PLS Calibration Model Research Articles

Partial least squares-near infrared determination of pesticides in commercial formulations

A solvent free, fast and environmentally friendly near infrared-based methodology (NIR) was developed for pesticide determination in commercially available formulations. This methodology was based on the direct measurement of the diffuse reflectance spectra of solid samples and a multivariate calibration model (partial least squares, PLS) to determine the active principle concentration in commercial formulations. The PLS calibration set was built on using the spiked samples by mixing different amounts of pesticide standards and powdered samples. Buprofezin, Diuron and Daminozide were used as test analytes. Concentration of Buprofezin in the samples was calculated employing a 4-factors PLS calibration using the spectral information in the range between 2231–2430 and 1657–1784 nm. For Diuron determination a 1-factor PLS calibration model using the spectral range 1110–2497 nm, after a linear removed correction. Daminozide determination was carried out employing a 4-factors PLS model using the spectral information in the ranges 1644–1772 and 2014–2607 nm without baseline correction. The root mean square errors of prediction (RMSEP) found were 1.1, 1.7 and 0.7% (w/w) for Buprofezin, Diuron and Daminozide determination, respectively. The developed PLS-NIR procedure allows the determination of 120 samples/h, does not require any sample pre-treatment and avoids waste generation.

Potentiometric bioelectronic tongue for the analysis of urea and alkaline ions in clinical samples

Urea biosensors based on urease covalently immobilized on to ammonium and hydrogen ion-selective electrodes were included in arrays together with ammonium, potassium, sodium, hydrogen and generic response to alkaline sensors. Response models based on artificial neural network (ANN) and partial least squares (PLS1) were built, tested and compared for the simultaneous determination of urea, ammonium, potassium and sodium. The results show that it is possible to obtain good ANN and PLS calibration models for simultaneous determination of these four species, but with better prediction capability when the ANN are used. The developed bioelectronic tongue was applied to multidetermination in urine samples. The ANN model showed again better agreement with reference methods, allowing a simple direct determination of urea in the real samples without the necessity of eliminating the alkaline interferences, or compensating endogenous ammonium.

Kernel PLS regression on wavelet transformed NIR spectra for prediction of sugar content of apple

A kernel PLS algorithm was implemented to estimate the sugar content of Golden Delicious apples based on NIR reflectance spectra in the range of 800–1690 nm. Covariance, Gaussian and polynomial kernel functions were considered. All kernels, except the covariance kernel, incorporate tuning parameters which were optimised by computer experiments. The calibration results were insensitive to the actual value of the tuning parameters over a wide range. No significant difference between the RMSEP values obtained with different kernels was obtained, irrespective of the applied transformation (none, log(1/ R), Kubelka–Munck) or first order derivative calculation. A wavelet compression procedure was implemented to speed up the computation of the kernel Gram matrices. It was shown that the kernel Gram matrix computed with the approximation and detail coefficients of the wavelet transformed spectra converges in norm to the real kernel Gram matrix. The time required for calculating the kernel Gram matrix is inversely proportional to the compression ratio. It was shown that a compression ratio of up to 25 did not affect the accuracy of the kernel PLS calibration models.

Use of NIR spectroscopy in the production of modified industrial resins

Natural resins are scarcely used, but after appropriate modification processes they acquire characteristics of viscosity, point of softening, stability, etc. that facilitate their application in fields such as paintings, varnishes, cosmetic, etc. The complexity of resins makes it very difficult to monitor the reactions involved in their modification, the extent of which is usually determined via more experimentally accessible parameters. However, the methods typically used to determine such parameters are slow and produce environmentally unfriendly waste. In this work, we assessed the potential of NIR spectroscopy, as an alternative to the traditional analytical methods, for monitoring the industrial processes involved in the production of modified resins. To this end, we developed PLS calibration models that were used to quantify physical (viscosity and cloud point) and chemical parameters (acid and hydroxyl numbers), with a view to characterize the evolution of the resins during the reaction that take place throughout the fabrication process. Samples were withdrawn at different times stages of the process for analysis with the proposed quantitation models; the data thus obtained were compared with those provided by reference methods. Based on the results, NIR spectroscopy is an effective choice for the accurate, expeditious monitoring of industrial resin modification processes.

