Evaluation of acquisition modes for semi‐quantitative analysis by targeted and untargeted mass spectrometry

Rationale Analyte quantitation by mass spectrometry underpins a diverse range of scientific endeavors. The fast‐growing field of mass spectrometer development has resulted in several targeted and untargeted acquisition modes suitable for these applications. By characterizing the acquisition methods available on an ion mobility (IM)‐enabled orthogonal acceleration time‐of‐flight (oa‐ToF) instrument, the optimum modes for analyte semi‐quantitation can be deduced. Methods Serial dilutions of commercial metabolite, peptide, or cross‐linked peptide analytes were prepared in matrices of human urine or Escherichia coli digest. Each analyte dilution was introduced into an IM separation‐enabled oa‐ToF mass spectrometer by reversed‐phase liquid chromatography and electrospray ionization. Data were acquired for each sample in duplicate using nine different acquisition modes, including four IM‐enabled acquisitions modes, available on the mass spectrometer. Results Five (metabolite) or seven (peptide/cross‐linked peptide) point calibration curves were prepared for analytes across each of the acquisition modes. A nonlinear response was observed at high concentrations for some modes, attributed to saturation effects. Two correction methods, one MS1 isotope‐correction and one MS2 ion intensity‐correction, were applied to address this observation, resulting in an up to twofold increase in dynamic range. By averaging the semi‐quantitative results across analyte classes, two parameters, linear dynamic range (LDR) and lower limit of quantification (LLOQ), were determined to evaluate each mode. Conclusion A comparison of the acquisition modes revealed that data‐independent acquisition and parallel reaction monitoring methods are most robust for semi‐quantitation when considering achievable LDR and LLOQ. IM‐enabled modes exhibited sensitivity increases, but a simultaneous reduction in dynamic range required correction methods to recover. These findings will assist users in identifying the optimum acquisition mode for their analyte quantitation needs, supporting a diverse range of applications and providing guidance for future acquisition mode developments.

Conclusion: A comparison of the acquisition modes revealed that data-independent acquisition and parallel reaction monitoring methods are most robust for semiquantitation when considering achievable LDR and LLOQ. IM-enabled modes exhibited sensitivity increases, but a simultaneous reduction in dynamic range required correction methods to recover. These findings will assist users in identifying the optimum acquisition mode for their analyte quantitation needs, supporting a diverse range of applications and providing guidance for future acquisition mode developments.

