Retention time prediction using neural networks increases identifications in crosslinking mass spectrometry

Artificial Intelligence in Metabolomics: A Current Review

Metabolomics and artificial intelligence (AI) form a synergistic partnership. Metabolomics generates large datasets comprising hundreds to thousands of metabolites with complex relationships. AI, aiming to mimic human intelligence through computational modeling, possesses extraordinary capabilities for big data analysis. In this review, we provide a recent overview of the methodologies and applications of AI in metabolomics studies in the context of systems biology and human health. We first introduce the AI concept, history, and key algorithms for machine learning and deep learning, summarizing their strengths and weaknesses. We then discuss studies that have successfully used AI across different aspects of metabolomic analysis, including analytical detection, data preprocessing, biomarker discovery, predictive modeling, and multi-omics data integration. Lastly, we discuss the existing challenges and future perspectives in this rapidly evolving field. Despite limitations and challenges, the combination of metabolomics and AI holds great promises for revolutionary advancements in enhancing human health.

Retention time dataset for heterogeneous molecules in reversed-phase liquid chromatography

Quantitative structure-property relationships have been extensively studied in the field of predicting retention times in liquid chromatography (LC). However, making transferable predictions is inherently complex because retention times are influenced by both the structure of the molecule and the chromatographic method used. Despite decades of development and numerous published machine learning models, the practical application of predicting small molecule retention time remains limited. The resulting models are typically limited to specific chromatographic conditions and the molecules used in their training and evaluation. Here, we have developed a comprehensive dataset comprising over 10,000 experimental retention times. These times were derived from 30 different reversed-phase liquid chromatography methods and pertain to a collection of 343 small molecules representing a wide range of chemical structures. These chromatographic methods encompass common LC setups for studying the retention behavior of small molecules. They offer a wide range of examples for modeling retention time with different LC setups.

Evaluating Imide-Based Mass Spectrometry-Cleavable Cross-Linkers for Structural Proteomics Studies

Disuccinimidyl dibutyric urea (DSBU) is a mass spectrometry (MS)-cleavable cross-linker that has multiple applications in structural biology, ranging from isolated protein complexes to comprehensive system-wide interactomics. DSBU facilitates a rapid and reliable identification of cross-links through the dissociation of its urea group in the gas phase. In this study, we further advance the structural capabilities of DSBU by remodeling the urea group into an imide, thus introducing a novel class of cross-linkers. This modification preserves the MS cleavability of the amide bond, granted by the two acyl groups of the imide function. The central nitrogen atom enables the introduction of affinity purification tags. Here, we introduce disuccinimidyl disuccinic imide (DSSI) as a prototype of this class of cross-linkers. It features a phosphonate handle for immobilized metal ion affinity chromatography enrichment. We detail DSSI synthesis and describe its behavior in solution and in the gas phase while cross-linking isolated proteins and human cell lysates. DSSI and DSBU cross-links are compared at the same enrichment depth to bridge these two cross-linker classes. We validate DSSI cross-links by mapping them in high-resolution structures of large protein assemblies. The cross-links observed yield insights into the morphology of intrinsically disordered proteins and their complexes. The DSSI linker might spearhead a novel class of MS-cleavable and enrichable cross-linkers.

Chemical cross-linking and mass spectrometry enabled systems-level structural biology

Structural information on protein-protein interactions (PPIs) is essential for improved understanding of regulatory interactome networks that confer various physiological and pathological responses. Additionally, maladaptive PPIs constitute desirable therapeutic targets due to inherently high disease state specificity. Recent advances in chemical cross-linking strategies coupled with mass spectrometry (XL-MS) have positioned XL-MS as a promising technology to not only elucidate the molecular architecture of individual protein assemblies, but also to characterize proteome-wide PPI networks. Moreover, quantitative in vivo XL-MS provides a new capability for the visualization of cellular interactome dynamics elicited by drug treatments, disease states, or aging effects. The emerging field of XL-MS based complexomics enables unique insights on protein moonlighting and protein complex remodeling. These techniques provide complimentary information necessary for in-depth structural interactome studies to better comprehend how PPIs mediate function in living systems.

Rescoring Peptide Spectrum Matches: Boosting Proteomics Performance by Integrating Peptide Property Predictors Into Peptide Identification

Rescoring of peptide spectrum matches originating from database search engines enabled by peptide property predictors is exceeding the performance of peptide identification from traditional database search engines. In contrast to the peptide spectrum match scores calculated by traditional database search engines, rescoring peptide spectrum matches generates scores based on comparing observed and predicted peptide properties, such as fragment ion intensities and retention times. These newly generated scores enable a more efficient discrimination between correct and incorrect peptide spectrum matches. This approach was shown to lead to substantial improvements in the number of confidently identified peptides, facilitating the analysis of challenging datasets in various fields such as immunopeptidomics, metaproteomics, proteogenomics, and single-cell proteomics. In this review, we summarize the key elements leading up to the recent introduction of multiple data-driven rescoring pipelines. We provide an overview of relevant post-processing rescoring tools, introduce prominent data-driven rescoring pipelines for various applications, and highlight limitations, opportunities, and future perspectives of this approach and its impact on mass spectrometry-based proteomics.

