Predictive Partial Least Squares Research Articles

Olfactory Evaluation of Geisha Coffee from Panama Using Electronic Nose

Geisha arabica coffee from Panama is featured in the world specialty coffee market. Its quality is assessed by sensory analysis with a panel of experts over several days. However, there is a risk of commercial fraud where cheaper coffees are mixed with pure specialty coffees. For these reasons, having an electronic nose (EN) device proves advantageous in supporting the cupping panel. It allows a greater number of fragrance and aroma analyses to be conducted per day, providing more objective results. In this study, an experimental EN equipped with a metal oxide semiconductor (MOS) gas sensor array was used. Olfactory evaluation of ground and infused Geisha coffee of different roast levels, brewing process, and purity was studied by EN, sensory analysis and chromatography. The sensory analysis perceived significant differences in fragrance and aromas in the light and dark roast levels of the samples. A total of 57 volatile organic compounds (VOC) were studied by gas chromatography. The EN data were analyzed chemometrically with principal component analysis (PCA) and predictive partial least squares (PLS). The data variances for two components were explained with values greater than 82%. The EN demonstrated its ability to differentiate the three levels of roasting, two production processes, and adulteration in the analyzed samples.

Prediction of quality traits in dry pepper powder using visible and near-infrared spectroscopy

Fruit quality phenotyping is a bottleneck in plant breeding. The present work aimed to investigate the applicability of visible (Vis) and near-infrared (NIR) spectroscopy for quality evaluation in dry red chili powder. We constructed prediction models for the American Spice Trade Association (ASTA)-colour and the Scoville Heat Unit (SHU)-pungency pepper traits using spectroscopy and multivariate statistical techniques. Predictive partial least squares (PLS) models were successfully achieved with high correlations (r) between the predicted and reference values for calibration and validation (r = 0.955 and 0.928 for ASTA-colour; r = 0.941 and 0.918 for SHU-pungency). Spectroscopy data from visible and short-wave radiation (Vis-SWNIR) provided the most robust (residual predictive deviation value) model for ASTA-colour (RPD = 2.84) and long-wave radiation (LWNIR) for SHU-pungency (RPD = 2.48). Spectral categories for wavelength range selection, variable importance for effective wavelength selection, and root mean press-statistic for factor selection were important criteria for PLS. Trait variance and distribution were also important criteria for the predictive capacity and power of the models. In conclusion, non-invasive spectroscopy was a promising tool in our study for dry red chili quality phenotyping.

Non-Invasive Method to Predict the Composition of Requeijão Cremoso Directly in Commercial Packages Using Time Domain NMR Relaxometry and Chemometrics

Low Field Time-Domain Nuclear Magnetic Resonance (TD-NMR) relaxometry was used to determine moisture, fat, and defatted dry matter contents in “requeijão cremoso” (RC) processed cheese directly in commercial packaged (plastic cups or tubes with approximately 200 g). Forty-five samples of commercial RC types (traditional, light, lactose-free, vegan, and fiber) were analyzed using longitudinal (T1) and transverse (T2) relaxation measurements in a wide bore Halbach magnet (0.23 T) with a 100 mm probe. The T1 and T2 analyses were performed using CWFP-T1 (Continuous Wave Free Precession) and CPMG (Carr-Purcell-Meiboom-Gill) single shot pulses. The scores of the principal component analysis (PCA) of CWFP-T1 and CPMG signals did not show clustering related to the RC types. Optimization by variable selection was carried out with ordered predictors selection (OPS), providing simpler and predictive partial least squares (PLS) calibration models. The best results were obtained with CWFP-T1 data, with root-mean-square errors of prediction (RMSEP) of 1.38, 4.71, 3.28, and 3.00% for defatted dry mass, fat in the dry and wet matter, and moisture, respectively. Therefore, CWFP-T1 data modeled with chemometrics can be a fast method to monitor the quality of RC directly in commercial packages.

FTIR-PAS coupled to partial least squares for prediction of ash content, volatile matter, fixed carbon and calorific value of coal

Fourier Transform Infrared Photoacoustic Spectroscopy (FT-IR PAS) and multivariate calibration using partial least squares (PLS) has been coupled for the development of predictive models of ash content, volatile matter, fixed carbon and calorific value of coal samples from several mines of Colombia. In the chemometric approach, a novel alternative is proposed to define the complexity and the selection of the models. According to tests and statistical criteria, at a 95% confidence level, the selected models demonstrate their linear character and a true correlation between the reference values and the predicted values. Results provide standard error of cross-validation (SECV) of 1.22 wt%, 0.78 wt%, 1.08 wt% and 0.75 MJ kg−1 predicting ash, volatile matter (VM), fixed carbon (FC) and calorific value, respectively.

