There is a need to better understand and handle the 'dark matter' of proteomics—the vast diversity of post-translational and chemical modifications that are unaccounted in a typical mass spectrometry–based analysis and thus remain unidentified. We present a fragment-ion indexing method, and its implementation in peptide identification tool MSFragger, that enables a more than 100-fold improvement in speed over most existing proteome database search tools. Using several large proteomic data sets, we demonstrate how MSFragger empowers the open database search concept for comprehensive identification of peptides and all their modified forms, uncovering dramatic differences in modification rates across experimental samples and conditions. We further illustrate its utility using protein–RNA cross-linked peptide data and using affinity purification experiments where we observe, on average, a 300% increase in the number of identified spectra for enriched proteins. We also discuss the benefits of open searching for improved false discovery rate estimation in proteomics.
This is a preview of subscription content
Subscribe to Journal
Get full journal access for 1 year
only $9.92 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Nesvizhskii, A.I. A survey of computational methods and error rate estimation procedures for peptide and protein identification in shotgun proteomics. J. Proteomics 73, 2092–2123 (2010).
Eng, J.K., Searle, B.C., Clauser, K.R. & Tabb, D.L. A face in the crowd: recognizing peptides through database search. Mol. Cel. Proteomics 10, R111.009522 (2011).
Skinner, O.S. & Kelleher, N.L. Illuminating the dark matter of shotgun proteomics. Nat. Biotechnol. 33, 717–718 (2015).
Chick, J.M. et al. A mass-tolerant database search identifies a large proportion of unassigned spectra in shotgun proteomics as modified peptides. Nat. Biotechnol. 33, 743–749 (2015).
Griss, J. et al. Recognizing millions of consistently unidentified spectra across hundreds of shotgun proteomics datasets. Nat. Methods 13, 651–656 (2016).
Nesvizhskii, A.I. et al. Dynamic spectrum quality assessment and iterative computational analysis of shotgun proteomic data: toward more efficient identification of post-translational modifications, sequence polymorphisms, and novel peptides. Mol. Cell. Proteomics 5, 652–670 (2006).
Nielsen, M.L., Savitski, M.M. & Zubarev, R.A. Extent of modifications in human proteome samples and their effect on dynamic range of analysis in shotgun proteomics. Mol. Cell. Proteomics 5, 2384–2391 (2006).
Ning, K., Fermin, D. & Nesvizhskii, A.I. Computational analysis of unassigned high-quality MS/MS spectra in proteomic data sets. Proteomics 10, 2712–2718 (2010).
Craig, R. & Beavis, R.C. TANDEM: matching proteins with tandem mass spectra. Bioinformatics 20, 1466–1467 (2004).
Creasy, D.M. & Cottrell, J.S. Error tolerant searching of uninterpreted tandem mass spectrometry data. Proteomics 2, 1426–1434 (2002).
Shortreed, M.R. et al. Global Identification of Protein Post-translational Modifications in a Single-Pass Database Search. J. Proteome Res. 14, 4714–4720 (2015).
Ahrné, E., Nikitin, F., Lisacek, F. & Müller, M. QuickMod: A tool for open modification spectrum library searches. J. Proteome Res. 10, 2913–2921 (2011).
Bandeira, N., Tsur, D., Frank, A. & Pevzner, P.A. Protein identification by spectral networks analysis. Proc. Natl. Acad. Sci. USA 104, 6140–6145 (2007).
Savitski, M.M., Nielsen, M.L. & Zubarev, R.A. ModifiComb, a new proteomic tool for mapping substoichiometric post-translational modifications, finding novel types of modifications, and fingerprinting complex protein mixtures. Mol. Cell. Proteomics 5, 935–948 (2006).
Ma, C.W. & Lam, H. Hunting for unexpected post-translational modifications by spectral library searching with tier-wise scoring. J. Proteome Res. 13, 2262–2271 (2014).
Tabb, D.L., Saraf, A. & Yates, J.R. III. GutenTag: high-throughput sequence tagging via an empirically derived fragmentation model. Anal. Chem. 75, 6415–6421 (2003).
Bern, M., Cai, Y. & Goldberg, D. Lookup peaks: a hybrid of de novo sequencing and database search for protein identification by tandem mass spectrometry. Anal. Chem. 79, 1393–1400 (2007).