Effect of refining on pulp surface charge accessible to polydadmac and FTIR characteristic bands of high yield kraft fibres

Studies were carried out to determine the surface charge and total charge of refined softwood high yield kraft pulps at different levels by titration methods. The results showed that the surface charge was highly dependent on the degree of refining. However, the total fibre charge was not affected by refining. The change of the fibre surface charge during refining monitored using ATR-FTIR characteristic bands was within 1700–1300 cm −1. Reliable PLS1 calibration models could be established to correlate the FTIR spectral data and the surface charge of the refined pulps.

Non-destructive measurement of bitter pit in apple fruit using NIR hyperspectral imaging

A hyperspectral NIR imaging system was developed to identify bitter pit lesions on apples. A discriminant PLS calibration model was constructed to discriminate between pixels of unaffected apple skin and bitter pit lesions. The calibration model was successfully validated on a different apple. The system was able to identify bitter pit lesions, even when not visible to the naked eye such as just after harvest, but could not discriminate between bitter pit lesions and corky tissue. The reduced luminosity at the boundary of the image caused in one image some misclassification errors.

Predicting intramuscular fat, moisture and Warner-Bratzler shear force in pork muscle using near infrared reflectance spectroscopy

AbstractPartial least-squares (PLS) models based on visible (Vis) and near infrared reflectance (NIR) spectroscopy data were explored to predict intramuscular fat (IMF), moisture and Warner Bratzler shear force (WBSF) in pork muscles (m. longissimus thoracis) using two sample presentations, namely intact and homogenized. Samples were scanned using a NIR monochromator instrument (NIRSystems 6500, 400 to 2500 nm). Due to the limited number of samples available, calibration models were developed and evaluated using full cross validation. The PLS calibration models developed using homogenized samples and raw spectra yielded a coefficient of determination in calibration (R2) and standard error of cross validation (SECV) for IMF (R2=0·87; SECV=1·8 g/kg), for moisture (R2=0·90; SECV=1·1 g/kg) and for WBSF (R2=0·38; SECV=9·0 N/cm). Intact muscle presentation gave poorer PLS calibration models for IMF and moisture (R2<0·70), however moderate good correlation was found for WBSF (R2=0·64; SECV=8·5 N/cm). Although few samples were used, the results showed the potential of Vis-NIR to predict moisture and IMF using homogenized pork muscles and WBSF in intact samples.

Estimating the influence of experimental parameters on the prediction error of PLS calibration models based on Raman spectra

AbstractPartial least squares (PLS) calibration is often the method of choice for making multivariate calibration models to predict analyte concentrations from Raman spectral measurements. In the development of such models, it is often difficult to assess beforehand what the prediction error will be, and whether instrumental or model factors limit the lower limit of the prediction error. Here, we present a method to assess the influence of experimental errors such as power fluctuations and spectral shifts, on the PLS prediction errors using simulated datasets. Assumptions that are implicit to PLS calibration and their implications with respect to the choice of experimental parameters for collecting a proper set of Raman spectra are discussed. The influence of various experimental parameters and signal pre‐processing steps on PLS prediction error is demonstrated by means of simulations. The results of simulations are compared with the outcome of PLS calibrations of an experimental dataset. Copyright © 2006 John Wiley & Sons, Ltd.

