Improving mass and liquid chromatography based identification of proteins using bayesian scoring

We present a method for peptide and protein identification based on LC-MS profiling. The method identified peptides at high-throughput without expending the sequencing time necessary for CID spectra based identification. The measurable peptide properties of mass and liquid chromatographic elution conditions are used to characterize and differentiate peptide features, and these peptide features are matched to a reference database from previously acquired and archived LC-MS/MS experiments to generate sequence assignments. The matches are scored according to the probability of an overlap between the peptide feature and the database peptides resulting in a ranked list of possible peptide sequences for each peptide submitted. This method resulted in 6 times more peptide sequence identifications from a single LC-MS analysis of yeast than from shotgun peptide sequencing using LC-MS/MS.     

Enhanced identification of peptides lacking basic residues by LC-ESI-MS/MS analysis of singly charged peptides

Peptide sequences lacking basic residues (arginine, lysine, or histidine, referred to as "base-less") are of particular importance in proteomic experiments targeting protein C-termini or employing nontryptic proteases such as GluC or chymotrypsin. We demonstrate enhanced identification of base-less peptides by focused analysis of singly charged precursors in liquid chromatography (LC) electrospray ionization (ESI) tandem mass spectrometry (MS/MS). Singly charged precursors are often excluded from fragmentation and sequence analysis in LC-MS/MS. We generated different pools of base-less and base-containing peptides by tryptic and nontryptic digestion of bacterial proteomes. Focused LC-MS/MS analysis of singly charged precursor ions yielded predominantly base-less peptide identifications. Similar numbers of base-less peptides were identified by LC-MS/M Sanalysis targeting multiply charged precursors. There was little redundancy between the base-less sequences derived by both MS/MS schemes. In the present experimental outcome, additional LC-MS/MS analysis of singly charged precursors substantially increased the identification rate of base-less sequences derived from multiply charged precursors. In conclusion, LC-MS/MS based identification of base-less peptides is substantially enhanced by additional focused analysis of singly charged precursors.     

An improved model for prediction of retention times of tryptic peptides in ion pair reversed-phase HPLC: its application to protein peptide mapping by off-line HPLC-MALDI MS

The proposed model is based on the measurement of the retention times of 346 tryptic peptides in the 560- to 4,000-Da mass range, derived from a mixture of 17 protein digests. These peptides were measured in HPLC-MALDI MS runs, with peptide identities confirmed by MS/MS. The model relies on summation of the retention coefficients of the individual amino acids, as in previous approaches, but additional terms are introduced that depend on the retention coefficients for amino acids at the N-terminal of the peptide. In the 17-protein mixture, optimization of two sets of coefficients, along with additional compensation for peptide length and hydrophobicity, yielded a linear dependence of retention time on hydrophobicity, with an R2 value about 0.94. The predictive capability of the model was used to distinguish peptides with close m/z values and for detailed peptide mapping of selected proteins. Its applicability was tested on columns of different sizes, from nano- to narrow-bore, and for direct sample injection, or injection via a pre-column. It can be used for accurate prediction of retention times for tryptic peptides on reversed-phase (300-A pore size) columns of different sizes with a linear water-ACN gradient and with TFA as the ion-pairing modifier.     

Tools for glycoproteomic analysis: size exclusion chromatography facilitates identification of tryptic glycopeptides with N-linked glycosylation sites

Proteomic techniques, such as HPLC coupled to tandem mass spectrometry (LC-MS/MS), have proved useful for the identification of specific glycosylation sites on glycoproteins (glycoproteomics). Glycosylation sites on glycopeptides produced by trypsinization of complex glycoprotein mixtures, however, are particularly difficult to identify both because a repertoire of glycans may be expressed at a particular glycosylation site, and because glycopeptides are usually present in relatively low abundance (2% to 5%) in peptide mixtures compared to nonglycosylated peptides. Previously reported methods to facilitate glycopeptide identification require either several pre-enrichment steps, involve complex derivatization procedures, or are restricted to a subset of all the glycan structures that are present in a glycoprotein mixture. Because the N-linked glycans expressed on tryptic glycopeptides contribute substantially to their mass, we demonstrate that size exclusion chromatography (SEC) provided a significant enrichment of N-linked glycopeptides relative to nonglycosylated peptides. The glycosylated peptides were then identified by LC-MS/MS after treatment with PNGase-F by the monoisotopic mass increase of 0.984 Da caused by the deglycosylation of the peptide. Analyses performed on human serum showed that this SEC glycopeptide isolation procedure results in at least a 3-fold increase in the total number of glycopeptides identified by LC-MS/MS, demonstrating that this simple, nonselective, rapid method is an effective tool to facilitate the identification of peptides with N-linked glycosylation sites.     

