An XIC-Centric Strategy for Improved Identification and Quantification in Proteomic Data Analyses.

Reproducibility is a "proteomic dream" yet to be fully realized. A typical data analysis workflow utilizing extracted ion chromatograms (XICs) often treats the information path from identification to quantification as a one-way street. Here, we propose an XIC-centric approach in which the data flow is bidirectional: identifications are used to derive XICs whose information is in turn applied to validate the identifications. In this study, we employed liquid chromatography-mass spectrometry data from glycoprotein and human hair samples to illustrate the XIC-centric concept. At the core of this approach was XIC-based monoisotope repicking. Taking advantage of the intensity information for all detected isotopes across the whole range of an XIC peak significantly improved the accuracy and uncovered misidentifications originating from monoisotope assignment mistakes. It could also rescue non-top-ranked glycopeptide hits. Identification of glycopeptides is particularly susceptible to precursor mass errors for their low abundances, large masses, and glycans differing by 1 or 2 Da easily confused as isotopes. In addition, the XIC-centric strategy significantly reduced the problem of one XIC peak associated with multiple unique identifications, a source of quantitative irreproducibility. Taken together, the proposed approach can lead to improved identification and quantification accuracy and, ultimately, enhanced reproducibility in proteomic data analyses.

Comparison of N-Glycopeptide to Released N-Glycan Abundances and the Influence of Glycopeptide Mass and Charge States on N-Linked Glycosylation of IgG Antibodies.

We report the comparison of mass-spectral-based abundances of tryptic glycopeptides to fluorescence abundances of released labeled glycans and the effects of mass and charge state and in-source fragmentation on glycopeptide abundances. The primary glycoforms derived from Rituximab, NISTmAb, Evolocumab, and Infliximab were high-mannose and biantennary complex galactosylated and fucosylated N-glycans. Except for Evolocumab, in-source ions derived from the loss of HexNAc or HexNAc-Hex sugars are prominent for other therapeutic IgGs. After excluding in-source fragmentation of glycopeptide ions from the results, a linear correlation was observed between fluorescently labeled N-glycan and glycopeptide abundances over a dynamic range of 500. Different charge states of human IgG-derived glycopeptides containing a wider variety of abundant attached glycans were also investigated to examine the effects of the charge state on ion abundances. These revealed a linear dependence of glycopeptide abundance on the mass of the glycan with higher charge states favoring higher-mass glycans. Findings indicate that the mass spectrometry-based bottom-up approach can provide results as accurate as those of glycan release studies while revealing the origin of each attached glycan. These site-specific relative abundances are conveniently displayed and compared using previously described glycopeptide abundance distribution spectra "GADS" representations. Mass spectrometry data are available from the MAssIVE repository (MSV000093562).

AIRI: Predicting Retention Indices and Their Uncertainties Using Artificial Intelligence.

The Kováts retention index (RI) is a quantity measured using gas chromatography and is commonly used in the identification of chemical structures. Creating libraries of observed RI values is a laborious task, so we explore the use of a deep neural network for predicting RI values from structure for standard semipolar columns. This network generated predictions with a mean absolute error of 15.1 and, in a quantification of the tail of the error distribution, a 95th percentile absolute error of 46.5. Because of the Artificial Intelligence Retention Indices (AIRI) network's accuracy, it was used to predict RI values for the NIST EI-MS spectral libraries. These RI values are used to improve chemical identification methods and the quality of the library. Estimating uncertainty is an important practical need when using prediction models. To quantify the uncertainty of our network for each individual prediction, we used the outputs of an ensemble of 8 networks to calculate a predicted standard deviation for each RI value prediction. This predicted standard deviation was corrected to follow the error between the observed and predicted RI values. The Z scores using these predicted standard deviations had a standard deviation of 1.52 and a 95th percentile absolute Z score corresponding to a mean RI value of 42.6.

Improved Sample Preparation Method for Protein and Peptide Identification from Human Hair.