Recent Developments in Machine Learning for Mass Spectrometry

Statistical analysis and modeling of mass spectrometry (MS) data have a long and rich history with several modern MS-based applications using statistical and chemometric methods. Recently, machine learning (ML) has experienced a renaissance due to advents in computational hardware and the development of new algorithms for artificial neural networks (ANN) and deep learning architectures. Moreover, recent successes of new ANN and deep learning architectures in several areas of science, engineering, and society have further strengthened the ML field. Importantly, modern ML methods and architectures have enabled new approaches for tasks related to MS that are now widely adopted in several popular MS-based subdisciplines, such as mass spectrometry imaging and proteomics. Herein, we aim to provide an introductory summary of the practical aspects of ML methodology relevant to MS. Additionally, we seek to provide an up-to-date review of the most recent developments in ML integration with MS-based techniques while also providing critical insights into the future direction of the field.

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.

AI-guided pipeline for protein-protein interaction drug discovery identifies a SARS-CoV-2 inhibitor

Protein-protein interactions (PPIs) offer great opportunities to expand the druggable proteome and therapeutically tackle various diseases, but remain challenging targets for drug discovery. Here, we provide a comprehensive pipeline that combines experimental and computational tools to identify and validate PPI targets and perform early-stage drug discovery. We have developed a machine learning approach that prioritizes interactions by analyzing quantitative data from binary PPI assays or AlphaFold-Multimer predictions. Using the quantitative assay LuTHy together with our machine learning algorithm, we identified high-confidence interactions among SARS-CoV-2 proteins for which we predicted three-dimensional structures using AlphaFold-Multimer. We employed VirtualFlow to target the contact interface of the NSP10-NSP16 SARS-CoV-2 methyltransferase complex by ultra-large virtual drug screening. Thereby, we identified a compound that binds to NSP10 and inhibits its interaction with NSP16, while also disrupting the methyltransferase activity of the complex, and SARS-CoV-2 replication. Overall, this pipeline will help to prioritize PPI targets to accelerate the discovery of early-stage drug candidates targeting protein complexes and pathways.

Generic and accurate prediction of retention times in liquid chromatography by post-projection calibration

Retention time predictions from molecule structures in liquid chromatography (LC) are increasingly used in MS-based targeted and untargeted analyses, providing supplementary evidence for molecule annotation and reducing experimental measurements. Nevertheless, different LC setups (e.g., differences in gradient, column, and/or mobile phase) give rise to many prediction models that can only accurately predict retention times for a specific chromatographic method (CM). Here, a generic and accurate method is present to predict retention times across different CMs, by introducing the concept of post-projection calibration. This concept builds on the direct projections of retention times between different CMs and uses 35 external calibrants to eliminate the impact of LC setups on projection accuracy. Results showed that post-projection calibration consistently achieved a median projection error below 3.2% of the elution time. The ranking results of putative candidates reached similar levels among different CMs. This work opens up broad possibilities for coordinating retention times between different laboratories and developing extensive retention databases.

Acquisition and Analysis of DIA-Based Proteomic Data: A Comprehensive Survey in 2023

Data-independent acquisition (DIA) mass spectrometry (MS) has emerged as a powerful technology for high-throughput, accurate, and reproducible quantitative proteomics. This review provides a comprehensive overview of recent advances in both the experimental and computational methods for DIA proteomics, from data acquisition schemes to analysis strategies and software tools. DIA acquisition schemes are categorized based on the design of precursor isolation windows, highlighting wide-window, overlapping-window, narrow-window, scanning quadrupole-based, and parallel accumulation-serial fragmentation-enhanced DIA methods. For DIA data analysis, major strategies are classified into spectrum reconstruction, sequence-based search, library-based search, de novo sequencing, and sequencing-independent approaches. A wide array of software tools implementing these strategies are reviewed, with details on their overall workflows and scoring approaches at different steps. The generation and optimization of spectral libraries, which are critical resources for DIA analysis, are also discussed. Publicly available benchmark datasets covering global proteomics and phosphoproteomics are summarized to facilitate performance evaluation of various software tools and analysis workflows. Continued advances and synergistic developments of versatile components in DIA workflows are expected to further enhance the power of DIA-based proteomics.

