Prediction Of Retention Times Research Articles

Streamlined MRM method transfer between instruments assisted with HRMS matching and retention-time prediction

Multiple reaction monitoring (MRM) mode using liquid-chromatography tandem mass spectrometry (e.g., LC-QqQ-MS/MS) has been extensively employed in the small molecule analysis with trace levels in complex samples owing to its high sensitivity. However, most of the reported MRM methods are developed using authentic standards, which are often costly yet not readily available. To address this question, a practical platform for the MRM method transfer between different LC-QqQ-MS/MS instruments, assisted by the high-resolution mass spectrometry (LC-HRMS) and retention time (RT) prediction, has been developed in this study. The reported platform can take advantage of both the high sensitivity of LC-MRM method and ion transition pairs from the previous publications. LC-HRMS can provide the accurate mass measurement of the compounds, though high-quality MS/MS fragments are usually difficult to obtain for chemicals at trace levels. Retention time matching and peaks matching between both instrumental platforms rule out isobaric candidates. With an additional retention time prediction filter from quantitative structure retention relationship (QSRR) model based on random forest feature selection (Pearson r2 = 0.63), identification of small molecules is achieved at a high confidence level without using authentic standards. The developed platform has been validated with robustness by examining spiked environmental chemicals in sludge water samples, biological urine, and cell extracts.

The METLIN small molecule dataset for machine learning-based retention time\xa0prediction

Machine learning has been extensively applied in small molecule analysis to predict a wide range of molecular properties and processes including mass spectrometry fragmentation or chromatographic retention time. However, current approaches for retention time prediction lack sufficient accuracy due to limited available experimental data. Here we introduce the METLIN small molecule retention time (SMRT) dataset, an experimentally acquired reverse-phase chromatography retention time dataset covering up to 80,038 small molecules. To demonstrate the utility of this dataset, we deployed a deep learning model for retention time prediction applied to small molecule annotation. Results showed that in 70% of the cases, the correct molecular identity was ranked among the top 3 candidates based on their predicted retention time. We anticipate that this dataset will enable the community to apply machine learning or first principles strategies to generate better models for retention time prediction.

Extension of the linear solvent strength retention model including a parameter that describes the elution strength changes in liquid chromatography

Modelling the retention behaviour of solutes in liquid chromatography, based on the composition of the mobile phase is a common task in the chromatographic practice. Along the development of liquid chromatography (LC), several models have been proposed to help in understanding the retention mechanisms, and especially, allow the prediction of retention times with optimisation purposes. Particular models are used for different LC modes, such as normal phase (NPLC), reversed phase (RPLC), hydrophilic interaction (HILIC), and micellar (MLC). In this work, a general equation is proposed that includes a parameter (the elution degree, g), which characterises the way the elution strength varies with the modifier concentration. The elution degree adopts the value g = 1 when the system follows the linear solvent strength (LSS) model, where the elution strength is constant. When g > 1, the elution strength decreases, and for g < 1, it increases with the modifier concentration. The proposed equation was applied to experimental retention data obtained for several chromatographic systems in RPLC, HILIC, MLC, and microemulsion LC. It was found that values in the 1 < g < 2 range are most usual. The general behaviour of the proposed equation was studied for isocratic and gradient elution. A general expression to calculate the compression factor of chromatographic peaks in gradient elution was also developed. It is shown that an increasing g value makes retention factors close to zero more difficult, since the elution strength decreases as the modifier concentration increases. For this reason, the larger the g value, the harder it is to reach significant peak compression. In contrast, an elution mode with g < 1 would yield increased elution strength with the modifier concentration, giving rise to significant peak compression.

Automated Isotopic Profile Deconvolution for High Resolution Mass Spectrometric Data (APGC-QToF) from Biological Matrices.