Optimization of informative spectral regions for the quantification of cholesterol, glucose and urea in control serum solutions using searching combination moving window partial least squares regression method with near infrared spectroscopy

Near infrared (NIR) spectra in the 10 000–4000 cm − 1 region were collected from control serum solutions mixed with cholesterol, glucose and urea solutions with various concentrations. The concentration ranges of cholesterol, glucose and urea were 46.08–395.39, 29.89–318.57 and 5.92–19.66 mg/dl, respectively, which cover the clinically important ranges. Moving window partial least squares regression (MWPLSR) method was applied to the NIR data to select informative regions for cholesterol, glucose and urea. Searching combination moving window partial least squares regression (SCMWPLSR) method was used to search for the optimized combinations of informative regions found by MWPLSR, and partial least squares calibration models were developed and compared for each spectral region proposed by MWPLSR and SCMWPLSR and whole region. The best PLS calibration model for each component was obtained from the spectral region optimized by SCMWPLSR. The best prediction results for cholesterol, glucose and urea have RMSEP of 6.68, 10.35 and 1.28 mg/dl, respectively. Of particular note is that the prediction result for urea is much better than that previously obtained by several research groups under similar conditions. SCMWPLSR can select the optimized combination spectral region for each blood component successfully within the complex blood matrix over the low concentration range.

Monitoring of technical oils in supercritical CO 2 under continuous flow conditions by NIR spectroscopy and multivariate calibration

Metal parts and residues from machining processes are usually polluted with cutting or grinding oil and have to be cleaned before further use. Supercritical carbon dioxide can be used for extraction processes and precision cleaning of metal parts, as developed at Forschungszentrum Karlsruhe. For optimizing and efficiently conducting the extraction process, in-line analysis of oil concentration is desirable. Therefore, a monitoring method using fiber-optic NIR spectroscopy in combination with PLS calibration has been developed. In an earlier paper we have described the instrumental set-up and a calibration model using the model compound squalane in the spectral range of the CH combination bands from 4900 to 4200 cm −1. With this model only poor prediction results were obtained if applied to technical oil samples in supercritical CO 2. In this paper we describe a new calibration model, which was set up for the squalane/carbon dioxide system covering the 323–353 K temperature and the 16–35.6 MPa pressure range. Here, calibration data in the spectral range from 6100 to 5030 cm −1 have been used. This range includes the 5100 cm −1 CO 2 band of the Fermi triad as well as the hydrocarbon 1st overtone C H stretching bands, where spectral features of oil compounds and squalane are more similar to each other. The root mean-squared error of prediction obtained with this model is 4 mg cm −3 for carbon dioxide and 0.4 mg cm −3 for squalane, respectively. The utilizability of the newly developed PLS calibration model for predicting the oil concentration and CO 2 density of solutions of technical oils in supercritical carbon dioxide has been tested. Three types of “real world” cutting and grinding oil formulations were used in these experiments. The calibration proved to be suitable for determining the technical oil concentration with an error of 1.1 mg cm −3 and the CO 2 density with an error of 6 mg cm −3. Therefore, it seems possible to apply this in-line analytical approach on the basis of a cost-effective and time-saving model compound calibration for the surveillance of real world de-oiling and other extraction process based on supercritical carbon dioxide, and furthermore to establish an automated process termination criterion based on this technique.

NON-DESTRUCTIVE ASSESSMENT OF PIGMENTS IN APPLE FRUIT AND CARROT BY LASER-INDUCED FLUORESCENCE SPECTROSCOPY (LIFS) MEASURED AT DIFFERENT TIME-GATE POSITIONS

Laser-induced fluorescence spectroscopy (LIFS) was non-destructively applied to apple fruit and carrot for determining changes in fruit and vegetable pigment contents. These changes were provoked by storing the products at low (0°C for carrots, 4°C for apples) or room (20°C) temperature. The samples were excited by short laser pulses emitting at 337 nm, and recordings of fluorescence spectra were carried out directly on the tissue surface and in fruit extracts in a wavelength range from 350 to 820 nm. Because of the different life-times of the excited state of the molecules of interest, spectra were taken with time-resolved readings to enable/allow separating overlaid fluorescence signals. The polyphenol content of apples was measured using high-performance liquid chromatography.It was found that changes in fluorescence intensity in the blue-green wavelength range can be attributed to variations in fruit polyphenol content. PLS calibration models were built and resulted in high correlation coefficients between LIFS data and polyphenol contents in apples, e.g. R=0.99 for caffeic acid. Furthermore, chlorophyll degradation during storage became visible in the red fluorescence of apple fruit. Furthermore, stored carrots showed increasing fluorescence intensities changes in the blue wavelength area. This might be due to diminishing quenching and reabsorption effects of the decreasing content of carotenoids, which are non-fluorescent molecules. These results show the potential use of LIFS for the rapid and non-destructively assessment of fruit and vegetable pigments such as carotenoids, chlorophylls and polyphenols which are considered as indicators to monitor the product quality along the entire supply chain.