| INTRODUCTION
Mass spectrometry (MS) is a powerful technique, which has developed to the point of being near-indispensable in several scientific environments. Every day MS is contributing to fundamental research in understanding the molecules of life, from metabolites and lipids, through to proteins and their complexes. 1 Furthermore, MS plays a vital role in a diverse range of industrial workflows, including quality control, drug detection, and clinical biomarker analysis. 2,3 It is the depth and breadth of these applications which make MS such a vital analytical tool. For several decades, qualitative MS analysis has been the key strand that cultivated these endeavors, because of the unique ability of MS to identify and characterize a wide array of molecules. MS is able to complete these discovery tasks with a high level of consistency, with an ever-growing number of technical improvements. In addition to the qualitative benefits of MS, there are now many workflows that focus not only on molecular characterization but also on quantitation. 4 Quantitative MS is able to measure the amount of each molecule in a sample and has therefore found favor across applications such as proteomics and pharmaceutical analysis. 5,6 However, quantitative high-resolution MS is still challenging to perform, not least because of the numerous technical considerations that must be considered.
One technical challenge that must be considered when performing a quantitative high-resolution MS workflow is selection of the quantitation method. Quantitative MS approaches can be broadly packaged into two distinct families, the first of which is termed the labeling approaches. This set of techniques involves quantitation using popular isotopic species, which comprises techniques such as tandem mass tags (TMT), stable isotope labeling by amino acids (SILAC), and isotopically labeled standards. [7][8][9][10][11] This suite of quantitation methods has been developed to the point that robust and reproducible workflows exist, and quantitation using these methods is generally considered to be highly accurate when performed correctly. There are, however, also limitations to the labeling techniques: they are expensive, can suffer from incomplete labeling efficiencies, and are limited in the number of samples that can be analyzed in parallel within a single experiment. These challenges in quantitation using labeling approaches have given rise to the second family of techniques, termed label-free quantitation. [12][13][14] In the case of the label-free methods no isotopic labels are required; instead the amount of a given analyte is determined by either chromatographic peak area or spectral counting. Compared to labeling approaches, label-free quantitation generally requires more straightforward sample preparation and can perform comparative analysis of a greater number of samples within a single experiment. However, given the reliance on peak area or spectral counting within label-free quantitation workflows, considerable care is required in selecting the sample preparation methodology and instrumental conditions for analysis.
Selection of a suitable acquisition mode is one important instrument consideration required for successful label-free quantitation. 15, 16 The importance of this requirement has been highlighted for several analyte classes by a number of excellent review articles. [4][5][6]17 The fast expanding field of mass spectrometer development has resulted in the existence of several such acquisition modes, each with their own set of strengths and weaknesses. These acquisition modes include both targeted methods, in which species of interest are predefined based on previous MS characterization, and untargeted acquisition modes, in which species are not defined in advance. Recent work has evaluated the qualitative and quantitative performance of these acquisition methods for the analysis of veterinary drug reference substances and histone posttranslational modifications on Orbitrap instrumentation. 18,19 The approach presented here parallels this recent work by evaluating semiquantitative performance of acquisition modes on an alternative popular class of instrumentation, an IM-enabled quadrupole orthogonal acceleration time-of-flight (oa-ToF) mass spectrometer.
Nine distinct acquisition modes available on this instrument, illustrated in Figure 1, were compared. These acquisition modes include the following: (a) a screening MS1-only method (MS); (b) the targeted parallel reaction monitoring (PRM) mode known as ToFMRM; (c) an untargeted method for data-dependent acquisition (DDA); and (d) two untargeted data-independent acquisition (DIA) methods, one being MS E , a broadband DIA method, and the other SONAR, a scanning quadrupole DIA mode. Furthermore, as the oa-ToF mass spectrometer used in this study contains an IM cell, these acquisition modes (with the exception of SONAR) were also assessed with IM enabled. These modes are denoted using HD in their acronym, known as HDMS, HDMRM, HDDDA, and HDMS E , respectively. Ion mobility-mass spectrometry (IM-MS) has previously been successful in improving the qualitative analysis of several analyte classes, including small molecules, tryptic peptides, protein complexes, and, for a handful of cases, on cross-linked peptides. [20][21][22][23][24][25] For readers requiring a more detailed description of the principles of each acquisition method, this information is provided in Supplementary Note 1. By evaluating these acquisition methods for two common analyte classes, metabolites and peptides, this study will support readers in selection of the optimum acquisition mode for their own label-free semi-quantitative analysis. Furthermore, a smallscale pilot analysis of cross-linked peptides using the most favorable acquisition modes for semi-quantitation suggests that the findings would also be applicable to more challenging analyte classes. Given the broad interest in quantitative and semi-quantitative analysis across the field of MS, these findings would be applicable to a range of users, from those in fundamental research through to clinical and other industrial applications.

| Metabolite sample preparation
The following eight metabolite standards were purchased and prepared at a concentration of 100 ng/μL in methanol: AKB-48 Apinaca 5-Hydroxypentyl metabolite, AKB-48 Apinaca 5-Hydroxypentyl metabolite-D 4 , JWH-073 3-Hydroxybutyl metabolite, F I G U R E 1 Schematic showing the processes taking place within each portion of the mass spectrometer for acquisition modes available on Synapt G2-Si mass spectrometer. The functions in each region are as follows: (i) the quadrupole can be used in three ways, allowing all ions to pass through (MS, HDMS, MS E , HDMS E ), performing precursor ion selection based on intensity (DDA, HDDDA) or a predefined m/z (ToFMRM, HDMRM), or as a scanning quadrupole (SONAR); (ii) CID fragmentation is applied in the trap for a subset of the acquisition modes (DDA, HDDDA, ToFMRM, HDMRM); (iii) for IM-enabled modes mobility separation is performed next either on precursor ions (HDMS, HDMS E ) or CID fragments (HDDDA, HDMRM); (iv) CID fragmentation can be performed in the transfer as an alternative to the trap; this is the case for the high energy MS2 experiment that makes up part of the MS E and HDMS E acquisition modes. The outcome of these steps is that PRM acquisition methods (TofMRM and HDMRM) provide MS2 data only, and MS modes (MS and HDMS) MS1 level data. Broadband DIA modes (MS E and HDMS E ), scanning quadrupole DIA (SONAR), and DDA provide both MS1 and MS2 data. Precursor/product ion color (orange/blue) denotes the ion m/z; and the precursor/product ion shape denotes collision cross section (

| Cross-linked peptide sample preparation
Cross-linking reactions were conducted as described previously. 24 In brief, 0.3 mg/mL BSA (Sigma-Aldrich) and 1 mg bis

