Interval Partial Least Squares Research Articles

Application of SIMPLISMA purity function for variable selection in multivariate regression analysis: A case study of protein secondary structure determination from infrared spectra

A novel approach for the pre-selection of wavelengths, to be used in combination with Partial Least Squares (PLS) or other multivariate regression techniques, is presented. This variable selection method makes use of the purity function, originally suggested in the SIMPLe-to-use Interactive Self-modeling Mixture Analysis (SIMPLISMA) algorithm, to map up the regions of potentially influential variables. The selected intervals are then individually tested in practical modeling and prediction, and an optimal subset of variables is obtained. The algorithm is simple and intuitive and does not rely on iterative variable searches. The method was tested on a set of infrared protein spectra in order to improve the quantitative determination of the fractions of two secondary structure elements, α-helices and β-strands (β-sheets) in the protein polypeptide chain. Comparable results to those obtained through interval PLS (iPLS), an exhaustive search-based algorithm, were achieved in this study. Our method was shown to be particularly beneficial in combination with variable weighting by their inverse standard deviation.

Quantification of the degree of blockiness in pectins using 1H NMR spectroscopy and chemometrics

The gelling properties of pectins are known not only to be closely related to the degree of esterification (DE), but also to the distribution of the ester groups. In this study we have examined an experimentally designed series of pectins originating from the same mother pectin and deesterified using combinations of two different enzymatic mechanisms. The DE and distribution patterns of methyl ester groups have been analyzed using high-resolution (HR) 1H nuclear magnetic resonance (NMR) spectroscopy on pectin solutions. Quantitative calibration models using partial least squares (PLS) regression were developed with the ability to predict DE as well as the specific enzyme treatment, expressed as amount of ester groups removed with random and block enzyme, respectively. NMR spectroscopy was able to distinguish between enzyme treatments in simple classification by principal component analysis (PCA). This was due to the spatial structure of pectin together with the methyl ester distribution. Nuclear Overhauser effect spectroscopy (NOESY) experiments confirmed all the general assignments with the expected nuclear Overhauser effect (NOE) correlations. Degree of random deesterification (R) was better predicted than the degree of block deesterification (B). The calibration models for prediction of R obtained on extended inverted signal correction (EISC) processed data gave a root mean square error (RMSE) of cross-validation (CV) of 2%p with 4 PLS components (latent variables, LV) and a correlation coefficient ( r) of 0.98. Spectral variable selection using interval PLS (iPLS) was shown to be valuable, as all the calibration models were improved.

Application of mid infrared spectroscopy and iPLS for the quantification of contaminants in lubricating oil

The feasibility of using a narrow spectral range in the mid infrared region for a rapid and conclusive determination of contaminants (gasoline, ethylene glycol and water) in lubricating oil was investigated. A graphically-oriented local multivariate calibration modeling procedure called interval partial least-squares (iPLS) was applied to select a spectral range that provides the lowest prediction error, in comparison to the full-spectrum model. The region from 650 to 784 cm −1, attributed to aromatics, was selected for gasoline determination. The region from 918 to 1052 cm −1, for C−O stretching from glycol, was selected for ethylene glycol determination. Finally, the region from 3598 to 3732 cm −1, absorption from hydrogen bonded OH groups, was chosen for water determination. In the determinations, a root mean square error of prediction (RMSEP) of 0.34% for gasoline, 0.037% for ethylene glycol and 0.023% for water, were obtained after interval selection.

Application of the local regression method interval partial least-squares to the elucidation of protein secondary structure

The infrared amide bands are sensitive to the conformation of the polypeptide backbone of proteins. Since the backbone of proteins folds in complex spatial arrangements, the amide bands of these proteins result from the superimposition of vibration modes corresponding to the different types of structural motifs (α helices, β sheets, etc.). Initially, band deconvolution techniques were applied to determine the secondary structure of proteins, i.e., the abundance of each structural motif in the polypeptide chain was directly related to the area of the suitable deconvolved vibration modes under the amide I band (1700–1600 cm −1). Recently, several multivariate regression methods have been used to predict the secondary structure of proteins as an alternative to the previous methods. They are based on establishing a relationship between a matrix of infrared protein spectra and another that includes their secondary structure, expressed as the fractions of the different structural motifs, determined from X-ray analysis. In this study, we investigated the use of the local regression method interval partial least-squares (iPLS) to seek improvements to the full-spectrum PLS and other regression methods. The local character of iPLS avoids the use of spectral regions that can introduce noise or that can be irrelevant for prediction and focuses on finding specific spectral ranges related to each secondary structure motif in all the proteins. This study has been applied to a representative protein data set with infrared spectra covering a large wavenumber range, including amides I–III bands (1700–1200 cm −1). iPLS has revealed new structural mode assignments related to less explored amide bands and has offered a satisfactory predictive ability using a small amount of selected specific spectral information.

