QSRR driven insight into retention in multimodal chromatography
Конференцијски прилог (Објављена верзија)
Метаподаци
Приказ свих података о документуАпстракт
Liquid chromatography system in which several separation mechanisms are integrated in the
composition of a single column is called multimodal or mixed-mode chromatography (MMC). MMC
systems are classified based on combined separation mechanisms as bimodal (RP/HILIC, RP/IEX,
HILIC/IEX) and trimodal (different combinations of RP/HILIC/IEX). The main benefit of MMC lies
in widening the spectra of properties of analytes that can be simultaneously chromatographed
(nonpolar, polar, organic, inorganic, ionized and / or non-ionized analytes). In this way, it is possible
to reduce the number of required analyses for one complex sample compared to unimodal
chromatographic systems. For that reason, the popularity of MMC has been growing fast in recent
years. However, in line with this achievement, MMC is characterized by large number of
intermolecular interactions governing separations which are related to the properties of the analyte
(charge and polarity) and chromatographic condition...s (ionic strength and the pH of the aqueous phase
and the content of the organic solvent) [1]. In order to get insight into relative contribution of
aforementioned factors to retention of selected group of analytes, preferred chemometric approach is
Quantitative Structure Retention Relationship (QSRR) study. The QSRR models relate the physicalchemical
properties of analytes reflected by assigned molecular descriptors with their retention
behaviour in predefined experimental space described by the range of chromatographic conditions
(instrumental and mobile phase composition related factors). Apart from its general purpose to assist
in the characterization of observed chromatographic system, the reliable predictions of retention
behaviour of so-called system blind analytes (analytes of known chemical structure but not subjected
to experimentations) can also be derived from a QSRR model. In such way, the development of MMC
based analytical method can be rationalized by saving time and other resources [2].
This research demonstrates the QSRR study performed on 31 pharmaceuticals covering wide range
of polarities, acid-base properties and divergent retention in RP/WCX system (Thermo Acclaim
Mixed Mode WCX-1 3 μm, 2.1x150 mm column). This system was subjected to variations of the
mobile phase composition (30-50% (v/v) of acetonitrile; 3.8-5.6 pH value and 20-40 mM ionic
strength of acetic buffer) and column temperature (30–38 °C) according to the plan of central
composite design of experiments. The machine learning algorithm based on Artificial Neural
Network was used for relating these independent variables to cube root transformed retention factors
of analytes as observed responses. The network comprising of 11-7-1 topology was trained through
1200 cycles with learning rate set at 0.3 and momentum set at 0.5. Cross validation and external
validation were used to prove good statistical performances of built model (Root Mean Square Error
values 0.131 and 0.147 and Squared Correlation vales 0.963 and 0.944, respectively). According to
the weighting scheme used, volume fraction of acetonitrile, pH of aqueous phase and descriptors
related to hydrophobicity and molecule size demonstrated the greatest impact towards retention in
MMC.
Извор:
12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event, 2021, 50-50Издавач:
- Aristotle University of Thessaloniki
- National Technical University of Athens
Напомена:
- Book of Abstracts
Институција/група
PharmacyTY - CONF AU - Otašević, Biljana AU - Svrkota, Bojana AU - Krmar, Jovana AU - Protić, Ana AU - Zečević, Mira PY - 2021 UR - https://farfar.pharmacy.bg.ac.rs/handle/123456789/4693 AB - Liquid chromatography system in which several separation mechanisms are integrated in the composition of a single column is called multimodal or mixed-mode chromatography (MMC). MMC systems are classified based on combined separation mechanisms as bimodal (RP/HILIC, RP/IEX, HILIC/IEX) and trimodal (different combinations of RP/HILIC/IEX). The main benefit of MMC lies in widening the spectra of properties of analytes that can be simultaneously chromatographed (nonpolar, polar, organic, inorganic, ionized and / or non-ionized analytes). In this way, it is possible to reduce the number of required analyses for one complex sample compared to unimodal chromatographic systems. For that reason, the popularity of MMC has been growing fast in recent years. However, in line with this achievement, MMC is characterized by large number of intermolecular interactions governing separations which are related to the properties of the analyte (charge and polarity) and chromatographic conditions (ionic strength and the pH of the aqueous phase and the content of the organic solvent) [1]. In order to get insight into relative contribution of aforementioned factors to retention of selected group of analytes, preferred chemometric approach is Quantitative Structure Retention Relationship (QSRR) study. The QSRR models relate the physicalchemical properties of analytes reflected by assigned molecular descriptors with their retention behaviour in predefined experimental space described by the range of chromatographic conditions (instrumental and mobile phase composition related factors). Apart from its general purpose to assist in the characterization of observed chromatographic system, the reliable predictions of retention behaviour of so-called system blind analytes (analytes of known chemical structure but not subjected to experimentations) can also be derived from a QSRR model. In such way, the development of MMC based analytical method can be rationalized by saving time and other resources [2]. This research demonstrates the QSRR study performed on 31 pharmaceuticals covering wide range of polarities, acid-base properties and divergent retention in