Dasari, S. et al. Sequence tagging reveals unexpected modifications in toxicoproteomics. Chem. Res. Toxicol. 24, 204–216 (2011).
Na, S., Bandeira, N. & Paek, E. Fast multi-blind modification search through tandem mass spectrometry. Mol. Cell. Proteomics 11, M111.010199 (2012).
Searle, B.C. et al. Identification of protein modifications using MS/MS de novo sequencing and the OpenSea alignment algorithm. J. Proteome Res. 4, 546–554 (2005).
Chen, Y., Chen, W., Cobb, M.H. & Zhao, Y. PTMap--a sequence alignment software for unrestricted, accurate, and full-spectrum identification of post-translational modification sites. Proc. Natl. Acad. Sci. USA 106, 761–766 (2009).
Tanner, S., Pevzner, P.A. & Bafna, V. Unrestrictive identification of post-translational modifications through peptide mass spectrometry. Nat. Protoc. 1, 67–72 (2006).
Fu, Y. in Statistical Analysis in Proteomics (ed. K. Jung) 265–275 (Springer New York, New York, NY, 2016).
Chi, H. et al. pFind-Alioth: A novel unrestricted database search algorithm to improve the interpretation of high-resolution MS/MS data. J. Proteomics 125, 89–97 (2015).
Cox, J. et al. Andromeda: a peptide search engine integrated into the MaxQuant environment. J. Proteome Res. 10, 1794–1805 (2011).
McIlwain, S. et al. Crux: rapid open source protein tandem mass spectrometry analysis. J. Proteome Res. 13, 4488–4491 (2014).
Eng, J.K., Jahan, T.A. & Hoopmann, M.R. Comet: an open-source MS/MS sequence database search tool. Proteomics 13, 22–24 (2013).
Fu, Y. & Qian, X. Transferred subgroup false discovery rate for rare post-translational modifications detected by mass spectrometry. Mol. Cell. Proteomics 13, 1359–1368 (2014).
Vaudel, M. et al. PeptideShaker enables reanalysis of MS-derived proteomics data sets. Nat. Biotechnol. 33, 22–24 (2015).
Diament, B.J. & Noble, W.S. Faster SEQUEST searching for peptide identification from tandem mass spectra. J. Proteome Res. 10, 3871–3879 (2011).
Tsou, C.C. et al. DIA-Umpire: comprehensive computational framework for data-independent acquisition proteomics. Nat. Methods 12, 258–264, 7, 264 (2015).
Houel, S. et al. Quantifying the impact of chimera MS/MS spectra on peptide identification in large-scale proteomics studies. J. Proteome Res. 9, 4152–4160 (2010).
Avtonomov, D.M., Raskind, A. & Nesvizhskii, A.I. BatMass: a Java software platform for LC-MS data visualization in proteomics and metabolomics. J. Proteome Res. 15, 2500–2509 (2016).
Zhang, B., Pirmoradian, M., Chernobrovkin, A. & Zubarev, R.A. DeMix workflow for efficient identification of cofragmented peptides in high resolution data-dependent tandem mass spectrometry. Mol. Cell. Proteomics 13, 3211–3223 (2014).
Bogdanow, B., Zauber, H. & Selbach, M. Systematic errors in peptide and protein identification and quantification by modified peptides. Mol. Cell. Proteomics 15, 2791–2801 (2016).
Nesvizhskii, A.I. Proteogenomics: concepts, applications and computational strategies. Nat. Methods 11, 1114–1125 (2014).
Sharma, K. et al. Ultradeep human phosphoproteome reveals a distinct regulatory nature of Tyr and Ser/Thr-based signaling. Cell Rep. 8, 1583–1594 (2014).
Lawrence, R.T. et al. The proteomic landscape of triple-negative breast cancer. Cell Rep. 11, 630–644 (2015).
Pozniak, Y. et al. System-wide clinical proteomics of breast cancer reveals global remodeling of tissue homeostasis. Cell Syst. 2, 172–184 (2016).
Metz, B. et al. Identification of formaldehyde-induced modifications in proteins: reactions with model peptides. J. Biol. Chem. 279, 6235–6243 (2004).
Huttlin, E.L. et al. The BioPlex network: a systematic exploration of the human interactome. Cell 162, 425–440 (2015).
Kabil, O. & Banerjee, R. Enzymology of H2S biogenesis, decay and signaling. Antioxid. Redox Signal. 20, 770–782 (2014).
