Elsevier

Analytica Chimica Acta

Volume 1197, 8 March 2022, 339463
Analytica Chimica Acta

Investigation of supercritical fluid chromatography retention behaviors using quantitative structure-retention relationships

https://doi.org/10.1016/j.aca.2022.339463Get rights and content

Highlights

  • Packed column SFC retention behaviors.

  • Stepwise-PLS QSRR model.

  • Bootstrapping and Monte Carlo resampling.

  • SFC local interactions profile.

Abstract

Supercritical Fluid Chromatography (SFC), a high-throughput separation technique, has been widely applied as a promising routine method in pharmaceutical, pesticides, and metabolome analysis in the same way as conventional liquid chromatography and gas chromatography. However, the retention behaviors of many compounds in SFC are not fully investigated. In this study, more than 500 pesticides were analyzed on several polar and nonpolar columns using SFC/MS/MS. Then, partial least squares regression (PLS) was used to explore the retention behaviors of pesticides and construct the quantitative structure-retention relationships under practical gradient elution. The optimized relationships between pesticide structures and pesticide retention were established and validated for predicting power using both internal- and external-validations; hence, several important factors affecting retention of the compounds were identified. In the best case, approximately almost all pesticides in the training set and nearly 80% of pesticides in the external validation set could be predicted with the prediction error of less than 0.5 min. Moreover, the proposed workflow successfully established the local interaction profiles, describing the possible interactions in the 8 studied chromatographic systems, and can be further applied for any groups of compounds under any system conditions.

Introduction

Supercritical fluid chromatography (SFC) is a chromatography separation technique, which employs supercritical mobile phases (e.g., supercritical CO2) and packed chromatography columns. SFC is considered to have several practical advantages relative to other conventional techniques (e.g., liquid chromatography (LC), gas chromatography (GC)) such as shorter analysis time, higher throughput, and faster equilibration, while the technique still shows comparable working range and sensitivity due to the lower viscosity and higher diffusivity of the super(sub)critical CO2 (scCO2) mobile phases [[1], [2], [3], [4]]. Almost all compounds that are soluble in methanol (MeOH) or a less polar solvent, can be subjected to analyze with SFC. In addition, the SFC uses almost the same instrument and software developed for HPLC, consumes less organic solvent, and can enrich the analytes in a quite small solvent volume due to the post-column CO2 evaporation. SFC has been widely applied in pharmaceutical industry, environmental analysis, food science, and in metabolomics, especially for the analysis of lipids. The technique used to be applied mainly for the analysis of non- and mid-polar compounds, however, recently the analysis of polar and very polar metabolites, which are poorly soluble in CO2, has also become the main concerns of many SFC studies [5,6]. Although SFC has been considered a powerful, high-throughput technique to solve many analytical challenges, unfortunately a complete understanding regarding the SFC retention mechanisms like those in LC and GC is still missing.