Informatics for peptide retention properties in proteomic LC-MS

Retention times in HPLC yield valuable information for the identification of various analytes and the prediction of peptide retention is useful for the identification of peptides/proteins in LC-MS-based proteomics. Informatics methods such as artificial neural networks and support vector machines capable of solving nonlinear problems made possible the accurate modeling of quantitative structure-retention relationships of peptides (including large polymers) up to 5 kDa to which classical linear models cannot be applied, as well as the proteome-wide prediction of peptide retention. Proteome-wide retention prediction and accurate mass-information facilitate the identification of peptides in complex proteomic samples. In this review, we address recent developments in solid informatics methods and their application to peptide-retention properties in 'bottom-up' shotgun proteomics. We also describe future prospects for the standardization and application of retention times.     

Identification of tryptic peptides from large databases using multiplexed tandem mass spectrometry: simulations and experimental results

Multiplexed tandem mass spectrometry (MS/MS) has recently been demonstrated as a means to increase the throughput of peptide identification in liquid chromatography (LC) MS/MS experiments. In this approach, a set of parent species is dissociated simultaneously and measured in a single spectrum (in the same manner that a single parent ion is conventionally studied), providing a gain in sensitivity and throughput proportional to the number of species that can be simultaneously addressed. In the present work, simulations performed using the Caenorhabditis elegans predicted proteins database show that multiplexed MS/MS data allow the identification of tryptic peptides from mixtures of up to ten peptides from a single dataset with only three "y" or "b" fragments per peptide and a mass accuracy of 2.5 to 5 ppm. At this level of database and data complexity, 98% of the 500 peptides considered in the simulation were correctly identified. This compares favorably with the rates obtained for classical MS/MS at more modest mass measurement accuracy. LC multiplexed Fourier transform-ion cyclotron resonance MS/MS data obtained from a 66 kDa protein (bovine serum albumin) tryptic digest sample are presented to illustrate the approach, and confirm that peptides can be effectively identified from the C. elegans database to which the protein sequence had been appended.     

Computational approach for identification and characterization of GPI-anchored peptides in proteomics experiments

Genes that encode glycosylphosphatidylinositol anchored proteins (GPI-APs) constitute an estimated 1-2% of eukaryote genomes. Current computational methods for the prediction of GPI-APs are sensitive and specific; however, the analysis of the processing site (omega- or omega-site) of GPI-APs is still challenging. Only 10% of the proteins that are annotated as GPI-APs have the omega-site experimentally verified. We describe an integrated computational and experimental proteomics approach for the identification and characterization of GPI-APs that provides the means to identify GPI-APs and the derived GPI-anchored peptides in LC-MS/MS data sets. The method takes advantage of sequence features of GPI-APs and the known core structure of the GPI-anchor. The first stage of the analysis encompasses LC-MS/MS based protein identification. The second stage involves prediction of the processing sites of the identified GPI-APs and prediction of the corresponding terminal tryptic peptides. The third stage calculates possible GPI structures on the peptides from stage two. The fourth stage calculates the scores by comparing the theoretical spectra of the predicted GPI-peptides against the observed MS/MS spectra. Automated identification of C-terminal GPI-peptides from porcine membrane dipeptidase, folate receptor and CD59 in complex LC-MS/MS data sets demonstrates the sensitivity and specificity of this integrated computational and experimental approach.     

Detection and identification of sub-nanogram levels of protein in a nanoLC-trypsin-MS system

Proteomic workflows involving liquid-based protein separations are an alternative to gel-based protein analysis, however the trypsin digestion procedure is usually difficult to implement, particularly when processing low abundance proteins from capillary column effluent. To convert the protein to peptides for the purpose of identification, current protocols require several sample handling steps, and sample losses become an issue. In this study, we present an improved system that conducts reversed-phase protein chromatography and rapid on-line tryptic digestion requiring sub-nanogram quantities of protein. This system employs a novel mirror-gradient concept that allows for dynamic titration of the column effluent to create optimal conditions for real-time tryptic digestion. The purpose behind this development was to improve the limits of detection of the online concept, to support flow-based alternatives to gel-based proteomics and to simplify the characterization of low abundance proteins. Using test mixtures of proteins, we show that peptide mass fingerprinting with high sequence representation can be easily achieved at the 20 fmol level, with detection limits down to 5 fmol (85 pg myoglobin). Limits of identification using standard data-dependent MS/MS experiments are as low as 10 fmol. These results suggest that the nanoLC-trypsin-MS/MS system could represent an alternative to the conventional "1D-gel to MS" proteomic strategy.     

