Enhanced methodology for porting ion chromatography retention data

doi:10.1016/j.chroma.2016.01.031

Journal of Chromatography A

Volume 1436, 4 March 2016, Pages 59-63

https://doi.org/10.1016/j.chroma.2016.01.031 Get rights and content

Highlights

•
Extension or porting to include six reference ions.
•
Very significant improvement in porting methodology.
•
Lifetime of existing IC databases can be extended considerably.

Abstract

Porting is a powerful methodology to recalibrate an existing database of ion chromatography (IC) retention times by reflecting the changes of column behavior resulting from either batch-to-batch variability in the production of the column or the manufacture of new versions of a column. This approach has been employed to update extensive databases of retention data of inorganic and organic anions forming part of the “Virtual Column” software marketed by Thermo Fisher Scientific, which is the only available commercial optimization tool for IC separation. The current porting process is accomplished by performing three isocratic separations with two representative analyte ions in order to derive a porting equation which expresses the relationship between old and new data. Although the accuracy of retention prediction is generally enhanced on new columns, errors were observed on some columns. In this work, the porting methodology was modified in order to address this issue, where the porting equation is now derived by using six representative analyte ions (chloride, bromide, iodide, perchlorate, sulfate, and thiosulfate). Additionally, the updated porting methodology has been applied on three Thermo Fisher Scientific columns (AS20, AS19, and AS11HC). The proposed approach showed that the new porting methodology can provide more accurate and robust retention prediction on a wide range of columns, where average errors in retention times for ten test anions under three eluent conditions were less than 1.5%. Moreover, the retention prediction using this new approach provided an acceptable level of accuracy on a used column exhibiting changes in ion-exchange capacity.

Introduction

Method translation or method transfer in chromatographic analysis has become an area of increasing interest. Numerous studies regarding method transfer have been reported in the area of liquid chromatography (LC) [1], [2] as well as gas chromatography (GC) [3], [4]. Method translation in GC is described as the rescaling of method parameters (temperature programs, pressures, etc.) as well as GC components (carrier gases, columns, detectors, etc.) without losing the peak elution pattern [3]. This can lead to the improvement of chromatographic performance in areas such as sample capacity, analysis time, and peak resolution through simple rescaling, along with the reduction of the development time and cost required for the creation of a desired chromatographic analysis method.

Recently, the concept of method transfer has been introduced to update extensive retention databases embedded in the “Virtual Column^®” software, using the so called “porting” methodology [5]. The Virtual Column software allows the simulation of ion chromatography (IC) separations performed under a wide range of experimental conditions (e.g. analytes of interest, eluent type, column type, temperature, flow-rate) in order to identify the optimal eluent and column conditions for a desired separation. This simulation is based on the application of mathematical retention models applied to an extensive database of experimentally-determined analyte retention data embedded in the software [6]. This database covers over 150 anions, cations, and carbohydrate species as analytes, 21 columns, 6 eluent types, 2 column diameters, and 2 temperatures, comprising in total 23,040 datapoints. It is the wide scope of this database which makes Virtual Column such a powerful tool for the development of IC separation methods, but in turn it is also the validity of these retention data which determines the accuracy of the simulated chromatograms. This retention database was constructed around 10 years ago, so the simulation and optimization for IC separations can cause errors when the predicted separations are applied on recently produced columns. These errors can result from changes in column behavior due to either batch-to-batch variability in the production process or the manufacture of new column versions. Errors can also result when the predicted separations are applied to used columns which may have a different ion-exchange capacity to the column on which the original retention data were acquired. With this in mind, a porting methodology to update the retention databases was developed to improve the accuracy of retention prediction by reflecting column-related changes, such as ion-exchange capacity [5].

According to our previously developed porting methodology [5], the retention data embedded into Virtual Column are recalibrated for the entire set of analytes on each particular column by using porting equations which relate existing (or embedded) data to new retention data. In this porting method, new retention data were obtained experimentally by conducting isocratic separations using two representative ions (chloride and thiosulfate) on a new column under three eluent concentrations. Porting equations describing the changes in retention data for these two ions are then derived and are applied to all ions in the database. The general principle of this approach is that any changes in retention observed for chloride and thiosulfate can be generalized across all ions in the database. It has been found that although the porting procedure generally improved the retention prediction accuracy on new columns such as AS20 and AS11HC columns, the retention prediction accuracy for some columns such as AS19 column was poorer than expected. We attribute this to some deficiencies in the porting procedure.