| Liquid chromatography
For metabolite analysis, an I-class LC system (Waters Corporation) was equipped with an HSS T3 1.8 μm 2.1 Â 100 mm column (Waters Corporation) operated at 400 μL/min. The gradient was held at 1% B for 0.3 min, followed by a 1%-50% increase in B from 0.3 to 7 min, and another step from 50% to 70% B in 1 min. Next, the solvent strength was increased to 99% B in 0.1 min, which was held for 1 min, and the column reconditioned for 1 min at initial gradient conditions.
Mobile phase A was 0.1% formic acid in water and mobile phase B 0.1% formic acid in acetonitrile. The column temperature was maintained at 45 C and the samples at 8 C. The injection volume equaled 5 μL.
All peptide separations, both linear and cross-linked, were conducted with a 1.7 μm CSH 130 C18 300 μm Â 100 mm column (Waters Corporation) operated at 7 μL/min using an M-class LC system (Waters Corporation). Here, the gradient was held first at 1% B for 2 min, followed by a 1%-30% increase from 2 to 30 min, which was held for 2 min. Next, the solvent strength was increased to 85% B in 1 min, which was held for another 2 min, followed by decreasing the solvent strength in 1 min and reconditioning of the column for 23 min at initial gradient concentration. Mobile phase A was 0.1% formic acid in water and mobile phase B 0.1% formic acid in acetonitrile. The column temperature was maintained at 55 C and the samples at 12 C. The injection volume equaled 4.5 μL.  Tables S1 and S2 (supporting   information).

| Mass spectrometry
For the IM-enabled acquisition methods, the Trap and Transfer T-Waves were pressurized with 2 mL/min of Ar. Gas-phase optimization for the separation of the analytes made use of N 2 . The He gate contained within the IM region was pressurized with 180 mL/min. The IM T-Wave was pressurized with 90 mL/min, the IM wave velocity was ramped, as described in Tables S1 and S2 (supporting information), and the pulse height held at 40 V during acquisition.
Parameters used for extracting data in the IM domain are detailed in Table S3 (supporting information).

| Data processing
Peak detection was carried out in Skyline and the multidimensional peak detected data exported as transition tables for down-stream analysis. [27][28][29] Explicit retention times were specified to aid peak detection and validate the integration. Match tolerances were ±0.1 min and ±0.2 min for the metabolite and (XL) peptide data sets, respectively. A resolution of 20 000 FWHM dimension was specified, from which tolerances are inferred that equal twice the expected peak FWHM in the m/z dimension, to extract chromatograms. Observed RMSE mass errors, averaged out over all concentration levels and analytes, equaled 4.1 (MS1 methods) and 3.3 (MS2 method), 2.8 (MS1 methods) and 2.6 (MS2 methods), and 6.4 (MS2 methods) ppm, respectively, for the metabolite, peptide, and XL peptides data sets.
To correctly describe whether signal is proportional to the concentration of analyte, modeling of the data is required to consider the effects of noise at low concentration and signal saturation at high concentration. Saturation effects for different mass analyzer types are not uncommon and described elsewhere. 30,31 Briefly, saturation effects are readily recognized by comparing the isotopic distributions of the highest intensity quartile detections with expected theoretical isotopic distributions. Intensity, both low and high, and isotopic (delta) mass shift errors can both be indicative of detection, that is, saturation, anomalies. This is typical in LC-MS analyses of biological samples where analytes differ greatly in concentration/amount and can be corrected for using various approaches, either post-acquisition or during the experiment. Here, two post-acquisition correction methods are applied and evaluated. As shown in Figure 2A, the data are modeled using exponential fits of type y = C(1-exp(Àx/R)) to detect the central linear range, with inflection points where the effects of saturation or noise become important.
Using the model, conceptually, two parameters of importance were determined in linear space: i. Lower limit of quantification (LLOQ) is recovered by finding the point at which the measured analyte intensity is equivalent to the level of background noise and one standard deviation as determined from multiple replicates and samples.
ii. Linear dynamic range (LDR), highlighted by the green arrow in Figure 2A, is determined as the ratio difference between the estimated LLOQ and the estimated upper limit of quantification (ULOQ). In this case the ULOQ is defined as the inflection point of the measure intensity curve, above which saturation is apparent and signal no longer scales linearly with concentration.

