A feed-forward artificial neural network (ANN) learned by error back-propagation is used to generate a retention predictive model for phenoxy acid herbicides in isocratic reversedphase high-performance liquid chromatography. The investigated solutes (18 compounds), apart from the most common herbicides of this class, include some derivatives of benzoic acid and phenylacetic acid structurally related to phenoxy acids, as a whole covering a pKa range between 2.3 and 4.3. A mixed model in terms of both solute descriptors and eluent attributes is built with the aim of predicting retention in water–acetonitrile mobile phases within a large range of composition (acetonitrile from 30% to 70%, v/v) and acidity (pH of water before mixing with acetonitrile ranging between 2 and 5). The set of input variables consists of solute pKa and quantum chemical molecular descriptors of both the neutral and dissociated form, %v/v of acetonitrile in the mobile phase and pH of aqueous phase before mixing with acetonitrile. After elimination of redundant variables, a ninedimensional model is identified and its prediction ability is evaluated by external validation based on three solutes not involved in model generation and by cross-validation. A multilinear counterpart in terms of the same descriptors is seen to provide a noticeably poorer retention prediction.

A feed-forward artificial neural network (ANN) learned by error back-propagation is used to generate a retention predictive model for phenoxy acid herbicides in isocratic reversed-phase high-performance liquid chromatography. The investigated solutes (18 compounds), apart from the most common herbicides of this class, include some derivatives of benzoic acid and phenylacetic acid structurally related to phenoxy acids, as a whole covering a pK(a) range between 2.3 and 4.3. A mixed model in terms of both solute descriptors and eluent attributes is built with the aim of predicting retention in water-acetonitrile mobile phases within a large range of composition (acetonitrile from 30% to 70%, v/v) and acidity (pH of water before mixing with acetonitrile ranging between 2 and S). The set of input variables consists of solute pK(a) and quantum chemical molecular descriptors of both the neutral and dissociated form, %v/v of acetonitrile in the mobile phase and pH of aqueous phase before mixing with acetonitrile. After elimination of redundant variables, a nine-dimensional model is identified and its prediction ability is evaluated by external validation based on three solutes not involved in model generation and by cross-validation. A multi-linear counterpart in terms of the same descriptors is seen to provide a noticeably poorer retention prediction. (C) 2008 Elsevier B.V. All rights reserved.

Modelling of the effect of solute structure and mobile phase pH and composition on the retention of phenoxy acid herbicides in reversed-phase high-performance liquid chromatography

ASCHI, MASSIMILIANO;D'ARCHIVIO, ANGELO ANTONIO
;
RUGGIERI, FABRIZIO
2008-01-01

Abstract

A feed-forward artificial neural network (ANN) learned by error back-propagation is used to generate a retention predictive model for phenoxy acid herbicides in isocratic reversed-phase high-performance liquid chromatography. The investigated solutes (18 compounds), apart from the most common herbicides of this class, include some derivatives of benzoic acid and phenylacetic acid structurally related to phenoxy acids, as a whole covering a pK(a) range between 2.3 and 4.3. A mixed model in terms of both solute descriptors and eluent attributes is built with the aim of predicting retention in water-acetonitrile mobile phases within a large range of composition (acetonitrile from 30% to 70%, v/v) and acidity (pH of water before mixing with acetonitrile ranging between 2 and S). The set of input variables consists of solute pK(a) and quantum chemical molecular descriptors of both the neutral and dissociated form, %v/v of acetonitrile in the mobile phase and pH of aqueous phase before mixing with acetonitrile. After elimination of redundant variables, a nine-dimensional model is identified and its prediction ability is evaluated by external validation based on three solutes not involved in model generation and by cross-validation. A multi-linear counterpart in terms of the same descriptors is seen to provide a noticeably poorer retention prediction. (C) 2008 Elsevier B.V. All rights reserved.
2008
A feed-forward artificial neural network (ANN) learned by error back-propagation is used to generate a retention predictive model for phenoxy acid herbicides in isocratic reversedphase high-performance liquid chromatography. The investigated solutes (18 compounds), apart from the most common herbicides of this class, include some derivatives of benzoic acid and phenylacetic acid structurally related to phenoxy acids, as a whole covering a pKa range between 2.3 and 4.3. A mixed model in terms of both solute descriptors and eluent attributes is built with the aim of predicting retention in water–acetonitrile mobile phases within a large range of composition (acetonitrile from 30% to 70%, v/v) and acidity (pH of water before mixing with acetonitrile ranging between 2 and 5). The set of input variables consists of solute pKa and quantum chemical molecular descriptors of both the neutral and dissociated form, %v/v of acetonitrile in the mobile phase and pH of aqueous phase before mixing with acetonitrile. After elimination of redundant variables, a ninedimensional model is identified and its prediction ability is evaluated by external validation based on three solutes not involved in model generation and by cross-validation. A multilinear counterpart in terms of the same descriptors is seen to provide a noticeably poorer retention prediction.
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11697/3417
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 8
  • ???jsp.display-item.citation.isi??? 8
social impact