-based, are capable of detecting the targeted species, and whether both analyzers can reproducibly and accurately measure the abundance of the targeted species. Given that PRM relies on an analyzer that is fundamentally less sensitive and slower than that used in SRM, investigation of these factors will reveal whether selectivity/specificity can overcome limitations in speed and sensitivity. In the following sections, we empirically investigate this issue through a systematic analysis of reproducibility, sensitivity, and linearity of both methods.
Detection criteria for PRM and SRM. The detection criteria we developed for PRM incorporate the benefits of high selectivity and specificity HR/AM mass analysis, as discussed above. By making use of full mass range MS/MS spectra and the high specificity of product ions when measured with high mass accuracy, we developed an automated detection algorithm that assigned a spectral score to each PRM spectrum based on the presence of b-or y-ions within ±5 ppm of expected target-specific product ions using equation 1, and then generated an "extracted score chromatogram" (XSC) for that peptide. In the design of our spectral score, we chose to weight b-ions less than equivalently numbered yions because all of the peptides targeted in this study were isotopically heavy-labeled at the c-terminus and analyzed in a background of endogenous yeast peptides. Thus, y-ions, containing the heavy label, were more specific to our target peptides than b-ions and were weighted as such. If the target peptides of interest were not heavy-labeled, one might consider equally weighting b-and y-ions for scoring purposes.
A positive detection event required that, in at least 2 of 3 replicates, the XSC met or exceeded a peptide specific threshold (equal to the length of the target peptide) within ±3σ min of the expected retention time (Supplemental Figure 1). Figure 2 shows the application of this score to exemplary data from the peptide AETLVQAr (#17) under neat and matrix-containing conditions over all 6 peptide concentrations.
The XSCs (grey) are overlaid with XICs (blue) generated using the summed intensity of all b-and y-ions by guest on May 6, 2020 present in each spectrum. Note that XSCs are used solely for establishing a detection event and not for quantification as the score is independent of ion intensity. Following a positive detection event, the target-specific ions that generated the detection event are extracted as an XIC for quantification. This

PRM measurement precision.
In quantitative studies, high measurement precision is critical to reliably distinguish differences between two analyses or samples. We assessed the degree of measurement precision for the PRM method, defined here as run-to-run area-under-the-curve (AUC) repeatability across technical replicates, for all concentrations and isolation widths. Overall, PRM exhibited high measurement precision with median percent relative standard deviation (%RSD) less than 10% in most cases (Supplemental Table 2A concentrations. In the matrix-containing PRM data, however, no significant differences in the measurement precision of adjacent concentrations were observed (Supplemental Table 2B). When grouped by peptide alone, only very hydrophilic, poorly-retained peptides, #3 and #7, had significantly decreased measurement precision when compared to the other 23 peptides (Supplemental Table 2C).
We find that PRM measurement precision is consistent with studies reporting run-to-run precision   Table 3B). The presence of matrix in PRM experiments resulted in a modest depression of the quantifiable dynamic range, ~0.5 orders-of-magnitude (matrix vs. neat, 10 2.4 vs. 10 3.0 ). While tighter isolation widths slightly mitigated the effects of matrix-induced sensitivity depression, at the expense of overall sensitivity, this difference was not significant by guest on May 6, 2020 (Supplemental Table 3A). These results indicate that, although an increase in selectivity due to gasphase enrichment would be expected by tighter isolation widths, the concomitant decrease in ion transmission at very tight isolation widths (±0.2 Th), in conjunction with Orbitrap detection, results in decreased sensitivity as too few ions are present for the target signal to exceed the Orbitrap's thermal noise band. However, with a HR/AM analyzer, the wider isolation width, which provides greater ion transmission, but also higher levels of chemical noise, can be used without a decrease in performance due to the high selectivity of the mass analysis.
Linearity of PRM measurement response. QqQ SRM measurement precision, dynamic range, and linearity. Like the PRM data described above, the QqQ SRM data, which queried a subset of 14 of the 25 peptides analyzed on the QqOrbi, also exhibited a high degree of run-to-run measurement precision, typically less than 15% RSD.
