Article pubs.acs.org/jpr
Measuring and Managing Ratio Compression for Accurate iTRAQ/ TMT Quantification Mikhail M. Savitski,†,* Toby Mathieson,† Nico Zinn,† Gavain Sweetman,† Carola Doce,† Isabelle Becher,† Fiona Pachl,‡ Bernhard Kuster,‡,§ and Marcus Bantscheff†,* †
Cellzome GmbH, Meyerhofstrasse 1, 69117 Heidelberg, Germany Chair of Proteomics and Bioanalytics, Technische Universität München, Emil Erlenmeyer Forum 5, 85354 Freising, Germany § Center for Integrated Protein Sciences Munich (CIPSM), Butenandtstrasse 5-13, 81377 Munich, Germany ‡
S Supporting Information *
ABSTRACT: Isobaric mass tagging (e.g., TMT and iTRAQ) is a precise and sensitive multiplexed peptide/protein quantification technique in mass spectrometry. However, accurate quantification of complex proteomic samples is impaired by cofragmentation of peptides, leading to systematic underestimation of quantitative ratios. Label-free quantification strategies do not suffer from such an accuracy bias but cannot be multiplexed and are less precise. Here, we compared protein quantification results obtained with these methods for a chemoproteomic competition binding experiment and evaluated the utility of measures of spectrum purity in survey spectra for estimating the impact of cofragmentation on measured TMT-ratios. While applying stringent interference filters enables substantially more accurate TMT quantification, this came at the expense of 30%−60% fewer proteins quantified. We devised an algorithm that corrects experimental TMT ratios on the basis of determined peptide interference levels. The quantification accuracy achieved with this correction was comparable to that obtained with stringent spectrum filters but limited the loss in coverage to 20 were used for quantification. (C) As for A but only peptides with S2I values below 0.9 were used for quantification. (D) Label-free (LF) experiment compared to the1D-TMT experiment; only peptides with S2I > 0.9 were used for quantification. (E) As for D but only peptides with S2I > 0.9 and P2T > 20 were used for quantification. (F) As for D but only peptides with S2I values below 0.9 were used for quantification. The blue line in parts A−F represents the median protein log2 fold change trend. (G) The protein log2 fold change trend lines from parts A−C plotted together with theoretical trend lines, indicating the deviation of measured protein log2 fold changes at a given percentage of interference (dotted lines). (H) As for part G using trend lines from parts D−F.
chemical labeling step in the TMT approach. For the 1D-TMT experiment (270 min gradient), an inclusion list-based targeted data acquisition approach was used.45 This method is commonly applied in order to maximize the coverage of those proteins in the chemoproteomic assay which specifically bind to the matrix and are likely to be competed with by free (HDAC) inhibitors.10,45 A systematic deviation in protein fold changes was observed when 1D-TMT experimental data were compared with those from 2D-TMT experiments indicating a higher degree of “ratio compression” in the 1D-TMT experiment (SI Figure S2).