Quantitative structure-retention relationships (QSRR), which is one of the popular approaches to study the SFC retention mechanisms, focuses on using the compound molecular descriptors (MDs) that can well reflect the molecular features (e.g., molecular physicochemical properties and basic intermolecular interactions) to establish the relationship with chromatographic retention data of these compounds [[7], [8], [9]]. A reliable QSRR model is expected to describe a good statistical relationship between the MDs (i.e., model variables) and the compound retention data (i.e., model responses) in the chromatographic systems by using several multivariate analysis techniques such as multiple linear regression (MLR) [10], partial least squares regression (PLS) [11,12], and machine learning (ML) [8,[13], [14], [15], [16], [17], [18]]. The choice of multivariate analysis techniques for the models is highly dependent on the characteristics of the datasets and the study purposes. A good prediction QSRR model indeed will add more reliability for the correct identification of unknown compounds in non-targeted analysis, benefit the development/optimization of the chromatographic methods, and provide more insight to the specific chromatography mechanisms [18,19]. Regarding the use of QSRR for the studying SFC retention mechanisms, Muteki et al. proposed a novel QSRR modeling strategy involving the combination of partial least squares discriminant analysis (PLS-DA) models, which classify compounds of interest based on the similar interactive relationships between the mobile-phase and stationary phase, and PLS-R model to predict the retention time of 11 compounds based on the mobile phase conditions, stationary phases, and analyte properties [20]. West and Lesellier have contributed great works regarding the investigation of SFC retention mechanisms under the isocratic modes by establishing several MLR models with the use of linear solvation energy relationships (LSER) descriptors [[21], [22], [23], [24], [25], [26]]. Results from the studies implied that the retention of the pharmaceutical compounds in the studied SFC condition was contributed by the combination of intermolecular interactions between the analytes, mobile phases, and stationary phases (i.e., the interactions from the presence of non-bonding and π electrons, the dipolarity/polarizability interaction, the overall solute hydrogen-bond acidity and basicity interactions, the endoergic cavity formation process and the exoergic dispersion interactions). Recently Gros et al. proposed a PLS- based LSER model using 2 more descriptors describing the shape features of compounds to investigate the retention mechanisms of 14 SFC-dedicated columns from the Shim-pack UC series [27]. The results suggested that the large binding ligands groups might prevent the insertion of bulky molecules, and the molecules having small, spherical, and rigid structures were better retained on the studied polar stationary phases. Despite many efforts of using QSRR, the SFC separation mechanisms have not been fully understood due to many reasons such as the highly specificities of the studies, small datasets, or lack of systematic investigation strategy [24,[28], [29], [30], [31], [32], [33], [34], [35], [36], [37]]. Another problem is the assessment of the model internal- and external-prediction abilities in many studies is neglected; hence, there is no way to confirm if the proposed models are under- or over-fitted.

In this study, a comprehensive investigation regarding the achiral retention behaviors of the pesticides on several packed columns under gradient elution was conducted using the PLS-based QSRR approach.

Section snippets

Chemicals and reagents

CO2 (99.9% grade, Iwatani Corporation, Osaka, Japan) was used as the mobile phase of SFC. LC/MS grade methanol was purchased from Wako Pure Chemical Ind. Ltd. (Osaka, Japan). LC/MS grade ammonium formate was purchased from Sigma-Aldrich (St. Louis, MO, USA). The pesticide standard mixtures (i.e., PL2005 Pesticide GC–MS Mix I, II, III, IV, V, VI, 7/PL2005 Pesticide LC–MS Mix I, II, III, 4, 5, 6, 7, 8, 9, 10/53 Polar pesticides Mix for STQ method) was purchased from Hayashi Pure Chemical Ind.

Retention time investigation

In this study, only compounds from the mixture of 509 pesticides that could be reliably retained, software extracted, and identified were used. The average reliable RT of all pesticides on all studied columns are provided in Supplementary Information (Table S1). There were more reliable identified compounds to be found on the ODS columns than on the polar columns. Moreover, results from PLS-DA using 3 latent components, which explain more than 75% variance of RT for all columns, showed that the

Conclusions

In this study, the compound behaviors under gradient elution in the SFC systems were investigated using the preliminary set of 509 pesticides. The proposed QSRR approach could provide useful information regarding the combination effects of interactions/properties affecting the compound retention under the studied gradient elution, and can be systematically applied, validated for the investigation in other systems. The use of highly diverse dataset as the “red-pill” allowed exploring all the

CRediT authorship contribution statement

Le Si-Hung: Conceptualization, Methodology, Data processing, Writing – original draft. Yoshihiro Izumi: Conceptualization, Investigation, Writing – review & editing. Motonao Nakao: Methodology. Masatomo Takahashi: Methodology, Validation. Takeshi Bamba: Conceptualization, Supervision, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

The author gratefully acknowledges Hedda Drexler (Institute of Materials Chemistry, TU Wien) and Okotta Hara, Kaan Georg Kutlucinar (Department of Analytical Chemistry, BOKU Vienna) for their helpful advices. This study was supported by the Development of Systems and Technology for Advanced Measurement and Analysis Project from Japan Science and Technology Agency (JST) [T. B.]; the Grant-in-Aid for Scientific Research on Innovative Areas (17H06304 & 17H06299) [T.B.] and Grant-in-Aid for

References (57)

  • A.A. D'Archivio et al.