An isotopic profile matching algorithm, the isotopic profile deconvoluted chromatogram (IPDC), was developed to screen for a wide variety of organic compounds in high-resolution mass spectrometry (HRMS) data acquired from instruments with resolution power as low as 22 000 fwhm. The algorithm initiates the screening process by generating a series of C/Br/Cl/S isotopic patterns consistent with the profiles of approximately 3 million molecular formulas for compounds with potentially persistent, bioaccumulative, and toxic (PBT) properties. To evaluate this algorithm, HRMS data were screened using these seed profiles to isolate relevant chlorinated and/or brominated compounds. Data reduction techniques included mass defect filtering and retention time prediction from estimated boiling points predicted using molecular formulas and reasonable elemental conformations. A machine learning classifier was also developed using spectrometric and chromatographic variables to minimize false positives. A scoring system was developed to rank candidate molecular formulas for an isotopic feature. The IPDC algorithm was applied to a Lake Michigan lake trout extract analyzed by atmospheric pressure gas chromatography-quadrupole time-of-flight (APGC-QToF) mass spectrometry in positive and negative modes. The IPDC algorithm detected isotopic features associated with legacy contaminants and a series of unknown halogenated features. The IPDC algorithm resolved 313 and 855 halogenated features in positive and negative modes, respectively, in Lake Michigan lake trout.

Target quantification and semi-target screening of halogenated carboxylic acids in drinking water using ultra-high performance liquid chromatography-quadrupole orbitrap high-resolution mass spectrometry

To monitor the existing and emerging halogenated carboxylic acids (HCAs) in drinking water, a sensitive and rapid ultra-high performance liquid chromatography-quadrupole orbitrap high-resolution mass spectrometry method for simultaneous target quantification of 10 haloacetic acids (HAAs) and semi-target screening of 19 HCAs was developed. After filtration, drinking water samples were injected into the instrument. HCAs were separated on an HSS T3 column and detected by a type of non-target scan in the electrospray ionization negative mode. For target quantification of 10 HAAs, good linearity was obtained and the correlation coefficients were higher than 0.995. The limits of detection were in the range of 0.050–2.0 µg/L. The recoveries were in the range of 89.7%-108%, 83.4%-121%, 77.1%-116% and 80.2%-104% at levels of 2.5, 5.0, 10 and 20 µg/L, respectively, with relative standard deviations of 1.26%-16.9%. For semi-target screening of 19 HCAs, several criteria including accurate m/z, predicted retention time, deduced fragment ions and simulated isotope pattern were used for identification. The method was applied to analyzing 41 drinking water samples successfully. Five HAAs were detected by target quantification, with dichloroacetic acid and trichloroacetic acid exceeding the limits suggested by the U.S. Environmental Protection Agency and the World Health Organization. Eight HCAs were preliminary identified by semi-target screening, and three of them were further confirmed with reference standards purchased later.

Retention time prediction of hydrocarbons in cryogenically modulated comprehensive two-dimensional gas chromatography: A method development and translation application

Thermodynamic modeling of GC × GC separations provides a tool for rapid method evaluation and optimization. Separations of 95 hydrocarbons on two cryogenically modulated GC × GC systems (atmospheric outlet and vacuum outlet) are modeled, displaying average second dimension retention time modeling absolute errors of 0.17 s and 0.12 s respectively, and generating modeled chromatograms which sufficiently represent experimental data. A web-based GC × GC modeling routine is presented which allows users to model separations, currently focused on hydrocarbons, with full control over all system parameters. The method translation capabilities of the application are further demonstrated by replicating Piotrowski et al.’s GC × GC-HRT temporal distribution plots of hydraulic fracturing flowback fluid hydrocarbons [28].

Utility of linear and nonlinear models for retention prediction in liquid chromatography

Linear solvation strength model in reversed-phase liquid chromatography assumes linear relationship between ln k and Φ. In this work we show that this assumption is true only in narrow range of mobile phase strength. The ln k versus Φ relationship could be more accurately described by three-parametric non-linear model in a wide range of eluent strength. We investigated the consequences of non-linearity on retention prediction accuracy and analyte retention behavior in reversed-phase chromatography. When the ln k versus Φ is measured in narrow range of mobile phase strength (ΔΦ ~ 0.1–0.2) both linear and nonlinear models provide comparable retention prediction results. We propose that the linear trend of ln k versus Φ relationship is obtained in the range flanking the elution factor ke (value of retention factor at the column end). We calculated and plotted changes of retention factor of analytes along the column. The visualization illustrates the ranges of retention factor values participating in separation during gradient. For typical gradient slopes employed in liquid chromatography practice and small molecules the elution factor ke value is between 2 and 8. As a simplified generalization for typical gradient slopes we propose using linear ln k versus Φ trend in the k range between 1 and 30. The spreadsheet was utilized to compare the retention prediction accuracy of linear and non-linear retention models. When fitting ln k versus Φ trend in k range 1–30 the simple linear model is in good agreement with nonlinear model with retention time prediction error 0.3–4.7% (for gradient slope 0.013–0.260).