| Data correction methods
Following on from data processing, two post-acquisition data correction methods were evaluated in an attempt to mitigate the known effects of mass analyzer saturation. 30,31 The correction methods are as follows: i. MS1 isotope correction-at the level where saturation is observed, the abundance of a saturated isotopic peak for a given species is corrected using the abundances of higher isotopes and their theoretical natural distributions. 32 ii. MS2 correction-at the level where saturation is observed, the abundance of the saturated MS1 isotopes for a given species is corrected using the abundance of the non-saturated precursor/ product ion relationships after collision-induced dissociation (CID) fragmentation for a given analyte. [33][34][35] Principles of the two correction methods are visually detailed in Figures 2B and 2C, respectively.

| Model development and corrections
Analyte mixtures were prepared to reflect two common sample types subjected to semi-quantitative MS analysis, namely metabolites and peptides. Accepted reversed-phase separation materials and methods were applied for both analyte classes. 36,37 The analyte mixtures comprised an eight-metabolite mixture diluted into urine matrix, and a four-protein tryptic digest mixture diluted into an E. coli digest background. Five (metabolites) or seven (peptide) serial dilutions of these samples were then analyzed in duplicate using each of the nine acquisition modes, detailed in Figure 1. Duplicate injection results were typically well within a 10% error of each other, suggesting sufficient precision between samples at the individual concentration levels; hence, precision limitations are not expected to affect generalization in any subsequent regression analysis. For example, the median MS1 abundance errors, averaged over all modes of acquisition and on-column levels, were found to equal to 3.7% and 8.3% for the metabolite and peptide standards, respectively. These figures of merit compare favorably with studies where peak area reproducibility values were reported ranging from 20% to 30% for the majority of the detected features within a metabolomics study pool QC sample (n = 17) and ranging from 5% to 15% for peptide standards spiked into a biological matrix (n = 5) using DIA methods of acquisition. 37,38 Example raw MS and MS/MS data for the different acquisition modes are shown in Figure S1 (supporting information), where the ability to conduct MS1, MS2, or combined semi-quantitation for each mode is also illustrated. Based on these data, multi-point calibration curves were created for each individual analyte for each of the nine acquisition modes, using MS1, MS2, and combined MS1/MS2 semiquantitation as relevant to the particular mode. It was determined from these serial dilution data that the calibration curve features were consistent across analytes and could therefore be modeled with the fit shown in Figure 2A. This model of the data also allowed two parameters, LLOQ and LDR, to be defined, providing metrics for assessing performance of the acquisition modes.
Interpretation of the serial dilution data and the semi-quantitative results revealed nonlinear response at high sample concentration in the case of some acquisition modes and analytes, attributed to saturation. To mitigate these effects and improve quantitative readout of the underlying data, two correction methods were applied.
It was expected that these corrections would increase linearity at higher sample concentrations, extending LDR for quantitation. The first correction, an isotope correction method, corrects the intensity of a saturated monoisotopic peak using the abundances of higher isotopes and their theoretical natural distributions. 32 The principles of this correction are detailed in Figure 2B. When considered for the F I G U R E 2 (A) The model used to characterize the calibration curves obtained from analyte quantitation in this study. A sigmoidal shape (red) is fitted to the data considering the effect of noise dominating over signal at low concentrations (orange), and detector saturation at high concentrations. Based on this fit it is possible to define the parameters of interest, LLOQ (solid blue line), ULOQ (dashed blue line), and LDR (green arrow). (B) and (C) show the principles of the isotope and MS2 correction methods, respectively, applied to combat saturation effects at high concentrations. The isotope correct method (B) recognizes a difference in the observed and theoretical ratios of isotope peaks, and a corrected intensity (Icorr) is determined by scaling the experimental value of the first isotope (I 1 exp) by the ratio of the theoretical monoisotopic peak (I 0 th) and first isotope (I 1 th). The result of this approach for one example peptide (NLAENISR) is compared for corrected (orange) vs. uncorrected (blue) data at the bottom of the panel. Equivalent data for the MS2 correction method are shown in (C). In this case, saturation is detected through an observed drop in the MS1/MS2 intensity ratio, and the Icorr is calculated using the sum of the MS2 peak intensities. In the case of both corrections, the effect is clear on the higher concentration scale where saturation effects occur, and the uncorrected and corrected curves diverge [Color figure can be viewed at wileyonlinelibrary.com] example doubly charged peptide NLAENSIR, data shown in the bottom panel of Figure 2B, the isotope correction method proved successful in increasing linear response about twofold. This increased linearity of the data at high concentration results in an improvement in the upper limit of quantitation (ULOQ) and thus LDR. The alternative correction method applied was an MS2-based correction, shown in Figure 2C, in which the abundance of the saturated species is corrected using the abundance of the non-saturated precursor/ product ion relationships after CID fragmentation for a given analyte. [33][34][35] This MS2 correction method improves linear response and LDR for peptide NLAENSIR to a similar but slightly smaller order of magnitude compared to the isotope correction method, as shown in bottom panels of Figure 2C. Analogous improvements are observed for metabolites, highlighted in Figure S2 (supporting information), with application of isotope and MS2 corrections to AKB-48 Apinaca 5-Hydroxypenytl metabolite showing a 7.1% and 8.4% LDR increase, respectively, compared to uncorrected data.
Dynamic range improvements obtained through applying these correction methods are particularly striking for the IM-enabled acquisition modes, shown in Table 1 (also depicted visually in Figure S3 [supporting information]). Averaging across both analyte types the net gain in LDR for these modes is equal to $43%, largely T A B L E 1 The semi-quantitative uncorrected and corrected summary figures of merit (average values summed over all analytes with acquisition, integration, and/or computational outliers excluded from the analysis when passing a modified [Iglewicz and Hoaglin] z-score threshold; errors represent difference in analyte response/ionization efficiency) for all acquisition methods are presented in the following tables for each analyte type

