Peptide Retention Time Prediction Research Articles

Intensity and retention time prediction improves the rescoring of protein-nucleic acid cross-links.

In protein-RNA cross-linking mass spectrometry, UV or chemical cross-linking introduces stable bonds between amino acids and nucleic acids in protein-RNA complexes that are then analyzed and detected in mass spectra. This analytical tool delivers valuable information about RNA-protein interactions and RNA docking sites in proteins, both in vitro and in vivo. The identification of cross-linked peptides with oligonucleotides of different length leads to a combinatorial increase in search space. We demonstrate that the peptide retention time prediction tasks can be transferred to the task of cross-linked peptide retention time prediction using a simple amino acid composition encoding, yielding improved identification rates when the prediction error is included in rescoring. For the more challenging task of including fragment intensity prediction of cross-linked peptides in the rescoring, we obtain, on average, a similar improvement. Further improvement in the encoding and fine-tuning of retention time and intensity prediction models might lead to further gains, and merit further research.

Toward an Ultimate Solution for Peptide Retention Time Prediction: The Effect of Column Temperature on Separation Selectivity.

We studied the effect of the column temperature on the selectivity of reversed-phase peptide separation in bottom-up proteomics. The number of peptide identifications from 2 h liquid chromatography with tandem mass spectrometry (LC-MS/MS) acquisitions reaches a plateau at 45-55 °C, driven simultaneously by improved separation efficiency, a gradual decrease in peptide retention, and possible on-column degradation of peptides at elevated temperatures. Performing 2D LC-MS/MS acquisitions at 25, 35, 45, and 55 °C resulted in the identification of ∼100,000 and ∼120,000 unique peptides for nonmodified and tandem mass tags (TMT)-labeled samples, respectively. These peptide collections were used to investigate the temperature-driven retention features. The latter is governed by the specific temperature response of individual residues, peptide hydrophobicity and length, and amphipathic helicity. On average, peptide retention decreased by 0.56 and 0.5% acetonitrile for each 10 °C increase for label-free and TMT-labeled peptides, respectively. This generally linear response of retention shifts allowed the extrapolation of predictive models beyond the studied temperature range. Thus, (trap) column cooling from room temperature to 0 °C will allow the retention of an additional 3% of detectable tryptic peptides. Meanwhile, the application of 90 °C would result in the loss of ∼20% of tryptic peptides that were amenable to MS/MS-based identification.

MARS an improved de novo peptide candidate selection method for non-canonical antigen target discovery in cancer

Understanding the nature and extent of non-canonical human leukocyte antigen (HLA) presentation in tumour cells is a priority for target antigen discovery for the development of next generation immunotherapies in cancer. We here employ a de novo mass spectrometric sequencing approach with a refined, MHC-centric analysis strategy to detect non-canonical MHC-associated peptides specific to cancer without any prior knowledge of the target sequence from genomic or RNA sequencing data. Our strategy integrates MHC binding rank, Average local confidence scores, and peptide Retention time prediction for improved de novo candidate Selection; culminating in the machine learning model MARS. We benchmark our model on a large synthetic peptide library dataset and reanalysis of a published dataset of high-quality non-canonical MHC-associated peptide identifications in human cancer. We achieve almost 2-fold improvement for high quality spectral assignments in comparison to de novo sequencing alone with an estimated accuracy of above 85.7% when integrated with a stepwise peptide sequence mapping strategy. Finally, we utilize MARS to detect and validate lncRNA-derived peptides in human cervical tumour resections, demonstrating its suitability to discover novel, immunogenic, non-canonical peptide sequences in primary tumour tissue.