Steroid identification via deep learning retention time predictions and two-dimensional gas chromatography-high resolution mass spectrometry

Untargeted steroid identification represents a great analytical challenge even when using sophisticated technology such as two-dimensional gas chromatography coupled to high resolution mass spectrometry (GC × GCHRMS) due to the chemical similarity of the analytes. Moreover, when analytical standards, mass spectral and retention index databases are not available, compound annotation is cumbersome. Hence, there is a need for the development of retention time prediction models in order to explore new annotation approaches. In this work, we evaluated the use of several in silico methods for retention time prediction in multidimensional gas chromatography. We use three classical machine learning (CML) algorithms (Partial Least Squares (PLS), Support Vector Regression (SVR) and Random Forest Regression (RFR)) and two deep learning approaches (dense neural network (DNN) and three-dimensional convolutional neural network (CNN)). Whereas molecular descriptors were utilized for the CLM and DNN algorithms, three-dimensional molecular representation based on the electrostatic potential (ESP) was studied as input data as is for the CNN. All the developed models showed similar performances with Q2 values over 0.9. However, among all CNN showed the best performance, resulting in average retention time prediction errors of 2% and 6% for the first and second separation dimension, respectively. Additionally, only the three-dimensional ESP representation coupled with CNN was able to extract the stereochemical information crucial for the separation of diastereomers. The combination of retention time prediction and high-resolution mass spectral data applied to clinical samples enabled the untargeted annotation of 12 steroid metabolites in the urine of new-borns.

Retention Time Prediction for Glycopeptides in Reversed-Phase Chromatography for Glycoproteomic Applications.

The sequence-specific retention calculator algorithm (SSRCalc) [ Krokhin , O. V. Anal. Chem. 2006 , 78 , 7785 ] was adapted for the prediction of retention times of N-glycopeptides separated by reversed-phase high performance liquid chromatography (RPLC). The retention time shifts (dHI = HIglyco - HIdeglyco, where HI is the hydrophobicity index, measured in percent acetonitrile units) used for modeling were measured for 602 glycopeptides versus 123 of their deglycosylated analogues. Our method used a tryptic digest of 12 purified glycoproteins, glycopeptide enrichment, deglycosylation with PNGaseF, and RPLC-MS/MS analysis of combined (deglycosylated and intact) peptide mixtures. On average, glycosylation yields a 0.79% acetonitrile unit decrease in retention, compared with the hydrophobicity indices of their deglycosylated analogues. These values, however, are drastically different for asialo (-1.37% acetonitrile units), monosialylated (-0.47% acetonitrile units), disialylated (+0.61% acetonitrile units), and trisialylated (+1.94% acetonitrile units) glycans. Peptide retention time shifts upon glycosylation (dHI) vary depending on the number of monosaccharide units, the presence or absence of sialic acid, peptide hydrophobicity, and the number of position-dependent features. These features are mostly driven by competing effects of acidic residues (aspartic acid and sialic acid) on ion-pair formation and by nearest-neighbor effects of hydrophilic glycans. The accuracy of the modified prediction model for glycopeptides approaches that of the prediction for nonmodified species (R2 = 0.97 vs 0.98). However, retention time prediction based on the experimental retention values of deglycosylated analogues (HIglyco = HIdeglyco + dHI, R2 = 0.995) is much more accurate, thus providing a solid support for glycopeptide identification in complex samples based on mass and retention time.

Retention Time Prediction and Protein Identification.