Unlike in PRM, however, measurement precision, when all peptides and concentrations were considered together, was significantly greater in the presence of matrix (Supplemental Table 5A). Lower peptide concentrations were again correlated with decreased measurement precision, though only significantly so under neat conditions (Supplemental Table 5B). Peptides were quantified on average over concentration if only data over the same dynamic range under both matrix-containing and neat conditions were considered, effectively truncating the neat data to the concentration range quantified in the matrixcontaining data, linearity differences due to matrix interferences were not significant. Adjusted %RSDs, however, still exhibited significantly decreased linearity in the neat case. Mean adjusted %RSDs were 13.0 and 8.0% for neat and matrix data, respectively (Supplemental Table 5C).
We postulate that the adjusted %RSD correction did not completely account for differences in linearity, as it did in the PRM data, because the SRM detection criteria were not sensitive enough to detect and exclude data at the lowest detected concentrations in neat experiments that were skewed due to by guest on May 6, 2020 small amounts of chemical interference. In the matrix-containing SRM experiments, on the other hand, more abundant matrix interferences were easily detected and excluded by the detection criteria (Supplemental Figure 4). Detection criteria incorporating HR/AM data, however, more sensitively detected and excluded aberrant responses and thus, deviations in linearity were predictable and correctable (as adjusted %RSDs) based on the detected dynamic range. This same rationale can be applied to the discussion of measurement precision above: when all neat data were considered, precision was diminished relative to the matrix-containing dataset due to the inclusion of background-skewed data not excluded by the detection criteria.  Table 6B).
SRM exhibited greater measurement precision likely due to the higher sampling rate of the QqQ (almost twice as many scans were acquired compared to the QqOrbi per 90 min chromatographic run).
The ability of the QqQ to sample more points over a given chromatographic peak provided a more accurate determination of the peak AUC and, in turn, greater run-to-run repeatability. The difference in sampling rate between the two methods is due to the characteristics of the instruments used and, to some extent, necessary aspects of the experimental design. Since the QqQ is a "beam-type" instrument (as opposed to a scanning instrument like the QqOrbi), it operates at a duty cycle nearing 100%; this means that there is very little "down time" where the instrument is not acquiring data. The Orbitrap, conversely, as a scanning instrument, has inefficiencies inherent to its design. The Orbitrap transients employed here  Table 6C).
Under neat conditions, PRM demonstrated significantly higher linearity over the quantified dynamic range compared to QqQ SRM ( Table 2). Given that PRM under matrix-containing conditions yielded quantitative data over a wider dynamic range than SRM, we calculated adjusted %RSDs, as before, to normalize the linearity metric for the dynamic range detected and permit a fair comparison of the data. With this consideration, PRM linearity was statistically no different from the linearity exhibited by guest on May 6, 2020 by SRM ( Table 2). Figure 4 plots the SRM linearity data for each of the targeted peptides as a function of PRM linearity and is stratified by the presence of matrix and isolation width. This data presentation demonstrates the effect of using the adjusted linearity metric on mean linearity estimates (shown as vertical and horizontal lines), as well as the relative distributions of the linearity metrics across the datasets.
With these data, we find that a high resolution and accurate mass MS can acquire data of similar quality to that collected on a QqQ in a targeted, quantitative proteomics context with 1) minimal upfront development time, 2) straightforward data analysis, and 3) similar performance metrics. Our PRM methods were designed and optimized on a method-level, rather than on the peptide-level as with SRM methods (i.e., optimization of individual collision energies and transition sets for each target peptide).
Thus, the PRM method has fewer parameters that require optimization (namely, only maximum injection time/AGC target, "global" collision energy, and, optionally, target scheduling) and can be performed with solely knowledge of the mass-to-charge ratios of the target peptides (and, optionally, the approximate retention time). These characteristics make PRM amenable to the "walk-up" instrument user who wishes to perform targeted, quantitative assays on a list of targets in a time-efficient mannercollection of MS/MS data and optimization of individual transition sets not required. We believe that the data presented here reflect the analytical performance a typical user might expect for PRM, however the analysis parameters used here may not represent the optimum conditions for all applications of PRM and would require application-specific investigation.