labeled sample (80%) was separated into 17 fractions using reversed-phase chromatography at pH 12 prior to LC-MS/MS analysis with data-dependent acquisition (2D-TMT sample, Figure 1); an additional technical replicate of the fractionated sample was also acquired. In the label-free experiments, approximately 500 proteins were identified in the six 130 min LC-MS/MS runs with at least four different peptides (replicate 1, 555 proteins, replicate 2, 481 proteins). This stringent threshold was applied for protein quantification in order to minimize the impact of precision on the accuracy achieved for relative protein quantification. Comparable thresholds were applied to the 2D-TMT samples that enabled quantification of approximately 1400 proteins in 17 runs of 130 min (replicate 1, 1350 proteins; replicate 2, 1498 proteins). Both methods showed a high degree of reproducibility of quantification results, each with comparable standard deviations (LF, 0.21; 2D-TMT, 0.18; Supporting Information (SI) Figure S1), indicating that the lower precision expected for a label-free approach in comparison to isobaric mass tagging is at least partially compensated by variation introduced by the
Comparison of Raw TMT and Label-Free Quantification Results
Next, we compared the relative quantification results achieved with the label-free and the isobaric tagging approaches. The comparison of log2 fold changes (compound treated vs vehicle control) of proteins quantified in both the 2D-TMT and the label-free experiments indicates good agreement for proteins that show little or no change when the pull-downs were 3591
dx.doi.org/10.1021/pr400098r | J. Proteome Res. 2013, 12, 3586−3598
Journal of Proteome Research
Article
performed in the presence of different concentrations of free inhibitor. However, an increasing discrepancy in quantification results between the two methods was observed for higher fold changes (Figure 2A). This trend was markedly stronger when LF data were compared with the 1D-TMT experiment (Figure 2B). These observations are in agreement with the assumption that the detected differences in label-free and TMT quantification results are due to interference caused by cofragmented peptides in TMT experiments. Since the vast majority of proteins detected in these experiments do not change, cofragmentation causes smaller absolute fold changes.19,38,40 This effect is expected to be more pronounced the higher the complexity of the sample and, conversely, less pronounced for samples that were fractionated prior to LC-MS analysis.38 In order to estimate the degree of interference in the 2D- and 1D-TMT samples, we compared the trend lines (moving average approach) determined from the abovedescribed scatter plots (Figure 2A, B) to simulations calculated for different levels of interference (Figure 2C). For the 2DTMT data the trend line shows a good agreement with the simulation, and a modest average interference of 10% was estimated. In contrast, for the 1D-TMT experiment the trend line indicates a significantly higher interference varying between 20 and 40%.
sample was separated using a 2D-chromatographic strategy. Similarly, 45% of the PSMs in the 1D-TMT experiment had a P2T value of below 10, whereas only 24% of the PSMs in the 2D-TMT experiments had such low P2T values (Figure 3D). Thus, on the global sample level, S2I and P2T measures correctly reflected where most ratio compression was to be expected. Next, we investigated the effect of S2I- and P2T-based spectrum filters on the accuracy of protein quantification. Experimental data of the TMT experiments were filtered according to the following criteria: (1) only PSMs with S2I > 0.9; (2) only PSMs with S2I > 0.9 and P2T > 20; (3) only PSMs with S2I < 0.9. Protein quantification results obtained upon application of these filters were compared to those obtained with the label-free approach (Figure 4). For the 2DTMT experiments, only a relatively small fraction of PSMs had strong interfering ions (Figure 3C); consequently, filters 1 and 2 had a moderate, but significant impact (Figure 4A, B). For both filters, comparison to label-free data suggested that interference was reduced from 10% (unfiltered data) to 7% (Figure 4G). However, if protein quantification was solely based on PSMs with low S2I (Figure 4C), a very pronounced shift from the label-free data was observed with an estimated interference of more than 20% (Figure 4G). The applied spectrum filters had more pronounced effects on TMT quantification when the chemoproteomic sample was only separated via 1-dimensional chromatography. Filter 1 led to a reduction of average interference on the protein level from approximately 25% to about 15% (Figure 4D, H). In contrast to the 2D-TMT data, filter 2 further improved the agreement with the label-free data by reducing the interference to below 10%, and the resulting trend line fit much better to the simulation (Figure 4E, H). This underscores that the systematic deviation of TMT quantification data from label-free data is better explained by considering spectrum interference and P2T levels in combination. This effect is more pronounced for samples where a lot of peptides were fragmented at abundances close to the threshold applied by the instrument acquisition software, such as in the 1D-TMT experiment (Figure 3D). In such experiments, the additional P2T measure makes it possible to reduce the impact of “hidden interference” to protein quantification that is caused by cofragmentation of peptides not visible in the survey spectrum. In agreement with the S2I distribution (Figure 3C), the subset of PSMs with S2I < 0.9 yielded a stronger disagreement from the label-free data in the 1D-TMT experiment (40% estimated interference, Figure 4F, H) than in the 2D-TMT experiments of the same sample. This again emphasizes the impact of sample complexity in individual chromatographic runs on the quantification accuracy with isobaric labels.38 While filters 1 and 2 improved quantification accuracy in 1Dand 2D-TMT experiments, the application of these stringent spectrum filters resulted in substantially fewer proteins being quantified. For example, filter 1 led to a 30% reduction in quantified proteins for the 2D-TMT experiments and to a more than 60% reduction in quantified proteins for the 1D-TMT experiment (SI Table S1).