Choi, H. et al. SAINT: probabilistic scoring of affinity purification-mass spectrometry data. Nat. Methods 8, 70–73 (2011).
Sardiu, M.E. & Washburn, M.P. Construction of protein interaction networks based on the label-free quantitative proteomics. Methods Mol. Biol. 781, 71–85 (2011).
Kramer, K. et al. Photo-cross-linking and high-resolution mass spectrometry for assignment of RNA-binding sites in RNA-binding proteins. Nat. Methods 11, 1064–1070 (2014).
Perez-Riverol, Y., Alpi, E., Wang, R., Hermjakob, H. & Vizcaíno, J.A. Making proteomics data accessible and reusable: current state of proteomics databases and repositories. Proteomics 15, 930–949 (2015).
Tan, M. et al. Identification of 67 histone marks and histone lysine crotonylation as a new type of histone modification. Cell 146, 1016–1028 (2011).
Yadav, M. et al. Predicting immunogenic tumour mutations by combining mass spectrometry and exome sequencing. Nature 515, 572–576 (2014).
Mommen, G.P.M. et al. Expanding the detectable HLA peptide repertoire using electron-transfer/higher-energy collision dissociation (EThcD). Proc. Natl. Acad. Sci. USA 111, 4507–4512 (2014).
van den Broek, I. et al. Quantifying protein measurands by peptide measurements: where do errors arise? J. Proteome Res. 14, 928–942 (2015).
Fenyö, D. & Beavis, R.C. A method for assessing the statistical significance of mass spectrometry-based protein identifications using general scoring schemes. Anal. Chem. 75, 768–774 (2003).
Deutsch, E.W. et al. A guided tour of the Trans-Proteomic Pipeline. Proteomics 10, 1150–1159 (2010).
Kryuchkov, F., Verano-Braga, T., Hansen, T.A., Sprenger, R.R. & Kjeldsen, F. Deconvolution of mixture spectra and increased throughput of peptide identification by utilization of intensified complementary ions formed in tandem mass spectrometry. J. Proteome Res. 12, 3362–3371 (2013).
Keller, A., Nesvizhskii, A.I., Kolker, E. & Aebersold, R. Empirical statistical model to estimate the accuracy of peptide identifications made by MS/MS and database search. Anal. Chem. 74, 5383–5392 (2002).
Nesvizhskii, A.I., Keller, A., Kolker, E. & Aebersold, R. A statistical model for identifying proteins by tandem mass spectrometry. Anal. Chem. 75, 4646–4658 (2003).
Choi, H., Ghosh, D. & Nesvizhskii, A.I. Statistical validation of peptide identifications in large-scale proteomics using the target-decoy database search strategy and flexible mixture modeling. J. Proteome Res. 7, 286–292 (2008).
Shanmugam, A.K., Yocum, A.K. & Nesvizhskii, A.I. Utility of RNA-seq and GPMDB protein observation frequency for improving the sensitivity of protein identification by tandem MS. J. Proteome Res. 13, 4113–4119 (2014).
Kong, A.T., Leprevost, F.V., Avtonomov, D.M., Mellacheruvu, D. & Nesvizhskii, A.I. Using MSFragger for ultrafast database searching. Protocol Exchange doi:10.1038/nprot.2017.032 (2017).
We thank R. Beavis for helpful discussions, N. Bandeira and S. Na for help with MODa software, and E. Huttlin for assisting with the transfer of raw MS data from the AP–MS study. This work was funded in part by grants from the NIH (R01GM94231 and U24CA210967 to A.I.N.).
The authors declare no competing financial interests.
Integrated supplementary information
The cost and efficiency of spectra similarity calculations can be approximated by the number of fragment comparisons required for each candidate peptide. In conventional strategies, tens to hundreds of comparisons are needed to compare an experimental spectrum to a theoretical spectrum. However, the vast majority of such fragment-fragment comparisons do not result in matches as the differences between their m/z is often far greater than the fragment mass tolerance. Using MSFragger’s fragment index, these comparisons are omitted as the binning strategy allows us to retrieve only the experimental-theoretical fragment pairs that are close in m/z – the majority of which falls within the fragment mass tolerance and are deemed relevant when they contribute to the score of a PSM. MSFragger’s alternative approach results in only a few fragments evaluated per candidate peptide across a variety of search scenarios. Reduction in the fragment bin width allows for fewer fragment comparisons to be performed at the expense of computational overheads associated with traversing a greater number of bins that overlap the fragment tolerance window. MSFragger dynamically selects a bin width appropriate for the search scenario (opting for smaller bins in open search where the number of comparisons is large, and larger bins in narrow window search, where the number of comparisons is small relative to the overhead costs). Hence, a greater number of fragments is evaluated per candidate and a lower percentage of comparisons are found relevant in narrow window searching due to this optimization.