RP/WCX system (Thermo Acclaim Mixed Mode WCX-1 3 μm, 2.1x150 mm column). This system was subjected to variations of the mobile phase composition (30-50% (v/v) of acetonitrile; 3.8-5.6 pH value and 20-40 mM ionic strength of acetic buffer) and column temperature (30–38 °C) according to the plan of central composite design of experiments. The machine learning algorithm based on Artificial Neural Network was used for relating these independent variables to cube root transformed retention factors of analytes as observed responses. The network comprising of 11-7-1 topology was trained through 1200 cycles with learning rate set at 0.3 and momentum set at 0.5. Cross validation and external validation were used to prove good statistical performances of built model (Root Mean Square Error values 0.131 and 0.147 and Squared Correlation vales 0.963 and 0.944, respectively). According to the weighting scheme used, volume fraction of acetonitrile, pH of aqueous phase and descriptors related to hydrophobicity and molecule size demonstrated the greatest impact towards retention in MMC. PB - Aristotle University of Thessaloniki PB - National Technical University of Athens C3 - 12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event T1 - QSRR driven insight into retention in multimodal chromatography SP - 50 EP - 50 UR - https://hdl.handle.net/21.15107/rcub_farfar_4693 ER -
@conference{ author = "Otašević, Biljana and Svrkota, Bojana and Krmar, Jovana and Protić, Ana and Zečević, Mira", year = "2021", abstract = "Liquid chromatography system in which several separation mechanisms are integrated in the composition of a single column is called multimodal or mixed-mode chromatography (MMC). MMC systems are classified based on combined separation mechanisms as bimodal (RP/HILIC, RP/IEX, HILIC/IEX) and trimodal (different combinations of RP/HILIC/IEX). The main benefit of MMC lies in widening the spectra of properties of analytes that can be simultaneously chromatographed (nonpolar, polar, organic, inorganic, ionized and / or non-ionized analytes). In this way, it is possible to reduce the number of required analyses for one complex sample compared to unimodal chromatographic systems. For that reason, the popularity of MMC has been growing fast in recent years. However, in line with this achievement, MMC is characterized by large number of intermolecular interactions governing separations which are related to the properties of the analyte (charge and polarity) and chromatographic conditions (ionic strength and the pH of the aqueous phase and the content of the organic solvent) [1]. In order to get insight into relative contribution of aforementioned factors to retention of selected group of analytes, preferred chemometric approach is Quantitative Structure Retention Relationship (QSRR) study. The QSRR models relate the physicalchemical properties of analytes reflected by assigned molecular descriptors with their retention behaviour in predefined experimental space described by the range of chromatographic conditions (instrumental and mobile phase composition related factors). Apart from its general purpose to assist in the characterization of observed chromatographic system, the reliable predictions of retention behaviour of so-called system blind analytes (analytes of known chemical structure but not subjected to experimentations) can also be derived from a QSRR model. In such way, the development of MMC based analytical method can be rationalized by saving time and other resources [2]. This research demonstrates the QSRR study performed on 31 pharmaceuticals covering wide range of polarities, acid-base properties and divergent retention in RP/WCX system (Thermo Acclaim Mixed Mode WCX-1 3 μm, 2.1x150 mm column). This system was subjected to variations of the mobile phase composition (30-50% (v/v) of acetonitrile; 3.8-5.6 pH value and 20-40 mM ionic strength of acetic buffer) and column temperature (30–38 °C) according to the plan of central composite design of experiments. The machine learning algorithm based on Artificial Neural Network was used for relating these independent variables to cube root transformed retention factors of analytes as observed responses. The network comprising of 11-7-1 topology was trained through 1200 cycles with learning rate set at 0.3 and momentum set at 0.5. Cross validation and external validation were used to prove good statistical performances of built model (Root Mean Square Error values 0.131 and 0.147 and Squared Correlation vales 0.963 and 0.944, respectively). According to the weighting scheme used, volume fraction of acetonitrile, pH of aqueous phase and descriptors related to hydrophobicity and molecule size demonstrated the greatest impact towards retention in MMC.", publisher = "Aristotle University of Thessaloniki, National Technical University of Athens", journal = "12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event", title = "QSRR driven insight into retention in multimodal chromatography", pages = "50-50", url = "https://hdl.handle.net/21.15107/rcub_farfar_4693" }
Otašević, B., Svrkota, B., Krmar, J., Protić, A.,& Zečević, M.. (2021). QSRR driven insight into retention in multimodal chromatography. in 12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event Aristotle University of Thessaloniki., 50-50. https://hdl.handle.net/21.15107/rcub_farfar_4693
Otašević B, Svrkota B, Krmar J, Protić A, Zečević M. QSRR driven insight into retention in multimodal chromatography. in 12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event. 2021;:50-50. https://hdl.handle.net/21.15107/rcub_farfar_4693 .
Otašević, Biljana, Svrkota, Bojana, Krmar, Jovana, Protić, Ana, Zečević, Mira, "QSRR driven insight into retention in multimodal chromatography" in 12th International Conference on Instrumental Methods of Analysis, Modern Trends and Applications, 20-23 September 2021, Virtual event (2021):50-50, https://hdl.handle.net/21.15107/rcub_farfar_4693 .