Adaptive prediction model for fluidized catalytic cracking processes based on the PLS method

AbstractBased on the partial least squares (PLS) method, an adaptive predictive PLS (AP‐PLS) method was developed for sensitive adaptation to process changes using large‐scale data (Big Data) from chemical processes. Utilizing data sets of fluidized catalytic cracking (FCC) and residue FCC (RFCC) processes as the basis, the AP‐PLS method was developed and its prediction ability was compared with the simple PLS method. The required parameters for the prediction model of the FCC and RFCC processes are readily available from reported process data using time‐varying model updating by the AP‐PLS method. The prediction results were compared with the simple PLS method in terms of average deviations from the real process data. This approach can be used for reasonably accurate prediction of product variables and can adapt to process changes in large‐scale chemical processes such as FCC and RFCC.

PLS regression-based chemometric modeling of odorant properties of diverse chemical constituents of black tea and coffee.

Tea and coffee are the most attractive non-alcoholic beverages used worldwide due to the odorant properties of diverse components present in these beverages. The aim of this work is to investigate the key structural features which regulate the odorant properties of constituents present in black tea and coffee using regression-based chemometric models. We have also investigated the key structural properties which create the odor difference between tea and coffee. We have employed different variable selection strategies to extract the most relevant variables prior to development of final partial least squares (PLS) models. The models were extensively validated using different validation metrics, and the results justify the reliability and usefulness of the developed predictive PLS models. The best PLS model captured the necessary structural information on relative hydrophobic surface area, heteroatoms with higher number of multiple bonds, hydrogen atoms connected to C3(sp3)/C2(sp2)/C3(sp2)/C3(sp) fragments, electron-richness, C–O atom pairs at the topological distance 10 and surface weighted charged partial negative surface areas for explaining the odorant properties of the constituents present in black tea. On the other hand, C–S atom pairs at the topological distance 1, C–C atom pairs at the topological distance 5, donor atoms like N and O for hydrogen bonds, hydrogen atoms connected to C3(sp3)/C2(sp2)/C3(sp2)/C3(sp) fragments and R–CX–X fragments (where, R represents any group linked through carbon and X represents any heteroatom (O, N, S, P, Se, and halogens)) are the key structural components captured by the PLS model developed from the constituents present in coffee. The developed models can thus be successfully utilized for in silico prediction of odorant properties of diverse classes of compounds and exploration of the structural information which creates the odor difference between black tea and coffee.

In‐line monitoring of hydrogen peroxide in two‐phase reactions using raman spectroscopy

Hydrogen peroxide is an environment‐friendly oxidizer, which is used in several chemical processes. However, safety necessitates the determination and control of the concentration of hydrogen peroxide during oxidation reactions. We propose a methodology to monitor hydrogen peroxide in disperse two‐phase reaction mixtures based on in‐line Raman spectroscopy. We compare indirect hard modeling (IHM), peak integration (PI), and partial least squares (PLS). Building predictive PLS and PI calibration models is challenging, whereas the IHM calibration is easy to develop. These methods show good accuracy for known samples (root mean square error of cross validation [RMSECV] of 0.3–0.7 wt %) compared to the classic titration method (RMSECV of 0.4 wt %). After calibration, inline monitoring during reaction is performed demonstrating that the concentration of hydrogen peroxide can be successfully monitored in a fast and reliable way by Raman spectroscopy. The IHM seems to give slightly better inline predictions. © 2017 American Institute of Chemical Engineers AIChE J, 63: 3994–4002, 2017

Monitoring and Quantification of Omeprazole Synthesis Reaction by In-Line Raman Spectroscopy and Characterization of the Reaction Components

The development of a quantitative in-line Raman spectroscopic method for the monitoring of the active pharmaceutical ingredient, omeprazole synthesis reaction, and characterization of the reaction components is described. In-line monitoring was performed both with Fourier transform and dispersive Raman spectrometers. Prior to reaction monitoring, the reaction components were characterized off-line by means of Raman and NMR spectroscopy, both in solution and in solid state. To unequivocally confirm the presence of each component in the reaction mixture, a state of the art LC-SPE/NMR methodology was also used. Owing to its higher sensitivity, dispersive Raman spectroscopy was further employed for quantification purposes. The spectroscopic measurements and the complementary HPLC analyses, used in the calibration development, were gathered from a set of experiments, performed at a 1 L scale. On the basis of the data set obtained from the calibration experiments, a predictive partial least-squares (PLS) regres...

