Improve Peptide Identification Research Articles

Koina: Democratizing machine learning for proteomics research.

Recent developments in machine-learning (ML) and deep-learning (DL) have immense potential for applications in proteomics, such as generating spectral libraries, improving peptide identification, and optimizing targeted acquisition modes. Although new ML/DL models for various applications and peptide properties are frequently published, the rate at which these models are adopted by the community is slow, which is mostly due to technical challenges. We believe that, for the community to make better use of state-of-the-art models, more attention should be spent on making models easy to use and accessible by the community. To facilitate this, we developed Koina, an open-source containerized, decentralized and online-accessible high-performance prediction service that enables ML/DL model usage in any pipeline. Using the widely used FragPipe computational platform as example, we show how Koina can be easily integrated with existing proteomics software tools and how these integrations improve data analysis.

UniSpec: Deep Learning for Predicting the Full Range of Peptide Fragment Ion Series to Enhance the Proteomics Data Analysis Workflow.

We present UniSpec, an attention-driven deep neural network designed to predict comprehensive collision-induced fragmentation spectra, thereby improving peptide identification in shotgun proteomics. Utilizing a training data set of 1.8 million unique high-quality tandem mass spectra (MS2) from 0.8 million unique peptide ions, UniSpec learned with a peptide fragmentation dictionary encompassing 7919 fragment peaks. Among these, 5712 are neutral loss peaks, with 2310 corresponding to modification-specific neutral losses. Remarkably, UniSpec can predict 73%-77% of fragment intensities based on our NIST reference library spectra, a significant leap from the 35%-45% coverage of only b and y ions. Comparative studies with Prosit elucidate that while both models are strong at predicting their respective fragment ion series, UniSpec particularly shines in generating more complex MS2 spectra with diverse ion annotations. The integration of UniSpec's predictions into shotgun proteomics data analysis boosts the identification rate of tryptic peptides by 48% at a 1% false discovery rate (FDR) and 60% at a more confident 0.1% FDR. Using UniSpec's predicted in-silico spectral library, the search results closely matched those from search engines and experimental spectral libraries used in peptide identification, highlighting its potential as a stand-alone identification tool. The source code and Python scripts are available on GitHub (https://github.com/usnistgov/UniSpec) and Zenodo (https://zenodo.org/records/10452792), and all data sets and analysis results generated in this work were deposited in Zenodo (https://zenodo.org/records/10052268).

Utilizing Precursor Ion Connectivity of Different Charge States to Improve Peptide and Protein Identification in MS/MS Analysis.

Tandem mass spectrometry (MS/MS) has become a key method for the structural analysis of biomolecules such as peptides and proteins. A pervasive problem in MS/MS analyses, especially for top-down proteomics, is the occurrence of chimeric spectra, when two or more precursor ions are co-isolated and fragmented, thus leading to complex MS/MS spectra that are populated with fragment ions originating from different precursor ions. This type of convoluted data typically results in low sequence database search scores due to the vast number of mixed-source fragment ions, of which only a fraction originates from a specific precursor ion. Herein, we present a novel workflow that deconvolutes the data of chimeric MS/MS spectra, improving the protein search scores and sequence coverages in database searching and thus providing a more confident peptide and protein identification. Previously misidentified proteins or proteins with insignificant search scores can be correctly and significantly identified following the presented data acquisition and analysis workflow with search scores increasing by a factor of 3-4 for smaller precursor ions (peptides) and >6 for larger precursor ions such as intact ubiquitin and cytochrome C.

Test-Time Training for Deep MS/MS Spectrum Prediction Improves Peptide Identification.