In this study, we have improved the accuracy of the porting procedure by increasing the number of marker anions used to derive the porting equations from two to six. Subsequently, we have validated the modified porting method (MPM) using three newly manufactured Thermo Fisher Scientific columns (AS20, AS19, and AS11-HC). For the validation, the values of mean absolute percentage errors (MAPEs) in the prediction of the retention times were compared in terms of the data types (original embedded data, ported data using the current porting method (CPM), and ported data using the MPM) employed by the mathematical retention model. The accuracy of the retention prediction was then illustrated by plots which show predicted versus measured retention times. Finally, isocratic separations for 13 ions were performed on an AS20 column which had been used for around more than 1500 runs, under three different eluent concentrations. The MPM resulted in a more precise and robust recalibration technique for the update of the retention databases on a wide range of columns compared to the CPM.

Section snippets

General

The isocratic retention data used in this work, which are embedded in the Virtual Column software, had been acquired previously as outlined in Ref. [7]. These isocratic data were collected at different times using different instruments and columns from different manufacturing batches. Therefore, any comparisons of data made between the isocratic measurements will include variability between instruments and column batches.

Reagents and solutions

Standard solutions of the test anions were prepared by dissolution of

Column void time

Since the equations used to predict retention in Virtual Column are based on retention factors, accurate measurement of the column void time was essential. Here, the void time with the columns connected to the IC system was measured using the water dip peak in the chromatograms of marker anions which had been selected to derive the porting equations. The void time without any column connected to the IC system was obtained by injecting water as a sample with the suppressor off where the water

Conclusions

In this study we have modified an existing porting methodology for updating an extensive retention database embedded in the Virtual Column software. Essentially, the number of marker anions used to derive the porting equation was increased from two to six, enabling greater diversity in the nature of the marker ions to be employed. With the modified porting procedure, good agreement between measured and predicted retention times were observed, with the MAPE <1.5% on all three columns tested

Acknowledgment

This research was supported by the Australian Research Council through Linkage Grant LP120200700.

References (12)

D. Guillarme et al.
Method transfer for fast liquid chromatography in pharmaceutical analysis: application to short columns packed with small particle. Part I: isocratic separation
Eur. J. Pharm. Biopharm.
(2007)
D. Guillarme et al.
Method transfer for fast liquid chromatography in pharmaceutical analysis: application to short columns packed with small particle. Part II: gradient experiments
Eur. J. Pharm. Biopharm.
(2008)
B.K. Ng et al.
Methodology for porting retention prediction data from old to new columns and from conventional-scale to miniaturised ion chromatography systems
J. Chromatogr. A
(2011)
B. Trathnigg et al.
Looking inside the pores of a chromatographic column—I. Variation of the pore volume with mobile phase composition
J. Chromatogr. A
(2006)
M. Wang et al.
Comparison of void volume, mobile phase volume and accessible volume determined from retention data for oligomers in reversed- phase liquid chromatographic systems
J. Chromatogr. A
(2011)
J.E. Madden et al.
Critical comparison of retention models for optimisation of the separation of anions in ion chromatography III. Anion chromatography using hydroxide eluents on a dionex AS11 stationary phase
J. Chromatogr. A
(1999)

There are more references available in the full text version of this article.

Cited by (10)