Deep graph convolutional network for small-molecule retention time prediction

The retention time (RT) is a crucial source of data for liquid chromatography-mass spectrometry (LCMS). A model that can accurately predict the RT for each molecule would empower filtering candidates with similar spectra but differing RT in LCMS-based molecule identification. Recent research shows that graph neural networks (GNNs) outperform traditional machine learning algorithms in RT prediction. However, all of these models use relatively shallow GNNs. This study for the first time investigates how depth affects GNNs' performance on RT prediction. The results demonstrate that a notable improvement can be achieved by pushing the depth of GNNs to 16 layers by the adoption of residual connection. Additionally, we also find that graph convolutional network (GCN) model benefits from the edge information. The developed deep graph convolutional network, DeepGCN-RT, significantly outperforms the previous state-of-the-art method and achieves the lowest mean absolute percentage error (MAPE) of 3.3% and the lowest mean absolute error (MAE) of 26.55 s on the SMRT test set. We also finetune DeepGCN-RT on seven datasets with various chromatographic conditions. The mean MAE of the seven datasets largely decreases 30% compared to previous state-of-the-art method. On the RIKEN-PlaSMA dataset, we also test the effectiveness of DeepGCN-RT in assisting molecular structure identification. By 30% lessening the number of potential structures, DeepGCN-RT is able to improve top-1 accuracy by about 11%.

Prediction of peptide mass spectral libraries with machine learning

The recent development of machine learning methods to identify peptides in complex mass spectrometric data constitutes a major breakthrough in proteomics. Longstanding methods for peptide identification, such as search engines and experimental spectral libraries, are being superseded by deep learning models that allow the fragmentation spectra of peptides to be predicted from their amino acid sequence. These new approaches, including recurrent neural networks and convolutional neural networks, use predicted in silico spectral libraries rather than experimental libraries to achieve higher sensitivity and/or specificity in the analysis of proteomics data. Machine learning is galvanizing applications that involve large search spaces, such as immunopeptidomics and proteogenomics. Current challenges in the field include the prediction of spectra for peptides with post-translational modifications and for cross-linked pairs of peptides. Permeation of machine-learning-based spectral prediction into search engines and spectrum-centric data-independent acquisition workflows for diverse peptide classes and measurement conditions will continue to push sensitivity and dynamic range in proteomics applications in the coming years.

inSPIRE: An Open-Source Tool for Increased Mass Spectrometry Identification Rates Using Prosit Spectral Prediction

Rescoring of mass spectrometry (MS) search results using spectral predictors can strongly increase peptide spectrum match (PSM) identification rates. This approach is particularly effective when aiming to search MS data against large databases, for example, when dealing with nonspecific cleavage in immunopeptidomics or inflation of the reference database for noncanonical peptide identification. Here, we present inSPIRE (in silico Spectral Predictor Informed REscoring), a flexible and performant open-source rescoring pipeline built on Prosit MS spectral prediction, which is compatible with common database search engines. inSPIRE allows large-scale rescoring with data from multiple MS search files, increases sensitivity to minor differences in amino acid residue position, and can be applied to various MS sample types, including tryptic proteome digestions and immunopeptidomes. inSPIRE boosts PSM identification rates in immunopeptidomics, leading to better performance than the original Prosit rescoring pipeline, as confirmed by benchmarking of inSPIRE performance on ground truth datasets. The integration of various features in the inSPIRE backbone further boosts the PSM identification in immunopeptidomics, with a potential benefit for the identification of noncanonical peptides.

Improved Peptide Backbone Fragmentation Is the Primary Advantage of MS-Cleavable Crosslinkers

Proteome-wide crosslinking mass spectrometry studies have coincided with the advent of mass spectrometry (MS)-cleavable crosslinkers that can reveal the individual masses of the two crosslinked peptides. However, recently, such studies have also been published with noncleavable crosslinkers, suggesting that MS-cleavability is not essential. We therefore examined in detail the advantages and disadvantages of using the commonly used MS-cleavable crosslinker, disuccinimidyl sulfoxide (DSSO). Indeed, DSSO gave rise to signature peptide fragments with a distinct mass difference (doublet) for nearly all identified crosslinked peptides. Surprisingly, we could show that it was not these peptide masses that proved the main advantage of MS cleavability of the crosslinker, but improved peptide backbone fragmentation which reduces the ambiguity of peptide identifications. This also holds true for another commonly used MS-cleavable crosslinker, DSBU. We show furthermore that the more intricate MS3-based data acquisition approaches lack sensitivity and specificity, causing them to be outperformed by the simpler and faster stepped higher-energy collisional dissociation (HCD) method. This understanding will guide future developments and applications of proteome-wide crosslinking mass spectrometry.