In bottom-up proteomics, peptide-spectrum matching is critical for peptide and protein identification. Recently, deep learning models have been used to predict tandem mass spectra of peptides, enabling the calculation of similarity scores between the predicted and experimental spectra for peptide-spectrum matching. These models follow the supervised learning paradigm, which trains a general model using paired peptides and spectra from standard data sets and directly employs the model on experimental data. However, this approach can lead to inaccurate predictions due to differences between the training data and the experimental data, such as sample types, enzyme specificity, and instrument calibration. To tackle this problem, we developed a test-time training paradigm that adapts the pretrained model to generate experimental data-specific models, namely, PepT3. PepT3 yields a 10-40% increase in peptide identification depending on the variability in training and experimental data. Intriguingly, when applied to a patient-derived immunopeptidomic sample, PepT3 increases the identification of tumor-specific immunopeptide candidates by 60%. Two-thirds of the newly identified candidates are predicted to bind to the patient's human leukocyte antigen isoforms. To facilitate access of the model and all the results, we have archived all the intermediate files in Zenodo.org with identifier 8231084.

RUPE-phospho: Rapid Ultrasound-Assisted Peptide-Identification-Enhanced Phosphoproteomics Workflow for Microscale Samples.

Global phosphoproteome profiling can provide insights into cellular signaling and disease pathogenesis. To achieve comprehensive phosphoproteomic analyses with minute quantities of material, we developed a rapid and sensitive phosphoproteomics sample preparation strategy based on ultrasound. We found that ultrasonication-assisted digestion can significantly improve peptide identification by 20% due to the generation of longer peptides that can be detected by mass spectrometry. By integrating this rapid ultrasound-assisted peptide-identification-enhanced proteomic method (RUPE) with streamlined phosphopeptide enrichment steps, we established RUPE-phospho, a fast and efficient strategy to characterize protein phosphorylation in mass-limited samples. This approach dramatically reduces the sample loss and processing time: 24 samples can be processed in 3 h; 5325 phosphosites, 4549 phosphopeptides, and 1888 phosphoproteins were quantified from 5 μg of human embryonic kidney (HEK) 293T cell lysate. In addition, 9219 phosphosites were quantified from 1-2 mg of OCT-embedded mouse brain with 120 min streamlined RUPE-phospho workflow. RUPE-phospho facilitates phosphoproteome profiling for microscale samples and will provide a powerful tool for proteomics-driven precision medicine research.

OpenProt 2.0 builds a path to the functional characterization of alternative proteins.

The OpenProt proteogenomic resource (https://www.openprot.org/) provides users with a complete and freely accessible set of non-canonical or alternative open reading frames (AltORFs) within the transcriptome of various species, as well as functional annotations of the corresponding protein sequences not found in standard databases. Enhancements in this update are largely the result of user feedback and include the prediction of structure, subcellular localization, and intrinsic disorder, using cutting-edge algorithms based on machine learning techniques. The mass spectrometry pipeline now integrates a machine learning-based peptide rescoring method to improve peptide identification. We continue to help users explore this cryptic proteome by providing OpenCustomDB, a tool that enables users to build their own customized protein databases, and OpenVar, a genomic annotator including genetic variants within AltORFs and protein sequences. A new interface improves the visualization of all functional annotations, including a spectral viewer and the prediction of multicoding genes. All data on OpenProt are freely available and downloadable. Overall, OpenProt continues to establish itself as an important resource for the exploration and study of new proteins.

MSBooster: improving peptide identification rates using deep learning-based features

Peptide identification in liquid chromatography-tandem mass spectrometry (LC-MS/MS) experiments relies on computational algorithms for matching acquired MS/MS spectra against sequences of candidate peptides using database search tools, such as MSFragger. Here, we present a new tool, MSBooster, for rescoring peptide-to-spectrum matches using additional features incorporating deep learning-based predictions of peptide properties, such as LC retention time, ion mobility, and MS/MS spectra. We demonstrate the utility of MSBooster, in tandem with MSFragger and Percolator, in several different workflows, including nonspecific searches (immunopeptidomics), direct identification of peptides from data independent acquisition data, single-cell proteomics, and data generated on an ion mobility separation-enabled timsTOF MS platform. MSBooster is fast, robust, and fully integrated into the widely used FragPipe computational platform.