In addition to the minimal upfront method development needed to successfully perform a quantitative PRM assay, a notable aspect of using HR/AM data for targeted proteomics is the ease with which data can be interpreted and data analysis can be automated. Within the course of this study, we Lastly, our analysis of the analytical performance characteristics of PRM suggests that targeted HR/AM methods can rival the performance of QqQ SRM in terms of dynamic range, linearity, and, to a lesser extent, precision. While SRM measurement precision was approximately 2-fold better (under matrix-containing experiments) likely due to differences in scan rate between the two mass analyzers, PRM yielded quantitative data over a wider dynamic range than SRM under matrix-containing conditions.
High mass accuracy and high resolution underlie this result: when high levels of matrix background are present, a single stage of isolation in combination with highly resolved data is statistically significantly more sensitive than two stages of isolation in combination with low resolution data, the "gold-standard" method for quantitative proteomic analyses. Additionally, achievable linearity over the quantifiable dynamic range was found to be statistically the same between SRM and PRM. Thus, in answer to the query posed in our theoretical comparison of SRM and PRM, our experimental data suggest that high selectivity/specificity can overcome limitations in speed and sensitivity to reliably provide lower detection limits and higher accuracy measurements. We conclude that the proposed PRM analysis by guest on May 6, 2020 paradigm holds promise as a viable, and accessible, alternative and/or complement to SRM for the quantitative proteomics toolbox.
In support of our conclusions, a report by Weisbrod, et al. (61), published during the preparation of this manuscript, described a method for data-independent discovery proteomics on an Oribtrap MS that reinforces the favorable performance of PRM relative to SRM for targeted proteomics. In their "Fourier Transform-All Reaction Monitoring" (FT-ARM) method performed on a QLT-Orbitrap MS, wide swaths of mass-to-charge space (i.e., 100 Th) were accumulated and isolated in the QLT using broadband waveforms, subjected in bulk to dissociation, and simultaneously mass analyzed to exploit the HR/AM properties of the Orbitrap. The authors noted the ability to quantify peptides with FT-ARM, using the product ions generated from in-bulk dissociation, with modest sensitivity, reproducibility, and precision despite quite significant interferences from other ions present in the 100 Th isolation swath. This further corroborates that HR/AM mass analysis enables real, quantitative information to be extracted from highly interference-riddled measurements with little upfront assay development or optimization. Since PRM involves significantly less co-isolated interference and uses a quadrupole mass filter-equipped platform better suited to accumulation and isolation of large quantities of ions (to improve sensitivity), these findings further support our conclusions that PRM yields high quality quantitative measurements, comparable to QqQ SRM, while simplifying method development.
Looking forward, the PRM paradigm also enables new modes of analysis that are currently unavailable on QqQ platforms. Because of the modular nature of hybrid HR/AM MS, these platforms provide an unprecedented amount of experimental flexibility. Possessing the capabilities of both high performance quantitative and high-throughput discovery proteomics instruments, one can envision mixed mode analysis types in which targeted and discovery experiments are performed simultaneously. For example, when not engaged in profiling an eluting target species, the instrument could be directed to perform conventional data-dependent or data-independent MS/MS, thereby maximizing instrument time, sample usage, and data density. Additionally, because of the high specificity of ions measured at high mass accuracy and consequently the reduced search space for potentially matching peptides, intelligent by guest on May 6, 2020 data acquisition strategies are possible that enable on-the-fly database searching, spectral matching, or spectral scoring (like that used here post-acquisition). Such strategies would enable the mass spectrometer to make decisions during acquisition, such as assessing whether a target peptide had been adequately identified, changing experimental parameters (CAD energies, isolation width, injection times, etc.) to improve performance for a particular target peptide, or determining its progress in a chromatographic run to make dynamic modifications to target peptide scheduling (67,68).