    Cross-column prediction of gas-chromatographic retention of polychlorinated biphenyls by artificial neural networks

    J. Chromatogr., A

    (2011)
  • J.P.M. Andries et al.

    Improvement of quantitative structure–retention relationship models for chromatographic retention prediction of peptides applying individual local partial least squares models

    Talanta

    (2020)
  • C. West et al.

    Characterization of stationary phases in subcritical fluid chromatography by the solvation parameter model: I. Alkylsiloxane-bonded stationary phases

    J. Chromatogr., A

    (2006)
  • C. West et al.

    Characterisation of stationary phases in subcritical fluid chromatography with the solvation parameter model IV. Aromatic stationary phases

    J. Chromatogr., A

    (2006)
  • C. West et al.

    Possibility of predicting separations in supercritical fluid chromatography with the solvation parameter model

    J. Chromatogr., A

    (2009)
  • C. West et al.

    Retention characteristics of porous graphitic carbon in subcritical fluid chromatography with carbon dioxide-methanol mobile phases

    J. Chromatogr., A

    (2004)
  • C. West et al.

    A unified classification of stationary phases for packed column supercritical fluid chromatography

    J. Chromatogr., A

    (2008)
  • Q. Gros et al.

    Characterization of stationary phases in supercritical fluid chromatography including exploration of shape selectivity

    J. Chromatogr., A

    (2021)
  • V. Grover et al.

    A thermodynamic approach for predicting retention time characteristics in supercritical fluid chromatographs

    J. Supercrit. Fluids

    (1990)
  • T.A. Berger et al.

    Correlation between column surface area and retention of polar solutes in packed-column supercritical fluid chromatography

    J. Chromatogr., A

    (1992)
  • C. Jiang et al.

    Study on retention factor and resolution of tocopherols by supercritical fluid chromatography

    J. Chromatogr., A

    (2003)
  • M. Enmark et al.

    Investigation of robustness for supercritical fluid chromatography separation of peptides: isocratic vs gradient mode

    J. Chromatogr., A

    (2018)
  • M. Roth

    Solute retention in supercritical fluid chromatography as a function of pressure: a thermodynamic analysis

    J. Supercrit. Fluids

    (1990)
  • E. Tyteca et al.

    Possibilities of retention modeling and computer assisted method development in supercritical fluid chromatography

    J. Chromatogr., A

    (2015)
  • M. Gevrey et al.

    Review and comparison of methods to study the contribution of variables in artificial neural network models

    Ecol. Model.

    (2003)
  • R. Goodacre et al.

    Metabolomics by numbers: acquiring and understanding global metabolite data

    Trends Biotechnol.

    (2004)
  • S. Wold et al.

    PLS-regression: a basic tool of chemometrics

  • M. Talebi et al.

    Performance comparison of partial least squares-related variable selection methods for quantitative structure retention relationships modelling of retention times in reversed-phase liquid chromatography

    J. Chromatogr., A

    (2015)
  • Cited by (2)

    • CORAL: Quantitative Structure Retention Relationship (QSRR) of flavors and fragrances compounds studied on the stationary phase methyl silicone OV-101 column in gas chromatography using correlation intensity index and consensus modelling

      2022, Journal of Molecular Structure
      Citation Excerpt :

      However, with today's computational capabilities, topological descriptors are simple to compute, but these aren't always connected to retention phenomena. The quantum-chemical descriptors have the advantage of correlating the retention phenomenon at a molecular level, but, the computation of these descriptors is time-consuming and requires high computation power [18–27]. Therefore, a QSRR model is needed that includes descriptors that are easy to calculate and do not depend on temperature.

    View full text