Determination of total nitrogen content, pH, density, refractive index, and brix in Thai fish sauces and their classification by near-infrared spectroscopy with searching combination moving window partial least squares

Near-infrared (NIR) transflectance spectra in the region of 1100-2500 nm were measured for 100 Thai fish sauces. Quantitative analyses of total nitrogen (TN) content, pH, refractive index, density and brix in the Thai fish sauces and their qualitative analyses were carried out by multivariate analyses with the aid of wavelength interval selection method named searching combination moving window partial least squares (SCMWPLS). The optimized informative region for TN selected by SCMWPLS was the region of 2264-2428 nm. A PLS calibration model, which used this region, yielded the lowest root mean square error of prediction (RMSEP) of 0.100% w/v for the PLS factor of 5. This prediction result is significantly better than those obtained by using the whole spectral region or informative regions selected by moving window partial least squares regression (MWPLSR). As for pH, density, refractive index and brix, the 1698-1722, and 2222-2258 nm regions, the 1358-1438 nm region, the 1774-1846, and 2078-2114 nm regions, and the 1322-1442, and 2000-2076 nm regions were selected by SCMWPLS as the optimized regions. The best prediction results were always obtained by use of the optimized regions selected by SCMWPLS. The lowest RMSEP for pH, density, refractive index and brix were 0.170, 0.007 g cm(-3), 0.0079 and 0.435 degrees Brix, respectively. Qualitative models were developed by using four supervised pattern recognitions, linear discriminant analysis (LDA), factor analysis-linear discriminant analysis (FA-LDA), soft independent modeling of class analog (SIMCA), and K neareat neighbors (KNN) for the optimized combination of informative regions of the NIR spectra of fish sauces to classify fish sauces into three groups based on TN. All the developed models can potentially classify the fish sauces with the correct classification rate of more than 82%, and the KNN classified model has the highest correct classification rate (95%). The present study has demonstrated that NIR spectroscopy combined with SCMWPLS is powerful for both the quantitative and qualitative analyses of Thai fish sauces.

On-Line Determination of the Intracellular Poly(β-hydroxybutyric acid) Content in Transformed Escherichia coli and Glucose during PHB Production Using Stopped-Flow Attenuated Total Reflection FT-IR Spectrometry

An automated approach for rapid on-line monitoring of the solid and liquid phases present in bioprocesses based on mid-IR Fourier transform spectrometry is introduced. The principles of this new approach are presented using the example of the quantitative and qualitative analysis of poly(beta-hydroxybutyric acid) (PHB) accumulated in living bacterial cells as well as dissolved glucose during two 27-h fermentation processes. FT-IR spectra were recorded on-line using a diamond-attenuated total reflection (ATR) cell connected to the fermentation broth by means of a computer-controlled flow system. For calibration of the FT-IR method, standard reference analysis procedures for PHB (gas chromatography) and glucose (HPLC) were used. While pumping the fermentation broth through the flow cell, the recorded spectra corresponded to the fermentation solution, which allowed the determination of glucose in a range from 3.8 to -10.3 g/L. Upon stopping the flow, the cells settled on the ATR surface, and the thus recorded spectra enabled the determination of the intracellular PHB content of Escherichia coli in a range from 0.005 to 0.766 g/L. Errors of cross-validation of 0.264 g/L for glucose and 0.037 g/L for PHB were obtained. Application of one PLS calibration model to another fermentation was possible with prediction errors of 0.493 g/L for glucose and 0.105 g/L for PHB. Furthermore, from the position and shape of the PHB carbonyl band, it could be concluded that the PHB granules inside the E. coli are predominantly amorphous.