Indexing and searching operations in MSFragger are designed for modern multi-core computers and are optimized to reduce pressures on memory bandwidth. Results are generated from open search times of a single LC-MS/MS run on a dual processor system with 14-cores in each processor. (a) MSFragger scales almost linearly in terms of overall search times on up to 8 cores. Reading of mass spectrometry data files and results compilation is not highly parallelizable resulting in reduced scalability beyond 8 cores. The jump from 14 to 28 threads causes non-local memory to be accessed by each processor, impacting scalability. (b) Fragment index searching by itself is efficiently parallelizable in MSFragger and scales to effectively utilize all cores.
MODa, run in single blind mode, generates a similar modification profile as that of an open search with differences that are likely due to the characteristics of the modification. Open searches (run in fully tryptic mode in both comparisons) are more likely to recover mass shifted peptides that have little discernible alterations in their tandem mass spectra (such as the modification near 302 Da) as it does not attempt to localize the modified mass. MODa is likely more effective for modifications that are more commonly found near the C-terminus (and disrupts the y-ions used in open search identification). MODa running in semi-tryptic mode (the mode of operation as recommended by its authors) recovers a greater number of PSMs at the expense of additional run time.
Supplementary Figure 4 Preferential boosting of unmodified peptides fails to rescue missing peptides.
Boosting recovers a greater percentage of the peptides found in narrow window search prior to FDR filtering. Note that not all peptides identified in narrow window search are recovered in open search with the boosting option enabled due to the presence of a default peptide probability filter of 0.05 in PeptideProphet (disabling this filter using the –p0 option results in near 100% recovery). However, after controlling for FDR, boosting does not improve the peptide overlap between open and narrow window search.
Supplementary Figure 5 Decreased sensitivity for common modifications in open searching can be overcome by specifying variable modifications.
Standard open searches tend to identify far fewer peptides modified with common modifications than narrow window searching specifying those modifications as variable modifications. This is due to decreased sensitivity when the shifted ions are no longer matched in open search. For the most abundant chemical modifications, this can result in a significant decrease in overall counts. The speed of MSFragger allows variable modifications to be specified in conjunction with open searching. Examining peptides with oxidized methionine reveals that standard open search recovers only 45.37% of the peptides originally identified with oxidized methionine in narrow window searching (with variably oxidized methionine). Specifying oxidized methionine as a variable modification in open search brings that percentage to 88.81%, close to the overall overlap in peptide identifications between narrow window and open searches.
Supplementary Figure 6 Complementary ions aid recovery of peptides with modifications near peptide C terminus.
(a) High intensity fragment ions are selected from the experimental spectrum and are assumed to be modified y-ions. Complementary ions based on the experimental precursor mass are inserted to form a modified spectrum that is subjected to open searching. (b) Evaluation of complementary ions using peptides containing a single oxidized methionine. 10, 20, and 30 complementary ions were inserted into each experimental spectrum and the counts of identified peptides were ordered by the distance of their oxidation site to the N or C-terminus. The addition of complementary ions decreased the number of identifications for peptides with oxidation near the N-terminus but greatly increased identification rates for peptides with oxidation near the C-terminus. For peptides with an oxidized methionine upstream of the tryptic cleavage site, the number of identified peptides increased by 48% when 20 complementary ions were added. The addition of more than 20 complementary ions was not found to be beneficial.
Supplementary Figure 7 Co-isolation of co-eluting precursors can result in mass differences that are not due to chemical modifications.