In bottom-up proteomics, proteins are typically identified by enzymatic digestion into peptides, tandem mass spectrometry and comparison of the tandem mass spectra with those predicted from a sequence database for peptides within measurement uncertainty from the experimentally obtained mass. Although now decreasingly common, isolated proteins or simple protein mixtures can also be identified by measuring only the masses of the peptides resulting from the enzymatic digest, without any further fragmentation. Separation methods such as liquid chromatography and electrophoresis are often used to fractionate complex protein or peptide mixtures prior to analysis by mass spectrometry. Although the primary reason for this is to avoid ion suppression and improve data quality, these separations are based on physical and chemical properties of the peptides or proteins and therefore also provide information about them. Depending on the separation method, this could be protein molecular weight (SDS-PAGE), isoelectric point (IEF), charge at a known pH (ion exchange chromatography), or hydrophobicity (reversed phase chromatography). These separations produce approximate measurements on properties that to some extent can be predicted from amino acid sequences. In the case of molecular weight of proteins without posttranslational modifications this is straightforward: simply add the molecular weights of the amino acid residues in the protein. For IEF, charge and hydrophobicity, the order of the amino acids, and folding state of the peptide or protein also matter, but it is nevertheless possible to predict the behavior of peptides and proteins in these separation methods to a degree which renders such predictions useful. This chapter reviews the topic of using data from separation methods for identification and validation in proteomics, with special emphasis on predicting retention times of tryptic peptides in reversed-phase chromatography under acidic conditions, as this is one of the most commonly used separation methods in bottom-up proteomics.

Localised quantitative structure–retention relationship modelling for rapid method development in reversed-phase high performance liquid chromatography

Quantitative structure–retention relationships (QSRR) predicting the values of solute “hydrophobicity” coefficient η′ in the approximate hydrophobic subtraction model (HSM) can be used to predict retention times of compounds on numerous reversed-phase (RP) columns, provided that column parameters on the corresponding stationary phases are available. In the present study, we propose a new dual clustering-based localised QSRR approach, combining P-ratio clustering (where P is the octanol–water partition coefficient) with second dominant interaction (SDI)-based clustering, to produce predictive models with an acceptable level of prediction accuracy for in silico column scoping in RP method development. QSRR models for η′ values were derived for 49 compounds out of 63 in a dataset extracted from the literature, where retention data were measured under one isocratic mobile phase condition (i.e., acetonitrile-water, 50:50 [v/v]). These models gave a predictive squared correlation coefficient Qext(F2)2 of 0.83 and a root mean square error of prediction (RMSEP) of 0.14. For the modelling, a genetic algorithm-partial least square regression (GA-PLS) approach was performed using the η′ values and their relevant molecular descriptors. The corresponding retention times were predicted by applying the predicted η′ values of the models and the stationary phase “hydrophobicity” parameter H values for the corresponding columns to the approximate HSM, resulting in excellent accuracy and predictability (Qext(F2)2 of 0.90 and RMSEP of 0.72 min). The established QSRR approach was experimentally verified for six Thermo Scientific columns (Acclaim™ 120 C18, Acclaim Polar Advantage, Acclaim Polar Advantage II, Accucore™ aQ, Accucore Phenyl-X, and Hypersil Gold C18 columns) using two types of datasets. The first dataset consisted of eight model compounds extracted from the original dataset and retention time predictions for those compounds were then evaluated on the above columns. The result showed good agreement between predicted and observed retention times with an acceptable error in retention time predictions (slope of 0.97, Qext(F2)2 of 0.95, a mean absolute error (MAE) of 0.43 min and RMSEP of 0.61 min). The second dataset included eight test compounds not included in the original dataset, which were all classified into the η′ cluster by applying a Tanimoto similarity (TS) threshold of 0.7. Similarly, predicted retention times of the test compounds were compared with their corresponding observed retention times, resulting in acceptable retention time predictions with the slope of 0.99, Qext(F2)2 of 0.93 and RMSEP of 0.52 min. Comparisons of resolution values between columns were utilised to select the most suitable columns for separations of the compounds in the respective test sets. Actual chromatograms obtained on the chosen columns showed the feasibility for effective column scoping without experimentation on numerous RP stationary phases available in the USP website, based on the predicted resolution values.