A fast and sensitive direct extraction (DE) method developed in our group can efficiently extract proteins in 30 min from a 5 cm-long hair strand. Previously, we coupled DE to downstream analysis using gel electrophoresis followed by in-gel digestion, which can be time-consuming. In searching for a better alternative, we found that a combination of DE with a bead-based method (SP3) can lead to significant improvements in protein discovery in human hair. Since SP3 is designed for general applications, we optimized it to process hair proteins following DE and compared it to several other in-solution digestion methods. Of particular concern are genetically variant peptides (GVPs), which can be used for human identification in forensic analysis. Here, we demonstrated improved GVP discovery with the DE and SP3 workflow, which was 3 times faster than the previous in-gel digestion method and required significantly less instrument time depending on the number of gel slices processed. Additionally, it led to an increased number of identified proteins and GVPs. Among the tested in-solution digestion methods, DE combined with SP3 showed the highest sequence coverage, with higher abundances of the identified peptides. This provides a significantly enhanced means for identifying proteins and GVPs in human hair.

Inferring the Nominal Molecular Mass of an Analyte from Its Electron Ionization Mass Spectrum.

The performance of three algorithms for predicting nominal molecular mass from an analyte's electron ionization mass spectrum is presented. The Peak Interpretation Method (PIM) attempts to quantify the likelihood that a molecular ion peak is contained in the mass spectrum, whereas the Simple Search Hitlist Method (SS-HM) and iterative Hybrid Search Hitlist Method (iHS-HM) leverage results from mass spectral library searching. These predictions can be employed in combination (recommended) or independently. The methods were tested on two sets of query mass spectra searched against libraries that did not contain the reference mass spectra of the same compounds: 19,074 spectra of various organic molecules searched against the NIST17 mass spectral library and 162 spectra of small molecule drugs searched against SWGDRUG version 3.3. Individually, each molecular mass prediction method had computed precisions (the fraction of positive predictions that were correct) of 91, 89, and 74%, respectively. The methods become more valuable when predictions are taken together. When all three predictions were identical, which occurred in 33% of the test cases, the predicted molecular mass was almost always correct (>99%).

Determining Site-Specific Glycan Profiles of Recombinant SARS-CoV-2 Spike Proteins from Multiple Sources.

Glycopeptide Abundance Distribution Spectra (GADS) were recently introduced as a means of representing, storing, and comparing glycan profiles of intact glycopeptides. Here, using that representation, an extensive analysis is made of multiple commercial sources of the recombinant SARS-CoV-2 spike protein, each containing 22 N-linked glycan sites (sequons). Multiple proteases are used along with variable energy fragmentation followed by ion trap confirmation. This enables a detailed examination of the reproducibility of the method across multiple types of variability. These results show that GADS are consistent between replicates and laboratories for sufficiently abundant glycopeptides. Derived GADS enable the examination and comparison of the glycan profiles between commercial sources of the spike protein. Multiple distinct glycopeptide distributions, generated by multiple proteases, confirm these profiles. Comparisons of GADS derived from 11 sources of recombinant spike protein reveal that sources for which protein expression methods were the same produced near-identical glycan profiles, thereby demonstrating the ability of this method to measure GADS of sufficient reliability to distinguish different glycoform distributions between commercial vendors and potentially to reliably determine and compare differences in glycosylation for any glycoprotein under different conditions of production. All mass spectrometry data files have been deposited in the MassIVE repository under the identifier MSV000091776.

AIomics: Exploring More of the Proteome Using Mass Spectral Libraries Extended by Artificial Intelligence.

The unbounded permutations of biological molecules, including proteins and their constituent peptides, present a dilemma in identifying the components of complex biosamples. Sequence search algorithms used to identify peptide spectra can be expanded to cover larger classes of molecules, including more modifications, isoforms, and atypical cleavage, but at the cost of false positives or false negatives due to the simplified spectra they compute from sequence records. Spectral library searching can help solve this issue by precisely matching experimental spectra to library spectra with excellent sensitivity and specificity. However, compiling spectral libraries that span entire proteomes is pragmatically difficult. Neural networks that predict complete spectra containing a full range of annotated and unannotated ions can be used to replace these simplified spectra with libraries of fully predicted spectra, including modified peptides. Using such a network, we created predicted spectral libraries that were used to rescore matches from a sequence search done over a large search space, including a large number of modifications. Rescoring improved the separation of true and false hits by 82%, yielding an 8% increase in peptide identifications, including a 21% increase in nonspecifically cleaved peptides and a 17% increase in phosphopeptides.

