Partial Least-squares Calibration Research Articles

Partial least squares–slice transform hybrid model for nonlinear calibration

A hybrid model (herein referred to as PLS-SLT) founded on partial least squares (PLS) and slice transform (SLT) is proposed to model nonlinear chemical systems with a wide range of response variable. In the modeling process, PLS predicted values of calibration set were taken as inputs for the subsequent SLT to further approximate to observed values by a least square criterion. The estimated optimal piecewise linear mapping function was then applied to test set to give the final prediction result. Theoretically, PLS-SLT can be proven to be equivalent to the PLS-based piecewise linear model in the y-space. PLS-SLT is compared with PLS and other calibration models on two spectral datasets. The Wilcoxon signed rank test is used to statistically compare predictive performance of two competing calibration models. Experimental results show that the performance of PLS-SLT is at least statistically not worse than PLS and other models.

Estimation of genetic and crossbreeding parameters of fatty acid concentrations in milk fat predicted by mid-infrared spectroscopy in New Zealand dairy cattle.

The objective of this study was to estimate heritability and crossbreeding parameters (breed and heterosis effects) of various fatty acid (FA) concentrations in milk fat of New Zealand dairy cattle. For this purpose, calibration equations to predict concentration of each of the most common FAs were derived with partial least squares (PLS) using mid-infrared (MIR) spectral data from milk samples (n=850) collected in the 2003-04 season from 348 second-parity crossbred cows during peak, mid and late lactation. The milk samples produced both, MIR spectral data and concentration of the most common FAs determined using gas chromatography (GC). The concordance correlation coefficients (CCC) between the concentration of a FA determined by GC and the PLS equation ranged from 0.63 to 0.94, suggesting that some prediction equations can be considered to have substantial predictive ability. The PLS calibration equations were then used to predict the concentration of each of the fatty acids in 26,769 milk samples from 7385 cows that were herd-tested during the 2007-08 season. Data were analysed using a single-trait repeatability animal model. Shorter chain FA (16:0 and below) were significantly higher (P<0.05) in Jersey cows, while longer chain, including unsaturated longer chain FA were higher in Holstein-Friesian cows. The estimates of heritabilities ranged from 0.17 to 0.41 suggesting that selective breeding could be used to ensure milk fat composition stays aligned to consumer, market and manufacturing needs.

IUPAC-consistent approach to the limit of detection in partial least-squares calibration.

There is currently no well-defined procedure for providing the limit of detection (LOD) in multivariate calibration. Defining an estimator for the LOD in this scenario has shown to be more complex than intuitively extending the traditional univariate definition. For these reasons, although many attempts have been made to arrive at a reasonable convention, additional effort is required to achieve full agreement between the univariate and multivariate LOD definitions. In this work, a novel approach is presented to estimate the LOD in partial least-squares (PLS) calibration. Instead of a single LOD value, an interval of LODs is provided, which depends on the variation of the background composition in the calibration space. This is in contrast with previously proposed univariate extensions of the LOD concept. With the present definition, the LOD interval becomes a parameter characterizing the overall PLS calibration model, and not each test sample in particular, as has been proposed in the past. The new approach takes into account IUPAC official recommendations, and also the latest developments in error-in-variables theory for PLS calibration. Both simulated and real analytical systems have been studied for illustrating the properties of the new LOD concept.

Potential of Near-Infrared Spectroscopy for Measurement of Heavy Metals in Soil as Affected by Calibration Set Size