Retention prediction in reversed phase high performance liquid chromatography using quantitative structure-retention relationships applied to the Hydrophobic Subtraction Model
2018, Journal of Chromatography A
Quantitative Structure-Retention Relationships (QSRR) methodology combined with the Hydrophobic Subtraction Model (HSM) have been utilized to accurately predict retention times for a selection of analytes on several different reversed phase liquid chromatography (RPLC) columns. This approach is designed to facilitate early prediction of co-elution of analytes, for example in pharmaceutical drug discovery applications where it is advantageous to predict whether impurities might be co-eluted with the active drug component. The QSRR model utilized VolSurf+ descriptors and a Partial Least Squares regression combined with a Genetic Algorithm (GA-PLS) to predict the solute coefficients in the HSM. It was found that only the hydrophobicity (η' H) term in the HSM was required to give the accuracy necessary to predict potential co-elution of analytes. Global QSRR models derived from all 148 compounds in the dataset were compared to QSRR models derived using a range of local modelling techniques based on clustering of compounds in the dataset by the structural similarity of compounds (as represented by the Tanimoto similarity index), physico-chemical similarity of compounds (represented by log D), the neutral, acidic, or basic nature of the compound, and the second dominant interaction between analyte and stationary phase after hydrophobicity. The global model showed reasonable prediction accuracy for retention time with errors of 30 s and less for up to 50% of modeled compounds. The local models for Tanimoto, nature of the compound and second dominant interaction approaches all exhibited prediction errors less than 30 s in retention time for nearly 70% of compounds for which models could be derived. Predicted retention times of five representative compounds on nine reversed-phase columns were compared with known experimental retention data for these columns and this comparison showed that the accuracy of the proposed modelling approach is sufficient to reliably predict the retention times of analytes based only on their chemical structures.
Towards a chromatographic similarity index to establish localised quantitative structure-retention relationships for retention prediction. II Use of Tanimoto similarity index in ion chromatography
2017, Journal of Chromatography A
Citation Excerpt :
The a- and b-values were obtained from Virtual Column software and recalibrated by the previously reported “porting” methodology [9]. Conditions for separation included an eluent flow-rate of 1.0 mL/min, a column temperature of 30 °C, and suppressed conductivity detection operated at 30 °C, for details please see reference [9]. The cation dataset (Table S2 in Supplementary Information) included 91 larger molecular weight organic monovalent ions (molecular mass from 32 [methylamine] up to 506 [dipyridamole] g mol−1).
Quantitative Structure-Retention Relationships (QSRR) are used to predict retention times of compounds based only on their chemical structures encoded by molecular descriptors. The main concern in QSRR modelling is to build models with high predictive power, allowing reliable retention prediction for the unknown compounds across the chromatographic space. With the aim of enhancing the prediction power of the models, in this work, our previously proposed QSRR modelling approach called “federation of local models” is extended in ion chromatography to predict retention times of unknown ions, where a local model for each target ion (unknown) is created using only structurally similar ions from the dataset. A Tanimoto similarity (TS) score was utilised as a measure of structural similarity and training sets were developed by including ions that were similar to the target ion, as defined by a threshold value. The prediction of retention parameters (a- and b-values) in the linear solvent strength (LSS) model in ion chromatography, log k = a – blog[eluent], allows the prediction of retention times under all eluent concentrations. The QSRR models for a- and b-values were developed by a genetic algorithm-partial least squares method using the retention data of inorganic and small organic anions and larger organic cations (molecular mass up to 507) on four Thermo Fisher Scientific columns (AS20, AS19, AS11HC and CS17). The corresponding predicted retention times were calculated by fitting the predicted a- and b-values of the models into the LSS model equation. The predicted retention times were also plotted against the experimental values to evaluate the goodness of fit and the predictive power of the models. The application of a TS threshold of 0.6 was found to successfully produce predictive and reliable QSRR models (Q_ext(F2)² > 0.8 and Mean Absolute Error < 0.1), and hence accurate retention time predictions with an average Mean Absolute Error of 0.2 min.
Retention prediction of low molecular weight anions in ion chromatography based on quantitative structure-retention relationships applied to the linear solvent strength model
2017, Journal of Chromatography A
Citation Excerpt :
The results obtained in this study show that QSRR models for a and b values can properly predict the retention times of inorganic and small organic anions in IC. Retention data of inorganic and small organic anions used in this work were obtained by recalibrating the retention data embedded in the Thermo Fisher Virtual Column™ software utilising the previously published “porting” methodology [27]. Using this porting strategy the retention data were updated to address possible changes in retention on newly produced columns due to batch-to-batch variability.