Unexpected Gas-Phase Nitrogen-Oxygen Smiles Rearrangement: Collision-Induced Dissociation of Deprotonated 2-(N-Methylanilino)ethanol and Morpholinylbenzoic Acid Derivatives.

A nitrogen-oxygen Smiles rearrangement was reported to occur after collisional activation of the PhN(R)CH2CH2O- (R = alkyl) anion, which undergoes a five-membered ring rearrangement to form a phenoxide ion C6H5O-. When R = H, such a Smiles rearrangement is unlikely since the negative charge is more favorably located on the nitrogen atom than the oxygen atom; hence, alternative neutral losses dominate the fragmentation. For example, collisional activation of deprotonated 2-anilinoethanol (PhN-CH2CH2OH) leads to the formation of an anilide anion (C6H5NH-, m/z 92) rather than a phenoxide ion (C6H5O-, m/z 93.0343). However, when the amino hydrogen of 2-anilinoethanol is substituted by a methyl group, i.e., 2-(N-methylanilino)ethanol, a Smiles rearrangement does occur, leading to the phenoxide ion, as the negative charge can only reside on the oxygen atom. To confirm the Smiles rearrangement mechanism, 2-(N-methylanilino)ethanol-18O was synthesized and subjected to collisional activation, leading to an intense peak at m/z 95.0385, which corresponds to the 18O phenoxide ion ([C6H518O]-). The abundance of the phenoxide ion is sensitive to substituents on the N atom, as demonstrated by the observation that an ethyl substituent results in the rearrangement ion with a much lower abundance. The nitrogen-oxygen Smiles rearrangement also occurs for various morpholinylbenzoic acid derivatives with a multistep mechanism, where the phenoxide ion is found to be predominantly formed after loss of CO2, proton transfers, breaking of the morpholine ring, and Smiles rearrangement. The Smiles mechanism is also supported by density functional theory calculations and other observations.

Mass Spectral Library Methods for Analysis of Site-Specific N-Glycosylation: Application to Human Milk Proteins.

We present a mass spectral library-based method for analyzing site-specific N-linked protein glycosylation. Its operation and utility are illustrated by applying it to both newly measured and available proteomics data of human milk glycoproteins. It generates two varieties of mass spectral libraries. One contains glycopeptide abundance distribution spectra (GADS). The other contains tandem mass spectra of the underlying glycopeptides. Both originate from identified glycopeptides in proteolytic digests of human milk and purified glycoproteins, which include tenascin, lactoferrin, and several antibodies. Analysis was also applied to digests of a NIST human milk standard reference material (SRM), leading to a GADS library of N-glycopeptides, enabling the direct comparison of glycopeptide distributions for individual proteins. Tandem spectra underlying each glycopeptide GADS peak are combined to create a second type of library that contains spectra of the underlying glycopeptide spectra. These were acquired by higher-energy (stepped) collision dissociation fragmentation followed by ion-trap fragmentation. Spectra are annotated using MS_Piano, recently reported annotation software. This data, with extensions of a widely used spectral library search and display software, provides accessible mass spectral libraries.

MS_Piano: A Software Tool for Annotating Peaks in CID Tandem Mass Spectra of Peptides and N-Glycopeptides.

Annotating product ion peaks in tandem mass spectra is essential for evaluating spectral quality and validating peptide identification. This task is more complex for glycopeptides and is crucial for the confident determination of glycosylation sites in glycoproteins. MS_Piano (Mass Spectrum Peptide Annotation) software was developed for reliable annotation of peaks in collision induced dissociation (CID) tandem mass spectra of peptides or N-glycopeptides for given peptide sequences, charge states, and optional modifications. The program annotates each peak in high or low resolution spectra with possible product ion(s) and the mass difference between the measured and theoretical m/z values. Spectral quality is measured by two major parameters: the ratio between the sum of unannotated vs all peak intensities in the top 20 peaks, and the intensity of the highest unannotated peak. The product ions of peptides, glycans, and glycopeptides in spectra are labeled in different class-type colors to facilitate interpretation. MS_Piano assists validating peptide and N-glycopeptide identification from database and library searches and provides quality control and optimizes search reliability in custom developed peptide mass spectral libraries. The software is freely available in .exe and .dll formats for the Windows operating system.