Characterization of the natural peptidome of four leeches by integrated proteogenomics and pseudotargeted peptidomics

Animal-derived drugs are an indispensable part of folk medicine worldwide. However, their chemical constituents are poorly approached, which leads to the low level of the quality standard system of animal-derived drugs and further causes a chaotic market. Natural peptides are ubiquitous throughout the organism, especially in animal-derived drugs. Thus, in this study, we used multi-source leeches, including Hirudo nipponica (HN), Whitmania pigra (WP), Whitmania acranulata (WA), and Poecilobdella manillensis (PM), as a model. A strategy integrating proteogenomics and novel pseudotargeted peptidomics was developed to characterize the natural peptide phenotype and screen for signature peptides of four leech species. First, natural peptides were sequenced against an in-house annotated protein database of closely related species constructed from RNA-seq data from the Sequence Read Archive (SRA) website, which is an open-sourced public archive resource. Second, a novel pseudotargeted peptidomics integrating peptide ion pair extraction and retention time transfer was established to achieve high coverage and quantitative accuracy of the natural peptides and to screen for signature peptides for species authentication. In all, 2323 natural peptides were identified from four leech species whose databases were poorly annotated. The strategy was shown to significantly improve peptide identification. In addition, 36 of 167 differential peptides screened by pseudotargeted proteomics were identified, and about one-third of them came from the leucine-rich repeat domain (LRR) proteins, which are widely distributed in organisms. Furthermore, six signature peptides were screened with good specificity and stability, and four of them were validated by synthetic standards. Finally, a dynamic multiple reaction monitoring (dMRM) method based on these signature peptides was established and revealed that one-half of the commercial samples and all of the Tongxinluo capsules were derived from WP. All in all, the strategy developed in this study was effective for natural peptide characterization and signature peptide screening, which could also be applied to other animal-derived drugs, especially for modelless species that are less studied in protein database annotation.

Alpha-Tri: a deep neural network for scoring the similarity between predicted and measured spectra improves peptide identification of DIA data.

Peptide identification of data-independent acquisition (DIA) mass spectrometry applying the peptide-centric approach heavily relies on the spectral library matching, such as the fragment intensity similarity. If the intensity similarity is calculated through all possible fragment ions of a targeted peptide instead of just a few fragment ions provided by the spectral library, the matching will be more comprehensive and reliable, and thus the identification will be more confident. In addition, the emergence of high precision spectrum predictors, like Prosit, also makes it possible to capitalize on the predicted spectrum, which contains all possible fragment ion intensities, to calculate the intensity similarity for DIA data. In this work, we propose Alpha-Tri, a neural-network-based model to calculate intensity similarity as a post-processing score using the predicted spectrum, measured spectrum and correlation spectrum (triple-spectrum). The predicted spectrum is generated by Prosit, the measured spectrum is retrieved from the apex of the chromatograms of all possible fragment ions and the correlation spectrum is used to indicate the present probabilities of these fragment ions as the link between the precursor and its fragment ions is lost in DIA. By adopting a data-driven method, Alpha-Tri is able to learn the intensity similarity from the triple-spectrum. This learned value is appended to initial scores from DIA-NN, allowing the ensuing statistical validation tool to report more peptides at the same false discovery rate (FDR). In our evaluation of the HeLa dataset with gradient lengths ranging from 0.5 to 2 h, Alpha-Tri delivered 3.0-7.2% gains in peptide detections at 1% FDR. On LFQbench dataset, a mixed-species dataset with known ratios, Alpha-Tri identified more peptides and proteins fell within the valid ratio ranges by up to 8.6% and 7.6%, respectively, compared with DIA-NN solely. The original datasets for benchmarks are downloaded from the ProteomeXchange with the identifiers PXD005573, PXD000954 and PXD002952. Source code is available at https://github.com/YuAirLab/Alpha-Tri.

Trapped Ion Mobility Spectrometry Reduces Spectral Complexity in Mass Spectrometry-Based Proteomics.