| Evaluation of acquisition modes
To evaluate the suitability of acquisition modes for semi-quantitative  should be noted that the PRM methods like ToFMRM require more knowledge of the system being studied and are more user intensive to set up. 40 To emphasize this particular aspect, providing further increased sensitivity and dynamic range, two dedicated TofMRM methods, both aimed at improving instrument duty cycle for a selected m/z value, a set of values, or range, were applied for the analysis of the peptide samples. These methods, TofMRM EDC and TofMRMsens, are detailed in Supplementary Note 1, and the results are described in Table S4 (supporting information). Moreover, aspects that are typically more associated with multiple reaction monitoring (MRM) analysis, such as interscan delay times and dwell time, affecting duty cycle and thus the number of points per peak, must also be considered. 41 Interestingly, the semi-quantification results for these methods suggest that high-resolution mass analyzers approach LDR and LLOQ levels normally obtained with MRM experiments conducted on tandem quadrupole instrumentation, which is seen as state-of-the-art. The main limitation appears to be the ability to discriminate signal from noise for a given acquisition method.
Comparison of the semi-quantitative performance of nonmobility acquisition modes with their IM-enabled counterparts was also carried out using the LLOQ and LDR metrics shown in Table 1.
Unsurprisingly, all IM-enabled modes exhibited a decreased LDR compared to their non-mobility counterparts, which can be partially mitigated by application of correction methods as previously discussed. The IM-enabled HDMS E and HDMRM modes show comparable LLOQ values relative to their non-mobility equivalents.
However, relative values vary between analyte types, attributed to differences in experimental conditions, which highlights that the metrics described are purely estimates to aid in assessing relative performance of the acquisition methods. IM-enabled modes do show an improved signal-to-noise ratio compared to their counterparts, likely because of the reduction in noise afforded by IM separation.   presented in this study to address this known issue, isotope correction and MS2 correction, were both found to assist in an up to twofold increase in the LDR for semi-quantitation. This increase will be important for applications working at the higher concentrations or over a wide concentration range, where users will benefit from applying these corrections. Selecting the correction method should be considered carefully; however, as while both are useful, the optimum correction is analyte specific and will therefore vary by application.
For any given quantitative MS experiment, a number of factors will need to be considered when selecting the acquisition mode for a study, specifically LLOQ and LDR, which were determined for each acquisition mode as part of this work. These values were determined to be agnostic to analyte type; therefore, the findings herein will be applicable to a number of future studies. Based on these metrics, DIA modes such as MS E and SONAR, along with PRM mode ToFMRM, were the best performing for label-free semi-quantitation by MS. This is consistent with findings from the wider community, in which these modes are the most commonly selected for quantitative and semiquantitative MS. IM-enabled modes also showed benefits for semiquantitative analysis within these framework metrics, particularly at low concentrations, because of their ability to improve the signal-tonoise ratio. Although there is some reduction in LDR, which can be repaired by applying the corrections discussed earlier, which may make use of IM-enabled modes challenging for working at the higher end of the concentration range. In addition to considering these metrics, when selecting an acquisition mode for quantitation, other factors must come into play. For example, instrument capabilities and throughput/speed of analysis may alter which modes can be practically incorporated into a workflow. Furthermore, limited knowledge of the analytes would require the user to select the DIA methods over the PRM workflow.
In addition to evaluating acquisition modes for semi-quantitation across two popular analyte types, this study applied selected semi-

PEER REVIEW
The peer review history for this article is available at https://publons. com/publon/10.1002/rcm.9308.