Representing and Comparing Site-Specific Glycan Abundance Distributions of Glycoproteins.

A method for representing and comparing distributions of N-linked glycans located at specific sites on proteins is presented. The representation takes the form of a simple mass spectrum for a given peptide sequence, with each peak corresponding to a different glycopeptide. The mass (in place of m/z) of each peak is that of the glycan mass, and its abundance corresponds to its relative abundance in the electrospray MS1 spectrum. This provides a facile means of representing all identifiable glycopeptides arising from a single protein "sequon" on a specific sequence, thereby enabling the comparison and searching of these distributions as routinely done for mass spectra. Likewise, these reference glycopeptide abundance distribution spectra (GADS) can be stored in searchable libraries. A set of such libraries created from available data is provided along with an adapted version of the widely used NIST-MS library-search software. Since GADS contain only MS1 abundances and identifications, they are equally suitable for expressing collision-induced fragmentation and electron-transfer dissociation determinations of glycopeptide identity. Comparisons of GADS for N-glycosylated sites on several proteins, especially the SARS-CoV-2 spike protein, demonstrate the potential reproducibility of GADS and their utility for comparing site-specific distributions.

CID Fragmentation of Deprotonated N-Acyl Aromatic Sulfonamides. Smiles-Type and Nitrogen-Oxygen Rearrangements.

The NIST tandem mass spectral library (2020 version) includes over 800 aromatic sulfonamides. In negative mode, upon collisional activation most benzenesulfonamides lose a neutral SO2 molecule leading to an anilide anion (C6H5NH-, m/z 92). However, for deprotonated N-benzoyl aromatic sulfonamides, the phenoxide ion (C6H5O-, m/z 93.0343) is the principal product ion. A variety of N-acylbenzenesulfonamide derivatives were also found to overwhelmingly produce the phenoxide ion as the most intense product ion. A mechanism is proposed in which, at low energy, a carbonyl oxygen atom (C═O) is transferred to a benzene ring, known as a Smiles-type rearrangement (the amide oxygen atom attacks the arylsulfonyl group at the ipso position), in parallel and determining the reaction at high energy a nitrogen-oxygen rearrangement mechanism leads to the formation of the phenoxide ion. Tandem mass spectra of deprotonated N-benzoyl-18O-benzenesulfonamide and N-thiobenzoyl-p-toluenesulfonamide confirmed the rearrangement since base peaks at m/z 95.0384 and 123.0270 which correspond to an 18O phenoxide ion ([C6H518O]-) and a 4-methylbenzenethiolate anion ([CH3C6H4S]-) were observed, respectively. The parallel mechanism is supported by the strong correlation between the observed product ion intensities and the corresponding activation energies obtained by Density Functional Theory calculations. This is an example of a relatively simple ion with a complex path to fragmentation, being a cautionary tale for indiscriminate use of in silico spectra in place of actual measurement.

Comprehensive Analysis of Tryptic Peptides Arising from Disulfide Linkages in NISTmAb and Their Use for Developing a Mass Spectral Library.