Spectrum Filter Approach for Improved Quantification Accuracy
We hypothesized that the quantification accuracy of the TMT samples would be improved if those spectra for which a high level of cofragmentation was observed were excluded from quantification. Interference levels for each peptide-to-spectrum match (PSM) were estimated using two measures: the signalto-interference (S2I) ratio and the precursor-to-threshold ratio (P2T).40,45,51 S2I is a measure of the signal-to-background ratio or the “purity” of a spectrum and is estimated by dividing the summed-up abundance of the isotopologues of the target ion by the total ion abundance obtained within the isolation width applied (Figure 3A). P2T is a measure of the signal-to-noise ratio observed for each precursor ion and is determined by dividing the abundance of the target isotope cluster by the noise threshold applied by the instrument control software (Figure 3B). When precursor ions are detected with a low P2T ratio, the proportion of interfering ions within the applied isolation window cannot be accurately determined because a significant portion might be hidden below the noise threshold (hidden interference). Hence, the ability of the S2I measure to estimate the interference in MS2 spectra is impaired by the noise cutoff in the survey spectrum. Conversely, for ion signals with high P2T, one would expect the S2I measure to more accurately determine the degree of interference. In previous reports, the ability of measures for spectrum purity to estimate the degree of cofragmentation has been questioned,52,55 and we hypothesized that considering the P2T ratios as an additional parameter might improve the situation. In the 1D-TMT experiment performed with the chemoproteomics sample, we found that only 37% of the PSMs had S2I values of above 0.9, suggesting that for more than 60% of all spectra the interference levels due to coeluting peptides were greater than 10% (Figure 3C). In the 2D-TMT experiments, however, on average 71% of the PSMs had S2I values above 0.9. These findings are in agreement with the results obtained by comparing label-free with TMT quantification, suggesting much lower interference levels when the chemoproteomic
S2I-Based Correction of TMT Ratios
In a next step we devised an S2I-based fold change correction algorithm to improve quantification accuracy with less impact on protein coverage. The algorithm was based on two main assumptions: (A) The relative signal abundances measured for 3592
dx.doi.org/10.1021/pr400098r | J. Proteome Res. 2013, 12, 3586−3598
Journal of Proteome Research
Article
Figure 5. Comparison of TMT and label-free protein quantification using log2 protein fold changes calculated for each of the five inhibitor doses compared against the vehicle (i.e., 5 log2 fold changes calculated for each protein). (A) Label-free, LF, experiment compared to the 2D-TMT experiment; only peptides with S2I values greater than 0.5 and P2T greater than 4 used for quantification. (B) As in part A, but S2I-based correction is performed on this data. (C) The protein log2 fold change trend lines from A and B plotted together with theoretical trend lines indicating the deviation of measured protein log2 fold changes at a given percentage of interference. (D) Label-free, LF, experiment compared to the 1D-TMT experiment; only peptides with S2I values greater than 0.5 and P2T greater than 4 used for quantification. (E) As in part D, but S2I-based correction is performed on this data. (F) The protein log2 fold change trend lines from parts D and E plotted together with theoretical trend lines which represent the deviation of measured protein log2 fold changes given a fixed percentage of interference.
which the interference likely dominates the observed reporter ion pattern. This soft filter (S2I > 0.5 and P2T > 4) had only a minor impact on the number of quantified proteins per experiment (less than 10% in 2D-TMT experiments; see SI Table S1) and had a negligible effect on the overall interference measured in the 2D- and 1D-TMT experiments (see Figures 2 and 5). The correction algorithm reduced the average interference observed in the 2D-TMT experiments from 10% to 5% (Figure 5A−C). Notably, the correction procedure performed slightly better than the stringent spectrum filter approach described above (Figure 4). An even more pronounced improvement was observed for the 1D-TMT experimental data, yielding a reduction in average interference from ≈25% to ≈15% (Figure 5D−F). Again, this improvement is comparable to what was achieved with the stringent filtering approach. It should be noted that the S2I-based fold change correction approach led to a slight decrease in the precision of protein fold changes; for the 1D experiment the calculated 95% protein confidence intervals40 increased by 16% from 0.12 to 0.14.
