Prediction Of Retention Times Research Articles

Graph Neural Networks for Improved Retention Time Predictions

In this extended special feature to celebrate the 35th anniversary edition of LCGC Europe, leading figures from the separation science community explore contemporary trends in separation science and identify possible future developments.

Top-Down Approach to Retention Time Prediction in Comprehensive Two-Dimensional Gas Chromatography-Mass Spectrometry.

In this contribution, we describe a novel modeling approach to predicting retention times (tr) in comprehensive two-dimensional gas chromatography coupled to time-of-flight mass spectrometry (GC × GC-ToF-MS) with a particular emphasis on the second-dimension (2D) retention time predictions (2tr). This approach is referred to as a "top-down" approach in that it breaks down the complete GC × GC separation into two independent one-dimensional gas chromatography separations (1D-GC). In this regard, both dimensions, that is, first dimension (1D) and second dimension (2D) are treated separately, and the cryogenic modulator is simply considered as a second consecutive injection device. Separate 1D-GC tr predictions are performed on both dimensions using the same flow rate as the one deployed in the conventional GC × GC system. The separate tr predictions are then combined to account for the two-dimensional separation. This model was applied to 24 analytes from 2 standard mixtures (Grob Test Mix and Fragrance Materials Test Mix) and assessed across 9 GC × GC chromatographic conditions. The experimental and predicted chromatographic retention space occupations were assessed by using the convex hull approach defined by the Delaunay triangulation. The predicted percentage of space occupation corresponded favorably with the experimental values. Furthermore, the top-down approach enabled an accurate prediction of the 2tr of all investigated analytes, providing an average 2tr modeling error of 0.26 ± 0.01 s.

User-Driven Strategy for In Silico Screening of Reversed-Phase Liquid Chromatography Conditions for Known Pharmaceutical-Related Small Molecules

In the pharmaceutical field, and more precisely in quality control laboratories, robust liquid chromatographic methods are needed to separate and analyze mixtures of compounds. The development of such chromatographic methods for new mixtures can result in a long and tedious process even while using the design of experiments methodology. However, developments could be accelerated with the help of in silico screening. In this work, the usefulness of a strategy combining response surface methodology (RSM) followed by multicriteria decision analysis (MCDA) applied to predictions from a quantitative structure-retention relationship (QSRR) model is demonstrated. The developed strategy shows that selecting equations for the retention time prediction models based on the pKa of the compound allows flexibility in the models. The MCDA developed is shown to help to make decisions on different criteria while being robust to the user's decision on the weights for each criterion. This strategy is proposed for the screening phase of the method lifecycle. The strategy offers the possibility to the user to select chromatographic conditions based on multiple criteria without being too sensitive to the importance given to them. The conditions with the highest desirability are defined as the starting point for further optimization steps.

AlphaPeptDeep: a modular deep learning framework to predict peptide properties for proteomics

Machine learning and in particular deep learning (DL) are increasingly important in mass spectrometry (MS)-based proteomics. Recent DL models can predict the retention time, ion mobility and fragment intensities of a peptide just from the amino acid sequence with good accuracy. However, DL is a very rapidly developing field with new neural network architectures frequently appearing, which are challenging to incorporate for proteomics researchers. Here we introduce AlphaPeptDeep, a modular Python framework built on the PyTorch DL library that learns and predicts the properties of peptides (https://github.com/MannLabs/alphapeptdeep). It features a model shop that enables non-specialists to create models in just a few lines of code. AlphaPeptDeep represents post-translational modifications in a generic manner, even if only the chemical composition is known. Extensive use of transfer learning obviates the need for large data sets to refine models for particular experimental conditions. The AlphaPeptDeep models for predicting retention time, collisional cross sections and fragment intensities are at least on par with existing tools. Additional sequence-based properties can also be predicted by AlphaPeptDeep, as demonstrated with a HLA peptide prediction model to improve HLA peptide identification for data-independent acquisition (https://github.com/MannLabs/PeptDeep-HLA).

Prediction of Retention Time and Collision Cross Section (CCSH+, CCSH-, and CCSNa+) of Emerging Contaminants Using Multiple Adaptive Regression Splines.