Sequential application of backward interval partial least squares and genetic algorithms for the selection of relevant spectral regions

AbstractIt is nowadays widely accepted that genetic algorithms (GAs) are powerful tools in variable selection and that after suitable modifications they can also be powerful in detecting the most relevant spectral regions for multivariate calibration. One of the main limitations of GAs is related to the fact that when spectral intensities are measured at a very large number of wavelengths the search domain increases correspondingly and therefore the detection of the relevant regions is much more difficult. A modification of interval partial least squares (iPLS), designated backward interval PLS (biPLS), is developed and studied such that it can detect and remove the least relevant regions, thereby reducing the search domain to a size that GAs can handle easily. In this paper the application to two different spectroscopic data sets will be shown: infrared spectroscopic analysis of polymer film additives and determination of the contents of erucic acid and total fatty acids in brassica seeds by near‐infrared spectroscopy. The developed method is compared with model performances based on expert selection of variables as well as with results from application of the previously developed GA‐PLS method. The sequential application of biPLS and GA‐PLS has proven successful, and comparable or better results have been obtained, introducing a more automatic region selection procedure and a substantial decrease in computation time. Copyright © 2005 John Wiley & Sons, Ltd.

Rapid Spectroscopic Analysis of Marzipan—Comparative Instrumentation

This study had the dual purpose of applying vibrational spectroscopy to measure marzipan composition and of comparing a number of infrared and near infrared (NIR) set-ups and sampling techniques. A set of 32 marzipan samples of nine different recipes were analysed chemically and measured on six different spectrometer set-ups. Calibrations were developed using partial least squares regression (PLSR) to moisture and sugar content. Interval PLS (iPLS) was applied to all the recorded data sets to find optimal spectral intervals (of 100 nm) for the calibrations. The root mean square error of cross-validation ( RMSECV) for the prediction of the moisture was found to be 0.36% w/w moisture (range 7–19%). The sugar content was predicted with a RMSECV of 1.43%w/w(range 33–68%). Despite large variations in spectral range, instrumental design and sampling devices the different NIR spectrometer set-ups yielded calibration results with only small differences. The results are discussed in terms of the robustness of NIR instrumentation for rapid non-destructive food analysis and underline the holistic information provided in the NIR spectral range.

Analysis of Contaminants in Lubricant Oil by near Infrared Spectroscopy and Interval Partial Least-Squares

The feasibility of using near infrared (NIR) transmission spectroscopy for rapid and conclusive determination of contaminants in lubricant oil was investigated. The NIR spectrum in the region from 1300 to 1700 nm was used to predict gasoline and ethylene glycol concentrations present in lubricant oil. A graphically-oriented local multivariate calibration modelling procedure called interval partial least-squares (iPLS) was applied to find variable intervals that featured the lowest prediction error. When compared with the full spectrum PLS model, better results were obtained through the iPLS program. High correlation coefficients and low root mean square errors of cross-validation ( RMSECV) were obtained for gasoline ( R = 0.98, RMSECV = 0.38%, range = 0.2–8.0% w/w) and ethylene glycol determinations ( R = 0.97, RMSECV = 0.04%, range = 0.06 to 0.7% w/w), indicating that the proposed methodology can be used for contaminant determinations in lubricant oil.

Interval Partial Least-Squares Regression (iPLS): A Comparative Chemometric Study with an Example from Near-Infrared Spectroscopy

A new graphically oriented local modeling procedure called interval partial least-squares ( iPLS) is presented for use on spectral data. The iPLS method is compared to full-spectrum partial least-squares and the variable selection methods principal variables (PV), forward stepwise selection (FSS), and recursively weighted regression (RWR). The methods are tested on a near-infrared (NIR) spectral data set recorded on 60 beer samples correlated to original extract concentration. The error of the full-spectrum correlation model between NIR and original extract concentration was reduced by a factor of 4 with the use of iPLS ( r = 0.998, and root mean square error of prediction equal to 0.17% plato), and the graphic output contributed to the interpretation of the chemical system under observation. The other methods tested gave a comparable reduction in the prediction error but suffered from the interpretation advantage of the graphic interface. The intervals chosen by iPLS cover both the variables found by FSS and all possible combinations as well as the variables found by PV and RWR, and iPLS is still able to utilize the first-order advantage.