Exploring the variable space of shallow machine learning models for reversed-phase retention time prediction

Peptide retention time (RT) prediction algorithms are tools to study and identify the physicochemical properties that drive the peptide-sorbent interaction. Traditional RT algorithms use multiple linear regression with manually curated parameters to determine the degree of direct contribution for each parameter and improvements to RT prediction accuracies relied on superior feature engineering. Deep learning led to a significant increase in RT prediction accuracy and automated feature engineering via chaining multiple learning modules. However, the significance and the identity of these extracted variables are not well understood due to the inherent complexity when interpreting “relationships-of-relationships” found in deep learning variables. To achieve both accuracy and interpretability simultaneously, we isolated individual modules used in deep learning and the isolated modules are the shallow learners employed for RT prediction in this work. Using a shallow convolutional neural network (CNN) and gated recurrent unit (GRU), we find that the spatial features obtained via the CNN correlate with real-world physicochemical properties namely cross-collisional sections (CCS) and variations of assessable surface area (ASA). Furthermore, we determined that the discovered parameters are “micro-coefficients” that contribute to the “macro-coefficient” – hydrophobicity. Manually embedding CCS and the variations of ASA to the GRU model yielded an R2 = 0.981 using only 525 variables and can represent 88% of the ∼110,000 tryptic peptides used in our dataset. This work highlights the feature discovery process of our shallow learners can achieve beyond traditional RT models in performance and have better interpretability when compared with the deep learning RT algorithms found in the literature.

DeepLC can predict retention times for peptides that carry as-yet unseen modifications.

The inclusion of peptide retention time prediction promises to remove peptide identification ambiguity in complex liquid chromatography-mass spectrometry identification workflows. However, due to the way peptides are encoded in current prediction models, accurate retention times cannot be predicted for modified peptides. This is especially problematic for fledgling open searches, which will benefit from accurate retention time prediction for modified peptides to reduce identification ambiguity. We present DeepLC, a deep learning peptide retention time predictor using peptide encoding based on atomic composition that allows the retention time of (previously unseen) modified peptides to be predicted accurately. We show that DeepLC performs similarly to current state-of-the-art approaches for unmodified peptides and, more importantly, accurately predicts retention times for modifications not seen during training. Moreover, we show that DeepLC's ability to predict retention times for any modification enables potentially incorrect identifications to be flagged in an open search of a wide variety of proteome data.

Boosting MS1-only Proteomics with Machine Learning Allows 2000 Protein Identifications in Single-Shot Human Proteome Analysis Using 5 min HPLC Gradient.

Proteome-wide analyses rely on tandem mass spectrometry and the extensive separation of proteolytic mixtures. This imposes considerable instrumental time consumption, which is one of the main obstacles in the broader acceptance of proteomics in biomedical and clinical research. Recently, we presented a fast proteomic method termed DirectMS1 based on ultrashort LC gradients as well as MS1-only mass spectra acquisition and data processing. The method allows significant reduction of the proteome-wide analysis time to a few minutes at the depth of quantitative proteome coverage of 1000 proteins at 1% false discovery rate (FDR). In this work, to further increase the capabilities of the DirectMS1 method, we explored the opportunities presented by the recent progress in the machine-learning area and applied the LightGBM decision tree boosting algorithm to the scoring of peptide feature matches when processing MS1 spectra. Furthermore, we integrated the peptide feature identification algorithm of DirectMS1 with the recently introduced peptide retention time prediction utility, DeepLC. Additional approaches to improve the performance of the DirectMS1 method are discussed and demonstrated, such as using FAIMS for gas-phase ion separation. As a result of all improvements to DirectMS1, we succeeded in identifying more than 2000 proteins at 1% FDR from the HeLa cell line in a 5 min gradient LC-FAIMS/MS1 analysis. The data sets generated and analyzed during the current study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD023977.

Confident Identification of Citrullination and Carbamylation Assisted by Peptide Retention Time Prediction.

The chromatographic behavior of peptides carrying citrulline and homocitrulline residues in proteomic two-dimensional (2D) liquid chromatography-mass spectrometry (LC-MS) experiments has been investigated. The primary goal of this study was to determine the chromatographic conditions that allow differentiating between arginine citrullination and deamidation of asparagine based on retention data, improving the confidence of MS-based identifications. Carbamylation was used as a reference point due to a high degree of similarity between modification products and anticipated changes in chromatographic behavior. We applied 2D LC-MS/MS (a high-pH-low-pH reversed phase (RP), hydrophilic interaction liquid chromatography (HILIC)-low-pH RP, and strong cation exchange (SCX)-low-pH RP) to acquire retention data for modified-nonmodified peptide pairs in the four separation modes. Modifications of a standard protein mixture were induced enzymatically (PAD-2) or chemically (urea) for citrullination and carbamylation, respectively. Deamidation occurs spontaneously. Similar retention shifts were observed for all three modifications in a high-pH RP (decrease) and a low-pH RP (increase), thus limiting the applicability of this 2D LC combination. HILIC on bare silica and strong cation exchange separations have been probed to amplify the effect of charge loss upon citrullination, with SCX demonstrating the most differentiating power: the elimination of basic residues upon citrullination/carbamylation results in an ∼58 mM KCl retention decrease, while retention of deamidated products decreases slightly.