Analytical control of an esterification batch reaction between glycerine and fatty acids by near-infrared spectroscopy

Near-infrared spectroscopy was used to control an esterification reaction between glycerine and middle- or long-chain fatty acids performed in a laboratory-scale reactor. The process involves the initial formation of monoglycerides, which is followed by that of di- and triglycerides as well as transesterification. Establishing the end point of the process is critical with a view to ensuring that the end product will have the composition required for its intended use. PLS calibration was applied to industrial and laboratory-scale batch samples, and laboratory samples were additionally used to extend calibration ranges and avoid correlation between the concentration of the batch samples. In this way, PLS calibration models for glycerine, fatty acids, water, and mono-, di- and triglycerides, were constructed. The proposed method allows the reaction to be monitored in real time, thereby avoiding long analysis times, excessive reagent consumption and the obtainment of out-of-specification products.

A escolha da faixa espectral no uso combinado de métodos espectroscópicos e quimiométricos

A method is presented for the choice of spectral regions when absorption measurements are coupled to chemometric tools to perform quantitative analyses. The method is based on the spectral distribution of the relative standard deviation of concentration (sc/c). It has been applied to the development of PLS-FTNIR calibration models for the determination of density and MON of gasoline, and ethanol content and density of ethanol fuel. The new method was also compared with the correlation (R2) method and has proved to generate PLS calibration models that present better accuracy and precision than those based on R2.

Influence of data pre-processing on the quantitative determination of the ash content and lipids in roasted coffee by near infrared spectroscopy

In near-infrared (NIR) measurements, some physical features of the sample can be responsible for effects like light scattering, which lead to systematic variations unrelated to the studied responses. These errors can disturb the robustness and reliability of multivariate calibration models. Several mathematical treatments are usually applied to remove systematic noise in data, being the most common derivation, standard normal variate (SNV) and multiplicative scatter correction (MSC). New mathematical treatments, such as orthogonal signal correction (OSC) and direct orthogonal signal correction (DOSC), have been developed to minimize the variability unrelated to the response in spectral data. In this work, these two new pre-processing methods were applied to a set of roasted coffee NIR spectra. A separate calibration model was developed to quantify the ash content and lipids in roasted coffee samples by PLS regression. The results provided by these correction methods were compared to those obtained with the original data and the data corrected by derivation, SNV and MSC. For both responses, OSC and DOSC treatments gave PLS calibration models with improved prediction abilities (4.9 and 3.3% RMSEP with corrected data versus 7.1 and 8.3% RMSEP with original data, respectively).

Detection of Brownheart in ‘Braeburn’ apple by transmission NIR spectroscopy

Abstract ‘Braeburn’ apples containing the internal browning disorder, Brownheart, were examined by transmission NIR spectroscopy to determine whether sample orientation and degree of browning were significant factors requiring consideration in the design of online detection systems. Two hundred and forty fruit were preselected using a clinical MRI system to have a uniform range of the disorder from 0 to 100%. This was later checked by dissecting fruit and calculating the degree of browning in transverse equatorial slices. Four arrangements of light source, fruit-orientation and detector were investigated. Pronounced spectral changes were observed in the range 700–900 nm as the percentage of browned tissue increased. Regression-based Ratio, MLR and PLS calibration models applied to independent prediction sets gave correlations (R2p) of between 0.69 and 0.91 for the different geometries across the full range of affected fruit. Corresponding root mean square error of prediction (RMSEP) statistics of area affected were between 7.9 and 15.4%, and were elevated, in part, because of the asymmetric distribution of browning in individual fruits. PLS models tended to produce lower error and better goodness-of-fit statistics than the simpler two- or three-wavelength Ratio or MLR models. The best individual model was obtained by averaging spectra from opposite sides of the fruit where the stem–calyx axis was horizontal and the light source and detector were located at right-angles to one another at the equator. Our results suggest that single NIR transmission measurements using the models above could lead to a worthwhile reduction in the incidence of Brownheart in commercial lines containing affected fruit. However, to produce an optimal grading system, techniques to reduce the prediction error that overcome the asymmetric distribution of affected flesh, such as averaging multiple views, are still required.