Ultra-high performance liquid chromatography coupled to ion mobility separation and high-resolution mass spectrometry instruments have proven very valuable for screening of emerging contaminants in the aquatic environment. However, when applying suspect or nontarget approaches (i.e., when no reference standards are available), there is no information on retention time (RT) and collision cross-section (CCS) values to facilitate identification. In silico prediction tools of RT and CCS can therefore be of great utility to decrease the number of candidates to investigate. In this work, Multiple Adaptive Regression Splines (MARS) were evaluated for the prediction of both RT and CCS. MARS prediction models were developed and validated using a database of 477 protonated molecules, 169 deprotonated molecules, and 249 sodium adducts. Multivariate and univariate models were evaluated showing a better fit for univariate models to the experimental data. The RT model (R2 = 0.855) showed a deviation between predicted and experimental data of ±2.32 min (95% confidence intervals). The deviation observed for CCS data of protonated molecules using the CCSH model (R2 = 0.966) was ±4.05% with 95% confidence intervals. The CCSH model was also tested for the prediction of deprotonated molecules, resulting in deviations below ±5.86% for the 95% of the cases. Finally, a third model was developed for sodium adducts (CCSNa, R2 = 0.954) with deviation below ±5.25% for 95% of the cases. The developed models have been incorporated in an open-access and user-friendly online platform which represents a great advantage for third-party research laboratories for predicting both RT and CCS data.

HormonomicsDB: a novel workflow for the untargeted analysis of plant growth regulators and hormones

Background: Metabolomics is the simultaneous determination of all metabolites in a system. Despite significant advances in the field, compound identification remains a challenge. Prior knowledge of the compound classes of interest can improve metabolite identification. Hormones are a small signaling molecules, which function in coordination to direct all aspects of development, function and reproduction in living systems and which also pose challenges as environmental contaminants. Hormones are inherently present at low levels in tissues, stored in many forms and mobilized rapidly in response to a stimulus making them difficult to measure, identify and quantify. Methods: An in-depth literature review was performed for known hormones, their precursors, metabolites and conjugates in plants to generate the database and an RShiny App developed to enable web-based searches against the database. An accompanying liquid chromatography – mass spectrometry (LC-MS) protocol was developed with retention time prediction in Retip. A meta-analysis of 14 plant metabolomics studies was used for validation. Results: We developed HormonomicsDB, a tool which can be used to query an untargeted mass spectrometry (MS) dataset against a database of more than 200 known hormones, their precursors and metabolites. The protocol encompasses sample preparation, analysis, data processing and hormone annotation and is designed to minimize degradation of labile hormones. The plant system is used a model to illustrate the workflow and data acquisition and interpretation. Analytical conditions were standardized to a 30 min analysis time using a common solvent system to allow for easy transfer by a researcher with basic knowledge of MS. Incorporation of synthetic biotransformations enables prediction of novel metabolites. Conclusions: HormonomicsDB is suitable for use on any LC-MS based system with compatible column and buffer system, enables the characterization of the known hormonome across a diversity of samples, and hypothesis generation to reveal knew insights into hormone signaling networks.

Retention Time Prediction with Message-Passing Neural Networks

Retention time prediction, facilitated by advances in machine learning, has become a useful tool in untargeted LC-MS applications. State-of-the-art approaches include graph neural networks and 1D-convolutional neural networks that are trained on the METLIN small molecule retention time dataset (SMRT). These approaches demonstrate accurate predictions comparable with the experimental error for the training set. The weak point of retention time prediction approaches is the transfer of predictions to various systems. The accuracy of this step depends both on the method of mapping and on the accuracy of the general model trained on SMRT. Therefore, improvements to both parts of prediction workflows may lead to improved compound annotations. Here, we evaluate capabilities of message-passing neural networks (MPNN) that have demonstrated outstanding performance on many chemical tasks to accurately predict retention times. The model was initially trained on SMRT, providing mean and median absolute cross-validation errors of 32 and 16 s, respectively. The pretrained MPNN was further fine-tuned on five publicly available small reversed-phase retention sets in a transfer learning mode and demonstrated up to 30% improvement of prediction accuracy for these sets compared with the state-of-the-art methods. We demonstrated that filtering isomeric candidates by predicted retention with the thresholds obtained from ROC curves eliminates up to 50% of false identities.

