Journal
TALANTA
Volume 107, Issue -, Pages 368-375Publisher
ELSEVIER
DOI: 10.1016/j.talanta.2013.01.035
Keywords
Local weighted-partial least squares regression (LW-PLSR); Human serum analysis; Vibrational spectroscopy; Infrared (IR); Chemometrics
Categories
Funding
- Ministerio de Educacion y Ciencia [CTQ2011-25743]
- Generalitat Valenciana [PROMETEO 2010-055]
- University of Valencia
- Instituto Carlos III (Ministry of Economy and Competitiveness) [Sara Borrell CD12/00667]
Ask authors/readers for more resources
Locally weighted partial least squares regression (LW-PLSR) has been applied to the determination of four clinical parameters in human serum samples (total protein, triglyceride, glucose and urea contents) by Fourier transform infrared (FTIR) spectroscopy. Classical LW-PLSR models were constructed using different spectral regions. For the selection of parameters by LW-PLSR modeling, a multi-parametric study was carried out employing the minimum root-mean square error of cross validation (RMSCV) as objective function. In order to overcome the effect of strong matrix interferences on the predictive accuracy of LW-PLSR models, this work focuses on sample selection. Accordingly, a novel strategy for the development of local models is proposed. It was based on the use of: (i) principal component analysis (PCA) performed on an analyte specific spectral region for identifying most similar sample spectra and (ii) partial least squares regression (PLSR) constructed using the whole spectrum. Results found by using this strategy were compared to those provided by PLSR using the same spectral intervals as for LW-PLSR. Prediction errors found by both, classical and modified LW-PLSR improved those obtained by PLSR. Hence, both proposed approaches were useful for the determination of analytes present in a complex matrix as in the case of human serum samples. (C) 2013 Elsevier B.V. All rights reserved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available