the precursor ions and the interfering ions in the survey spectra are predictive for the relative amounts of reporter ions generated for these ions. Hence, the S2I measure would quantitatively reflect the ratio compression effect. (B) In relative quantification experiments, the majority of proteins and peptides do not change significantly per condition. Therefore, the interfering ions likely shift observed fold changes toward the median fold change observed for each condition. This is a reasonable assumption for many proteomics experiments such as the one in this study. Of the 6,750 fold changes calculated for the 1350 quantified proteins in the 2D-TMT experiments, only 83 (1.2%) showed a reduction in matrix binding greater than 2fold. Consequently, the corrected relative fold change of a peptide in the condition labeled with TMT126 would be represented by FCcorr(126) = Aprec_126/Aprec_131, where Aprec_126 and Aprec_131 are the interference corrected abundances of the 126 and the 131 reporter ions calculated as described in the method section. Previous reports have indicated a lack of correlation between measures of signal interference on the MS1 level and observed cofragmentation in MS2 spectra which might be explained by cofragmentation of peptides not visible in preceding MS1 spectra, differences in reporter ion yield for individual peptides ions at any given collision energy,52,55 and shortcomings in measuring interference from survey spectra.40 These previous findings already indicate that an exact correction for each PSM can likely not be achieved. At the protein level, however, we expected that the individual errors in correction of peptide fold changes would cancel upon aggregation and produce more accurate protein fold changes. Further, we decided not to attempt correction on PSMs for
Evaluation of Filtering and Correction on a Spike-in Sample
In order to estimate the accuracy of TMT quantification with an independent approach, YOL086C (yeast alcohol dehydrogenase) and HBA1 (alpha 1 hemoglobin) were labeled with TMT reagents, combined to achieve the following proportions, TMT126:0, TMT127:1, TMT128:2, TMT129:4, TMT130:8, and TMT131:1, and spiked into another chemoproteomic sample using mixed kinase inhibitor beads.9 The TMT130/ TMT127, TMT130/TMT128, and TMT130/TMT129 fold changes were calculated (Figure 6) based on a total of 64 3593
dx.doi.org/10.1021/pr400098r | J. Proteome Res. 2013, 12, 3586−3598
Journal of Proteome Research
Article
bottom of the competitive binding curve. We have shown above that the fold changes measured in the label-free and TMT-based quantitative experiments were very similar if multidimensional separation and either stringent S2I/P2T filters or an S2I-based correction algorithm were employed. In TMT samples for which peptide separation was performed using only a single chromatographic dimension, the fold change correction algorithm provided a significant improvement of the cofragmentation induced ratio compression effects; however, estimated interference levels were still above those obtained for the 2D-TMT sample (Figure 5). In order to evaluate if the applied correction is sufficient to enable accurate IC 50 determination, we fitted the quantitative data obtained with TMT experiments to competitive binding curves and compared the obtained half-maximal binding values to those obtained in the label-free strategy (Figure 7 and SI Figures S3 and S4). Competitive binding was observed for class I and IIb histone deacetylases as well as for those proteins forming tight protein complexes with these enzymes (see also ref 10). Components of a particular protein complex are expected to yield very similar competitive binding curves. For example, for the CoRest deacetylase complex, 2D-TMT and label-free data demonstrated very good agreement in IC50’s of the interaction partners with both methods (SI Figure S3) with average pIC50 (−log IC50) of 7.71 and standard deviation of 0.04 for 2D-TMT and average pIC50 of 7.75 and standard deviation of 0.10 for experiments quantified with the label-free strategy. More generally, very good agreement was found for IC50 ’s determined from LF and 2D-TMT experiments regardless of whether S2I-based fold change correction was employed. The standard deviation of pIC50 differences was 0.2 in each case (SI Figure S3). The agreement of IC50 values between these different experimental approaches was similar to that obtained from technical replicates performed with either approach. The standard deviation of pIC50’s for 2D-TMT replicates was 0.14 and 0.18 for replicate label-free analyses (SI Figure S4). This suggests that, for chemoproteomics experiments, extensive peptide fractionation can sufficiently reduce peptide interference to allow for accurate IC50 determination based on isobaric mass tagging. Additional application of S2I-based fold change correction algorithms provides a higher dynamic range in quantification but little or no additional benefit for IC50 determination when compared to a label-free strategy. However, the situation is very different if TMT-labeled samples are analyzed in a targeted 1D-LC-MS/MS approach. Compared