Quantitative Structure-Retention Relationships with Non-Linear Programming for Prediction of Chromatographic Elution Order.

In this work, we employed a non-linear programming (NLP) approach via quantitative structure–retention relationships (QSRRs) modelling for prediction of elution order in reversed phase-liquid chromatography. With our rapid and efficient approach, error in prediction of retention time is sacrificed in favor of decreasing the error in elution order. Two case studies were evaluated: (i) analysis of 62 organic molecules on the Supelcosil LC-18 column; and (ii) analysis of 98 synthetic peptides on seven reversed phase-liquid chromatography (RP-LC) columns with varied gradients and column temperatures. On average across all the columns, all the chromatographic conditions and all the case studies, percentage root mean square error (%RMSE) of retention time exhibited a relative increase of 29.13%, while the %RMSE of elution order a relative decrease of 37.29%. Therefore, sacrificing %RMSE(tR) led to a considerable increase in the elution order predictive ability of the QSRR models across all the case studies. Results of our preliminary study show that the real value of the developed NLP-based method lies in its ability to easily obtain better-performing QSRR models that can accurately predict both retention time and elution order, even for complex mixtures, such as proteomics and metabolomics mixtures.

Wide-scope target and suspect screening methodologies to investigate the occurrence of new psychoactive substances in influent wastewater from Athens

Almost all licit and illicit drugs consumed by the society end up either unchanged or as a mixture of metabolites in the sewage systems. The analysis of influent wastewater samples and the estimation of drug consumption is the field of wastewater-based epidemiology (WBE). A new trend of WBE is the estimation of the consumption of New Psychoactive Substances (NPS), which are legal replacements of established narcotic and psychotropic drugs with slightly modified chemical structures and similar or new effects. To investigate the occurrence of NPS, 30 composite daily influent wastewater samples from the wastewater treatment plant (WWTP) of Athens (Greece) were collected in a four-year sampling campaign (2015–2018). A generic four-sorbent solid-phase extraction (SPE) sample preparation protocol able to retain compounds with wide physicochemical properties was used. Extracts were analyzed by liquid-chromatography coupled to quadrupole-time-of-flight mass spectrometry (LC-QTOF-MS) using target screening for 278 NPS and suspect screening for 451 NPS. Target screening method was validated for a subset of 49 representative NPS and illicit drugs with similar structures with the NPS. 24 NPS and related compounds were detected by target screening and two compounds were tentatively identified based on mass accuracy, prediction of retention time using in-house QSRR prediction models, isotopic pattern and HRMS/MS fragmentation, whereas the excreted mass loads were also calculated. The results indicated an occasional and low occurrence of NPS in wastewater during the week and over the years, whereas the estimation of the exact sources and the evaluation of the patterns in wastewater were critically discussed.

Prosit: proteome-wide prediction of peptide tandem mass spectra by deep learning.

In mass-spectrometry-based proteomics, the identification and quantification of peptides and proteins heavily rely on sequence database searching or spectral library matching. The lack of accurate predictive models for fragment ion intensities impairs the realization of the full potential of these approaches. Here, we extended the ProteomeTools synthetic peptide library to 550,000 tryptic peptides and 21 million high-quality tandem mass spectra. We trained a deep neural network, termed Prosit, resulting in chromatographic retention time and fragment ion intensity predictions that exceed the quality of the experimental data. Integrating Prosit into database search pipelines led to more identifications at >10× lower false discovery rates. We show the general applicability of Prosit by predicting spectra for proteases other than trypsin, generating spectral libraries for data-independent acquisition and improving the analysis of metaproteomes. Prosit is integrated into ProteomicsDB, allowing search result re-scoring and custom spectral library generation for any organism on the basis of peptide sequence alone.

DynaStI: A Dynamic Retention Time Database for Steroidomics.