We investigate the potential of near-infrared (NIR) spectroscopy to predict some heavy metals content (Zn, Cu, Pb, Cr and Ni) in several soil types in Stara Zagora Region, South Bulgaria, as affected by the size of calibration set using partial least squares (PLS) regression models. A total of 124 soil samples from the 0–20 and 20–40 cm layers were collected from fields with different cropping systems. Total Zn, Cu, Pb, Cr and Ni concentrations were determined by Atomic Absorption Spectrometry. Spectra of air dried soil samples were obtained using an FT-NIR Spectrometer (spectral range 700–2,500 nm). PLS calibration models were developed with full-cross-validation using calibration sets of 90 %, 80 %, 70 % and 60 % of the 124 samples. These models were validated with the same prediction set of 12 samples. The validation of the NIR models showed Cu to be best predicted with NIR spectroscopy. Less accurate prediction was observed for Zn, Pb and Ni, which was classified as possible to distinguish between high and low concentrations and as approximate quantitative. The worst model performance in cross-validation and prediction was for Cr. Results also showed that values of root mean square error in cross-validation (RMSEcv) increased with decreasing number of samples in calibration sets, which was particularly clear for Cu, Pb, Ni and Cr content. A similar tendency was observed in the prediction sets, where RMSEP values increased with a decrease in the number of samples, particularly for Pb, Ni and Cr content. This tendency was not clear for Zn, while even an increase in RMSEP for Cu with the sample size was observed. It can be concluded that NIR spectroscopy can be used to measure heavy metals in a sample set with different soil type, when sufficient number of soil samples (depending on variability) is used in the calibration set.

Application of near-infrared spectroscopy for the rapid analysis ofLonicerae Japonicae Flossolution extracted by water

A rapid quantitative analytical method for three components of Lonicerae Japonicae Flos solution (Lonicera Japonica Thumb.) extracted by water was developed using near-infrared (NIR) spectroscopy and the partial least-squares (PLS) method. The NIR spectra of 81 samples collected from a production line were obtained. The concentrations of secologanic acid, chlorogenic acid and galuteolin were determined by using high-performance liquid chromatography-diode array detection as the reference method. Several pretreatment methods for the NIR spectra were used during PLS calibration. The most appropriate latent variable number of the PLS factor was selected based on the standard error of cross-validation (SECV). The performance of the final PLS models was evaluated according to SECV, standard error of prediction (SEP) and determination coefficient (R2). The compounds secologanic acid, chlorogenic acid and galuteolin had SEP values of 0.030, 0.061 and 1.668 μg/mL, respectively and R2values over 0.85. This work shows that NIR spectroscopy is a rapid and convenient method for the analysis of Lonicerae Japonicae Flos solution extracted by water. The proposed method can help in the application of process analytical technology in the pharmaceutical industry, particularly in traditional Chinese medicine injections.

The use of FTIR spectroscopy in combination with chemometrics for the authentication of red fruit (Pandanus conoideus Lam) oil from sunflower and palm oils

Abstract In this study, Fourier transform infrared (FTIR) spectroscopy in combination with multivariate calibration based on partial least squares (PLS) has been developed for quantitative analysis of sunflower oil (SFO) and palm oil (PO) as oil adulterants in red fruit oil (RFO) for the authentication study. The binary mixtures of SFO and PO in RFO were infrared spectroscopically scanned using horizontal attenuated total reflectance at mid infrared region (4000–650 cm −1 ). Some frequency regions where the variations were observed among RFO, SFO, and PO were optimized in order to seek the best frequency regions giving the best PLS calibration models. Besides, the derivatization of FTIR spectra (fisrt and second derivatives) was also investigated in order to look the types of FTIR spectra giving the best spectral performance in the model. The results showed that SFO was better determined using FTIR normal spectra at the frequency region of 1200–1000 cm −1 with coefficient of determination ( R 2 ) of >0.99 and with root mean square error of calibration (RMSEC) of 1.42% v/v. FTIR normal spectra were also more preferred than its derivative for the quantification of PO in RFO. The R 2 value obtained for the relationship between actual value (reference value) and predicted value obtained during PLS modeling is very high (0.9991) with the low level of error (RMSEC of 0.98% v/v). It can be concluded that FTIR in combination with PLS is a reliable method for quantitative analysis of SFO and PO as adulterant in RFO for authenticity verification.