A novel method for efficient screening and annotation of important pathway-associated metabolites based on the modified metabolome and probe molecules

植物次生代谢物在抵御生物/非生物胁迫、生物间互作以及信息传递等方面发挥重要作用,次生代谢途径解析对植物分子育种、天然产物合成等方面具有重要意义。液相色谱-高分辨串联质谱(LC-HRMS/MS)为次生代谢物鉴定及途径表征提供了技术手段。非靶向LC-HRMS/MS方法可获得丰富的质谱信号,包括一级质谱和二级质谱(MS, MS/MS),但受质谱数据库规模以及次生代谢物复杂性的制约,次生代谢物注释十分困难。该研究以玉米叶片中苯丙烷途径代谢物为例,发展用于非靶向代谢组数据中重要途径代谢物的高效筛选和注释新方法。首先,利用公共代谢途径数据库及文献获取参与苯丙烷代谢途径的61种修饰反应类型,进而从非靶向实验数据中筛选出修饰代谢组。其次,获取开源串联质谱数据中的苯丙烷类化合物作为探针分子,构建探针分子质谱数据库。将探针分子与修饰代谢组共建分子网络,锁定目标途径代谢物并注释结构。该方法在正、负离子模式下分别筛选出玉米叶片中392个和417个苯丙烷途径候选代谢物,去冗余后共注释出129个代谢物,涉及苯丙烷代谢的主要分支途径,如黄酮途径的8个类黄酮、19个氧苷类黄酮和32个碳苷类黄酮,31个羟基肉桂酸途径代谢物以及22个木脂素途径代谢物;其中26个在PubChem和SciFinder数据库中未见收录。该研究利用探针分子结合修饰组可快速锁定途径代谢物,且有助于快速、准确的网络传播注释,可显著提高目标途径代谢物筛选与注释效率,为植物次生代谢途径的深入解析提供分析手段。

Comparison of hydrophobicity scales for predicting biophysical properties of antibodies.

While antibody-based therapeutics have grown to be one of the major classes of novel medicines, some antibody development candidates face significant challenges regarding expression levels, solubility, as well as stability and aggregation, under physiological and storage conditions. A major determinant of those properties is surface hydrophobicity, which promotes unspecific interactions and has repeatedly proven problematic in the development of novel antibody-based drugs. Multiple computational methods have been devised for in-silico prediction of antibody hydrophobicity, often using hydrophobicity scales to assign values to each amino acid. Those approaches are usually validated by their ability to rank potential therapeutic antibodies in terms of their experimental hydrophobicity. However, there is significant diversity both in the hydrophobicity scales and in the experimental methods, and consequently in the performance of in-silico methods to predict experimental results. In this work, we investigate hydrophobicity of monoclonal antibodies using hydrophobicity scales. We implement several scoring schemes based on the solvent-accessibility and the assigned hydrophobicity values, and compare the different scores and scales based on their ability to predict retention times from hydrophobic interaction chromatography. We provide an overview of the strengths and weaknesses of several commonly employed hydrophobicity scales, thereby improving the understanding of hydrophobicity in antibody development. Furthermore, we test several datasets, both publicly available and proprietary, and find that the diversity of the dataset affects the performance of hydrophobicity scores. We expect that this work will provide valuable guidelines for the optimization of biophysical properties in future drug discovery campaigns.

A method for the sensitive targeted screening of synthetic cannabinoids and opioids in whole blood by LC-QTOF with simultaneous suspect screening using HighResNPS.com