Steroidomics studies face the challenge of separating analytical compounds with very similar structures (i.e., isomers). Liquid chromatography (LC) is commonly used to this end, but the shared core structure of this family of compounds compromises effective separations among the numerous chemical analytes with comparable physico-chemical properties. Careful tuning of the mobile phase gradient and an appropriate choice of the stationary phase can be used to overcome this problem, in turn modifying the retention times in different ways for each compound. In the usual workflow, this approach is suboptimal for the annotation of features based on retention times since it requires characterizing a library of known compounds for every fine-tuned configuration. We introduce a software solution, DynaStI, that is capable of annotating liquid chromatography-mass spectrometry (LC–MS) features by dynamically generating the retention times from a database containing intrinsic properties of a library of metabolites. DynaStI uses the well-established linear solvent strength (LSS) model for reversed-phase LC. Given a list of LC–MS features and some characteristics of the LC setup, this software computes the corresponding retention times for the internal database and then annotates the features using the exact masses with predicted retention times at the working conditions. DynaStI is able to automatically calibrate its predictions to compensate for deviations in the input parameters. The database also includes identification and structural information for each annotation, such as IUPAC name, CAS number, SMILES string, metabolic pathways, and links to external metabolomic or lipidomic databases.

Retention prediction using quantitative structure-retention relationships combined with the hydrophobic subtraction model in reversed-phase liquid chromatography.

The hydrophobic subtraction model (HSM) combined with quantitative structure-retention relationships (QSRR) methodology was utilized to predict retention times in reversed-phase liquid chromatography (RPLC). A selection of new analytes and new RPLC columns that had never been used in the QSRR modeling process were used to verify the proposed approach. This work is designed to facilitate early prediction of co-elution of analytes in pharmaceutical drug discovery applications where it is advantageous to predict whether impurities might be co-eluted with the active drug component. The QSRR models were constructed through partial least squares regression combined with a genetic algorithm (GA-PLS) which was employed as a feature selection method to choose the most informative molecular descriptors calculated using VolSurf+ software. The analyte hydrophobicity coefficient of the HSM was predicted for subsequent calculation of retention. Clustering approaches based on the local compound type and the local second dominant interaction were investigated to select the most appropriate training set of analytes from a larger database. Predicted retention times of five new compounds on five new RPLC C18 columns were compared with their measured retention times with percentage root-mean-square errors of 15.4 and 24.7 for the local compound type and local second dominant interaction clustering methods, respectively.

Thermodynamics-based modelling of gas chromatography separations across column geometries and systems, including the prediction of peak widths.

Thermodynamics-based models have been demonstrated to be useful for predicting retention time and peak widths in gas chromatography and two-dimensional gas chromatography separations. However, the collection of data to train the models can be time consuming, which lessens the practical utility of the method. In this contribution, a method for obtaining thermodynamic-based data to predict peak widths in temperature-programmed gas chromatography is presented. Experimental work to collect data for peak width prediction is identical to that required to collect data for retention time prediction using approaches that we have presented previously. Using this combined approach, chromatograms including retention times and peak widths are predicted with very high accuracy. Typical errors in retention time are<0.5%, while errors in peak width are typically<5% as demonstrated using polycycic aromatic hydrocarbons and a mixture containing compounds with aldehyde, ketone, alkene, alkane, alcohol, and ester functionalities.

A Refined Nontarget Workflow for the Investigation of Metabolites through the Prioritization by in Silico Prediction Tools.

The application of nontargeted strategies based on high-resolution mass spectrometry (HRMS) directed toward the discovery of metabolites of known contaminants in fish is an interesting alternative to true nontarget screening. To reduce prolonged and costly laboratory experiments, recent advances in computing power have permitted the development of comprehensive knowledge-based software to predict the metabolic fate of chemicals. In addition, machine-based learning tools allow the prediction of chromatographic retention times (RT) or collision cross section (CCS) values when using ion mobility spectrometry (IMS). These tools can ease data evaluation and strengthen the confidence in the identification of compounds. The current work explores the capabilities of in silico prediction tools, refined by the use of RT and CCS prediction, to prioritize and facilitate nontarget liquid chromatography (LC)-IMS-HRMS data processing. The fate of the insecticide pirimiphos-methyl (PM) in farmed Atlantic salmon exposed to contaminated feed was used as a case study. The theoretical prediction of 60 potentially relevant biological PM metabolites permitted the prioritization of screening in different salmon tissues (liver, kidney, bile, muscle, and fat) of known and unknown PM metabolites. An average of 43 potential positives was found in the sample matrixes based on the accurate mass of protonated molecules (mass error ≤5 ppm). The application of different tolerance filters for RT (Δ ≤ 2 min) and CCS (Δ ≤ 6%) based on predicted values permitted us to reduce this number up to 66% of the features. Finally, five PM metabolites could be identified; two known metabolites (2-DAMP and N-desethyl PM) were confirmed with a standard, whereas three previously unknown metabolites (2-DAMP glucuronide, didesethyl PM, and hydroxy-2-DAMP glucuronide) were tentatively identified in different matrixes, allowing the first proposition of a metabolic pathway in fish.