FTIR Spectroscopy Combined with Partial Least Square for Analysis of Red Fruit Oil in Ternary Mixture System

FTIR spectroscopy is a promising method for quantification of edible oils. Three edible oils, namely, red fruit oil (RFO), corn oil (CO), and soybean oil (SO), in ternary mixture system were quantitatively analyzed using FTIR spectroscopy in combination with partial least square (PLS). FTIR spectra of edible oils in ternary mixture were subjected to several treatments including normal spectra and their derivative. Using PLS calibration, the first derivative FTIR spectra can be exploited for determination of RFO; meanwhile, the second derivative spectra were preferred for determination of CO and SO. The R2 values obtained for the relationship between actual and FTIR predicted values of RFO, CO, and SO were 0.9863, 0.9276, and 0.9693, respectively. The root mean square error of calibration (RMSEC) values obtained were 1.59, 1.72, and 1.60% (v/v) for RFO, CO, and SO, respectively. The result showed that FTIR spectroscopy offers accurate and reliable technique for quantitative analysis of RFO, CO, and SO simultaneously in ternary mixture systems. Besides, the developed method can be extended for analysis of CO and SO as adulterants in RFO.

Predicting polycyclic aromatic hydrocarbon concentrations in resident aquatic organisms using passive samplers and partial least-squares calibration.

The current work sought to develop predictive models between time-weighted average polycyclic aromatic hydrocarbon (PAH) concentrations in the freely dissolved phase and those present in resident aquatic organisms. We deployed semipermeable membrane passive sampling devices (SPMDs) and collected resident crayfish (Pacifastacus leniusculus) at nine locations within and outside of the Portland Harbor Superfund Mega-site in Portland, OR. Study results show that crayfish and aqueous phase samples collected within the Mega-site had PAH profiles enriched in high molecular weight PAHs and that freely dissolved PAH profiles tended to be more populated by low molecular weight PAHs compared to crayfish tissues. Results also show that of several modeling approaches, a two-factor partial least-squares (PLS) calibration model using detection limit substitution provided the best predictive power for estimating PAH concentrations in crayfish, where the model explained ≥72% of the variation in the data set and provided predictions within ∼3× of measured values. Importantly, PLS calibration provided a means to estimate PAH concentrations in tissues when concentrations were below detection in the freely dissolved phase. The impact of measurements below detection limits is discussed.

Mid-infrared spectroscopy and partial least-squares regression to estimate soil arsenic at a highly variable arsenic-contaminated site

The potential of mid-infrared spectroscopy in combination with partial least-squares regression was investigated to estimate total and phosphate-extractable arsenic contents in soil samples collected from a highly variable arsenic-contaminated disused cattle-dip site. Principal component analysis was performed prior to mid-infrared partial least-squares analysis to identify spectral outliers in the absorbance spectra of soil samples. The mid-infrared partial least-squares calibration model (n = 149) excluding spectral outliers showed an acceptable reliability (coefficient of determination, $$R_{\text{c}}^{2}$$ = 0.75 (P < 0.01); ratio of performance to interquartile distance, RPIQc = 2.20) to estimate total soil arsenic. For total soil arsenic, the validation of final calibration model using 149 unknown samples also resulted in a good acceptability with $$R_{\text{v}}^{2}$$ = 0.67 (P < 0.05) and RPIQv = 2.01. However, the mid-infrared partial least-squares calibration model based on phosphate-extractable arsenic was not acceptable to estimate the extractable (bioavailable) arsenic content in soil ( $$R_{\text{c}}^{2}$$ = 0.13 (P > 0.05); RPIQc = 1.37; n = 149). The results show that the mid-infrared partial least-squares prediction model based on total arsenic can provide a rapid estimate of soil arsenic content by taking into account the integrated effects of adsorbed arsenic, arsenic-bearing minerals and arsenic associated with organic components in the soils. This approach can be useful to estimate total soil arsenic in situations, where analysis of a large number of samples is required for a single soil type and/or to monitor changes in soil arsenic content following (phyto)remediation at a particular site.