Peptide retention time prediction in hydrophilic interaction liquid chromatography: Zwitter-ionic sulfoalkylbetaine and phosphorylcholine stationary phases

Peptide retention time prediction models have been developed for zwitter-ionic ZIC-HILIC and ZIC-cHILIC stationary phases (pH 4.5 eluents) using proteomics-derived retention datasets of ~30 thousand tryptic peptides each. Overall, hydrophilicity of these stationary phases was found to be similar to the previously studied Amide HILIC phase, but lower compared to bare silicas. Peptide retention is driven by interactions of all charged (hydrophilic) residues at pH 4.5 (Asp, Glu, Arg, Lys, His), but shows specificity according to orientation of functional groups in zwitter-ionic pair. Thus, ZIC-cHILIC exhibits an increased contribution of negatively charged Asp and Glu due to the distal positioning of positively charged quaternary amines on the stationary phase. These findings confirm that HILIC interactions are driven by both peptide distribution between water layer adsorbed on the stationary phase and by interactions specific to functional groups of the packing material. Sequence-Specific Retention Calculator HILIC models were optimized for these columns showing 0.967–0.976 R2-values between experimental and predicted retention values. ZIC-HILIC separations represent a good choice as a first dimension in 2D LC-MS of peptide mixtures with correlations between retention values of ZIC-HILIC against RPLC found at 0.197 (ZIC-HILIC) and 0.137 (ZIC-cHILIC) R2-values, confirming a good orthogonality.

Improved Peptide Retention Time Prediction in Liquid Chromatography through Deep Learning.

The accuracy of peptide retention time (RT) prediction model in liquid chromatography (LC) is still not sufficient for wider implementation in proteomics practice. Herein, we propose deep learning as an ideal tool to considerably improve this prediction. A new peptide RT prediction tool, DeepRT, was designed using a capsule network model, and the public data sets containing peptides separated by reverse-phase liquid chromatography were used to evaluate the DeepRT performance. Compared with other prevailing RT predictors, DeepRT attained overall improvement in the prediction of peptide RTs with an R2 of ∼0.994. Moreover, DeepRT was able to accommodate to the peptides that were separated by different types of LC, such as strong cation exchange (SCX) and hydrophilic interaction liquid chromatography (HILIC) and to reach the RT prediction with R2 values of ∼0.996 for SCX and ∼0.993 for HILIC, respectively. If a large peptide data set is available for one type of LC, DeepRT can be promoted to DeepRT(+) using transfer learning. Based on a large peptide data set gained from SWATH, DeepRT(+) further elevated the accuracy of RT prediction for peptides in a small data set and enabled a satisfactory prediction upon limited peptides approximating hundreds. Further, DeepRT automatically learns retention-related properties of amino acids under different separation mechanisms, which are well consistent with retention coefficients (Rc) of the amino acids. DeepRT was thus proven to be an improved RT predictor with high flexibility and efficiency. DeepRT is available at https://github.com/horsepurve/DeepRTplus .

Peptide retention time prediction for immobilized artificial membrane phosphatidylcholine stationary phase: method development and preliminary observations

Development of the first peptide retention prediction model for immobilized artificial membrane phosphatidylcholine (IAM.PC) stationary phase is reported. 2D LC-MS/MS analysis of a whole cell lysate of S. cerevisiae yielded a retention dataset of ~29,500 tryptic peptides; sufficient for confident assignment of retention coefficients which determine the contribution of individual amino acids in peptide retention. Retention data from the first dimension was used for the modeling: IAM.PC DD2 column, pH 7.4 ammonium bicarbonate, and water/acetonitrile gradient. Peptide separation using IAM.PC was compared to a standard C18 phase (Luna C18(2)). There was a significant reduction in peptide retention (~14% acetonitrile on average), indicating that the phosphatidylcholine stationary phase is significantly more hydrophilic. In comparison to the C18 phase, we found a substantial increase in the relative retention contribution for the positively charged Arg and Lys, and the aromatic Tyr, Trp and His residues. We also observed a decrease in retention contribution for the negatively charged Asp and Glu. This indicates an involvement of electrostatic interactions with the glycerophosphate functional groups, and possibly, delocalization effects from a hydrogen bond between the phosphate group and aromatic side chains in the separation mechanism.