In bottom-up mass spectrometry-based proteomics, deep proteome coverage is limited by high cofragmentation rates. Cofragmentation occurs when more than one analyte is isolated by the quadrupole and the subsequent fragmentation event produces fragment ions of heterogeneous origin. One strategy to reduce cofragmentation rates is through effective peptide separation techniques such as chromatographic separation and, the more recently popularized, ion mobility (IM) spectrometry, which separates peptides by their collisional cross section. Here, we use a computational model to investigate the capability of the trapped IM spectrometry (TIMS) device at effectively separating peptide ions and quantify the separation power of the TIMS device in the context of a parallel accumulation-serial fragmentation (PASEF) workflow. We found that TIMS separation increases the number of interference-free MS1 peptide features 9.2-fold, while decreasing the average peptide density in precursor spectra 6.5-fold. In a data-dependent acquisition PASEF workflow, IM separation increases the number of spectra without cofragmentation by a factor of 4.1 and the number of high-quality spectra 17-fold. Using a categorical model, we estimate that this observed decrease in spectral complexity results in an increased likelihood for peptide spectral matches, which may improve peptide identification rates. In the context of a data-independent acquisition workflow, the reduction in spectral complexity resulting from IM separation is estimated to be equivalent to a 4-fold decrease in the isolation window width (from 25 to 6.5 Da). Our study demonstrates that TIMS separation decreases spectral complexity by reducing cofragmentation rates, suggesting that TIMS separation may contribute toward the high identification rates observed in PASEF workflows.

Identification and Mitigation of Defensins in the Immunopurification of Peptide MHC-I Antigens from Lung Tissue.

The class I major histocompatibility (MHC-I) complex is a set of diverse cell surface receptors encoded by the human leukocyte antigen gene complex. These receptors present intracellular antigens to cytotoxic T cells providing information on the state and health of cells. Changes in the immunopeptidome during cancer may provide novel targets for therapeutic intervention. To understand how the tumor immunopeptidome is altered, we developed a mass spectrometry (MS) based platform for isolating and identifying MHC-I peptide antigens in lung tumors. In the course of our work, we encountered several large unknown peptide contaminants which had not been previously reported. To understand the source of these major contaminants, we isolated them using offline fractionation and identified them by liquid chromatography-tandem mass spectrometry (LC-MS/MS) as members of the host defense protein family known as the defensins. To mitigate their detrimental effects, we modified our "Original" data-dependent acquisition (DDA) MS method to narrowly target the MHC-I peptides based on their physical properties including charge state and molecular weight ("z state" DDA), evaluated field asymmetric ion mobility spectrometry to attempt gas-phase separation prior to MS analysis, and developed an immunodepletion approach using defensin specific antibodies. This modified approach improves peptide identification and reduces the impact of defensin contamination in lung tissue samples.

CsoDIAq Software for Direct Infusion Shotgun Proteome Analysis.

Direct infusion shotgun proteome analysis (DISPA) is a new paradigm for expedited mass spectrometry-based proteomics, but the original data analysis workflow was onerous. Here, we introduce CsoDIAq, a user-friendly software package for the identification and quantification of peptides and proteins from DISPA data. In addition to establishing a complete and automated analysis workflow with a graphical user interface, CsoDIAq introduces algorithmic concepts to spectrum-spectrum matching to improve peptide identification speed and sensitivity. These include spectra pooling to reduce search time complexity and a new spectrum-spectrum match score called match count and cosine, which improves target discrimination in a target-decoy analysis. Fragment mass tolerance correction also increased the number of peptide identifications. Finally, we adapt CsoDIAq to standard LC-MS DIA and show that it outperforms other spectrum-spectrum matching software.

Alpha-XIC: a deep neural network for scoring the coelution of peak groups improves peptide identification by data-independent acquisition mass spectrometry.

The peptide-centric identification methodologies of data-independent acquisition (DIA) data mainly rely on scores for the mass spectrometric signals of targeted peptides. Among these scores, the coelution scores of peak groups constructed by the chromatograms of peptide fragment ions have a significant influence on the identification. Most of the existing coelution scores are achieved by artificially designing some functions in terms of the shape similarity, retention time shift of peak groups. However, these scores cannot characterize the coelution robustly when the peak group is in the circumstance of interference. On the basis that the neural network is more powerful to learn the implicit features of data robustly from a large number of samples, and thus minimizing the influence of data noise, in this work, we propose Alpha-XIC, a neural network-based model to score the coelution. By learning the characteristics of the coelution of peak groups derived from the being analyzed DIA data, Alpha-XIC is capable of yielding robust coelution scores even for peak groups with interference. With this score appending to initial scores generated by the accompanying identification engine DIA-NN, the ensuing statistical validation can report the identification result and recover the misidentified peptides. In our evaluation of the HeLa dataset with gradient lengths ranging from 0.5 to 2 h, Alpha-XIC delivered 9.4-16.2% improvements in the number of identified precursors at 1% false discovery rate. Furthermore, Alpha-XIC was tested on LFQbench, a mixed-species dataset with known ratios, and increased the number of peptides and proteins fell within valid ratios by up to 16.4% and 17.8%, respectively, compared to the initial identification by DIA-NN. Source code is available at https://github.com/YuAirLab/Alpha-XIC.