Quantitative structure-retention relationship for chromatographic behaviour of anthraquinone derivatives through considering organic modifier features in micellar liquid chromatography

A simple and informative quantitative structure-retention relationship (QSRR) model has been introduced for prediction of retention times in some anthraquinone derivatives using reversed-phase micellar liquid chromatography (MLC) technique. In the developed multiple linear regression model, the structural descriptors of analytes as well as the empirical parameters of organic modifiers in the applied MLC systems have been considered. Retention times of 77 chromatographic samples (16 anthraquinones were evaluated by using 6 different organic modifiers) were experimentally determined and utilized as the independent variables of the QSRR model. Five small-chain alcohols (methanol, ethanol, propanol, butanol, and pentanol) as well as acetonitrile were used as the eluent modifiers. A five-parametric model was attained for the logarithm of the retention time values which covered about 96 and 95% variance of the chromatographic data in training and cross-validation, respectively. The presence of an excellent correlation coefficient for external validation test (= 0.94) and a well-applicable domain proved the prediction ability of the constructed model. Both validity and reliability of the formulated model were examined through its application on diverse random-selected training and test sets. Moreover, quantum chemical calculations were performed in the framework of density functional theory to simulate the interactions between AQs and modifiers and gain mechanistic details about the retention behavior in the MLC system.

Determining gradient conditions for peptide purification in RPLC with machine-learning-based retention time predictions

A strategy for determining a suitable solvent gradient in silico in preparative peptide separations is presented. The strategy utilizes a machine-learning–based method, called ELUDE, for peptide retention time predictions based on the amino acid sequences of the peptides. A suitable gradient is calculated according to linear solvent strength theory by predicting the retention times of the peptides being purified at three different gradient slopes. The advantage of this strategy is that fewer experiments are needed to develop a purification method, making it useful for labs conducting many separations but with limited resources for method development.The preparative separation of met-enkephalin and leu-enkephalin was used as model solutes on two stationary phases: XBridge C18 and CSH C18. The ELUDE algorithm contains a support vector regression and is pre-trained, meaning that only 10–50 peptides are needed to calibrate a model for a certain stationary phase and gradient. The calibration is done once and the model can then be used for new peptides similar in size to those in the calibration set. We found that the accuracy of the retention time predictions is good enough to usefully estimate a suitable gradient and that it was possible to compare the selectivity on different stationary phases in silico. The absolute relative errors in retention time for the predicted gradients were 4.2% and 3.7% for met-enkephalin and leu-enkephalin, respectively, on the XBridge C18 column and 2.0% and 2.8% on the CSH C18 column. The predicted retention times were also used as initial values for adsorption isotherm parameter determination, facilitating the numerical calculation of overloaded elution profiles. Changing the trifluoroacetic acid (TFA) concentration from 0.05% to 0.15% in the eluent did not seriously affect the error in the retention time predictions for the XBridge C18 column, an increase of 1.0 min (in retention factor, 1.3). For the CSH C18 column the error was, on average, 2.6 times larger. This indicates that the model needs to be recalibrated when changing the TFA concentration for the CSH column.Studying possible scale-up complications from UHPLC to HPLC such as pressure, viscous heating (i.e., temperature gradients), and stationary-phase properties (e.g., packing heterogeneity and surface chemistry) revealed that all these factors were minor to negligible. The pressure effect had the largest effect on the retention, but increased retention by only 3%. In the presented case, method development can therefore proceed using UHPLC and then be robustly transferred to HPLC.