FractionOptimizer: a method for optimal peptide fractionation in bottom-up proteomics.

Recent advances in mass spectrometry and separation technologies created the opportunities for deep proteome characterization using shotgun proteomics approaches. The "real world" sample complexity and high concentration range limit the sensitivity of this characterization. The common strategy for increasing the sensitivity is sample fractionation prior to analysis either at the protein or the peptide level. Typically, fractionation at the peptide level is performed using linear gradient high-performance liquid chromatography followed by uniform fraction collection. However, this way of peptide fractionation results in significantly suboptimal operation of the mass spectrometer due to the non-uniform distribution of peptides between the fractions. In this work, we propose an approach based on peptide retention time prediction allowing optimization of chromatographic conditions and fraction collection procedures. An open-source software implementing the approach called FractionOptimizer was developed and is available at http://hg.theorchromo.ru/FractionOptimizer . The performance of the developed tool was demonstrated for human embryonic kidney (HEK293) cell line lysate. In these experiments, we improved the uniformity of the peptides distribution between fractions. Moreover, in addition to 13,492 peptides, we found 6787 new peptides not identified in the experiments without fractionation and up to 800 new proteins (or 25%). Graphical abstract The analysis workflow employing FractionOptimizer software.

Peptide retention time prediction in hydrophilic interaction liquid chromatography. Comparison of separation selectivity between bare silica and bonded stationary phases

Peptide separation selectivity of four different HILIC sorbents was compared through the development of sequence-specific retention prediction algorithms for the datasets of 36–40 thousands peptides each. Hydrophilicity of these sorbents at pH 4.5 (peptide retention under acetonitrile:water gradients) increases in the following order: Luna HILIC < XBridge Amide < Atlantis Silica ∼ Luna HILIC Silica. Bare silica phases are characterized by higher retention coefficients for basic residues (Arg, Lys, His), while interactions with neutral HILIC phases is driven by interaction with the charged residues (Asp, Glu, Arg, Lys, His). Such difference is caused by electrostatic repulsion of negatively charged side chains of Asp and Glu from silanol groups on bare silica sorbents. The sequence-specific features characteristic for both bonded and bare-silica supports were established. Variation of retention coefficients at N- and C-termini showed a peptide orientation effect on silica matrices. The negatively charged surface of silica attracts positively charged groups, thus increasing interaction of N-terminal residues. The effect of peptide helicity is also different between two types of sorbents. The contribution of amphipathic helicity was found to be higher for more hydrophilic silica matrices, causing a reduced predictive accuracy of the models: all four Sequence-Specific Retention Calculator HILIC models had R2-values in 0.973–0.98 range. Development of accurate prediction models provides a solid support for retention prediction filtering of false positive identifications in high-throughput proteomics analyses, guided development of 2D LCMS procedures ((HILIC-RP)). Our analysis also shows that all phases possess a high degree of orthogonality to RPLC (formic acid) with Luna HILIC being the best choice in this regard.

Predictive Liquid Chromatography of Peptides Based on Hydrophilic Interactions for Mass Spectrometry-Based Proteomics