Retrieve and Revise: Improving Peptide Identification with Similar Mass Spectra

Tandem mass spectrometry is an indispensable technology for identification of proteins from complex mixtures. Accurate and sensitive analysis of large amounts of mass spectra data is a principal challenge in proteomics. Conventional deep learning-based peptide identification models usually adopt an encoder-decoder framework and generate target sequence from left to right without fully exploiting the global information. A few recent approaches seek to employ two-pass decoding, yet have limitations when facing the spectra filled with noise. In this paper, we propose a new paradigm for improved peptide identification, which first retrieves a similar mass spectrum from the database as a reference and then revise the matched sequence according to the difference information between the referenced spectrum and current context. The inspiration of design comes that the retrieved peptide-spectrum pair provides a good start point and indirect access to both past and future information, such that each revised amino acid can be produced with better noise perception and global understanding. Moreover, a disturb-based optimization process is introduced to sharpen the attention for difference vector with reinforcement learning before fed to decoder. Experimental results on several public datasets demonstrate that prominent performance boost is obtained with the proposed method. Remarkably, we achieve new state-of-the-art identification results on these datasets.

Effect of Phosphorylation on the Collision Cross Sections of Peptide Ions in Ion Mobility Spectrometry.

The insertion of ion mobility spectrometry (IMS) between LC and MS can improve peptide identification in both proteomics and phosphoproteomics by providing structural information that is complementary to LC and MS, because IMS separates ions on the basis of differences in their shapes and charge states. However, it is necessary to know how phosphate groups affect the peptide collision cross sections (CCS) in order to accurately predict phosphopeptide CCS values and to maximize the usefulness of IMS. In this work, we systematically characterized the CCS values of 4,433 pairs of mono-phosphopeptide and corresponding unphosphorylated peptide ions using trapped ion mobility spectrometry (TIMS). Nearly one-third of the mono-phosphopeptide ions evaluated here showed smaller CCS values than their unphosphorylated counterparts, even though phosphorylation results in a mass increase of 80 Da. Significant changes of CCS upon phosphorylation occurred mainly in structurally extended peptides with large numbers of basic groups, possibly reflecting intramolecular interactions between phosphate and basic groups.

Front Cover: DeepRescore: Leveraging Deep Learning to Improve Peptide Identification in Immunopeptidomics

Front Cover: DeepRescore: Leveraging Deep Learning to Improve Peptide Identification in Immunopeptidomics

DeepRescore: Leveraging Deep Learning to Improve Peptide Identification in Immunopeptidomics.

The identification of major histocompatibility complex (MHC)-binding peptides in mass spectrometry (MS)-based immunopeptideomics relies largely on database search engines developed for proteomics data analysis. However, because immunopeptidomics experiments do not involve enzymatic digestion at specific residues, an inflated search space leads to a high false positive rate and low sensitivity in peptide identification. In order to improve the sensitivity and reliability of peptide identification, a post-processing tool named DeepRescore is developed. DeepRescore combines peptide features derived from deep learning predictions, namely accurate retention timeand MS/MS spectra predictions, with previously used features to rescore peptide-spectrum matches. Using two public immunopeptidomics datasets, it is shown that rescoring by DeepRescore increases both the sensitivity and reliability of MHC-binding peptide and neoantigen identifications compared to existing methods. It is also shown that the performance improvement is, to a large extent, driven by the deep learning-derived features. DeepRescore is developed using NextFlow and Docker and is available at https://github.com/bzhanglab/DeepRescore.