Determination of the iodine value and the free fatty acid content of waste animal fat blends using FT-NIR

This study determined iodine value (IV) and free fatty acids (FFA) content of four different animal fat wastes and their blends using Fourier transform near-infrared spectroscopy (FT-NIR). Chemometric analysis by partial least squares (PLS) regression was used to correlate spectral data with IV and FFA reference values of the samples. The effects of four spectra pre-processing (first derivative (FD), second derivative (SD), multiplicative scatter correction (MSC) and vector normalization (VN)) methods were investigated to predict the reproducibility and robustness of the PLS-NIR model developed. A set of 70% of animal fat wastes and their blends were used for developing PLS calibration models for measuring IV and FFA content using the remaining 30% samples as an independent test set validation. The coefficient of determination (R2), the root mean square error estimation (RMSEE), and the residual prediction deviation (RPD) were used as indicators for the predictability of the PLS models. PLS-NIR models developed using first derivative and second derivative spectral preprocessing methods were the best for both IV and FFA content analysis (For IV, FD; R2=0.9870, RMSEE=1.40gI2/100g, RPD=8.76, SD; R2=0.9892, RMSEE=1.28gI2/100g, RPD=9.64 while For FFA, FD; R2=0.9991, RMSEE=0.195%, RPD=34.00, SD; R2=0.9993, RMSEE=0.182%, RPD=36.8). Overall, the results of this study demonstrate the suitability of FT-NIR spectroscopy for the quality control analysis of feedstocks for biodiesel production.

Simultaneous determination of the impurity and radial tensile strength of reduced glutathione tablets by a high selective NIR-PLS method

This paper establishes a high-throughput and high selective method to determine the impurity named oxidized glutathione (GSSG) and radial tensile strength (RTS) of reduced glutathione (GSH) tablets based on near infrared (NIR) spectroscopy and partial least squares (PLS). In order to build and evaluate the calibration models, the NIR diffuse reflectance spectra (DRS) and transmittance spectra (TS) for 330 GSH tablets were accurately measured by using the optimized parameter values. For analyzing GSSG or RTS of GSH tablets, the NIR-DRS or NIR-TS were selected, subdivided reasonably into calibration and prediction sets, and processed appropriately with chemometric techniques. After selecting spectral sub-ranges and neglecting spectrum outliers, the PLS calibration models were built and the factor numbers were optimized. Then, the PLS models were evaluated by the root mean square errors of calibration (RMSEC), cross-validation (RMSECV) and prediction (RMSEP), and by the correlation coefficients of calibration (Rc) and prediction (Rp). The results indicate that the proposed models have good performances. It is thus clear that the NIR-PLS can simultaneously, selectively, nondestructively and rapidly analyze the GSSG and RTS of GSH tablets, although the contents of GSSG impurity were quite low while those of GSH active pharmaceutical ingredient (API) quite high. This strategy can be an important complement to the common NIR methods used in the on-line analysis of API in pharmaceutical preparations. And this work expands the NIR applications in the high-throughput and extraordinarily selective analysis.

Exploring computer simulation to assess counseling skills amongst pharmacy undergraduates

Abstract: This study illustrated the genetic algorithms (GA) is useful tool in wavelength selection for the simultaneous spectrophotometric determination of paracetamol (PAR) and chlorzoxazone (CZX) in a multicomponent pharmaceutical dosage form by partial least-squares calibration procedure. Artificial, a training set of the mixtures of PAR and CZX corresponding to central composite design was built to be used as calibration set for PLS and GA-PLS. Synthetic mixtures were used to verify the models. Reasonably good recoveries were obtained and it leads to 32% improvement of predictive ability for PAR and 35% for CZX, respectively. A commercial sample containing two drugs was assayed, and there was statistically no significant difference (P < 0.05) between the results with GA-PLS method and those obtain with the official method. The proposed method has been proved to be beneficial for the determination of both drugs in pharmaceutical tablets without prior extraction. Keywords : Genetic algorithms, Partial least-squares, Paracetamol, Chlorzoxazone

Performance evaluation of preprocessing techniques utilizing expert information in multivariate calibration