Highly multiplexed targeted proteomics using precise control of peptide retention time

Large-scale proteomics applications using SRM analysis on triple quadrupole mass spectrometers present new challenges to LC-MS/MS experimental design. Despite the automation of building large-scale LC-SRM methods, the increased numbers of targeted peptides can compromise the balance between sensitivity and selectivity. To facilitate large target numbers, time-scheduled SRM transition acquisition is performed. Previously published results have demonstrated incorporation of a well-characterized set of synthetic peptides enabled chromatographic characterization of the elution profile for most endogenous peptides. We have extended this application of peptide trainer kits to not only build SRM methods but to facilitate real-time elution profile characterization that enables automated adjustment of the scheduled detection windows. Incorporation of dynamic retention time adjustments better facilitate targeted assays lasting several days without the need for constant supervision. This paper provides an overview of how the dynamic retention correction approach identifies and corrects for commonly observed LC variations. This adjustment dramatically improves robustness in targeted discovery experiments as well as routine quantification experiments.     

Definition and characterization of a "trypsinosome" from specific peptide characteristics by nano-HPLC-MS/MS and in silico analysis of complex protein mixtures

Although HPLC-ESI-MS/MS is rapidly becoming an indispensable tool for the analysis of peptides in complex mixtures, the sequence coverage it affords is often quite poor. Low protein expression resulting in peptide signal intensities that fall below the limit of detection of the MS system in combination with differences in peptide ionization efficiency plays a significant role in this. A second important factor stems from differences in physicochemical properties of each peptide and how these properties relate to chromatographic retention and ultimate detection. To identify and understand those properties, we compared data from experimentally identified peptides with data from peptides predicted by in silico digest of all corresponding proteins in the experimental set. Three different complex protein mixtures extracted were used to define a training set to evaluate the amino acid retention coefficients based on linear regression analysis. The retention coefficients were also compared with other previous hydrophobic and retention scale. From this, we have constructed an empirical model that can be readily used to predict peptides that are likely to be observed on our HPLC-ESI-MS/MS system based on their physicochemical properties. Finally, we demonstrated that in silico prediction of peptides and their retention coefficients can be used to generate an inclusion list for a targeted mass spectrometric identification of low abundance proteins in complex protein samples. This approach is based on experimentally derived data to calibrate the method and therefore may theoretically be applied to any HPLC-MS/MS system on which data are being generated.     

A systematical analysis of tryptic peptide identification with reverse phase liquid chromatography and electrospray ion trap mass spectrometry

In this study we systematically analyzed the elution condition of tryptic peptides and the characteristics of identified peptides in reverse phase liquid chromatography and electrospray tandem mass spectrometry (RPLC-MS/MS) analysis. Following protein digestion with trypsin, the peptide mixture was analyzed by on-line RPLC-MS/MS. Bovine serum albumin (BSA) was used to optimize acetonitrile (ACN) elution gradient for tryptic peptides, and Cytochrome C was used to retest the gradient and the sensitivity of LC-MS/MS. The characteristics of identified peptides were also analyzed. In our experiments, the suitable ACN gradient is 5% to 30% for tryptic peptide elution and the sensitivity of LC-MS/MS is 50 fmol.Analysis of the tryptic peptides demonstrated that longer (more than 10 amino acids) and multi-charge state ( 2, 3) peptides are likely to be identified, and the hydropathicity of the peptides might not be related to whether it is more likely to be identified or not. The number of identified peptides for a protein might be used to estimate its loading amount under the same sample background. Moreover, in this study the identified peptides present three types of redundancy, namely identification, charge, and sequence redundancy, which may repress low abundance protein identification.     

Protein identification assisted by the prediction of retention time in liquid chromatography/tandem mass spectrometry

Two-dimensional liquid chromatography (2D-LC) coupled on-line with electrospray ionization tandem mass spectrometry (2D-LC-ESI-MS/MS) is a new platform for analysis and identification of proteome. Peptides are separated by 2D-LC and then performed MS/MS analysis by tandem MS/MS. The MS/MS data are searched against database for protein identification. In one 2D-LC-ESI-MS/MS run, we obtained not only the structural information of peptides directly from MS/MS, but also the retention time of peptides eluted from LC. Information on the chromatographic behavior of peptides can assist protein identification in the new platform for proteomics. The retention time of the matching peptides of the identified protein was predicted by the hydrophobic contribute of each amino acid on reversed-phase liquid chromatography (RPLC). By using this strategy proteins were identified by four types of information: peptide mass fingerprinting (PMF), sequence query, and MS/MS ions searched and the predicted retention time. This additional information obtained from LC could assist protein identification with no extra experimental cost.     