Full-Spectrum Prediction of Peptides Tandem Mass Spectra using Deep Neural Network.

The ability to predict tandem mass (MS/MS) spectra from peptide sequences can significantly enhance our understanding of the peptide fragmentation process and could improve peptide identification in proteomics. However, current approaches for predicting high-energy collisional dissociation (HCD) spectra are limited to predict the intensities of expected ion types, that is, the a/b/c/x/y/z ions and their neutral loss derivatives (referred to as backbone ions). In practice, backbone ions only account for <70% of total ion intensities in HCD spectra, indicating many intense ions are ignored by current predictors. In this paper, we present a deep learning approach that can predict the complete spectra (both backbone and nonbackbone ions) directly from peptide sequences. We made no assumptions or expectations on which kind of ions to predict but instead predicting the intensities for all possible m/z. Training this model needs no annotations of fragment ion nor any prior knowledge of the fragmentation rules. Our analyses show that the predicted 2+ and 3+ HCD spectra are highly similar to the experimental spectra, with average full-spectrum cosine similarities of 0.820 (±0.088) and 0.786 (±0.085), respectively, very close to the similarities between the experimental replicated spectra. In contrast, the best-performed backbone only models can only achieve an average similarity below 0.75 and 0.70 for 2+ and 3+ spectra, respectively. Furthermore, we developed a multitask learning (MTL) approach for predicting spectra of insufficient training samples, which allows our model to make accurate predictions for electron transfer dissociation (ETD) spectra and HCD spectra of less abundant charges (1+ and 4+).

Accurate Precursor Mass Assignment Improves Peptide Identification in Data-Independent Acquisition Mass Spectrometry

Proteomics research today no longer simply seeks exhaustive protein identification; increasingly, it is also desirable to obtain robust, large-scale quantitative information. To accomplish this, data-independent acquisition (DIA) has emerged as a promising strategy largely owing to developments in advanced mass spectrometers and sophisticated data analysis methods. Nevertheless, the highly complex multiplexed MS/MS spectra produced by DIA remain challenging to interpret. Here, we present a novel strategy to analyze DIA data, based on unambiguous precursor mass assignment through the mPE-MMR (multiplexed post-experimental monoisotopic mass refinement) procedure and combined with complementary multistage database searching. Compared to conventional spectral library searching, the accuracy and sensitivity of peptide identification were significantly increased by incorporating precise precursor masses in DIA data. We demonstrate identification of additional peptides absent from spectral libraries, including sample-specific mutated peptides and post-translationally modified peptides using MS-GF+ and MODa/MODi multistage database searching. This first use of unambiguously determined precursor masses to mine DIA data demonstrates considerable potential for further exploitation of this type of experimental data.

A Simple Light Isotope Metabolic Labeling (SLIM-labeling) Strategy: A Powerful Tool to Address the Dynamics of Proteome Variations In Vivo

Many quantitative proteomics strategies rely on in vivo metabolic incorporation of amino acids with modified stable isotope profiles into proteins. These methods give rise to multiple ions for each peptide, with possible distortion of the isotopolog distribution, making the overall analytical process complex. We validated an alternative strategy, simple light isotope metabolic labeling (SLIM-labeling), which alleviates many of these problems. SLIM-labeling is based on the in vivo reduction of the isotopic composition of proteins using metabolic precursors with a unique light isotope composition to label all amino acids. This brings a new dimension to in-depth, high resolution MS-based quantitative proteomics. Here, we describe a 12C-based SLIM-labeling strategy using U-[12C]-glucose as the metabolic precursor of all amino acids in the pathogenic yeast Candida albicans Monoisotopic ion intensity increased exponentially following 12C enrichment, substantially improving peptide identification scores and protein sequence coverage in bottom-up analyses. Multiplexing samples of 12C composition varying from natural abundance (98.93%) to 100% makes it possible to address relative quantification issues, keeping all the critical information for each peptide within a single isotopolog cluster. We applied this method to measure, for the first time, protein turnover at the proteome scale in Candida albicans and its modulation by inhibitors of the proteasome and vacuolar protein degradation systems.