Partial Least Squares (PLS) regression is one of the most used methods for extracting chemical information from Near Infrared (NIR) spectroscopic measurements. The success of a PLS calibration relies largely on the representativeness of the calibration data set. This is not trivial, because not only the expected variation in the analyte of interest, but also the variation of other contributing factors (interferents) should be included in the calibration data. This also implies that changes in interferent concentrations not covered in the calibration step can deteriorate the prediction ability of the calibration model. Several researchers have suggested that PLS models can be robustified against changes in the interferent structure by incorporating expert knowledge in the preprocessing step with the aim to efficiently filter out the spectral influence of the spectral interferents. However, these methods have not yet been compared against each other. Therefore, in the present study, various preprocessing techniques exploiting expert knowledge were compared on two experimental data sets. In both data sets, the calibration and test set were designed to have a different interferent concentration range. The performance of these techniques was compared to that of preprocessing techniques which do not use any expert knowledge. Using expert knowledge was found to improve the prediction performance for both data sets. For data set-1, the prediction error improved nearly 32% when pure component spectra of the analyte and the interferents were used in the Extended Multiplicative Signal Correction framework. Similarly, for data set-2, nearly 63% improvement in the prediction error was observed when the interferent information was utilized in Spectral Interferent Subtraction preprocessing.

SB-ATR FTIR Spectroscopic Monitoring of Free Fatty Acids in Commercially AvailableNigella sativa(Kalonji) Oil

Free fatty acids (FFA) inNigella sativa(N. sativa) commercial and seed oil were determined using single-bounce attenuated total reflectance (SB-ATR) Fourier transform infrared (FTIR) spectroscopy. Gravimetrical mixing was done by adding 0.1–40% oleic acids in neutralizedN. sativaoil containing 0.1% FFA. FTIR spectroscopy technique and partial least square (PLS) calibration were used to detect the absorption region of carbonyl (C=O) which is in the range of 1690–1727 cm−1. The results of PLS calibration model and root mean square error of calibration (RMSEC) are 0.999 and 0.449, respectively. Comparing the FFA obtained inN. sativaoil by using FTIR with the FFA obtained using AOCS titrimetric method shows a positive correlation and confirms that the described method is a useful procedure.

Evaluation of Fourier Transform-Near Infrared Spectroscopic Measurements for the Quantification of Curcumin in Turmeric Herbal Medicines

In this investigation, Fourier transform-near infrared (FT-NIR) spectroscopy was used to determine the curcumin content in commercial turmeric herbal medicines. Two different sample presentations originating from a turmeric capsule and powder were evaluated and compared using an FT-NIR spectrometer equipped with a 30-capsule sample plate in the wavenumber region of 11,500–6000 cm−1 (transmittance mode) and a 6-vial sample plate in the wavenumber region of 10,000–4000cm−1 (reflectance mode), respectively. One hundred and forty samples of turmeric herbal medicines were divided into two sets: a training set with 110 samples and a prediction set with 30 samples. High-quality FT-NIR spectra for the turmeric powder containing the informative regions of curcumin (6850, 6000 and 4600 cm−1) were obtained while collecting FT-NIR data on the powder samples in the glass vials. A successful partial least-squares calibration model for curcumin content in commercial turmeric herbal medicine powders was then obtained by utilising the entire spectral region (10,000–4000 cm−1) as well as a second derivative spectral pretreatment containing seven principal components. The best model for predicting curcumin in 30 test samples returned a coefficient of determination ( r2) of 0.98, a root mean squares error of prediction ( RMSEP) of 0.13% that was only 1.35 times the standard error of laboratory and a ratio of RMSEP to the standard deviation of curcumin content in the prediction set ( RPD) of 7.76. These statistical values confirmed the high predictive performance of this model. All these results confirm that the use of turmeric herbal medicine powder and a wide informative wavenumber range of 10,000–4000 cm−1 can increase the efficacy of FT-NIR spectroscopy for the rapid and non-destructive analysis of curcumin content in commercial turmeric herbal medicines. By contrast, models developed for determining curcumin in turmeric contained in capsules were not suitable for analytical purposes.