This work presents methods for identifying and then creating a mass spectral library for disulfide-linked peptides originating from the NISTmAb, a reference material of the humanized IgG1k monoclonal antibody (RM 8671). Analyses involved both partially reduced and non-reduced samples under neutral and weakly basic conditions followed by nanoflow liquid chromatography tandem mass spectrometry (LC-MS/MS). Spectra of peptides containing disulfide bonds are identified by both MS1 ion and MS2 fragment ion data in order to completely map all the disulfide linkages in the NISTmAb. This led to the detection of 383 distinct disulfide-linked peptide ions, arising from fully tryptic cleavage, missed cleavage, irregular cleavage, complex Met/Trp oxidation mixtures, and metal adducts. Fragmentation features of disulfide bonds under low-energy collision dissociation were examined. These include (1) peptide bond cleavage leaving disulfide bonds intact; (2) disulfide bond cleavage, often leading to extensive fragmentation; and (3) double cleavage products resulting from breakages of two peptide bonds or both peptide and disulfide bonds. Automated annotation of various complex MS/MS fragments enabled the identification of disulfide-linked peptides with high confidence. Peptides containing each of the nine native disulfide bonds were identified along with 86 additional disulfide linkages arising from disulfide bond shuffling. The presence of shuffled disulfides was nearly completely abrogated by refining digest conditions. A curated spectral library of 702 disulfide-linked peptide spectra was created from this analysis and is publicly available for free download. Since all IgG1 antibodies have the same constant regions, the resulting library can be used as a tool for facile identification of "hard-to-find" disulfide-bonded peptides. Moreover, we show that one may identify such peptides originating from IgG1 proteins in human serum, thereby serving as a means of monitoring the completeness of protein reduction in proteomics studies. Data are available via ProteomeXchange with identifier PXD023358.

Creation and filtering of a recurrent spectral library of CHO cell metabolites and media components.

This paper reports the first implementation of a new type of mass spectral library for the analysis of Chinese hamster ovary (CHO) cell metabolites that allows users to quickly identify most compounds in any complex metabolite sample. We also describe an annotation methodology developed to filter out artifacts and low-quality spectra from recurrent unidentified spectra of metabolites. CHO cells are commonly used to produce biological therapeutics. Metabolic profiles of CHO cells and media can be used to monitor process variability and look for markers that discriminate between batches of product. We have created a comprehensive library of both identified and unidentified metabolites derived from CHO cells that can be used in conjunction with tandem mass spectrometry to identify metabolites. In addition, we present a workflow that can be used for assigning confidence to a NIST MS/MS Library search match based on prior probability of general utility. The goal of our work is to annotate and identify (when possible), all liquid chromatography-mass spectrometry generated metabolite ions as well as create automatable library building and identification pipelines for use by others in the field.

Increasing the Coverage of a Mass Spectral Library of Milk Oligosaccharides Using a Hybrid-Search-Based Bootstrapping Method and Milks from a Wide Variety of Mammals.

This study significantly expands both the scope and method of identification for construction of a previously reported tandem mass spectral library of 74 human milk oligosaccharides (HMOs) derived from results of combined LC-MS/MS experiments and comprehensive structural analysis of HMOs. In the present work, a hybrid search "bootstrap" identification method was employed that substantially broadens the coverage of milk oligosaccharides and thereby increases utility use of a spectrum library-based method for the rapid tentative identification of all distinguishable glycans in milk. This involved hybrid searching of the previous library, which was itself constructed using the hybrid search of oligosaccharide spectra in the NIST 17 Tandem MS Library. The general approach appears applicable to library construction of other classes of compounds. The coverage of oligosaccharides was significantly extended using milks from a variety of mammals, including bovine, Asian buffalo, African lion, and goat. This new method led to the identification of another 145 oligosaccharides, including an additional 80 HMOs from reanalysis of human milk. The newly identified compounds were added to a freely available mass spectral reference database of 219 milk oligosaccharides. We also provide suggestions to overcome several limitations and pitfalls in the interpretation of spectra of unknown oligosaccharides.

Mass Spectral Library of Acylcarnitines Derived from Human Urine.