Aligning LC peaks by converting gradient retention times to retention index of peptides in proteomic experiments

MOTIVATION: Liquid chromatography-tandem mass spectrometry (LC-MS/MS) is a powerful tool in proteomics studies, but when peptide retention information is used for identification purposes, it remains challenging to compare multiple LC-MS/MS runs or to match observed and predicted retention times, because small changes of LC conditions unavoidably lead to variability in retention times. In addition, non-contiguous retention data obtained with different LC-MS instruments or in different laboratories must be aligned to confirm and utilize rapidly accumulating published proteomics data. RESULTS: We have developed a new alignment method for peptide retention times based on linear solvent strength (LSS) theory. We found that log k(0) (logarithm of retention factor for a given organic solvent) in the LSS theory can be utilized as a 'universal' retention index of peptides (RIP) that is independent of LC gradients, and depends solely on the constituents of the mobile phase and the stationary phases. We introduced a machine learning-based scheme to optimize the conversion function of gradient retention times (t(g)) to log k(0). Using the optimized function, t(g) values obtained with different LC-MS systems can be directly compared with each other on the RIP scale. In an examination of Arabidopsis proteomic data, the vast majority of retention time variability was removed, and five datasets obtained with various LC-MS systems were successfully aligned on the RIP scale.     

The importance of peptide detectability for protein identification, quantification, and experiment design in MS/MS proteomics

Peptide detectability is defined as the probability that a peptide is identified in an LC-MS/MS experiment and has been useful in providing solutions to protein inference and label-free quantification. Previously, predictors for peptide detectability trained on standard or complex samples were proposed. Although the models trained on complex samples may benefit from the large training data sets, it is unclear to what extent they are affected by the unequal abundances of identified proteins. To address this challenge and improve detectability prediction, we present a new algorithm for the iterative learning of peptide detectability from complex mixtures. We provide evidence that the new method approximates detectability with useful accuracy and, based on its design, can be used to interpret the outcome of other learning strategies. We studied the properties of peptides from the bacterium Deinococcus radiodurans and found that at standard quantities, its tryptic peptides can be roughly classified as either detectable or undetectable, with a relatively small fraction having medium detectability. We extend the concept of detectability from peptides to proteins and apply the model to predict the behavior of a replicate LC-MS/MS experiment from a single analysis. Finally, our study summarizes a theoretical framework for peptide/protein identification and label-free quantification.     

Comprehensive peptidome analysis of mouse livers by size exclusion chromatography prefractionation and nanoLC-MS/MS identification

Peptidome analysis has received increasing attention in recent years. Cancer diagnosis by serum peptidome has also been reported by peptides' profiling for discovery of peptide biomarkers. Tissue, which may have a higher biomarker concentration than blood, has not been investigated extensively by means of peptidome analysis. Here, a method for the peptidome analysis of mouse liver was developed by the combination of size exclusion chromatography (SEC) prefractionation with nano-liquid chromatography-tamdem mass spectrometry (nanoLC-MS/MS) analysis. The extracted peptides from mouse liver were separated according to their molecular weight using a size exclusion column. MALDI-TOF MS was used to characterize the molecular weight distribution of the peptides in fractions eluted from the SEC column. The low molecular weight (LMW) (MW < 3000 Da) peptides in the collected fractions were directly analyzed by LC-MS/MS which resulted in the identification of 1181 unique peptides (from 371 proteins). The high molecular weight (HMW) (MW > 3000 Da) peptides in the early two fractions from the SEC column were first digested with trypsin, and the resulted digests were then analyzed by LC-MS/MS, which led to the identification of 123 and 127 progenitor proteins of the HMW peptides in fractions 1 and 2, respectively. Analysis of the peptides' cleavage sites showed that the peptides are cleaved in regulation, which may reflect the protease activity and distribution in body, and also represent the biological state of the tissue and provide a fresh source for biomarker discovery.     