High-performance liquid chromatography (HPLC) is widely used for separation of complex peptide mixtures before mass spectrometry-based proteome analysis. In this analysis, reversed phase HPLC (RPHPLC) using non-polar stationary phases such as surface-modified silica containing alkyl groups (e.g., C18) is typically employed. Because of the high heterogeneity of proteomic samples, multidimensional separation approaches gained increasing attention recently to tackle this complexity and extremely high range of concentrations. In two-dimensional liquid chromatography, hydrophilic interaction chromatography (HILIC) is often a method of choice for combination with RP-HPLC because it uses reversed-phase type eluents and allows efficient separation of polar peptides. Due to the high degree of orthogonality in this two-dimensional separation space, it is tempting to develop approaches for predicting peptide retention times for HILIC-based separations similar to the ones for RP-HPLC. Recent successful efforts in this area were focused on developing retention coefficient (RC)-based approaches. Herein, we explored the feasibility of using a statistical thermodynamic model for prediction of peptide retention times in HILIC separations and determined the phenomenological parameters of the model for a bare silica column. The performance of the developed model was tested using HPLC-MS analysis of a set of synthetic peptides, as well as a tryptic peptide mixture.

Peptide Retention Time Prediction in Hydrophilic Interaction Liquid Chromatography: Data Collection Methods and Features of Additive and Sequence-Specific Models.

The development of a peptide retention prediction model for hydrophilic interaction liquid chromatography (XBridge Amide column) is described for a collection of ∼40 000 tryptic peptides. Off-line 2D LC-MS/MS analysis (HILIC-RPLC) of S. cerevisiae whole cell lysate has been used to acquire retention information for a HILIC separation. The large size of the optimization data set (more than 2 orders of magnitude compared to previous reports) permits the accurate assignment of hydrophilic retention coefficients of individual amino acids, establishing both the effects of amino acid position relative to peptide termini and the influence of peptide secondary structure in HILIC. The accuracy of a simple additive model with peptide length correction (R2 value of ∼0.96) was found to be much higher compared to an algorithm of similar complexity applied to RPLC (∼0.91). This indicates significantly smaller influence of peptide secondary structure and interactions with counterions in HILIC. Nevertheless, sequence-specific features were found. Helical peptides that tend to retain stronger than predicted in RPLC exhibit negative prediction errors using an additive HILIC model. N-cap helix stabilizing motifs, which increase retention of amphipathic helical peptides in RP, reduce peptide retention in HILIC independently of peptide amphipathicity. Peptides carrying multiple Pro and Gly (residues with lowest helical propensity) retain stronger than predicted. We conclude that involvement of the peptide backbone's carbonyl and amide groups in hydrogen-bond stabilization of helical structures is a major factor, which determines sequence-dependent behavior of peptides in HILIC. The incorporation of observed sequence dependent features into our Sequence-Specific Retention Calculator HILIC model resulted in 0.98 R2 value correlation, significantly exceeding the predictive performance of all RP and HILIC models developed for complex mixtures of tryptic peptides.

Locus-specific Retention Predictor (LsRP): A Peptide Retention Time Predictor Developed for Precision Proteomics

The precision prediction of peptide retention time (RT) plays an increasingly important role in liquid chromatography–tandem mass spectrometry (LC–MS/MS) based proteomics. Owing to the high reproducibility of liquid chromatography, RT prediction provides promising information for both identification and quantification experiment design. In this work, we present a Locus-specific Retention Predictor (LsRP) for precise prediction of peptide RT, which is based on amino acid locus information and Support Vector Regression (SVR) algorithm. Corresponding to amino acid locus, each peptide sequence was converted to a featured locus vector consisting of zeros and ones. With locus vector information from LC-MS/MS data sets, an SVR computational process was trained and evaluated. LsRP finally provided a prediction correlation coefficient of 0.95~0.99. We compared our method with two common predictors. Results showed that LsRP outperforms these methods and tracked up to 30% extra peptides in an extraction RT window of 2 min. A new strategy by combining LsRP and calibration peptide approach was then proposed, which open up new opportunities for precision proteomics.

Reply to 'Comparison of peptide retention prediction algorithm in reversed-phase chromatography. Comment on "Predictive chromatography of peptides and proteins as a complementary tool for proteomics"', by O. V. Krokhin, Analyst, 2017, 142, DOI: 10.1039/c6an02510b.

On the applicability of peptide retention time prediction software for data acquired using different LC conditions.

A proteogenomic approach for protein-level evidence of genomic variants in cancer cells.