Comparison of partial least squares calibration models of viscosity, acid number and asphaltene content in petroleum, based on GC and IR data

Partial Least Squares (PLS) calibration models for viscosity, asphaltene content and Total Acid Number (TAN) based on Gas Chromatography (GC) and Infra Red (IR) spectroscopy have been compared based on predictive quality, and also based on chemical significance. Crude oils and condensates, 20 in total, originating mainly from the North Sea, have been analysed and PLS calibration models have been built. The models based on IR data gives somewhat better predictions. Biodegradation is an important factor for the variation in TAN values, and there are indications that biodegradation has some effect on the variation of asphaltene content as well.

Analysis of chicken fat as adulterant in butter using fourier transform infrared spectroscopy and chemometrics

Butter may be adulterated with cheaper animal fats, such as chicken fat (CF). Thus, the detection and quantification of butter adulteration with CF was monitored using Fourier transform infrared (FTIR) spectroscopy, combined with chemometric of partial least square (PLS) at the frequency regions of 1200-1000cm–1. FTIR measurements were made on pure butter and that adulterated with varying concentrations of CF (0-100% w/w in butter). PLS calibration exhibits a good relationship between actual and FTIR predicted values of CF with a coefficient of determination (R2) of 0.981. The root means standard error of calibration (RMSEC) and during cross validation (RMSECV) obtained using six principal components (PCs) are 2.08 and 4.33% v/v, respectively.

An integrated approach to the simultaneous selection of variables, mathematical pre-processing and calibration samples in partial least-squares multivariate calibration

A new optimization strategy for multivariate partial-least-squares (PLS) regression analysis is described. It was achieved by integrating three efficient strategies to improve PLS calibration models: (1) variable selection based on ant colony optimization, (2) mathematical pre-processing selection by a genetic algorithm, and (3) sample selection through a distance-based procedure. Outlier detection has also been included as part of the model optimization. All the above procedures have been combined into a single algorithm, whose aim is to find the best PLS calibration model within a Monte Carlo-type philosophy. Simulated and experimental examples are employed to illustrate the success of the proposed approach.

QUANTITATIVE EVALUATION OF MULTIVARIATE ANALYSIS METHODS FOR EXCITATION–EMISSION SPECTROSCOPY

Multivariable Analysis Methods have been used widely in Spectroscopy Analysis. Partial least square (PLS) and principal component analysis (PCA) are the two most popular methods due to their excellent ability in data components analysis and results prediction. This work derived 1D/2D PLS and 1D/2D PCA based on the viewpoint of three-layer artificial neural networks, and uses theoretical proving to figure out the essences of these two methods. Two 2D experimental dataset was used to verify the calibration and prediction ability, furthermore, the similarity and dissimilarity of PLS and PCA. The finding showed that both the 1D/2D PLS and 1D/2D PCA methods use maximum covariance and minimum sum of square error to figure out the relationship between independent and dependent variables. The dissimilarity is that, weight vectors calibration of PLS is cross-correlation, and that of PCA is autocorrelation. The difference causes that the PCA method would keep more principal characters than PLS under insufficient sample among and provides better calibration ability. PLS would provide higher performance in prediction under sufficient sample among. For the needs of system implementation of spectroscopic measurement and analysis, this study designed "Multivariate Analysis Toolkits for LabVIEW" for the convenient implementation in automatic measurement system integration.

Direct Determination of Danofloxacin and Flumequine in Milk by Use of Fluorescence Spectrometry in Combination with Partial Least-Squares Calibration

A new method for the simultaneous determination of danofloxacin and flumequine in milk samples was developed by using the nonlinear variable-angle synchronous fluorescence technique to acquire data and a partial least-squares chemometric algorithm to process them. A calibration set of standard samples was designed by combination of a factorial design with two levels per factor and a central star design. Whey was used as the third component of the calibration matrix. In order to assess the goodness of the proposed method, a prediction set of 11 synthetic samples was analyzed, obtaining recovery percentages between 96.1% and 104.0%. Limits of detection, calculated by means of a new criterion, were 0.90 and 12.4 ng mL(-1) for danofloxacin and flumequine, respectively. Finally, the simultaneous determination of both fluoroquinoles in milk samples containing the analytes was successfully carried out, obtaining an average recovery percentage of 99.3 ± 4.4 for danofloxacin and 100.7 ± 4.4.