Simulating and validating proteomics data and search results

The computational simulation of complete proteomic data sets and their utility to validate detection and interpretation algorithms, to aid in the design of experiments and to assess protein and peptide false discovery rates is presented. The simulation software has been developed for emulating data originating from data-dependent and data-independent LC-MS workflows. Data from all types of commonly used hybrid mass spectrometers can be simulated. The algorithms are based on empirically derived physicochemical liquid and gas phase models for proteins and peptides. Sample composition in terms of complexity and dynamic range, as well as chromatographic, experimental and MS conditions, can be controlled and adjusted independently. The effect of on-column amounts, gradient length, mass resolution and ion mobility on search specificity will be demonstrated using tryptic peptides from human and yeast cellular lysates simulated over five orders of magnitude in dynamic range. Initial justification of the simulated data sets is achieved by comparing and contrasting the in silico simulated data to experimentally derived results from a 48 protein mixture, spanning a similar magnitude of five orders of magnitude. Additionally, experimental data from replicate and dilutions series experiments will be utilized to determine error rates at the peptide and protein level with respect to mass, area, retention and drift time. The data presented reveal a high degree of similarity at the ion detection, peptide and protein level when analyzed under similar conditions.     

Identification and quantification of DNA repair proteins by liquid chromatography/isotope-dilution tandem mass spectrometry using their fully 15N-labeled analogues as internal standards

Oxidatively induced DNA damage is implicated in disease, unless it is repaired by DNA repair. Defects in DNA repair capacity may be a risk factor for various disease processes. Thus, DNA repair proteins may be used as early detection and therapeutic biomarkers in cancer and other diseases. For this purpose, the measurement of the expression level of these proteins in vivo will be necessary. We applied liquid chromatography/isotope-dilution tandem mass spectrometry (LC-MS/MS) for the identification and quantification of DNA repair proteins human 8-hydroxyguanine-DNA glycosylase (hOGG1) and Escherichia coli formamidopyrimidine DNA glycosylase (Fpg), which are involved in base-excision repair of oxidatively induced DNA damage. We overproduced and purified (15)N-labeled analogues of these proteins to be used as suitable internal standards to ensure the accuracy of quantification. Unlabeled and (15)N-labeled proteins were digested with trypsin and analyzed by LC-MS/MS. Numerous tryptic peptides of both proteins were identified on the basis of their full-scan mass spectra. These peptides matched the theoretical peptide fragments expected from trypsin digestion and provided statistically significant protein scores that would unequivocally identify these proteins. We also recorded the product ion spectra of the tryptic peptides and defined the characteristic product ions. Mixtures of the analyte proteins and their (15)N-labeled analogues were analyzed by selected-reaction monitoring on the basis of product ions. The results obtained suggest that the methodology developed would be highly suitable for the positive identification and accurate quantification of DNA repair proteins in vivo as potential biomarkers for cancer and other diseases.     

MSSimulator: Simulation of mass spectrometry data

Mass spectrometry coupled to liquid chromatography (LC-MS and LC-MS/MS) is commonly used to analyze the protein content of biological samples in large scale studies, enabling quantitation and identification of proteins and peptides using a wide range of experimental protocols, algorithms, and statistical models to analyze the data. Currently it is difficult to compare the plethora of algorithms for these tasks. So far, curated benchmark data exists for peptide identification algorithms but data that represents a ground truth for the evaluation of LC-MS data is limited. Hence there have been attempts to simulate such data in a controlled fashion to evaluate and compare algorithms. We present MSSimulator, a simulation software for LC-MS and LC-MS/MS experiments. Starting from a list of proteins from a FASTA file, the simulation will perform in-silico digestion, retention time prediction, ionization filtering, and raw signal simulation (including MS/MS), while providing many options to change the properties of the resulting data like elution profile shape, resolution and sampling rate. Several protocols for SILAC, iTRAQ or MS(E) are available, in addition to the usual label-free approach, making MSSimulator the most comprehensive simulator for LC-MS and LC-MS/MS data.     

More than 100,000 detectable peptide species elute in single shotgun proteomics runs but the majority is inaccessible to data-dependent LC-MS/MS

Shotgun proteomics entails the identification of as many peptides as possible from complex mixtures. Here we investigate how many peptides are detectable by high resolution MS in standard LC runs of cell lysate and how many of them are accessible to data-dependent MS/MS. Isotope clusters were determined by MaxQuant and stringently filtered for charge states and retention times typical of peptides. This resulted in more than 100,000 likely peptide features, of which only about 16% had been targeted for MS/MS. Three instrumental attributes determine the proportion of additional peptides that can be identified: sequencing speed, sensitivity, and precursor ion isolation. In our data, an MS/MS scan rate of 25/s would be necessary to target all peptide features, but this drops to less than 17/s for reasonably abundant peptides. Sensitivity is a greater challenge, with many peptide features requiring long MS/MS injection times (>250 ms). The greatest limitation, however, is the generally low proportion of the target peptide ion intensity in the MS/MS selection window (the "precursor ion fraction" or PIF). Median PIF is only 0.14, making the peptides difficult to identify by standard MS/MS methods. Our results aid in developing strategies to further increase coverage in shotgun proteomics.