Wavelength interval selection in multicomponent spectral analysis by moving window partial least-squares regression with applications to mid-infrared and near-infrared spectroscopic data.

A new wavelength interval selection procedure, moving window partial least-squares regression (MWPLSR), is proposed for multicomponent spectral analysis. This procedure builds a series of PLS models in a window that moves over the whole spectral region and then locates useful spectral intervals in terms of the least complexity of PLS models reaching a desired error level. Based on a proposed theory demonstrating the necessity of wavelength selection, it is shown that MWPLSR provides a viable approach to eliminate the extra variability generated by non-composition-related factors such as the perturbations in experimental conditions and physical properties of samples. A salient advantage of MWPLSR is that the calibration model is very stable against the interference from non-composition-related factors. Moreover, the selection of spectral intervals in terms of the least model complexity enables the reduction of the size of a calibration sample set in calibration modeling. Two strategies are suggested for coupling the MWPLSR procedure with PLS for multicomponent spectral analysis: One is the inclusion of all selected intervals to develop a PLS calibration model, and the other is the combination of the PLS models built separately in each interval. The combination of multiple PLS models offers a novel potential tool for improving the performance of individual models. The proposed procedures are evaluated using two open-path Fourier transform infrared data sets and one near-infrared data set, each having different noise characteristics. The results reveal that the proposed procedures are very promising for vibrational spectroscopy-based multicomponent analyses and give much better prediction than the full-spectrum PLS modeling.

Orthogonal signal correction in near infrared calibration

The different physical characteristics and treatments of the solid samples are responsible for the spectral variability that takes place in near infrared (NIR) measures. These changes, not related to the analyte concentration, may yield in complex and not very robust calibration models. Mathematical treatments are usually applied for the correction of this variability, being the most common derivation, standard normal variate (SNV) and multiplicative scatter correction (MSC). Orthogonal signal correction (OSC) is a new mathematical treatment designed to minimize, in a set of spectral data, the variability not related with the concentration of the analyte. In this work the application of this new treatment to minimize the spectral differences of two types of samples: production samples and laboratory samples, is evaluated. A method is developed for the determination of the content of the active component in a pharmaceutical preparation by means of PLS calibration. Results obtained by OSC are compared with those obtained with the original data and with those corrected by derivation, SNV and MSC. OSC treatment leads to PLS calibration models with good prediction ability and simpler than those obtained using other pretreatments.

Determination of polymorphic purity by near infrared spectrometry

Differences in physical and chemical properties between polymorphic compounds (e.g. solubility, dissolution rate, chemical reactivity, resistance to degradation, bioavailability) are highly significant for the pharmaceutical industry, which requires effective methods to control the polymorph of interest in the active principle and also to detect and determine the undesirable form. This paper reports on the characterization and determination of a crystalline and an amorphous polymorphic compound of the antibiotic miokamycin by near infrared spectrometry (NIRS). Based on spectral differences between the two, a new analytical method for determining the crystalline form was developed. The underlying study was approached in both qualitative (classification by binary logistic regression) and quantitative terms (quantitation of the crystalline miokamycin content by multivariate calibration). Logistic regression allows one to accurately classify all samples of amorphous miokamycin and also those contaminated with more than 0.3% of the crystalline form. Low contents (<10%) of the crystalline form were determined by testing various PLS1 calibration models involving different spectral modes and ranges, with relative standard errors (RSEs) in the region of 3.5%. The limit of detection (LOD) thus achieved is 0.37% and coincides with the limit of contamination established by logistic regression. The simplicity, expeditiousness and reliability of the proposed method make it an effective tool for controlling polymorphic purity.