(a) A MS/MS event was triggered at m/z 685.84 (green arrow) resulting in the identification of the peptide LGPALATGNVVVMK with a mass difference of 0.878. The parent survey scan reveals a co-eluting precursor with m/z 685.40 (cyan arrow). The difference in m/z at charge 2+ matches the observed mass difference suggesting that the co-eluting precursor is identified instead of the target precursor in this chimeric spectrum. (b) BatMass visualization of the MS/MS event described in (a) with MS/MS isolations marked by the purple line segments. The cyan arrow indicates the monoisotopic peak of the target precursor while the red arrow indicates the monoisotopic peak of the identified precursor. (c) The peptide RESVELALK was identified with a mass difference of -349.185 at m/z 348.21 (green arrow). Parent survey scan reveals a co-eluting precursor with m/z 348.87 (cyan arrow). While the target precursor ion is of charge 2+, the co-eluting precursor is of charge 3+, which transforms this 0.66 difference in m/z between these co-eluting precursors into the observed mass difference of -349.185. (d) Similar BatMass visualization of the MS/MS event described in (c). Note how the isolation window of the charge 2+ target precursor (cyan) crosses the monoisotopic peak of the charge 3+ co-eluting precursor (red).
Supplementary Figure 8 MS1-based correction of precursor masses and identification-based calibration helps delineate modifications in close mass proximity.
Identified number of PSMs with mass differences in the range of 0.98 Da to 1.01 Da from a single HEK293 LC-MS/MS run. Expected mass differences in this range are due to deamidation (with a delta mass of 0.984 Da) and C12/C13 error (with a delta mass of 1.003 Da). (a) Prior to correction a broad peak with no coherent shape is observed with a center around 1.005 Da. Knowledge of expected mass differences may lead to the calling of a peak near 0.986 Da. (b) Two cleanly resolved peaks are observed after mass correction. Expected peaks corresponding to deamidation and C12/C13 error are resolved with mean mass accurate to 1/1000 Da. The ability to determine such peaks from a single LC-MS/MS run demonstrates the accuracy of modern instruments and the power of our mass correction procedure.
Common modifications were selected and amino acid localization enrichment was calculated separately for each dataset. Amino acid localizations were largely consistent across each dataset despite the differences in modification rates.
Supplementary Figure 10 Highly similar spectra pair for peptide LEAEIATYR with precursor mass difference of 284.126.
3214 PSMs (corresponding to 1087 unique peptides) were identified in the mass difference bin of 284.126 Da. These PSMs were predominantly observed in the HeLa dataset and were shown to have a spectral similarity score of 0.90 (indicating that the spectra of mass shifted peptides are highly similar to that of corresponding unmodified peptides). Here, we selected a pair of PSMs that were both identified to be the peptide LEAEIATYR in the same LC-MS/MS run. Despite their highly similar fragmentation patterns and few unmatched fragments, they were observed with precursor masses that differ by 284.1251 Da. The full y-ion series was successfully matched, which when overlapped with the matched b-2, b-3, and b-4 ions, rules out the possibility of a modified residue in the fragmentation spectrum.
Supplementary Figures 1–10 (PDF 1483 kb)
MSFragger Manual (PDF 611 kb)
Analysis times for a single file (b1906_293T_proteinID_01A_QE3_122212) in HEK293 dataset using different search engines. (XLSX 12 kb)
List of mass spectrometry data files analyzed from each dataset and their corresponding number of MS/MS spectra. (XLSX 146 kb)
List of top 500 detected features in mass shift histogram with potential explanations. (XLSX 116 kb)
Mass shift localization by data set. (XLSX 112 kb)
Number of peptide ions and PSMs identified in narrow-window and open searching by bait protein in AP-MS dataset. (XLSX 187 kb)
Peptide identifications in ETHE1 AP–MS experiments using narrow-window and mass-tolerant searches. (XLSX 571 kb)
List of genes associated with 'small molecule metabolic process' that have a large increase in identified bait peptide ions. (XLSX 9 kb)
List of identified peptides in RNA–protein cross-linking experiment. (XLSX 28 kb)
About this article
Cite this article
Kong, A., Leprevost, F., Avtonomov, D. et al. MSFragger: ultrafast and comprehensive peptide identification in mass spectrometry–based proteomics. Nat Methods 14, 513–520 (2017). https://doi.org/10.1038/nmeth.4256
BMC Bioinformatics (2021)
Nature Computational Science (2021)
Tutorial: best practices and considerations for mass-spectrometry-based protein biomarker discovery and validation
Nature Protocols (2021)
Octenidine-based hydrogel shows anti-inflammatory and protease-inhibitory capacities in wounded human skin
Scientific Reports (2021)
Nature Methods (2021)