Figure 6. Measured log2 fold changes of TMT130/TMT127, TMT130/TMT128, and TMT130/TMT129 for the spiked in YOL086C and HBA1 proteins compared to the true, theoretical log2 fold changes. Trend lines represent the deviation of measured protein log2 fold changes given a fixed percentage of interference. Red circles represent log2 fold changes obtained using all data, purple circles represent log2 fold changes obtained using only peptides with S2I greater than 0.9 for quantification, blue circles represent log2 fold changes obtained using only peptides with S2I less than 0.9 for quantification, and green circles represent log2 fold changes obtained using all peptides and applying S2I-based correction. Note that purple circles are largely covered by green circles.
matching spectra, and results suggested ≈8% interference. Partitioning the PSMs into two groups according to their S2I values [group 1: S2I > 0.9 (30 spectra), group 2: S2I < 0.9 (34 spectra)] yielded markedly different fold changes. Fold changes derived from group 1 had roughly a 2% interference level, while fold changes derived from group 2 peptides exhibited 15% interference. When the S2I-based correction approach was applied (64 quantified spectra), the average interference was low and virtually identical to the S2I > 0.9 filtered data (Figure 6). Similarly to the previous example, the correction algorithm led to an increase in the standard error of the calculated fold changes compared to the uncorrected data by, on average, 30% (SI Table S2). IC50 Determination
The ultimate goal of chemoproteomic competition binding experiments is the determination of competitive binding curves of free inhibitors for those targets that are captured on the matrix (see Figure 1) and, upon curve fitting, determination of binding potencies (IC50’s).56−58 Accurate protein quantification is a prerequisite for determination of correct half-maximal binding concentrations (IC50’s), in particular for defining the
Figure 7. IC50 values determined in chemoproteomics experiments. (A) Comparison of pIC50 (−log IC50) values obtained from the chemoproteomic label-free experiment and from the 1D-TMT experiment. (B) Comparison of pIC50 values obtained from the label-free experiment and from the 1D-TMT experiment after S2I-based fold change correction. Standard deviations in A and B are calculated for the distributions of differences between the label-free and TMT obtained pIC50 values. Dashed blue lines indicate 2-fold deviation from equality, and dashed purple lines indicate 3-fold deviation from equality. 3594
dx.doi.org/10.1021/pr400098r | J. Proteome Res. 2013, 12, 3586−3598
Journal of Proteome Research
Article
Figure 8. Evaluation of the S2I-based fold change correction approach for very complex samples. A human chemoproteomic sample displaying a range of fold changes was analyzed in the presence of different amounts of E. coli tryptic digest. (A) Distribution of protein abundances, determined as the average of the logarithm-transformed MS1 signal abundances of the top three peptides: blue line, human proteins from chemoproteomic spike sample; red line, 0.4 μg E. coli tryptic digest. (B) Comparison of log2 protein fold changes calculated after S2I-based ratio correction of the pure chemoproteomic spike sample (x-axis) and the chemoproteomic sample spiked into a 0.4 μg E. coli 1:1 background: blue line, experimental trend line; black line, equality line. (C) Protein fold change trend lines (log2 transformed) in comparison to the pure chemoproteomic spike sample after S2I correction (x-axis): blue line, pure spike sample without S2I correction; green line, chemoproteomic sample spiked into a 0.4 μg of E. coli 1:1 background, no S2I correction applied; red line, chemoproteomic sample spiked into a 0.4 μg of E. coli 1:1 background, with S2I correction. Dashed lines refer to modeled trend lines at indicated interference levels. (D) Distribution of protein abundances, determined as average of the logarithmtransformed MS1 signal abundances of the top three peptides: blue line, human proteins from chemoproteomic spike sample; red line, 1.6 μg E. coli tryptic digest. (E) As in part B with the exception that the sample was spiked into a 1.6 μg of E. coli 1:1 background. (F) As in part C with the exception that green and red lines refer to the chemoproteomic sample spiked into a 1.6 μg E. coli 1:1 background.