We describe the creation of a mass spectral library of acylcarnitines and conjugated acylcarnitines from the LC-MS/MS analysis of six NIST urine reference materials. To recognize acylcarnitines, we conducted in-depth analyses of fragmentation patterns of acylcarnitines and developed a set of rules, derived from spectra in the NIST17 Tandem MS Library and those identified in urine, using the newly developed hybrid search method. Acylcarnitine tandem spectra were annotated with fragments from carnitine and acyl moieties as well as neutral loss peaks from precursors. Consensus spectra were derived from spectra having similar retention time, fragmentation pattern, and the same precursor m/z and collision energy. The library contains 157 different precursor masses, 586 unique acylcarnitines, and 4 332 acylcarnitine consensus spectra. Furthermore, from spectra that partially satisfied the fragmentation rules of acylcarnitines, we identified 125 conjugated acylcarnitines represented by 987 consensus spectra, which appear to originate from Phase II biotransformation reactions. To our knowledge, this is the first report of conjugated acylcarnitines. The mass spectra provided by this work may be useful for clinical screening of acylcarnitines as well as for studying relationships among fragmentation patterns, collision energies, structures, and retention times of acylcarnitines. Further, these methods are extensible to other classes of metabolites.

Disparate Metabolomics Data Reassembler: A Novel Algorithm for Agglomerating Incongruent LC-MS Metabolomics Datasets.

In the past decade, the field of LC-MS-based metabolomics has transformed from an obscure specialty into a major "-omics" platform for studying metabolic processes and biomolecular characterization. However, as a whole the field is still very fractured, as the nature of the instrumentation and the information produced by the platform essentially creates incompatible "islands" of datasets. This lack of data coherency results in the inability to accumulate a critical mass of metabolomics data that has enabled other -omics platforms to make impactful discoveries and meaningful advances. As such, we have developed a novel algorithm, called Disparate Metabolomics Data Reassembler (DIMEDR), which attempts to bridge the inconsistencies between incongruent LC-MS metabolomics datasets of the same biological sample type. A single "primary" dataset is postprocessed via traditional means of peak identification, alignment, and grouping. DIMEDR utilizes this primary dataset as a progenitor template by which data from subsequent disparate datasets are reassembled and integrated into a unified framework that maximizes spectral feature similarity across all samples. This is accomplished by a novel procedure for universal retention time correction and comparison via identification of ubiquitous features in the initial primary dataset, which are subsequently utilized as endogenous internal standards during integration. For demonstration purposes, two human and two mouse urine metabolomics datasets from four unrelated studies acquired over 4 years were unified via DIMEDR, which enabled meaningful analysis across otherwise incomparable and unrelated datasets.

Sensitive Method for the Confident Identification of Genetically Variant Peptides in Human Hair Keratin.

Recent reports have demonstrated that genetically variant peptides derived from human hair shaft proteins can be used to differentiate individuals of different biogeographic origins. We report a method involving direct extraction of hair shaft proteins more sensitive than previously published methods regarding GVP detection. It involves one step for protein extraction and was found to provide reproducible results. A detailed proteomic analysis of this data is presented that led to the following four results: (i) A peptide spectral library was created and made available for download. It contains all identified peptides from this work, including GVPs that, when appropriately expanded with diverse hair-derived peptides, can provide a routine, reliable, and sensitive means of analyzing hair digests; (ii) an analysis of artifact peptides arising from side reactions is also made using a new method for finding unexpected modifications; (iii) detailed analysis of the gel-based method employed clearly shows the high degree of cross-linking or protein association involved in hair digestion, with major GVPs eluting over a wide range of high molecular weights while others apparently arise from distinct non-cross-linked proteins; and (v) finally, we show that some of the specific GVP identifications depend on the sample preparation method.

Hybrid Search: A Method for Identifying Metabolites Absent from Tandem Mass Spectrometry Libraries.

Metabolomics has a critical need for better tools for mass spectral identification. Common metabolites may be identified by searching libraries of tandem mass spectra, which offers important advantages over other approaches to identification. But tandem libraries are not nearly complete enough to represent the full molecular diversity present in complex biological samples. We present a novel hybrid search method that can help identify metabolites not in the library by similarity to compounds that are. We call it "hybrid" searching because it combines conventional, direct peak matching with the logical equivalent of neutral-loss matching. A successful hybrid search requires the library to contain "cognates" of the unknown: similar compounds with a structural difference confined to a single region of the molecule, that does not substantially alter its fragmentation behavior. We demonstrate that the hybrid search is highly likely to find similar compounds under such circumstances.