Ad hoc learning of peptide fragmentation from mass spectra enables an interpretable detection of phosphorylated and cross-linked peptides

Mass spectrometry-based proteomics provides a holistic snapshot of the entire protein set of living cells on a molecular level. Currently, only a few deep learning approaches exist that involve peptide fragmentation spectra, which represent partial sequence information of proteins. Commonly, these approaches lack the ability to characterize less studied or even unknown patterns in spectra because of their use of explicit domain knowledge. Here, to elevate unrestricted learning from spectra, we introduce ‘ad hoc learning of fragmentation’ (AHLF), a deep learning model that is end-to-end trained on 19.2 million spectra from several phosphoproteomic datasets. AHLF is interpretable, and we show that peak-level feature importance values and pairwise interactions between peaks are in line with corresponding peptide fragments. We demonstrate our approach by detecting post-translational modifications, specifically protein phosphorylation based on only the fragmentation spectrum without a database search. AHLF increases the area under the receiver operating characteristic curve (AUC) by an average of 9.4% on recent phosphoproteomic data compared with the current state of the art on this task. Furthermore, use of AHLF in rescoring search results increases the number of phosphopeptide identifications by a margin of up to 15.1% at a constant false discovery rate. To show the broad applicability of AHLF, we use transfer learning to also detect cross-linked peptides, as used in protein structure analysis, with an AUC of up to 94%.

Two-Dimensional Fractionation Method for Proteome-Wide Cross-Linking Mass Spectrometry Analysis

Cross-linking mass spectrometry (XL-MS) is an emergent technology for studying protein-protein interactions (PPIs) and elucidating architectures of protein complexes. The development of various MS-cleavable cross-linkers has facilitated the identification of cross-linked peptides, enabling XL-MS studies at the systems level. However, the scope and depth of cellular networks revealed by current XL-MS technologies remain limited. Due to the inherently broad dynamic range and complexity of proteomes, interference from highly abundant proteins impedes the identification of low-abundance cross-linked peptides in complex samples. Thus, peptide enrichment prior to MS analysis is necessary to enhance cross-link identification for proteome-wide studies. Although chromatographic techniques including size exclusion (SEC) and strong cation exchange (SCX) have been successful in isolating cross-linked peptides, new fractionation methods are still needed to further improve the depth of PPI mapping. Here, we present a two-dimensional (2D) separation strategy by integrating peptide SEC with tip-based high pH reverse-phase (HpHt) fractionation to expand the coverage of proteome-wide XL-MS analyses. Combined with the MS-cleavable cross-linker DSSO, we have successfully mapped in vitro PPIs from HEK293 cell lysates with improved identification of cross-linked peptides compared to existing approaches. The method developed here is effective and can be generalized for cross-linking studies of complex samples.

Label-free visual proteomics: Coupling MS- and EM-based approaches in structural biology

Combining diverse experimental structural and interactomic methods allows for the construction of comprehensible molecular encyclopedias of biological systems. Typically, this involves merging several independent approaches that provide complementary structural and functional information from multiple perspectives and at different resolution ranges. A particularly potent combination lies in coupling structural information from cryoelectron microscopy or tomography (cryo-EM or cryo-ET) with interactomic and structural information from mass spectrometry (MS)-based structural proteomics. Cryo-EM/ET allows for sub-nanometer visualization of biological specimens in purified and near-native states, while MS provides bioanalytical information for proteins and protein complexes without introducing additional labels. Here we highlight recent achievements in protein structure and interactome determination using cryo-EM/ET that benefit from additional MS analysis. We also give our perspective on how combining cryo-EM/ET and MS will continue bridging gaps between molecular and cellular studies by capturing and describing 3D snapshots of proteomes and interactomes.

Accurate Retention Time Prediction Based on Monolinked Peptide Information to Confidently Identify Cross-Linked Peptides

Cross-linking mass spectrometry methods have not been successfully applied to protein-protein interaction discovery at a proteome-wide level mainly due to the computation complexity (O (n²)) issue. In a previous report, we proposed a decision tree searching strategy (DTSS), which can reduce complexity by orders of magnitude. In this study, we further found that the monolinked peptides carry out the information on the retention time of the corresponding cross-linked pairs; therefore, the retention time of cross-linked peptide pairs can be predicted accurately. By utilizing the retention time as an extra filter, the false positive rate can be reduced by around 86% with a sensitivity loss of 10%. The method combined with DTSS (T-DTSS) not only benefits improving identification confidence but also leads to lower cutoff scores and facilitates substantially increasing inter-cross-link identification. T-DTSS was successfully applied to the identification of inter-cross-links obtained from Escherichia coli cell lysate cross-linked by a newly synthesized enrichable cross-linker, pDSBE. The approach can be applicable to both cleavable and noncleavable methods.