To develop and validate a qualitative screening method capable of detecting the low concentrations found in cases of synthetic cannabinoid and opioid consumption using a targeted scope, while simultaneously performing efficient suspect screening against a much larger database. Screening for synthetic cannabinoids and opioids in blood samples remains challenging due to the large number of substances available and the constantly changing market, as well as the low limits of detection required for some compounds. Many laboratories employ LC-QQQ when screening for these compound classes due to the ubiquity and sensitivity of those instruments but LC-HRMS instruments are similarly able to detect a large number of pre-defined analytes with high sensitivity and specificity, as well as acquiring full scan data which enables both suspect screening and retrospective screening if required. Analytes were extracted from 0.5 mL of whole blood using alkaline liquid-liquid extraction and data were acquired using data dependent acquisition on an Agilent 6545 QTOF. 28 opioids and 23 cannabinoids were validated for limit of detection, recovery, matrix effects, selectivity, extract stability and carryover. Data processing occurred in two stages; first a targeted screen was performed using an in-house database containing retention times (RT), accurate masses and MS/MS spectra for the validated compounds plus an additional 16 opioids and 34 cannabinoids (101 total compounds). Suspect screening was then performed using a database downloaded from the crowdsourced NPS data website HighResNPS.com which contains mass, consensus MS/MS spectra and laboratory specific predicted retention times for over 1400 compounds. The suspect screening workflow was assessed by determining the accuracy of the retention time prediction and MS/MS library matching against consensus spectra. For this workflow, qualifying parameters for initial presumptive identification included mass accuracy of ±4mDa, MS/MS match score greater than 20 and retention time difference to predicted RT of ±2 min. In practice, results of this suspect screen would be verified following purchase of authentic reference material for comparison. The method was applied to 66 forensic cases where synthetic cannabinoid or opioid screening was requested by the client or their use was suspected due to case information. Limits of detection ranged from 0.03–0.27 μg/L (median, 0.04 μg/L) for the 28 validated opioids and 0.04–0.5 (median, 0.07 μg/L) for the 23 synthetic cannabinoids. Recoveries were greater than 70% for all compounds except MDMB-CHMCZCA (63%) and desomorphine (42%). Synthetic cannabinoids or opioids were detected in seven of the 66 forensic cases. CUMYL-PEGACLONE and etodesnitazine were each detected in two cases and 5F-MDMB-PICA, 4-Cyano-CUMYL-BUTINACA and carfentanil were detected in one case each. These compounds were within the targeted scope of the method but were also detected through the suspect screening workflow. The developed method has similar limits of detections to other published methods using LC-QQQ instruments but has several additional advantages. The suspect screening workflow enables efficient, automated screening of data against the HighResNPS database in the absence of reference materials. The suspect screening parameters adopted were verified through the successful detection of compounds in spiked validation samples, as well as compounds known to be present in authentic samples. The data acquired is also suitable for retrospective screening as new compounds emerge. The dual approach of simultaneous targeted and suspect screening allows laboratories certainty in their coverage of a defined scope, as well as a “best attempt” to cover a much large number of compounds. This provides a good compromise, especially for laboratories where the cost of purchasing hundreds of NPS reference standards is prohibitive. The advantages of testing for these compound classes by HRMS over QQQ mean this instrumentation should be selected preferentially wherever possible when developing new methods.

Retention time prediction boosts confidence in identification of tumorspecific antigens in HLA immunopeptidomics

Retention time prediction boosts confidence in identification of tumorspecific antigens in HLA immunopeptidomics

Identification of Nonvolatile Migrates from Food Contact Materials Using Ion Mobility-High-Resolution Mass Spectrometry and in Silico Prediction Tools.

The identification of migrates from food contact materials (FCMs) is challenging due to the complex matrices and limited availability of commercial standards. The use of machine-learning-based prediction tools can help in the identification of such compounds. This study presents a workflow to identify nonvolatile migrates from FCMs based on liquid chromatography–ion mobility–high-resolution mass spectrometry together with in silico retention time (RT) and collision cross section (CCS) prediction tools. The applicability of this workflow was evaluated by screening the chemicals that migrated from polyamide (PA) spatulas. The number of candidate compounds was reduced by approximately 75% and 29% on applying RT and CCS prediction filters, respectively. A total of 95 compounds were identified in the PA spatulas of which 54 compounds were confirmed using reference standards. The development of a database containing predicted RT and CCS values of compounds related to FCMs can aid in the identification of chemicals in FCMs.

Utilizing Skyline to analyze lipidomics data containing liquid chromatography, ion mobility spectrometry and mass spectrometry dimensions.

Lipidomics studies suffer from analytical and annotation challenges because of the great structural similarity of many of the lipid species. To improve lipid characterization and annotation capabilities beyond those afforded by traditional mass spectrometry (MS)-based methods, multidimensional separation methods such as those integrating liquid chromatography, ion mobility spectrometry, collision-induced dissociation and MS (LC-IMS-CID-MS) may be used. Although LC-IMS-CID-MS and other multidimensional methods offer valuable hydrophobicity, structural and mass information, the files are also complex and difficult to assess. Thus, the development of software tools to rapidly process and facilitate confident lipid annotations is essential. In this Protocol Extension, we use the freely available, vendor-neutral and open-source software Skyline to process and annotate multidimensional lipidomic data. Although Skyline ( https://skyline.ms/skyline.url ) was established for targeted processing of LC-MS-based proteomics data, it has since been extended such that it can be used to analyze small-molecule data as well as data containing the IMS dimension. This protocol uses Skyline's recently expanded capabilities, including small-molecule spectral libraries, indexed retention time and ion mobility filtering, and provides a step-by-step description for importing data, predicting retention times, validating lipid annotations, exporting results and editing our manually validated 500+ lipid library. Although the time required to complete the steps outlined here varies on the basis of multiple factors such as dataset size and familiarity with Skyline, this protocol takes ~5.5 h to complete when annotations are rigorously verified for maximum confidence.