The “double cross-validation” software tool for MLR QSAR model development

Quantitative structure–activity relationship (QSAR) modeling is a widely used computational technique applied in various fields including rational drug design, toxicity and property prediction of chemicals and pharmaceuticals, environmental risk assessment and fate modeling. External validation is generally considered as the gold standard in evaluating the model predictivity performance, at least to a group of QSAR practitioners. External validation is commonly performed by employing the hold-out method, where the original dataset is divided into training and test sets; the training set is employed for model building and model selection, while the test set is solely used in model assessment. However, since the composition of the training set remains the same in this method, it is not certain that the resultant model is optimal as there may be a bias in descriptor selection. This problem is more evident for the multiple linear regression (MLR) models than more robust and generalized partial least squares (PLS) and principal component regression (PCR) models. Thus, employing double cross-validation technique could be a better choice, in which the training set is further divided into ‘n’ calibration and validation sets resulting in diverse compositions. In the present work, we have developed an open access “Double Cross-Validation (DCV)” software tool which can be used to perform multiple linear regression (MLR) model development by employing the double cross-validation technique. Two variable selection methods, namely, stepwise MLR (S-MLR) and genetic algorithm MLR (GA-MLR) are incorporated in this tool and optionally, this tool also performs a data-pretreatment prior to the application of double cross-validation. Also, we have performed a study using the “Double Cross-Validation” tool on three different datasets in order to find out which technique among the hold-out and double cross-validation performs better in the selection of an optimal model in terms of model predictive performance checked on the test set. The performance of the tool in generating predictive PLS models is also compared. The “Double Cross-Validation” (version 2.0) tool is freely available to download from the sites http://teqip.jdvu.ac.in/QSAR_Tools/ and http://dtclab.webs.com/software-tools.

The elephant in the room: Predictive performance of PLS models

Attempts to introduce predictive performance metrics into partial least squares (PLS) path modeling have been slow and fall short of demonstrating impact on either practice or scientific development in PLS. This study contributes to PLS development by offering a comprehensive framework that identifies different dimensions of prediction and their effect on predictive performance evaluation with PLS. This framework contextualizes prior efforts in PLS and prediction and highlights potential opportunities and challenges. A second contribution to PLS development lies in proposed procedures to generate and evaluate different types of predictions from PLS models. These procedures account for the best practices that the new framework identifies. An outline of the many powerful ways in which predictive PLS methodologies can strengthen theory-building research constitutes a third contribution to PLS development. The framework, procedures, and research guidelines hopefully form the basis for a more informed and unified development of the rigorous theoretical and practical applications of PLS.

The Elephant in the Room: Evaluating the Predictive Performance of Partial Least Squares (PLS) Path Models

Attempts to introduce predictive performance metrics into Partial Least Squares (PLS) path modeling have been slow and fall short of demonstrating impact on both practice and scientific development in PLS. This study contributes to PLS development by offering a comprehensive framework that identifies different dimensions of prediction and their effect on predictive performance evaluation with PLS. We use this framework to contextualize prior efforts in PLS and prediction, and to highlight potential opportunities and challenges. Our second contribution is to formally propose procedures to generate and evaluate different types of predictions from PLS models. The procedures account for the best practices identified in our framework. Our third contribution is outlining the many powerful ways in which predictive PLS methodologies can strengthen theory­ building research. Our framework, procedures, and research guidelines will hopefully form the basis for a more informed and unified development of the rigorous theoretical and practical applications of PLS.

