Modelling highly co-eluted peaks of analytes with high spectral similarity

Modelling co-eluted peaks has always been a main keystone in chemometric applications on chromatographic data. This interest is nowadays increased due to the higher capability of modern chromatographic devices coupled with multichannel detectors and obtaining complex data structures. Techniques like...

Descripción completa

Detalles Bibliográficos
Autores: Bordagaray Eizaguirre, Ane, Amigo Rubio, José Manuel
Tipo de recurso: artículo
Fecha de publicación:2015
País:España
Institución:Universidad del País Vasco
Repositorio:Addi. Archivo Digital para la Docencia y la Investigación
OAI Identifier:oai:addi.ehu.eus:10810/71157
Acceso en línea:http://hdl.handle.net/10810/71157
Access Level:acceso abierto
Palabra clave:chemometrics
chromatography
co-eluted peak
co-elution
curve resolution
enantiomer
multivariate Curve Resolution (MCR)
parallel pactor analysis with linear dependence (PARALIND)
parallel factor analysis 2 (PARAFAC2)
spectral similarity
Descripción
Sumario:Modelling co-eluted peaks has always been a main keystone in chemometric applications on chromatographic data. This interest is nowadays increased due to the higher capability of modern chromatographic devices coupled with multichannel detectors and obtaining complex data structures. Techniques like Multivariate Curve Resolution (MCR) or Parallel Factors Analysis 2 (PARAFAC2) have been widely used as curve resolution methods to solve the co-elution problem (among others). The main advantage of these curve resolution techniques is that they profit the property of uniqueness of the spectrum for each co-eluted analyte. Nevertheless, there are cases where these curve resolution approaches may fail. That is when the same analyte gives two co-eluted peaks or when two analytes have highly correlated spectra, leading to a rank deficiency/co-linearity problem. In this paper we put forward the usefulness of a recent multi-way technique, parallel factor analysis with linear dependence (PARALIND) that is able to handle overlapping peaks when the spectral profile of the analytes are highly correlated. To illustrate the problem, two different situations are thoughtfully studied here: first, the co-elution of two peaks, belonging to L-proline produced in a wrong derivatization step and; second, an overlapped mixture of triazole fungicides with a high similarity in their spectra