MS2Rescore: Data-Driven Rescoring Dramatically Boosts Immunopeptide Identification Rates.

Immunopeptidomics aims to identify major histocompatibility complex (MHC)-presented peptides on almost all cells that can be used in anti-cancer vaccine development. However, existing immunopeptidomics data analysis pipelines suffer from the nontryptic nature of immunopeptides, complicating their identification. Previously, peak intensity predictions by MS2PIP and retention time predictions by DeepLC have been shown to improve tryptic peptide identifications when rescoring peptide-spectrum matches with Percolator. However, as MS2PIP was tailored toward tryptic peptides, we have here retrained MS2PIP to include nontryptic peptides. Interestingly, the new models not only greatly improve predictions for immunopeptides but also yield further improvements for tryptic peptides. We show that the integration of new MS2PIP models, DeepLC, and Percolator in one software package, MS2Rescore, increases spectrum identification rate and unique identified peptides with 46% and 36% compared to standard Percolator rescoring at 1% FDR. Moreover, MS2Rescore also outperforms the current state-of-the-art in immunopeptide-specific identification approaches. Altogether, MS2Rescore thus allows substantially improved identification of novel epitopes from existing immunopeptidomics workflows.

Nonlinear Predictive Modelling Enables In Silico Optimization of Chromatographic Methods for Complex Stationary Phase‑Analyte Interactions

The development of robust analytical assays for separation and analysis of complex multicomponent mixtures can often be challenging, reflecting the increased complexity of new medicine and vaccine processes. In silico liquid chromatography (LC) method development strategies for small molecules have reached a mature stage across the pharmaceutical industry. However, a straightforward approach for large molecules remains elusive because of conformational changes that can often influence chromatographic retention. Nonetheless, an excellent correlation between experimental and predicted retention time is possible by deploying the correct regression retention models in terms of ln k vs. %B and ln k vs. 1/T (ΔtR < 0.1%). Excellent outcomes generated through in silico chromatographic method development of large molecules using different chaotropic and denaturing mobile phases are illustrated. Linear and nonlinear (polynomial regression) retention models using readily available software were deployed as a function of several chromatographic parameters (gradient slope and column temperature) for a variety of proteins (12–670 kDa) and peptides.

Prediction of Intact N-Glycopeptide Retention Time Windows in Hydrophilic Interaction Liquid Chromatography.

Analysis of protein glycosylation is challenging due to micro- and macro-heterogeneity of the attached glycans. Hydrophilic interaction liquid chromatography (HILIC) is a mode of choice for separation of intact glycopeptides, which are inadequately resolved by reversed phase chromatography. In this work, we propose an easy-to-use model to predict retention time windows of glycopeptides in HILIC. We constructed this model based on the parameters derived from chromatographic separation of six differently glycosylated peptides obtained from tryptic digests of three plasma proteins: haptoglobin, hemopexin, and sex hormone-binding globulin. We calculated relative retention times of different glycoforms attached to the same peptide to the bi-antennary form and showed that the character of the peptide moiety did not significantly change the relative retention time differences between the glycoforms. To challenge the model, we assessed chromatographic behavior of fetuin glycopeptides experimentally, and their retention times all fell within the calculated retention time windows, which suggests that the retention time window prediction model in HILIC is sufficiently accurate. Relative retention time windows provide complementary information to mass spectrometric data, and we consider them useful for reliable determination of protein glycosylation in a site-specific manner.

Probabilistic metabolite annotation using retention time prediction and meta-learned projections

Retention time information is used for metabolite annotation in metabolomic experiments. But its usefulness is hindered by the availability of experimental retention time data in metabolomic databases, and by the lack of reproducibility between different chromatographic methods. Accurate prediction of retention time for a given chromatographic method would be a valuable support for metabolite annotation. We have trained state-of-the-art machine learning regressors using the 80, 038 experimental retention times from the METLIN Small Molecule Retention Tim (SMRT) dataset. The models included deep neural networks, deep kernel learning, several gradient boosting models, and a blending approach. 5, 666 molecular descriptors and 2, 214 fingerprints (MACCS166, Extended Connectivity, and Path Fingerprints fingerprints) were generated with the alvaDesc software. The models were trained using only the descriptors, only the fingerprints, and both types of features simultaneously. Bayesian hyperparameter search was used for parameter tuning. To avoid data-leakage when reporting the performance metrics, nested cross-validation was employed. The best results were obtained by a heavily regularized deep neural network trained with cosine annealing warm restarts and stochastic weight averaging, achieving a mean and median absolute errors of 39.2 pm 1.2; s and 17.2 pm 0.9;s, respectively. To the best of our knowledge, these are the most accurate predictions published up to date over the SMRT dataset. To project retention times between chromatographic methods, a novel Bayesian meta-learning approach that can learn from just a few molecules is proposed. By applying this projection between the deep neural network retention time predictions and a given chromatographic method, our approach can be integrated into a metabolite annotation workflow to obtain z-scores for the candidate annotations. To this end, it is enough that just as few as 10 molecules of a given experiment have been identified (probably by using pure metabolite standards). The use of z-scores permits considering the uncertainty in the projection when ranking candidates, and not only the accuracy. In this scenario, our results show that in 68% of the cases the correct molecule was among the top three candidates filtered by mass and ranked according to z-scores. This shows the usefulness of this information to support metabolite annotation. Python code is available on GitHub at https://github.com/constantino-garcia/cmmrt.