Quantitative Structure-Retention Relationships (QSRRs) represent a popular technique to predict the retention times of analytes, based on molecular descriptors encoding the chemical structures of the analytes. The linear solvent strength (LSS) model relating the retention factor, k to the eluent concentration (log k = a − blog [eluent]), is a well-known and accurate retention model in ion chromatography (IC). In this work, QSRRs for inorganic and small organic anions were used to predict the regression parameters a and b in the LSS model (and hence retention times) for these analytes under a wide range of eluent conditions, based solely on their chemical structures. This approach was performed on retention data of inorganic and small organic anions from the “Virtual Column” software (Thermo Fisher Scientific). These retention data were recalibrated via a “porting” methodology on three columns (AS20, AS19, and AS11HC), prior to the QSRR modeling. This provided retention data more applicable on recently produced columns which may exhibit changes of column behavior due to batch-to-batch variability. Molecular descriptors for the analytes were calculated with Dragon software using the geometry-optimized molecular structures, employing the AM1 semi-empirical method. An optimal subset of molecular descriptors was then selected using an evolutionary algorithm (EA). Finally, the QSRR models were generated by multiple linear regression (MLR). As a result, six QSRR models with good predictive performance were successfully derived for a- and b-values on three columns (R² > 0.98 and RMSE < 0.11). External validation showed the possibility of using the developed QSRR models as predictive tools in IC (Q_ext(F3)² > 0.7 and RMSEP < 0.4). Moreover, it was demonstrated that the obtained QSRR models for the a- and b-values can predict the retention times for new analytes with good accuracy and predictability (R² of 0.98, RMSE of 0.89 min, Q_ext(F3)² of 0.96 and RMSEP of 1.18 min).
Towards a chromatographic similarity index to establish localized quantitative structure-retention models for retention prediction: Use of retention factor ratio
2017, Journal of Chromatography A
Citation Excerpt :
The effectiveness of these filters was compared with a reference case in which the whole database, except for the unknown compound, was used as the training set to generate the predictive QSRR model (further referred to as global QSRR modelling). The IC retention dataset (Table S-1), including 49 low molecular weight inorganic and organic anions on two columns (4 mm I.D. Dionex IonPac AS20 and AS19, with void volumes of 2.68 and 2.61 mL respectively) at 30 °C, was obtained by recalibrating the retention data embedded in the Virtual Column software using “porting” methodology [31]. The mobile phases consisted of 35 and 25 mM KOH for the AS20 and AS19 columns, respectively and the flow-rate was 1 mL/min.
Quantitative Structure-Retention Relationships (QSRR) have the potential to speed up the screening phase of chromatographic method development as the initial exploratory experiments are replaced by prediction of analyte retention based solely on the structure of the molecule. The present study offers further proof-of-concept of localized QSRR modelling, in which the retention of any given compound is predicted using only the most chromatographically similar compounds in the available dataset. To this end, each compound in the dataset was sequentially removed from the database and individually utilized as a test analyte. In this study, we propose the retention factor k as the most relevant chromatographic similarity measure and compare it with the Tanimoto index, the most popular similarity measure based on chemical structure. Prediction error was reduced by up to 8 fold when QSRR was based only on chromatographically similar compounds rather than using the entire dataset. The study therefore shows that the design of a practically useful structural similarity index should select the same compounds in the dataset as does the k-similarity filter in order to establish accurate predictive localized QSRR models. While low average prediction errors (Mean Absolute Error (MAE) < 0.5 min) and slopes of the regression lines through the origin close to 1.00 were obtained using k-similarity searching, the use of the structural Tanimoto similarity index, considered as the gold standard in Quantitative Structure-Activity Relationships (QSAR) studies, generally resulted in much higher prediction errors (MAE > 1 min) and significant deviations from the reference slope of 1.0. The Tanomoto similarity index therefore appears to have limited general utility in QSRR studies. Future studies therefore aim at designing a more appropriate chromatographic similarity index that can then be applied for unknown compounds (that is, compounds which have not been tested previously on the chromatographic system used, but for which the chemical structures are known).
Facilitating Structural Elucidation of Small Environmental Solutes in Rplc-Hrms by Retention Index Prediction
2023, SSRN
Prediction of Analyte Retention Time in Liquid Chromatography
2021, Analytical Chemistry

View all citing articles on Scopus

View full text

Enhanced methodology for porting ion chromatography retention data

Highlights

Abstract

Introduction

Section snippets

General

Reagents and solutions

Column void time

Conclusions

Acknowledgment

Eur. J. Pharm. Biopharm.

Eur. J. Pharm. Biopharm.

J. Chromatogr. A

J. Chromatogr. A

J. Chromatogr. A

J. Chromatogr. A