A new method for assessment of bunch compactness using automated image analysis

Background and Aims Bunch compactness is a key feature determining grape and wine composition because tight bunches show a less homogeneous ripening, and are prone to greater fungal disease incidence. The Organisation Internationale de la Vigne et du Vin descriptor, the most recent method for the assessment of bunch compactness, requires visual inspection and trained evaluators, and provides subjective and qualitative values. The aim of this work was to develop a methodology based on image analysis to determine bunch compactness in a non-invasive, objective and quantitative way. Methods and Results Ninety bunches of nine different red cultivars of Vitis vinifera L. were photographed with a colour camera, and their bunch compactness was determined by visual inspection. A predictive partial least squares (PLS) model was developed in order to estimate bunch compactness from the morphological features extracted by automated image analysis, after the supervised segmentation of the images. The PLS model showed a capability of 85.3% for predicting correctly the rating of bunch compactness. The most discriminant variables of the model were highly correlated with the tightness of the berries in the bunch (proportion of visibility of berries, rachis and holes) and with the shape of the bunch (roundness, compactness shape factor and aspect ratio). Conclusions The non-invasive, image analysis methodology presented here enables the quantitative assessment of bunch compactness, thereby providing precise objective information for this key parameter. Significance of the Study A quantitative, objective and accurate system based on image analysis was developed as an alternative to current visual methods for the estimation of bunch compactness. This novel method could be applied to the classification of table grapes and/or at the receival point of wineries for sorting and assessment of wine grapes before vinification.

A Fourier Transform Infrared Spectroscopy Method for Analysis of Palm Oil Adulterated with Lard in Pre-Fried French fries

Fourier transform infrared spectroscopy with attenuated total reflectance accessory was used to detect the presence of lard in French fries pre-fried in palm oil adulterated with lard. A Fourier transform infrared calibration model was obtained using partial least squares for prediction of lard in a blend mixture of lard and palm oil. The coefficient of determination (R2) of 0.9791 was obtained with 0.5% of detection limit. The error in calibration expressed with root mean square error of calibration was 0.979%. In addition, the error obtained during cross validation was 2.45%. A discriminant analysis test was able to distinguish between fries samples adulterated with lard and samples, which were pre-fried with palm oils. Fourier transform infrared spectroscopy is a fast and powerful technique for quantification of lard present in French fries.

Calibration of artificial neural network and partial least squares regression models for the prediction of secoisolariciresinol diglucoside contents in microwave-assisted extracts of various flaxseed (Linum usitatissimum L.) samples

Abstract Partial least squares (PLS) and artificial neural network (ANN) regression models were calibrated for predicting the content of secoisolariciresinol diglucoside (SDG) in six flaxseed cultivars, defatted flaxseed meal and flax hulls extracts. The SDG was quantified by HPLC after microwave-assisted extraction (MAE) from flaxseed; the data were used in conjunction with the light absorption of the extracts measured after Folin–Ciocalteu’s assay at 289, 298, 343 and 765 nm, in order to calibrate the predictive PLS and ANN models. The accuracy and the predictive ability of the models ranged from good to excellent as indicated by RPD values (the ratio of the standard deviation of the reference values to the standard error of prediction) of 5.03–13.7. The PLS and ANN predictive models are useful to the flaxseed processing industry for rapidly and accurately predicting the SDG contents of various flaxseed samples based on their UV–Vis light absorption.

FTIR spectroscopy can be used as a screening tool for organic matter quality in regenerating cutover peatlands

Vegetational changes during the restoration of cutover peatlands leave a legacy in terms of the organic matter quality of the newly formed peat. Current efforts to restore peatlands at a large scale therefore require low cost and high throughput techniques to monitor the evolution of organic matter. In this study, we assessed the merits of using Fourier transform infrared (FTIR) spectra to predict the organic matter composition in peat samples at various stages of peatland regeneration from five European countries. Using predictive partial least squares (PLS) analyses, we were able to reconstruct peat C:N ratio and carbohydrate signatures with reasonable accuracy, but not the micromorphological composition of vegetation remains. Despite utilising different size fractions, both carbohydrate (<200 μm fraction) and FTIR (bulk soil) analyses report on the composition of plant cell wall constituents in the peat and therefore essentially reveal the composition of the parent vegetational material. The accuracy of the FTIR-based PLS models for C:N ratios and carbohydrate signatures was adequate to allow for their use as initial screening tools in the evaluation of the present and future organic matter composition of peat during monitoring of restoration efforts.

Quantitative Analysis of Red Wine Tannins Using Fourier-Transform Mid-Infrared Spectrometry

Tannin content and composition are critical quality components of red wines. No spectroscopic method assessing these phenols in wine has been described so far. We report here a new method using Fourier transform mid-infrared (FT-MIR) spectroscopy and chemometric techniques for the quantitative analysis of red wine tannins. Calibration models were developed using protein precipitation and phloroglucinolysis as analytical reference methods. After spectra preprocessing, six different predictive partial least-squares (PLS) models were evaluated, including the use of interval selection procedures such as iPLS and CSMWPLS. PLS regression with full-range (650-4000 cm(-1)), second derivative of the spectra and phloroglucinolysis as the reference method gave the most accurate determination for tannin concentration (RMSEC = 2.6%, RMSEP = 9.4%, r = 0.995). The prediction of the mean degree of polymerization (mDP) of the tannins also gave a reasonable prediction (RMSEC = 6.7%, RMSEP = 10.3%, r = 0.958). These results represent the first step in the development of a spectroscopic methodology for the quantification of several phenolic compounds that are critical for wine quality.