Evaluation of Machine Learning Models for Proteoform Retention and Migration Time Prediction in Top-Down Mass Spectrometry.

Reversed-phase liquid chromatography (RPLC) and capillary zone electrophoresis (CZE) are two primary proteoform separation methods in mass spectrometry (MS)-based top-down proteomics. Proteoform retention time (RT) prediction in RPLC and migration time (MT) prediction in CZE provide additional information for accurate proteoform identification and quantification. While existing methods are mainly focused on peptide RT and MT prediction in bottom-up MS, there is still a lack of methods for proteoform RT and MT prediction in top-down MS. We systematically evaluated eight machine learning models and a transfer learning method for proteoform RT prediction and five models and the transfer learning method for proteoform MT prediction. Experimental results showed that a gated recurrent unit (GRU)-based model with transfer learning achieved a high accuracy (R = 0.978) for proteoform RT prediction and that the GRU-based model and a fully connected neural network model obtained a high accuracy of R = 0.982 and 0.981 for proteoform MT prediction, respectively.

A simple mathematical treatment for predicting linear solvent strength behavior in gradient elution: Application to biomolecules.

This paper describes an approach to rapidly and easily calculate the linear solvent strength parameters, namely log k 0 and S, under reversed‐phase liquid chromatography conditions. This approach, which requires two preliminary gradient experiments to determine the retention parameters, was applied to various representative compounds including small molecules, peptides, and proteins. The retention time prediction errors were compared to the ones obtained with a commercial HPLC modeling software, and a good correlation was found between the values. However, two important constraints have to be accounted for to maintain good predictions with this new approach: i) the retention factor at the initial composition of the preliminary gradient series have to be large enough (i.e., log k i above 2.1) and ii) the retention models have to be sufficiently linear. While these two conditions are not always met with small molecules or even peptides, the situation is different with large biomolecules. This is why our simple calculation method should be preferentially applied to calculate the linear solvent strength parameters of protein samples.

Diagnostic Fragmentation Pathways for Identification of Phthalate Metabolites in Nontargeted Analysis Studies.

Phthalates have been studied due to their linkages with adverse developmental effects; however, metabolites of this class of compounds are undercharacterized and are poorly captured by traditional targeted analysis. In this study, we developed a nontargeted analysis approach for identifying and classifying phthalate metabolites based on a comprehensive study of their fragmentation pathways in electrospray ionization (ESI) quadrupole-time-of-flight mass spectrometry (QTOF-MS). This approach identifies molecular features in the data as phthalate metabolites via the detection of three structurally significant fragment ions. Then phthalate metabolites are classified into four types based on the presence of additional fragment ions specific to each type. Cleavage mechanisms for each class of phthalate metabolite are proposed based on fragmentation patterns generated at various collision energies (CE). All of the tested phthalate metabolites including oxidative and nonoxidative metabolites produced a fragment ion at m/z 121.0295, representing the deprotonated benzoate ion [C6H5COO]-. Most tested phthalate metabolites can produce a specific ion at m/z 147.0088, the deprotonated o-phthalic anhydride ion. However, phthalate carboxylate metabolites can only produce the [M-H-R]- ion at m/z 165.0193 and do not produce the fragment at m/z 147.0088. Other phthalate oxidative metabolites (hydroxyl- and oxo-) follow a different fragmentation pathway than nonoxidative metabolites. With this workflow, eight unknown phthalate metabolites were putatively identified in pooled urine, with one identified as a previously unreported metabolite by a combination of the MS/MS spectrum and the predicted retention time. Method detection limits for phthalate metabolites in urine were also estimated.