Variations in protein coding sequence may sometimes play important roles in cancer development. However, since variants may not express into proteins due to various cellular quality control systems, it is important to get protein-level evidence of the genomic variations. We present a proteogenomic strategy getting protein-level evidence of genomic variants, which we call sequential targeted LC-MS/MS based on prediction of peptide pI and Retention time (STaLPIR). Our approach shows improved peptide identification, and has the potential for the unbiased analysis of variant sequence as well as corresponding reference sequence. Integrated analysis of DNA, mRNA and protein suggests that protein expression level of the nonsynonymous variant is regulated either before or after translation, according to influence of the variant on protein function. In conclusion, our data provides an excellent approach getting direct evidence for the expression of variant protein forms from genome sequence data.

The program Proteocat as a tool for planning of proteomic experiments

ProteoCat is a computer program that has been designed to help researchers in the planning of large-scale proteomic experiments. The central part of this program is the unit of hydrolysis simulation that supports 4 proteases (trypsin, lysine C, endoproteinases Asp-N and GluC). For peptides obtained after virtual hydrolysis or loaded from data files a number of properties important in mass-spectrometric experiments can be calculated and predicted; the resultant data can be analyzed or filtered (to reduce a set of peptides). The program is using new and improved modifications of own earlier developed methods for pI prediction, which can be also predicted by means of popular pKa scales proposed by other reseachers. The algorithm for prediction of peptide retention time has been realized similarly to the algorithm used in the SSRCalc program. Using ProteoCat it is possible to estimate the coverage of amino acid sequences of analyzed proteins under defined limitation on peptides detection, as well as the possibility of assembly of peptide fragments with user-defined minimal sizes of “sticky” ends. The program has a graphical user interface, written on JAVA and available at http://www.ibmc.msk.ru/LPCIT/ProteoCat.

ProteoСat: a tool for planning of proteomic experiments

ProteoCat is a computer program has been designed to help researchers in the planning of large-scale proteomic experiments. The central part of this program is the subprogram of hydrolysis simulation that supports 4 proteases (trypsin, lysine C, endoproteinases AspN and GluC). For the peptides obtained after virtual hydrolysis or loaded from data file a number of properties important in mass-spectrometric experiments can be calculated or predicted. The data can be analyzed or filtered to reduce a set of peptides. The program is using new and improved modification of our methods developed to predict pI and probability of peptide detection; pI can also be predicted for a number of popular pKa's scales, proposed by other investigators. The algorithm for prediction of peptide retention time was realized similar to the algorithm used in the program SSRCalc. ProteoCat can estimate the coverage of amino acid sequences of proteins under defined limitation on peptides detection, as well as the possibility of assembly of peptide fragments with user-defined size of "sticky" ends. The program has a graphical user interface, written on JAVA and available at http://www.ibmc.msk.ru/LPCIT/ProteoCat.

A new peptide retention time prediction method for mass spectrometry based proteomic analysis by a serial and parallel support vector machine model

The online reversed-phase liquid chromatography (RPLC) contributes a lot for the large scale mass spectrometry based protein identification in proteomics. Retention time (RT) as an important evidence can be used to distinguish the false positive/true positive peptide identifications. Because of the nonlinear concentration curve of organic phase in the whole range of run time and the interactions among peptides, the sequence based RT prediction of peptides has low accuracy and is difficult to generalize in practice, and thus is less effective in the validation of peptide identifications. A serial and parallel support vector machine (SP-SVM) method was proposed to characterize the nonlinear effect of organic phase concentration and the interactions among peptides. The SP-SVM contains a support vector regression (SVR) only for model training (named as p-SVR) and 4 SVM models (named as C-SVM, 1-SVR, s-SVR and n-SVR) for the RT prediction. After distinguishing the peptide chromatographic behavior by C-SVM, 1-SVR and s-SVR were used to predict the peptide RT specifically to improve the accuracy. Then the peptide RT was normalized by n-SVR to characterize the peptide interactions. The prediction accuracy was improved significantly by applying this method to the processing of the complex sample dataset. The coefficient of the determination between predictive and experimental RTs reaches 0. 95, the prediction error range was less than 20% of the total LC run time for more than 95% cases, and less than 10% of the total LC run time for more than 70% cases. The performance of this model reaches the best of known so far. More important, the SP-SVM method provides a framework to take into account the interactions among peptides in chromatographic separation, and its performance can be improved further by introducing new data processing and experiment strategy.