to the label-free strategy, differences in IC50’s of up to 10-fold were observed and the standard deviation of pIC50’s was 0.33 (Figure 7A). Application of the S2I-based fold change correction algorithm substantially improved the agreement with the LF data. The standard deviation was reduced to 0.18, and all data points were within a 3.1-fold window (Figure 7B). Hence, the described fold change correction strategy enabled significantly more accurate determination of inhibitor binding potencies using isobaric mass tag-based quantitative strategies and minimal sample separation, thus providing higher throughput as compared to strategies employing more extensive fractionation.
from the pure chemoproteomic sample, 633 were quantified with at least 4 PSMs and used for comparison of quantification accuracy (SI Table S3). Due to the higher sampling rate of the applied experimental strategy compared to the CID/HCD dual scan approach applied for the experiments described above, a higher fraction of peptides was detected with poor S2I or P2T values, resulting in a 19% reduction of quantified proteins when a S2I-based fold change correction was applied. Less than 10% of detected proteins displayed pronounced log2 fold changes (>1, 10% to approximately 3% (Figure 8). In line with the observations detailed above, we observed only a minor impact of the S2I-based correction approach on the precision of protein fold change determination. 95% confidence intervals of protein fold changes (log2 scale) increased by 17% (0.18−0.21) for the pure chemoproteomics sample, by 19% (0.16−0.19) in the presence of 0.4 μg of E. coli digest, and by 31% (0.146−0.191) in the presence of 0.4 μg of E. coli digest. Taken together, this data demonstrates that the described computation approach for S2I-based correction substantially improves the accuracy of isobaric mass tag-based quantitative proteomics experiments and is applicable for a wide range of different samples.
■
CONCLUSIONS
■
ASSOCIATED CONTENT
Article
AUTHOR INFORMATION
Corresponding Author
*M.M.S.: tel, +49-6221-13757-318; fax, +49-6221-13757-201; e-mail,
[email protected]. M.B.: tel, +49-622113757-310; fax, +49-6221-13757-201; e-mail, Marcus.x. Bantscheff@gsk.com. Notes
The authors declare no competing financial interest.