Correlating sensory attributes to gas chromatography–mass spectrometry profiles and e-nose responses using partial least squares regression analysis

Sensory scores for 15 attributes identified in soy sauce aroma by quantitative descriptive analysis were correlated with purge and trap gas chromatography–mass spectrometry (GC–MS) profiles and electronic nose (e-nose) responses using partial least squares (PLS) regression analysis. Highly predictive PLS models were obtained for every attribute based on whole GC–MS profiles. However, the predictability has been greatly improved in the models calculated from 20 selected peaks that showed higher contribution to each attribute in the first PLS analysis. Contrarily, except for alcoholic and fishy notes, predictability of PLS models calculated from e-nose responses was poor. The correlation between GC–MS profiles and e-nose responses was unsatisfactory due to high similarity in sensor responses.

Correlating sensory attributes to gas chromatography–mass spectrometry profiles and e-nose responses using partial least squares regression analysis

Sensory scores for 15 attributes identified in soy sauce aroma by quantitative descriptive analysis were correlated with purge and trap gas chromatography-mass spectrometry (GC-MS) profiles and electronic nose (e-nose) responses using partial least squares (PLS) regression analysis. Highly predictive PLS models were obtained for every attribute based on whole GC-MS profiles. However, the predictability has been greatly improved in the models calculated from 20 selected peaks that showed higher contribution to each attribute in the first PLS analysis. Contrarily, except for alcoholic and fishy notes, predictability of PLS models calculated from e-nose responses was poor. The correlation between GC-MS profiles and e-nose responses was unsatisfactory due to high similarity in sensor responses.

Rapid analysis of grape aroma glycoconjugates using Fourier-transform infrared spectrometry and chemometric techniques

The varietal aroma of most wines from non-aromatic grapes is partly dependent on the qualitative and quantitative composition of glycosidic precursors in grapes. The only rapid method to assess these glycoconjugates (Red-Free Glycosyl–Glucose (G–G) method) allows only their total quantitation. We developed a new method using Fourier-transform infrared spectrometry (FT-IR) and chemometric techniques allowing these glycosidic precursors to be determined more finely. Vitis vinifera cv. Melon B. grapes grown in different areas of Muscadet vineyard (Northwest France) and harvested at different maturity stages were used to demonstrate the potentiality of this analysis method. Predictive partial least squares (PLS) regressions were established using 39 samples, representative of the glycoside variability. These models allowed the total levels of C13-norisoprenoidic and monoterpenic glycoconjugates, the most relevant aroma glycoconjugates for Muscadet wines, to be determined with a predictive error of 14 and 15%, respectively, and the total levels of the other classes of glycoconjugates with predictive errors ranging from 22% for volatile phenols to 36% for aliphatic and aromatic alcohols.

Rapid analysis of grape aroma glycoconjugates using Fourier-transform infrared spectrometry and chemometric techniques

The varietal aroma of most wines from non-aromatic grapes is partly dependent on the qualitative and quantitative composition of glycosidic precursors in grapes. The only rapid method to assess these glycoconjugates (Red-Free Glycosyl–Glucose (G–G) method) allows only their total quantitation. We developed a new method using Fourier-transform infrared spectrometry (FT-IR) and chemometric techniques allowing these glycosidic precursors to be determined more finely. Vitis vinifera cv. Melon B. grapes grown in different areas of Muscadet vineyard (Northwest France) and harvested at different maturity stages were used to demonstrate the potentiality of this analysis method. Predictive partial least squares (PLS) regressions were established using 39 samples, representative of the glycoside variability. These models allowed the total levels of C13-norisoprenoidic and monoterpenic glycoconjugates, the most relevant aroma glycoconjugates for Muscadet wines, to be determined with a predictive error of 14 and 15%, respectively, and the total levels of the other classes of glycoconjugates with predictive errors ranging from 22% for volatile phenols to 36% for aliphatic and aromatic alcohols.