■ ■
ACKNOWLEDGMENTS The authors thank Frank Weisbrodt for help with preparing the figures and Gerard Drewes for helpful discussions. REFERENCES
(1) Beck, M.; Schmidt, A.; Malmstroem, J.; Claassen, M.; Ori, A.; Szymborska, A.; Herzog, F.; Rinner, O.; Ellenberg, J.; Aebersold, R. The quantitative proteome of a human cell line. Mol. Syst. Biol. 2011, 7, 549. (2) Nagaraj, N.; Wisniewski, J. R.; Geiger, T.; Cox, J.; Kircher, M.; Kelso, J.; Paabo, S.; Mann, M. Deep proteome and transcriptome mapping of a human cancer cell line. Mol. Syst. Biol. 2011, 7, 548. (3) Schwanhausser, B.; Busse, D.; Li, N.; Dittmar, G.; Schuchhardt, J.; Wolf, J.; Chen, W.; Selbach, M. Global quantification of mammalian gene expression control. Nature 2011, 473 (7347), 337−42. (4) Pflieger, D.; Junger, M. A.; Muller, M.; Rinner, O.; Lee, H.; Gehrig, P. M.; Gstaiger, M.; Aebersold, R. Quantitative proteomic analysis of protein complexes: concurrent identification of interactors and their state of phosphorylation. Mol. Cell Proteomics 2008, 7 (2), 326−46. (5) Przybylski, C.; Junger, M. A.; Aubertin, J.; Radvanyi, F.; Aebersold, R.; Pflieger, D. Quantitative analysis of protein complex constituents and their phosphorylation states on a LTQ-Orbitrap instrument. J. Proteome Res. 2010, 9 (10), 5118−32. (6) Iwai, L. K.; Benoist, C.; Mathis, D.; White, F. M. Quantitative phosphoproteomic analysis of T cell receptor signaling in diabetes prone and resistant mice. J. Proteome Res. 2010, 9 (6), 3135−45. (7) Olsen, J. V.; Blagoev, B.; Gnad, F.; Macek, B.; Kumar, C.; Mortensen, P.; Mann, M. Global, in vivo, and site-specific phosphorylation dynamics in signaling networks. Cell 2006, 127 (3), 635−48. (8) Rubbi, L.; Titz, B.; Brown, L.; Galvan, E.; Komisopoulou, E.; Chen, S. S.; Low, T.; Tahmasian, M.; Skaggs, B.; Muschen, M.; Pellegrini, M.; Graeber, T. G. Global phosphoproteomics reveals crosstalk between Bcr-Abl and negative feedback mechanisms controlling Src signaling. Sci. Signaling 2011, 4 (166), ra18. (9) Bantscheff, M.; Eberhard, D.; Abraham, Y.; Bastuck, S.; Boesche, M.; Hobson, S.; Mathieson, T.; Perrin, J.; Raida, M.; Rau, C.; Reader, V.; Sweetman, G.; Bauer, A.; Bouwmeester, T.; Hopf, C.; Kruse, U.; Neubauer, G.; Ramsden, N.; Rick, J.; Kuster, B.; Drewes, G. Quantitative chemical proteomics reveals mechanisms of action of clinical ABL kinase inhibitors. Nat. Biotechnol. 2007, 25 (9), 1035−44. (10) Bantscheff, M.; Hopf, C.; Savitski, M. M.; Dittmann, A.; Grandi, P.; Michon, A. M.; Schlegl, J.; Abraham, Y.; Becher, I.; Bergamini, G.; Boesche, M.; Delling, M.; Dumpelfeld, B.; Eberhard, D.; Huthmacher, C.; Mathieson, T.; Poeckel, D.; Reader, V.; Strunk, K.; Sweetman, G.; Kruse, U.; Neubauer, G.; Ramsden, N. G.; Drewes, G. Chemoproteomics profiling of HDAC inhibitors reveals selective targeting of HDAC complexes. Nat. Biotechnol. 2011, 29 (3), 255−65. (11) Bantscheff, M.; Scholten, A.; Heck, A. J. Revealing promiscuous drug-target interactions by chemical proteomics. Drug Discovery Today 2009, 14 (21−22), 1021−9. (12) Patricelli, M. P.; Nomanbhoy, T. K.; Wu, J.; Brown, H.; Zhou, D.; Zhang, J.; Jagannathan, S.; Aban, A.; Okerberg, E.; Herring, C.; Nordin, B.; Weissig, H.; Yang, Q.; Lee, J. D.; Gray, N. S.; Kozarich, J. W. In situ kinase profiling reveals functionally relevant properties of native kinases. Chem. Biol. 2011, 18 (6), 699−710.
We have demonstrated that measures for spectrum purity in survey spectra (S2I and P2T) can be used to improve the accuracy of protein quantification using isobaric mass tags. This can be achieved either by applying strict filters or by applying an S2I-based protein fold change correction algorithm in combination with soft spectrum filters. While both approaches achieved similar quantification accuracy, the fold change correction approach enabled quantification of a substantially higher fraction of proteins. At the same time the impact of the global correction procedure on the precision of protein quantification was relatively modest (
